SCHOOL OF BUSINESS AND MANAGEMENT

CHRIST (DEEMED TO BE UNIVERSITY),

BENGALURU

Data Mining - CIA 3

A comparative analysis of association Rule
Mining techniques: Support and confidence
vs. FP growth Algorithm

Submitted by : Anuj Golchha (2324406)

Class : 3BBA FMA

Section : B
Under the supervision of
Dr Dhanalakshmi K
 DATASETS

i) Numerical Dataset

ii) Text Dataset

FP-Growth (Frequent Pattern Growth) and Association Rules are two key techniques in data
mining, often used to discover patterns and relationships in large datasets. Here's a brief
overview of both:

1. FP-Growth (Frequent Pattern Growth Algorithm)

FP-Growth is an efficient algorithm for mining frequent item sets without candidate
generation. It is often used to identify frequent item sets in transaction datasets and is faster
than Apriori due to its ability to compress the data.

How FP-Growth works:

• Step 1: Create the FP-Tree:

o The dataset is scanned to count the frequency of each item.
o Items are then sorted in descending order of their frequency, and a compact
tree structure, the FP-Tree, is built. Each branch of the tree represents a pattern
of items from the dataset.
• Step 2: Mine the FP-Tree:
o The tree is recursively divided into conditional FP-Trees, focusing on frequent
patterns (item sets) that occur together.
o By traversing this tree, frequent patterns are discovered without needing to
generate all possible item sets.

Benefits:

• Requires fewer database scans.

• Avoids generating candidate itemsets.
• Compact representation of transactions using FP-Tree.

2. Association Rules

Association rules are used to discover interesting relationships or patterns between items in
transactional data. These rules take the form of “if-then” statements (e.g., if a customer buys
item X, they are likely to buy item Y).

Key terms:

• Support: The frequency with which items appear together in a dataset. It is used to
identify how often a rule applies to the data.
• Confidence: The likelihood that a rule is true. It measures how often the consequent
(Y) appears in transactions containing the antecedent (X).
• Lift: The ratio of the observed support to that expected if the two items were
independent. Lift helps identify how much more likely the consequent is given the
antecedent.

Example: For a grocery store dataset:

• Rule: {Bread} → {Butter}

o Support: 30% (30% of transactions contain both bread and butter).
 o Confidence: 80% (80% of the transactions that include bread also include
butter).
o Lift: 1.5 (Customers who buy bread are 1.5 times more likely to buy butter
than by random chance).

FP-Growth and Association Rules Workflow:

1. Find Frequent Itemsets: Use the FP-Growth algorithm to find item sets that occur
frequently in the dataset.
2. Generate Association Rules: From the frequent item sets, create association rules by
calculating support, confidence, and lift.
Detailed Steps to Use FP-Growth in RapidMiner (For Numerical Data)

Here’s a step-by-step guide to implementing FP-Growth with numerical data in RapidMiner:

Step 1: Import Data into RapidMiner

• Start by loading your dataset into RapidMiner.

o Drag and drop the data file into the Design view, or use the Read CSV
operator if you're working with a CSV file.
o Connect the Read CSV operator to the Process panel to load the data.

Step 2: Add the “Select Attributes” Operator

• To focus on specific attributes, drag the Select Attributes operator into your process.
o Configure it to select only the columns relevant to your analysis (like item IDs
or transaction-related data).
o This step ensures you're only working with necessary attributes.

Step 3: Add the “Set Role” Operator

• Include the Set Role operator in the process to designate roles for your selected
attributes.
o Set the appropriate columns as ID or label depending on whether they're
identifiers or target variables.

Step 4: Add the “Numerical to Binomial” Operator

• Since FP-Growth requires categorical (binomial) data, drag the Numerical to

Binomial operator into the process.
o Configure it to convert relevant numerical columns into binary values (e.g.,
whether an item was purchased in a transaction or not).

Step 5: Add the “FP-Growth” Operator

• Now, add the FP-Growth operator to extract frequent itemsets from the processed
data.
o Adjust the settings, such as minimum support, to control how frequently an
itemset must appear to be considered.

Step 6: Add the “Create Association Rules” Operator

• Finally, add the Create Association Rules operator to generate association rules from
the frequent itemsets.
o Set the minimum confidence level, which determines how strong the rules
should be, and link this operator to the FP-Growth results.

After completing these steps, run the process to discover frequent patterns and generate
association rules based on your numerical dataset.
• Importing data into RapidMiner
• Add “Select Attributes” operator to the spline
• Add “Set Role” operator to the spline
• Add “Numerical to Binomial” operator to the spline

• Add “FP-Growth” to the spline

• Add “Create Association Rules” to the spline
RESULT

interpretation of the FP-Growth results

1. Frequent Itemsets: The frequent itemsets represent attributes that frequently appear
together in the dataset.
2. Support Value (1.000): All itemsets have a support of 1.000, meaning these items or
combinations of items occur in 100% of the dataset transactions.
3. Single Item Interpretation:
o Attributes like "Attitude," "Employee Engagement," "Moral Support," and
"Attrition" individually appear in all dataset entries.
4. Pairwise Combinations:
o Combinations like {Attitude, Attrition} or {Employee Engagement, Employee
Resilience} always occur together in every record, indicating strong
associations between these attributes.
5. Strong Association: Since all itemsets have a support of 1.000, there is a perfect co-
occurrence of these attributes in the dataset, meaning these relationships are very
significant and reliable.
Interpretation of the key elements

1. Graphical Representation:
o Each box labeled as "Rule X" (e.g., Rule 25, Rule 16) represents an
association rule discovered during the mining process. These rules show
relationships between different variables within a dataset.
o The numbers in parentheses (e.g., 1.000 / 1.000) represent the support and
confidence metrics. In this case, they both show 1.000, meaning that the rule
has full support (it applies to all transactions or data points) and perfect
confidence (every instance of the antecedent leads to the consequent).
2. Nodes and Connections:
o The large label nodes such as "Attrition", "Reporting Manager",
"Performance Appraisal", etc., represent the consequent of rules. The rules
connect the antecedent (other variables that lead to this consequence) to these
main nodes.
o The connections between nodes indicate how one concept, such as Employee
Engagement or Incentives, is related to outcomes like Attrition or Moral
Support.
3. Filter Panel:
o The panel on the left allows for filtering of the association rules. You can
narrow down the rules by selecting specific conclusions like Attrition,
Performance Appraisal, etc., or by setting a minimum confidence value.
4. Rule Criteria:
o The algorithm used the confidence measure to evaluate the strength of the
rules. A confidence of 1 means that the association rules are very strong,
indicating that whenever the antecedent occurs, the consequent also happens.

In summary, the visualization displays the relationships between various factors like
Attitude, Employee Engagement, and Attrition through high-confidence association rules.
You can use these insights to understand how different employee-related variables correlate,
such as how changes in employee engagement or management structure may predict attrition
or performance issues

ASSOCIATION RULES:
Association Rules
[Attitude] --> [Attrition] (confidence: 1.000)
[Attrition] --> [Attitude] (confidence: 1.000)
[Attitude] --> [Employee Engagement] (confidence: 1.000)
[Employee Engagement] --> [Attitude] (confidence: 1.000)
[Attitude] --> [EmployeeResilence] (confidence: 1.000)
[EmployeeResilence] --> [Attitude] (confidence: 1.000)
[Attitude] --> [Incentives] (confidence: 1.000)
[Incentives] --> [Attitude] (confidence: 1.000)
[Attitude] --> [Moral Support] (confidence: 1.000)
[Moral Support] --> [Attitude] (confidence: 1.000)
[Attitude] --> [Performance Appriasal] (confidence: 1.000)
[Performance Appriasal] --> [Attitude] (confidence: 1.000)
[Attitude] --> [Reporting Manager] (confidence: 1.000)
[Reporting Manager] --> [Attitude] (confidence: 1.000)
[Attrition] --> [Employee Engagement] (confidence: 1.000)
[Employee Engagement] --> [Attrition] (confidence: 1.000)
[Attrition] --> [EmployeeResilence] (confidence: 1.000)
[EmployeeResilence] --> [Attrition] (confidence: 1.000)
[Attrition] --> [Incentives] (confidence: 1.000)
[Incentives] --> [Attrition] (confidence: 1.000)
[Attrition] --> [Moral Support] (confidence: 1.000)
[Moral Support] --> [Attrition] (confidence: 1.000)
[Attrition] --> [Performance Appriasal] (confidence: 1.000)
[Performance Appriasal] --> [Attrition] (confidence: 1.000)
[Attrition] --> [Reporting Manager] (confidence: 1.000)
[Reporting Manager] --> [Attrition] (confidence: 1.000)
[Employee Engagement] --> [EmployeeResilence] (confidence: 1.000)
[EmployeeResilence] --> [Employee Engagement] (confidence: 1.000)
[Employee Engagement] --> [Incentives] (confidence: 1.000)
[Incentives] --> [Employee Engagement] (confidence: 1.000)
[Employee Engagement] --> [Moral Support] (confidence: 1.000)
[Moral Support] --> [Employee Engagement] (confidence: 1.000)
[Employee Engagement] --> [Performance Appriasal] (confidence: 1.000)
[Performance Appriasal] --> [Employee Engagement] (confidence: 1.000)
[Employee Engagement] --> [Reporting Manager] (confidence: 1.000)
[Reporting Manager] --> [Employee Engagement] (confidence: 1.000)
[EmployeeResilence] --> [Incentives] (confidence: 1.000)
[Incentives] --> [EmployeeResilence] (confidence: 1.000)
[EmployeeResilence] --> [Moral Support] (confidence: 1.000)
[Moral Support] --> [EmployeeResilence] (confidence: 1.000)
[EmployeeResilence] --> [Performance Appriasal] (confidence: 1.000)
[Performance Appriasal] --> [EmployeeResilence] (confidence: 1.000)
[EmployeeResilence] --> [Reporting Manager] (confidence: 1.000)
[Reporting Manager] --> [EmployeeResilence] (confidence: 1.000)
[Incentives] --> [Moral Support] (confidence: 1.000)
[Moral Support] --> [Incentives] (confidence: 1.000)
[Incentives] --> [Performance Appriasal] (confidence: 1.000)
[Performance Appriasal] --> [Incentives] (confidence: 1.000)
[Incentives] --> [Reporting Manager] (confidence: 1.000)
[Reporting Manager] --> [Incentives] (confidence: 1.000)
[Moral Support] --> [Performance Appriasal] (confidence: 1.000)
[Performance Appriasal] --> [Moral Support] (confidence: 1.000)
[Moral Support] --> [Reporting Manager] (confidence: 1.000)
[Reporting Manager] --> [Moral Support] (confidence: 1.000)
[Performance Appriasal] --> [Reporting Manager] (confidence: 1.000)
[Reporting Manager] --> [Performance Appriasal] (confidence: 1.000)
[Attitude] --> [Attrition, Employee Engagement] (confidence: 1.000)
[Attrition] --> [Attitude, Employee Engagement] (confidence: 1.000)
[Attitude, Attrition] --> [Employee Engagement] (confidence: 1.000)
[Employee Engagement] --> [Attitude, Attrition] (confidence: 1.000)
[Attitude, Employee Engagement] --> [Attrition] (confidence: 1.000)
[Attrition, Employee Engagement] --> [Attitude] (confidence: 1.000)
[Attitude] --> [Attrition, EmployeeResilence] (confidence: 1.000)
[Attrition] --> [Attitude, EmployeeResilence] (confidence: 1.000)
[Attitude, Attrition] --> [EmployeeResilence] (confidence: 1.000)
[EmployeeResilence] --> [Attitude, Attrition] (confidence: 1.000)
[Attitude, EmployeeResilence] --> [Attrition] (confidence: 1.000)
[Attrition, EmployeeResilence] --> [Attitude] (confidence: 1.000)
[Attitude] --> [Attrition, Incentives] (confidence: 1.000)
[Attrition] --> [Attitude, Incentives] (confidence: 1.000)
[Attitude, Attrition] --> [Incentives] (confidence: 1.000)
[Incentives] --> [Attitude, Attrition] (confidence: 1.000)
[Attitude, Incentives] --> [Attrition] (confidence: 1.000)
[Attrition, Incentives] --> [Attitude] (confidence: 1.000)
[Attitude] --> [Attrition, Moral Support] (confidence: 1.000)
[Attrition] --> [Attitude, Moral Support] (confidence: 1.000)
[Attitude, Attrition] --> [Moral Support] (confidence: 1.000)
[Moral Support] --> [Attitude, Attrition] (confidence: 1.000)
[Attitude, Moral Support] --> [Attrition] (confidence: 1.000)
[Attrition, Moral Support] --> [Attitude] (confidence: 1.000)
[Attitude] --> [Attrition, Performance Appriasal] (confidence: 1.000)
[Attrition] --> [Attitude, Performance Appriasal] (confidence: 1.000)
[Attitude, Attrition] --> [Performance Appriasal] (confidence: 1.000)
[Performance Appriasal] --> [Attitude, Attrition] (confidence: 1.000)
[Attitude, Performance Appriasal] --> [Attrition] (confidence: 1.000)
[Attrition, Performance Appriasal] --> [Attitude] (confidence: 1.000)
[Attitude] --> [Attrition, Reporting Manager] (confidence: 1.000)
[Attrition] --> [Attitude, Reporting Manager] (confidence: 1.000)
[Attitude, Attrition] --> [Reporting Manager] (confidence: 1.000)
[Reporting Manager] --> [Attitude, Attrition] (confidence: 1.000)
[Attitude, Reporting Manager] --> [Attrition] (confidence: 1.000)
[Attrition, Reporting Manager] --> [Attitude] (confidence: 1.000)
[Attitude] --> [Employee Engagement, EmployeeResilence] (confidence: 1.000)
[Employee Engagement] --> [Attitude, EmployeeResilence] (confidence: 1.000)
[Attitude, Employee Engagement] --> [EmployeeResilence] (confidence: 1.000)
[EmployeeResilence] --> [Attitude, Employee Engagement] (confidence: 1.000)
[Attitude, EmployeeResilence] --> [Employee Engagement] (confidence: 1.000)
[Employee Engagement, EmployeeResilence] --> [Attitude] (confidence: 1.000)
[Attitude] --> [Employee Engagement, Incentives] (confidence: 1.000)
[Employee Engagement] --> [Attitude, Incentives] (confidence: 1.000)
[Attitude, Employee Engagement] --> [Incentives] (confidence: 1.000)
... 5950 other rules ...
SUPPORT LEVELAND CONFIDENCE LEVEL

Interpretation:

• Strong Association Rules: Since all the rules have a support and confidence of 1, it
implies that the conditions (premises) always lead to the outcomes (conclusions) in
the dataset. This suggests very strong relationships between the premises and
conclusions.
• Rule Examples:
o Rule 1: When the premise is Attitude, the conclusion is Attrition. This means
that certain attitudes are fully associated with employee attrition.
o Rule 3: Attitude leads to Employee Engagement, indicating that employees'
attitudes are strongly connected with their engagement levels.
o Rule 5: Attitude again leads to Employee Resilience, suggesting a close link
between attitudes and how resilient employees are in the workplace.
• Overall Pattern: The rules show that Attitude as a premise is linked to a variety of
key outcomes, such as Attrition, Employee Engagement, Incentives, Moral
Support, and more. This could indicate that employee attitude plays a central role in
multiple aspects of organizational behavior.
Detailed Steps to Use FP-Growth in RapidMiner (For Text Data)

Here’s a step-by-step guide to implementing FP-Growth with numerical data in RapidMiner:

Step 1: Import Data into RapidMiner

• Start by loading your dataset into RapidMiner.

o Drag and drop the data file into the Design view or use the Read CSV
operator if you're working with a CSV file.
o Connect the Read CSV operator to the Process panel to load the data.

Step 2: Add the “Select Attributes” Operator

• To focus on specific attributes, drag the Select Attributes operator into your process.
o Configure it to select only the columns relevant to your analysis (like item IDs
or transaction-related data).
o This step ensures you're only working with necessary attributes.

Step 3: Add the “Set Role” Operator

• Include the Set Role operator in the process to designate roles for your selected
attributes.
o Set the appropriate columns as ID or label depending on whether they're
identifiers or target variables.

Step 4: Add the “Numerical to Binomial” Operator

• Since FP-Growth requires categorical (binomial) data, drag the Numerical to

Binomial operator into the process.
o Configure it to convert relevant numerical columns into binary values (e.g.,
whether an item was purchased in a transaction or not).

Step 5: Add the “FP-Growth” Operator

• Now, add the FP-Growth operator to extract frequent itemsets from the processed
data.
o Adjust the settings, such as minimum support, to control how frequently an
itemset must appear to be considered.

Step 6: Add the “Create Association Rules” Operator

• Finally, add the Create Association Rules operator to generate association rules from
the frequent itemsets.
o Set the minimum confidence level, which determines how strong the rules
should be, and link this operator to the FP-Growth results.

After completing these steps, run the process to discover frequent patterns and generate
association rules based on your numerical dataset.
Detailed Steps to Use FP-Growth in RapidMiner (For Textual Data)

Here’s a step-by-step guide to implementing FP-Growth with numerical data in RapidMiner:

Step 1: Import Data into RapidMiner

• Start by loading your dataset into RapidMiner.

o Drag and drop the data file into the Design view, or use the Read CSV
operator if you're working with a CSV file.
o Connect the Read CSV operator to the Process panel to load the data.

Step 2: Add the “Select Attributes” Operator

• To focus on specific attributes, drag the Select Attributes operator into your process.
o Configure it to select only the columns relevant to your analysis (like item IDs
or transaction-related data).
o This step ensures you're only working with necessary attributes.

Step 3: Add the “Set Role” Operator

• Include the Set Role operator in the process to designate roles for your selected
attributes.
o Set the appropriate columns as ID or label depending on whether they're
identifiers or target variables.

Step 4: Add the “Numerical to Binomial” Operator

• Since FP-Growth requires categorical (binomial) data, drag the Numerical to

Binomial operator into the process.
o Configure it to convert relevant numerical columns into binary values (e.g.,
whether an item was purchased in a transaction or not).

Step 5: Add the “FP-Growth” Operator

• Now, add the FP-Growth operator to extract frequent itemsets from the processed
data.
o Adjust the settings, such as minimum support, to control how frequently an
itemset must appear to be considered. (set minimum support as 0.3 (30%))

Step 6: Add the “Create Association Rules” Operator

• Finally, add the Create Association Rules operator to generate association rules from
the frequent itemsets.
o Set the minimum confidence level, which determines how strong the rules
should be, and link this operator to the FP-Growth results. set minimum
confidence as 0.7 (70%))

After completing these steps, run the process to discover frequent patterns and generate
association rules based on your Textual Data dataset.
• Import data into RapidMiner
Add “Select Attributes” to the spline
• Add “Set Role” to the spline
• Add “Numerical to Binomial” operator to the spline

• Add “FP Growth” operator to the spline, set minimum support as 0.3 (30%)
• Add “Create Association” operator to the spline, set minimum confidence as 0.7
(70%) and connect it in the following way
 Result

Interpretation

• Frequent Itemsets:

• The FP-Growth algorithm identifies frequent itemsets that occur together across
transactions. These frequent itemsets are essentially groups of products that are often
bought together.
• For example, we may discover itemsets like:
o Laptop and USB Drive: Customers who buy a laptop are highly likely to buy
a USB drive.
o Smartphone and External Hard Drive: Customers purchasing a Smartphone
might also purchase an External Hard Drive, possibly for storage purposes.
o Smart TV and Bluetooth Speaker: The Smart TV and Bluetooth Speaker
pair might indicate that customers upgrading their TV often purchase
complementary audio equipment.
interpretation of the FP-Growth results

1. Frequent Itemsets: The frequent itemsets represent attributes that frequently appear
together in the dataset.
2. Support Value (1.000): All itemsets have a support of 1.000, meaning these items or
combinations of items occur in 100% of the dataset transactions.
3. Single Item Interpretation:
o Attributes like "Attitude," "Employee Engagement," "Moral Support," and
"Attrition" individually appear in all dataset entries.
4. Pairwise Combinations:
o Combinations like {Attitude, Attrition} or {Employee Engagement, Employee
Resilience} always occur together in every record, indicating strong
associations between these attributes.
5. Strong Association: Since all itemsets have a support of 1.000, there is a perfect co-
occurrence of these attributes in the dataset, meaning these relationships are very
significant and reliable.
SUPPORT LEVELAND CONFIDENCE LEVEL

Association Rules

[Price per Unit] --> [Quantity] (confidence: 1.000)

[Quantity] --> [Price per Unit] (confidence: 1.000)
[Price per Unit] --> [Shipping Cost] (confidence: 1.000)
[Shipping Cost] --> [Price per Unit] (confidence: 1.000)
[Price per Unit] --> [Total Amount] (confidence: 1.000)
[Total Amount] --> [Price per Unit] (confidence: 1.000)
[Price per Unit] --> [Transaction ID] (confidence: 1.000)
[Transaction ID] --> [Price per Unit] (confidence: 1.000)
[Discount Applied] --> [Price per Unit] (confidence: 1.000)
[Quantity] --> [Shipping Cost] (confidence: 1.000)
[Shipping Cost] --> [Quantity] (confidence: 1.000)
[Quantity] --> [Total Amount] (confidence: 1.000)
[Total Amount] --> [Quantity] (confidence: 1.000)
[Quantity] --> [Transaction ID] (confidence: 1.000)
[Transaction ID] --> [Quantity] (confidence: 1.000)
[Discount Applied] --> [Quantity] (confidence: 1.000)
[Shipping Cost] --> [Total Amount] (confidence: 1.000)
[Total Amount] --> [Shipping Cost] (confidence: 1.000)
[Shipping Cost] --> [Transaction ID] (confidence: 1.000)
[Transaction ID] --> [Shipping Cost] (confidence: 1.000)
[Discount Applied] --> [Shipping Cost] (confidence: 1.000)
[Total Amount] --> [Transaction ID] (confidence: 1.000)
[Transaction ID] --> [Total Amount] (confidence: 1.000)
[Discount Applied] --> [Total Amount] (confidence: 1.000)
[Discount Applied] --> [Transaction ID] (confidence: 1.000)
[Price per Unit] --> [Quantity, Shipping Cost] (confidence: 1.000)
[Quantity] --> [Price per Unit, Shipping Cost] (confidence: 1.000)
[Price per Unit, Quantity] --> [Shipping Cost] (confidence: 1.000)
[Shipping Cost] --> [Price per Unit, Quantity] (confidence: 1.000)
[Price per Unit, Shipping Cost] --> [Quantity] (confidence: 1.000)
[Quantity, Shipping Cost] --> [Price per Unit] (confidence: 1.000)
[Price per Unit] --> [Quantity, Total Amount] (confidence: 1.000)
[Quantity] --> [Price per Unit, Total Amount] (confidence: 1.000)
[Price per Unit, Quantity] --> [Total Amount] (confidence: 1.000)
[Total Amount] --> [Price per Unit, Quantity] (confidence: 1.000)
[Price per Unit, Total Amount] --> [Quantity] (confidence: 1.000)
[Quantity, Total Amount] --> [Price per Unit] (confidence: 1.000)
[Price per Unit] --> [Quantity, Transaction ID] (confidence: 1.000)
[Quantity] --> [Price per Unit, Transaction ID] (confidence: 1.000)
[Price per Unit, Quantity] --> [Transaction ID] (confidence: 1.000)
[Transaction ID] --> [Price per Unit, Quantity] (confidence: 1.000)
[Price per Unit, Transaction ID] --> [Quantity] (confidence: 1.000)
[Quantity, Transaction ID] --> [Price per Unit] (confidence: 1.000)
[Discount Applied] --> [Price per Unit, Quantity] (confidence: 1.000)
[Price per Unit, Discount Applied] --> [Quantity] (confidence: 1.000)
[Quantity, Discount Applied] --> [Price per Unit] (confidence: 1.000)
[Price per Unit] --> [Shipping Cost, Total Amount] (confidence: 1.000)
[Shipping Cost] --> [Price per Unit, Total Amount] (confidence: 1.000)
[Price per Unit, Shipping Cost] --> [Total Amount] (confidence: 1.000)
[Total Amount] --> [Price per Unit, Shipping Cost] (confidence: 1.000)
[Price per Unit, Total Amount] --> [Shipping Cost] (confidence: 1.000)
[Shipping Cost, Total Amount] --> [Price per Unit] (confidence: 1.000)
[Price per Unit] --> [Shipping Cost, Transaction ID] (confidence: 1.000)
[Shipping Cost] --> [Price per Unit, Transaction ID] (confidence: 1.000)
[Price per Unit, Shipping Cost] --> [Transaction ID] (confidence: 1.000)
[Transaction ID] --> [Price per Unit, Shipping Cost] (confidence: 1.000)
[Price per Unit, Transaction ID] --> [Shipping Cost] (confidence: 1.000)
[Shipping Cost, Transaction ID] --> [Price per Unit] (confidence: 1.000)
[Discount Applied] --> [Price per Unit, Shipping Cost] (confidence: 1.000)
[Price per Unit, Discount Applied] --> [Shipping Cost] (confidence: 1.000)
[Shipping Cost, Discount Applied] --> [Price per Unit] (confidence: 1.000)
[Price per Unit] --> [Total Amount, Transaction ID] (confidence: 1.000)
[Total Amount] --> [Price per Unit, Transaction ID] (confidence: 1.000)
[Price per Unit, Total Amount] --> [Transaction ID] (confidence: 1.000)
[Transaction ID] --> [Price per Unit, Total Amount] (confidence: 1.000)
[Price per Unit, Transaction ID] --> [Total Amount] (confidence: 1.000)
[Total Amount, Transaction ID] --> [Price per Unit] (confidence: 1.000)
[Discount Applied] --> [Price per Unit, Total Amount] (confidence: 1.000)
[Price per Unit, Discount Applied] --> [Total Amount] (confidence: 1.000)
[Total Amount, Discount Applied] --> [Price per Unit] (confidence: 1.000)
[Discount Applied] --> [Price per Unit, Transaction ID] (confidence: 1.000)
[Price per Unit, Discount Applied] --> [Transaction ID] (confidence: 1.000)
[Transaction ID, Discount Applied] --> [Price per Unit] (confidence: 1.000)
[Quantity] --> [Shipping Cost, Total Amount] (confidence: 1.000)
[Shipping Cost] --> [Quantity, Total Amount] (confidence: 1.000)
[Quantity, Shipping Cost] --> [Total Amount] (confidence: 1.000)
[Total Amount] --> [Quantity, Shipping Cost] (confidence: 1.000)
[Quantity, Total Amount] --> [Shipping Cost] (confidence: 1.000)
[Shipping Cost, Total Amount] --> [Quantity] (confidence: 1.000)
[Quantity] --> [Shipping Cost, Transaction ID] (confidence: 1.000)
[Shipping Cost] --> [Quantity, Transaction ID] (confidence: 1.000)
[Quantity, Shipping Cost] --> [Transaction ID] (confidence: 1.000)
[Transaction ID] --> [Quantity, Shipping Cost] (confidence: 1.000)
[Quantity, Transaction ID] --> [Shipping Cost] (confidence: 1.000)
[Shipping Cost, Transaction ID] --> [Quantity] (confidence: 1.000)
[Discount Applied] --> [Quantity, Shipping Cost] (confidence: 1.000)
[Quantity, Discount Applied] --> [Shipping Cost] (confidence: 1.000)
[Shipping Cost, Discount Applied] --> [Quantity] (confidence: 1.000)
[Quantity] --> [Total Amount, Transaction ID] (confidence: 1.000)
[Total Amount] --> [Quantity, Transaction ID] (confidence: 1.000)
[Quantity, Total Amount] --> [Transaction ID] (confidence: 1.000)
[Transaction ID] --> [Quantity, Total Amount] (confidence: 1.000)
[Quantity, Transaction ID] --> [Total Amount] (confidence: 1.000)
[Total Amount, Transaction ID] --> [Quantity] (confidence: 1.000)
[Discount Applied] --> [Quantity, Total Amount] (confidence: 1.000)
[Quantity, Discount Applied] --> [Total Amount] (confidence: 1.000)
[Total Amount, Discount Applied] --> [Quantity] (confidence: 1.000)
[Discount Applied] --> [Quantity, Transaction ID] (confidence: 1.000)
[Quantity, Discount Applied] --> [Transaction ID] (confidence: 1.000)
[Transaction ID, Discount Applied] --> [Quantity] (confidence: 1.000)
[Shipping Cost] --> [Total Amount, Transaction ID] (confidence: 1.000)
... 291 other rules ...
COMPARISON
1. Comparing different confidence levels
i) when the confidence level is 0.5 (50%)

Results
Interpretation

General Interpretation:

• Support for the Rule: When the confidence level is 0.5, the conclusion (consequent)
happens 50% of the time whenever the premise (antecedent) is true. In other words,
there is an equal chance that the rule holds and doesn't hold. The relationship is
moderate, meaning that while there is some association between the premise and the
conclusion, it's not strong enough to make confident predictions or major strategic
decisions.
• Uncertainty in Relationship: The fact that only half the instances follow the rule
implies that there may be several other factors at play that influence the outcome.
Therefore, the relationship suggested by this rule needs further investigation. It’s
possible that the premise is only one of many conditions leading to the conclusion,
and this confidence level indicates a higher degree of variability.

Specific Example Interpretation:

Let’s say you have the rule Attitude → Employee Attrition with a confidence of 0.5. This
means:

• 50% Reliability: In half the cases where there is a specific attitude (perhaps
dissatisfaction, disengagement, or negativity), attrition follows. This would mean that
while attitude is a factor influencing employees leaving, it is not always the case.
Other factors—such as salary, workplace conditions, leadership, or personal
reasons—may also play an important role in influencing whether an employee stays
or leaves.
• Business Decision Impact: At this level of confidence, you wouldn’t want to
implement any significant policies based solely on this rule. However, it suggests that
monitoring employee attitudes is still important, though you should be cautious
about over-relying on this factor for predicting or preventing attrition. You might use
this insight to initiate further qualitative or quantitative research to explore what other
variables contribute to attrition when attitudes don’t lead to turnover.

Actionable Insights:

• Exploration Required: Rules with 0.5 confidence are good for hypothesis generation
rather than decisive action. For instance, you could start tracking attitudes more
carefully, but you would also need to investigate what additional factors are
influencing attrition to gain a more complete picture.
ii)When confidence Level is 0.7(70%)

Results
Interpretation

Confidence 0.7: Moderate-High Confidence

General Interpretation:

• Stronger Support: With a confidence of 0.7, the conclusion follows the premise 70%
of the time. This indicates a much stronger association compared to a 0.5 confidence
level. While it still isn’t an absolute relationship, the premise is a fairly reliable
indicator of the outcome. There is still some room for exceptions, but the premise is a
significant predictor.
• Reliable Association: At this confidence level, you can start to see that the premise
has a notable influence on the outcome. While it doesn’t hold in all cases, it holds
enough that you can make fairly reliable inferences. This means that in 70% of
situations, the premise contributes strongly to the conclusion, indicating that this is an
actionable insight for business decisions.

Specific Example Interpretation:

For the rule Attitude → Employee Engagement with a confidence of 0.7:

• 70% Reliability: In 70% of the cases, when employees have a certain attitude
(positive, motivated, or satisfied), they tend to be more engaged in their work. This
suggests that attitude is a key driver of engagement, though there is still a 30% chance
that employees might be engaged for other reasons—such as meaningful work, team
culture, or career growth opportunities.
• Business Decision Impact: At this level, you can confidently focus on improving
employee attitudes to increase engagement. Since this relationship holds true for
most cases, it could be worthwhile to implement programs aimed at fostering positive
attitudes—such as recognition programs, wellness initiatives, or training on emotional
intelligence—to increase engagement levels. However, you would still need to
monitor other factors that could impact engagement.

Actionable Insights:

• Strategic Initiatives: Rules with a confidence of 0.7 are actionable but still leave
room for exceptions. You might decide to prioritize attitude-related interventions as a
central strategy for improving employee engagement, while also exploring what other
aspects (work-life balance, leadership) might explain engagement in the remaining
30% of cases.
iii)When Confidence Level is 0.9

Results
Interpretation

Confidence 0.9: High Confidence

General Interpretation:

• Very Strong Association: When the confidence is 0.9, the relationship between the
premise and the conclusion is very strong, meaning that the conclusion happens 90%
of the time when the premise is true. This suggests a near-deterministic relationship,
where the premise almost always leads to the outcome. Only in very few instances
(10%) does the premise not lead to the conclusion, making these rules highly reliable
for predictive purposes.
• Predictive Power: A confidence of 0.9 means you can confidently predict the
outcome if the premise holds. There’s very little uncertainty here, which means that
the premise is almost certainly a dominant factor influencing the conclusion. The
relationship is strong enough that you can make strategic decisions and interventions
with confidence.

Specific Example Interpretation:

Let’s consider the rule Attitude → Employee Resilience with a confidence of 0.9:

• 90% Reliability: In 90% of the cases, when employees have a certain attitude (likely
positive, adaptive, or growth-oriented), they also exhibit resilience in the workplace.
This means attitude is a key determinant of resilience, making it a crucial focus for
organizations looking to foster a resilient workforce.
• Business Decision Impact: This is a highly reliable relationship. At this level, you
should prioritize programs that cultivate positive attitudes as a key strategy for
enhancing resilience. Since this relationship is almost guaranteed, you can be
confident that improving attitude through leadership, coaching, or support programs
will directly improve resilience. The 10% exceptions are small enough that they can
likely be explained by unique or external factors (personal challenges, etc.).

Actionable Insights:

• Focus on Attitude: Rules with a confidence of 0.9 are critical for decision-making. In
this case, fostering a positive employee attitude would be a priority intervention to
build resilience. Since this rule holds true almost all the time, you can confidently
implement measures based on this insight, knowing that it will yield tangible results.
Let's expand the interpretation of your association rule analysis across the different
confidence levels (0.5, 0.7, and 0.9) in a detailed way, considering how each level affects the
strength of the insights and business decisions you might make. I'll take the key points and
elaborate on the implications for both the relationships in the data and potential actions based
on these confidence thresholds.

Confidence 0.5: Moderate-Low Confidence

General Interpretation:

• Support for the Rule: When the confidence level is 0.5, the conclusion (consequent)
happens 50% of the time whenever the premise (antecedent) is true. In other words,
there is an equal chance that the rule holds and doesn't hold. The relationship is
moderate, meaning that while there is some association between the premise and the
conclusion, it's not strong enough to make confident predictions or major strategic
decisions.
• Uncertainty in Relationship: The fact that only half the instances follow the rule
implies that there may be several other factors at play that influence the outcome.
Therefore, the relationship suggested by this rule needs further investigation. It’s
possible that the premise is only one of many conditions leading to the conclusion,
and this confidence level indicates a higher degree of variability.

Specific Example Interpretation:

Let’s say you have the rule Attitude → Employee Attrition with a confidence of 0.5. This
means:

• 50% Reliability: In half the cases where there is a specific attitude (perhaps
dissatisfaction, disengagement, or negativity), attrition follows. This would mean that
while attitude is a factor influencing employees leaving, it is not always the case.
Other factors—such as salary, workplace conditions, leadership, or personal
reasons—may also play an important role in influencing whether an employee stays
or leaves.
• Business Decision Impact: At this level of confidence, you wouldn’t want to
implement any significant policies based solely on this rule. However, it suggests that
monitoring employee attitudes is still important, though you should be cautious
about over-relying on this factor for predicting or preventing attrition. You might use
this insight to initiate further qualitative or quantitative research to explore what other
variables contribute to attrition when attitudes don’t lead to turnover.

Actionable Insights:

• Exploration Required: Rules with 0.5 confidence are good for hypothesis generation
rather than decisive action. For instance, you could start tracking attitudes more
carefully, but you would also need to investigate what additional factors are
influencing attrition to gain a more complete picture.
Confidence 0.7: Moderate-High Confidence

General Interpretation:

• Stronger Support: With a confidence of 0.7, the conclusion follows the premise 70%
of the time. This indicates a much stronger association compared to a 0.5 confidence
level. While it still isn’t an absolute relationship, the premise is a fairly reliable
indicator of the outcome. There is still some room for exceptions, but the premise is a
significant predictor.
• Reliable Association: At this confidence level, you can start to see that the premise
has a notable influence on the outcome. While it doesn’t hold in all cases, it holds
enough that you can make fairly reliable inferences. This means that in 70% of
situations, the premise contributes strongly to the conclusion, indicating that this is an
actionable insight for business decisions.

Specific Example Interpretation:

For the rule Attitude → Employee Engagement with a confidence of 0.7:

• 70% Reliability: In 70% of the cases, when employees have a certain attitude
(positive, motivated, or satisfied), they tend to be more engaged in their work. This
suggests that attitude is a key driver of engagement, though there is still a 30% chance
that employees might be engaged for other reasons—such as meaningful work, team
culture, or career growth opportunities.
• Business Decision Impact: At this level, you can confidently focus on improving
employee attitudes as a way to increase engagement. Since this relationship holds
true for the majority of cases, it could be worthwhile to implement programs aimed at
fostering positive attitudes—such as recognition programs, wellness initiatives, or
training on emotional intelligence—to increase engagement levels. However, you
would still need to monitor other factors that could impact engagement.

Actionable Insights:

• Strategic Initiatives: Rules with a confidence of 0.7 are actionable but still leave
room for exceptions. You might decide to prioritize attitude-related interventions as a
central strategy for improving employee engagement, while also exploring what other
aspects (work-life balance, leadership) might explain engagement in the remaining
30% of cases.

Confidence 0.9: High Confidence

General Interpretation:

• Very Strong Association: When the confidence is 0.9, the relationship between the
premise and the conclusion is very strong, meaning that the conclusion happens 90%
of the time when the premise is true. This suggests a near-deterministic relationship,
where the premise almost always leads to the outcome. Only in very few instances
 (10%) does the premise not lead to the conclusion, making these rules highly reliable
for predictive purposes.
• Predictive Power: A confidence of 0.9 means you can confidently predict the
outcome if the premise holds. There’s very little uncertainty here, which means that
the premise is almost certainly a dominant factor influencing the conclusion. The
relationship is strong enough that you can make strategic decisions and interventions
with confidence.

Specific Example Interpretation:

Let’s consider the rule Attitude → Employee Resilience with a confidence of 0.9:

• 90% Reliability: In 90% of the cases, when employees have a certain attitude (likely
positive, adaptive, or growth-oriented), they also exhibit resilience in the workplace.
This means attitude is a key determinant of resilience, making it a crucial focus for
organizations looking to foster a resilient workforce.
• Business Decision Impact: This is a highly reliable relationship. At this level, you
should prioritize programs that cultivate positive attitudes as a key strategy for
enhancing resilience. Since this relationship is almost guaranteed, you can be
confident that improving attitude through leadership, coaching, or support programs
will directly improve resilience. The 10% exceptions are small enough that they can
likely be explained by unique or external factors (personal challenges, etc.).

Actionable Insights:

• Focus on Attitude: Rules with a confidence of 0.9 are critical for decision-making. In
this case, fostering a positive employee attitude would be a priority intervention to
build resilience. Since this rule holds true almost all the time, you can confidently
implement measures based on this insight, knowing that it will yield tangible results.

Comparing the Three Confidence Levels:

1. Confidence 0.5:
o Moderate-low confidence suggests potential relationships but with significant
variability. The rules are useful for exploration and hypothesis-building but
are not strong enough to guide critical decisions. You should continue
investigating additional factors to explain the variability in outcomes.
2. Confidence 0.7:
o Moderate-high confidence provides a stronger and more actionable
relationship. While not perfect, these rules are good predictors and should be
considered in decision-making. They suggest strong trends but still require
monitoring for exceptions or additional factors.
3. Confidence 0.9:
o High confidence offers nearly guaranteed relationships between the premise
and conclusion. These rules are highly reliable and should be prioritized in
strategic planning and interventions. The small number of exceptions does not
undermine the strength of these rules, making them critical for predictive
decision-making.
Overall Conclusion:

As you analyze the rules with varying confidence levels (0.5, 0.7, and 0.9), you will see a
clear progression in the reliability of the relationships. Lower confidence levels indicate
exploratory insights, whereas higher confidence levels suggest solid trends that can guide
important business strategies. Each confidence level helps you understand how much weight
to give to the rule when considering actions, whether for employee retention, engagement, or
resilience programs.
2. Comparing different Support levels
i) When Support Level is 0.5

Results
Interpretation

Support 0.5: Moderate Support

General Interpretation:

• Frequency of Occurrence: A support level of 0.5 means that the combination of the
premise and conclusion happens in 50% of the dataset. In other words, half the
records in the dataset follow the rule, while the other half do not. This indicates that
the rule has a moderate relevance within the data. It’s not ubiquitous, but it still
occurs often enough to be considered significant.
• Implication of Moderate Frequency: At this level, the rule captures a significant
but not overwhelming portion of the dataset. This suggests that while the
relationship is important, there are also other factors in play that influence the
outcome for the remaining 50% of the cases.

Specific Example Interpretation:

Let’s take the rule Attitude → Employee Engagement with a support of 0.5:

• 50% Presence: This rule suggests that in half the cases, an employee’s attitude is
associated with their engagement level. So, attitudes such as positivity, commitment,
or enthusiasm are contributing to employee engagement, but this factor is not
universal. For the remaining half of the employees, other factors (such as team
dynamics, job satisfaction, or work environment) might play a stronger role.
• Business Decision Impact: Since this rule applies to 50% of the dataset, it is
important enough to be taken into consideration when designing engagement
strategies, but you cannot rely solely on this factor. This insight indicates that attitude
is a key driver of engagement for a large portion of employees, but there is still
room to investigate other influences.

Actionable Insights:

• Moderate Relevance: Rules with 0.5 support are moderately relevant. You should
use this information to guide engagement strategies but remain flexible and consider
other factors for the rest of the employee population. For instance, while attitude is a
major factor for half your workforce, other initiatives—like improving work-life
balance or career development—might help engage the other half.
ii) When Support Level is 0.7

Results
Interpretation

Support 0.7: High Support

General Interpretation:

• High Frequency of Occurrence: With a support level of 0.7, the rule holds in 70%
of the dataset, meaning that the premise and conclusion occur together very
frequently. This indicates a strong association that holds true for a large majority of
the dataset. Rules with high support are much more representative of the dataset and
are likely to have a broad impact.
• Implication of High Frequency: A support level of 0.7 indicates that the rule is
highly relevant and can be generalized across most of the data. This means the
relationship between the premise and the conclusion is significant enough to influence
major decisions across a wide range of cases.

Specific Example Interpretation:

For the rule Attitude → Employee Resilience with a support of 0.7:

• 70% Presence: In 70% of cases, employee attitudes (such as optimism, adaptability,

or grit) are linked to resilience. This means that attitude is a major contributing
factor to how resilient employees are, making it a critical area of focus for fostering
resilience across your workforce. Attitudes like determination and positivity are
driving resilience in the majority of employees.
• Business Decision Impact: At this level of support, you can confidently implement
attitude-building programs to enhance resilience for the vast majority of employees.
Initiatives like resilience training, mental health programs, or motivational coaching
should have a broad, meaningful impact on improving employee resilience, since the
relationship holds for most of your workforce.

Actionable Insights:

• Broad Applicability: Rules with 0.7 support are highly relevant and should be used
as the foundation for broad, organization-wide initiatives. Since the rule applies to a
large portion of employees, you can be confident that investing in positive attitude
development will have a major impact on resilience. However, for the 30% of
employees not covered by this rule, additional resilience-building factors (such as
personal support or coping strategies) may be needed.
iii) When Support Level is 0.9

Results
Support 0.9: Very High Support

General Interpretation:

• Near-Universal Occurrence: With a support level of 0.9, the rule applies to 90% of
the dataset. This means that the combination of the premise and conclusion occurs
almost universally, and the relationship is extremely representative of the entire
population. Rules with very high support are highly significant, as they apply to
nearly everyone in the dataset.
• Implication of Very High Frequency: At this level, the rule can be treated as a
nearly universal truth within the data. It indicates that the premise is almost always
associated with the conclusion, making this rule highly reliable for broad decision-
making. The few exceptions (10%) might be anomalies or specific cases where
external factors influence the outcome.

Specific Example Interpretation:

Let’s analyze the rule Attitude → Moral Support with a support of 0.9:

• 90% Presence: This means that in 90% of cases, an employee’s attitude (likely
positive or supportive) is linked to them receiving or offering moral support in the
workplace. This suggests that employees with strong attitudes are almost always
involved in creating a supportive work environment, whether through helping
colleagues, fostering team spirit, or contributing positively to the work culture.
• Business Decision Impact: With this level of support, you can confidently say that
attitudes play a critical role in moral support. Fostering positive attitudes within
your organization will not only improve individual performance but will also enhance
the overall team dynamic, as moral support is tightly linked to attitudes. Programs that
focus on cultivating supportive attitudes will be highly effective and have a
widespread impact.

Actionable Insights:

• Near-Certain Relevance: Rules with 0.9 support are extremely relevant and should
be considered as key drivers in decision-making. Since this rule applies to nearly the
entire dataset, focusing on attitude-building initiatives will yield consistent and
powerful results. You can confidently prioritize such initiatives, knowing that they
will have a near-universal effect on improving moral support within the organization.

Comparing the Three Support Levels:

1. Support 0.5:
o Moderate support means the rule applies to half the dataset. This indicates that
the relationship is important but not ubiquitous. Rules with 0.5 support are
helpful for understanding part of the population but should be supplemented
by other factors.
2. Support 0.7:
 oHigh support means the rule holds for a large majority (70%) of the dataset.
These rules are highly relevant and can be generalized across most of the
population. Business strategies based on these rules will be effective for most
employees, though there is still some room for exceptions.
3. Support 0.9:
o Very high support means the rule applies to almost the entire dataset. Rules
with 0.9 support are nearly universal and should be prioritized in decision-
making, as they represent key relationships that apply to almost everyone in
the population.

Overall Conclusion:

Support levels give you insight into how often certain relationships appear in your dataset. As
support increases from 0.5 to 0.9, the rule becomes more relevant to a larger portion of your
population. While rules with moderate support (0.5) are still useful for guiding specific
strategies, rules with high or very high support (0.7 and 0.9) should be considered as central
pillars for decision-making, as they represent trends that hold true for most of your
workforce.
3) Comparing Different Gain levels

Results

Interpretation

Gain 0.5: Moderate-Low Information Gain

General Interpretation:

• Modest Predictive Improvement: A gain value of 0.5 suggests that the rule offers a
moderate improvement over the default probability of the conclusion. The rule adds
 some new information but does not drastically increase the likelihood of the
conclusion happening.
• Implication of Moderate-Low Gain: This means that while the premise (antecedent)
helps predict the conclusion (consequent), the increase in predictive power is only
moderate. Other factors may play a more significant role, and the rule only
moderately improves the prediction of the outcome.

Specific Example Interpretation:

Let’s take the rule Attitude → Employee Engagement with a gain of 0.5:

• Moderate Predictive Value: This rule indicates that knowing an employee’s attitude
provides some value in predicting their engagement but is not highly influential. It
suggests that attitudes such as positivity or motivation are moderately linked to
engagement, but other factors like team dynamics or job satisfaction are likely just as,
if not more, important.
• Business Decision Impact: A gain of 0.5 suggests that attitude alone is not a strong
enough predictor of engagement. While improving employee attitudes can have a
positive effect on engagement, it is not sufficient on its own. You may need to explore
additional strategies (like creating a collaborative environment or offering career
development opportunities) to boost engagement.

Actionable Insights:

• Moderate Influence: Rules with a gain of 0.5 offer some predictive power but aren’t
highly impactful. You should consider attitude-building programs as one part of a
broader strategy to improve engagement but focus on multiple factors to drive
significant change.
ii) When the gain level is changed to 0.7

Interpretation

Gain 0.7: Moderate-High Information Gain

General Interpretation:

• Significant Predictive Improvement: A gain value of 0.7 indicates that the rule
brings a considerable increase in predictive power. The premise (antecedent)
strongly increases the likelihood of the conclusion happening compared to the default
probability.
• Implication of Moderate-High Gain: At this level, the rule provides valuable new
information and is a reliable indicator of the conclusion. The premise plays an
important role in predicting the outcome, and this relationship is strong enough to
guide business decisions with confidence.

Specific Example Interpretation:

Let’s analyze the rule Attitude → Employee Resilience with a gain of 0.7:

• Reliable Predictive Power: This rule suggests that knowing an employee’s attitude
provides a strong improvement in predicting their resilience. Positive attitudes, such
as adaptability and determination, are important indicators of resilience, though some
external factors (like workplace stress or team support) might still play a role.
 • Business Decision Impact: A gain of 0.7 means that focusing on improving attitudes
will significantly enhance employee resilience. You can confidently implement
attitude-focused training or mental resilience programs knowing that they will
likely result in more resilient employees. While there are still some other factors in
play, attitude is a key contributor.

Actionable Insights:

• Strong Influence: Rules with a gain of 0.7 are strong predictors. Attitude-building
programs should be a high priority in resilience strategies, as they provide substantial
added value. Initiatives aimed at cultivating positive attitudes will likely have a
meaningful impact on improving resilience for the majority of employees.
iii) When gain level is changed 0.9

Interpretation

Gain 0.9: High Information Gain

General Interpretation:

• Near-Certain Predictive Improvement: A gain value of 0.9 means that the rule
provides significant new information about the outcome. The premise almost
completely determines the conclusion, offering nearly complete predictive value over
the baseline probability.
• Implication of High Gain: At this level, the rule is highly reliable and provides
crucial information for predicting the conclusion. The premise is a dominant factor
in the outcome, with very little left to chance or other variables.

Specific Example Interpretation:

For the rule Attitude → Moral Support with a gain of 0.9:

• Highly Predictive: This rule suggests that knowing an employee’s attitude almost
fully predicts whether they will give or receive moral support in the workplace.
Employees with positive or supportive attitudes are very likely to contribute to or
benefit from a supportive work environment. There is little room for doubt or external
factors; attitude is a key determinant of moral support.
• Business Decision Impact: A gain of 0.9 means that attitude is nearly a guarantee
of moral support in the workplace. You can confidently implement programs that
focus on cultivating positive attitudes, knowing that this will lead to a more
supportive and cohesive team environment. Since the relationship is almost
deterministic, such programs will have a high likelihood of success.

Actionable Insights:

• Critical Influence: Rules with a gain of 0.9 offer nearly complete predictive power.
Programs aimed at improving attitudes should be the top priority in fostering a
 supportive workplace, as this relationship leaves little room for other factors. You can
be confident that focusing on attitude will lead to positive outcomes in team dynamics
and support.

Comparing the Three Gain Levels:

1. Gain 0.5:
o Moderate-Low Gain means the rule adds some predictive power, but it is not
highly influential. These rules are useful but require additional factors to make
stronger predictions. The rule helps, but it’s not enough to base decisions
solely on this information.
2. Gain 0.7:
o Moderate-High Gain shows that the rule provides significant added value.
You can confidently use this information to guide strategies, as the premise is
a reliable predictor of the conclusion. These rules suggest that the relationship
is strong enough to influence key business decisions.
3. Gain 0.9:
o High Gain means the rule offers nearly complete predictive power. The
premise almost guarantees the conclusion, making these rules critical for
decision-making. You should prioritize these rules when creating
interventions, as they leave very little room for error or other influencing
factors.

Overall Conclusion:

As you increase the gain from 0.5 to 0.9, the rule provides more added value in predicting the
outcome. Low gain values (0.5) indicate that the premise slightly improves the likelihood of
the conclusion, but it’s not strong enough to rely on exclusively. Moderate-high gain (0.7)
suggests that the rule adds significant new information, making it a reliable predictor for
decision-making. High gain (0.9) represents a near-guarantee that the premise leads to the
conclusion, making the rule highly actionable and influential in shaping business strategies.
4. Changing the lift levels
I) When the lift level is 0.5

Results
Interpretation

Lift = 0.3: Strong Negative Association

General Interpretation:

• Decreased Likelihood: A lift of 0.3 means that the presence of the premise reduces
the likelihood of the conclusion occurring. In this case, the conclusion is 70% less
likely to occur when the premise is present compared to random chance.
• Implication: A lift of 0.3 indicates a strong negative association, meaning that the
premise actively works against the conclusion. The premise and conclusion rarely
happen together, suggesting that the premise discourages the outcome.

Specific Example Interpretation:

Let’s take the rule Attitude → Employee Resilience with a lift of 0.3:

• Negative Association: This rule suggests that when employees exhibit certain
attitudes (perhaps negative or disengaged attitudes), they are 70% less likely to show
resilience in the workplace. This means that poor attitudes have a significant
negative impact on resilience.
• Business Decision Impact: A lift of 0.3 indicates that negative attitudes are highly
detrimental to resilience. Organizations should focus on addressing negative
attitudes, as they strongly decrease employees' ability to cope with stress and
challenges. Employee coaching, wellness programs, or emotional intelligence training
might be critical to improving resilience in this case.

Actionable Insights:

• Focus on Reducing Negative Attitudes: Rules with a lift of 0.3 point to situations
where the premise harms the outcome. In this case, you should prioritize addressing
negative attitudes through targeted interventions, as they significantly reduce
resilience.
ii) When the Lift level is 0.4

Results
Lift = 0.4: Moderate Negative Association

General Interpretation:

• Moderately Reduced Likelihood: A lift of 0.4 means that the premise moderately
reduces the likelihood of the conclusion. The conclusion is 60% less likely to occur
when the premise is present compared to random chance.
• Implication: A lift of 0.4 indicates a negative relationship between the premise and
the conclusion. The premise makes the conclusion less likely, though the effect is not
as extreme as with a lift of 0.3.

Specific Example Interpretation:

For the rule Attitude → Moral Support with a lift of 0.4:

• Moderately Negative Impact: This rule suggests that certain attitudes (perhaps
apathy or detachment) make employees 60% less likely to engage in or receive moral
support in the workplace. This could mean that employees with poor attitudes are less
inclined to contribute to a supportive work environment, or they might be less open to
receiving support from others.
• Business Decision Impact: A lift of 0.4 means that attitudes have a moderately
negative impact on moral support. Companies should work to identify and mitigate
these negative attitudes to ensure a more supportive and cohesive work culture.
Focusing on fostering positive attitudes through training and team-building activities
could help reverse this trend.

Actionable Insights:

• Counter Negative Attitudes: Rules with a lift of 0.4 suggest that the premise is
moderately detrimental to the outcome. It’s important to counteract negative
attitudes through cultural initiatives that promote openness, communication, and
mutual support.
iii) When lift level changes to 0.5

Results
Interpretation

Lift = 0.5: Weak Negative Association

General Interpretation:

• Weak Negative Association: A lift of 0.5 means that the premise weakly reduces the
likelihood of the conclusion. The conclusion is 50% less likely to occur when the
premise is present compared to random chance.
• Implication: A lift of 0.5 indicates a mild negative relationship between the premise
and conclusion. While the premise still discourages the conclusion, the effect is not
very strong.

Specific Example Interpretation:

Let’s consider the rule Attitude → Employee Engagement with a lift of 0.5:

• Slight Negative Impact: This rule suggests that employees with certain attitudes
(such as negativity or disengagement) are 50% less likely to be engaged at work.
While this is not an overwhelming negative effect, it shows that attitudes still play a
role in reducing engagement.
• Business Decision Impact: A lift of 0.5 means that negative attitudes moderately
decrease engagement. Organizations should monitor employee attitudes, but this
relationship suggests that there are other important factors influencing engagement.
Additional strategies, like improving the work environment or offering professional
development, might be needed to fully engage employees.

Actionable Insights:

• Address Negative Influences on Engagement: Rules with a lift of 0.5 suggest a

weak negative association, meaning the premise isn’t the only factor driving the
outcome. You should still work on improving attitudes but consider that other aspects
of the workplace (e.g., leadership, career growth) might be just as important in
improving engagement.

Comparing the Three Lift Levels:

1. Lift = 0.3 (Strong Negative Association):

o Significant Negative Impact: A lift of 0.3 means the premise strongly
reduces the likelihood of the conclusion happening. These rules should be
prioritized for interventions that seek to eliminate negative behaviors or
influences.
2. Lift = 0.4 (Moderate Negative Association):
o Moderate Negative Impact: A lift of 0.4 indicates a moderately negative
relationship between the premise and the conclusion. While not as severe as a
lift of 0.3, it still suggests that the premise negatively affects the outcome,
requiring action to counter this trend.
3. Lift = 0.5 (Weak Negative Association):
 o Mild Negative Impact: A lift of 0.5 means the premise slightly reduces the
likelihood of the conclusion. While the effect is not strong, it’s still worth
addressing the premise to prevent further reduction in the desired outcome.

Overall Conclusion:

When interpreting low lift values (0.3, 0.4, and 0.5), the focus is on negative associations—
situations where the presence of the premise actually decreases the likelihood of the
conclusion. The lower the lift value, the more detrimental the premise is to the conclusion.

• A lift of 0.3 suggests a strong negative relationship, meaning the premise significantly
reduces the outcome, and it requires urgent attention.
• A lift of 0.4 indicates a moderate negative relationship, which still requires addressing
but is less severe.
• A lift of 0.5 shows a weak negative relationship, where the premise has a mild impact
on the conclusion but is still worth managing.

In all these cases, the goal should be to mitigate the negative effects of the premise through
targeted interventions, such as improving employee attitudes, fostering better engagement, or
enhancing workplace support structures.
Overall Conclusion for Association Rule Interpretation (Support, Confidence,
Gain, Lift):

The interpretation of association rule results is an essential step in understanding the

relationships between variables in your dataset. By analyzing support, confidence, gain, and
lift, we gain insights into how strong, reliable, and meaningful these relationships are. Here's
a breakdown of the overall conclusions from analyzing all four metrics:

1. Support:

• What it Measures: Support shows how frequently an itemset (a combination of

attributes) appears in the dataset.
• Low Support (e.g., 0.5): Indicates that the rule applies to 50% of the data. This
means the itemset is common but not universally present. Rules with low support are
useful for specific contexts but may not always be reliable for general decision-
making.
• Moderate to High Support (e.g., 0.7, 0.9): Indicates that the itemset is present in a
large portion of the dataset (70% or more). Higher support makes the rule more
relevant to the majority of the dataset and thus more useful for strategic decision-
making. It suggests that the identified relationship is more generalizable to the
broader population.

2. Confidence:

• What it Measures: Confidence represents how likely the conclusion is to occur when
the premise is true. Higher confidence indicates a stronger relationship.
• Low Confidence (e.g., 0.5): Suggests that the relationship between the premise and
the conclusion is weak or uncertain. For example, if attitude leads to employee
engagement with a confidence of 0.5, it means there’s only a 50% chance that
improving attitude will lead to higher engagement. This suggests that other factors are
also influencing engagement.
• High Confidence (e.g., 0.9): Indicates a strong and reliable relationship. For
instance, if attitude → resilience has a confidence of 0.9, this shows that employee
attitude is a strong predictor of resilience, and interventions aimed at improving
attitude will likely lead to a high increase in resilience.

3. Gain:

• What it Measures: Gain measures the additional predictive power the rule provides
over the default probability of the conclusion occurring.
• Low Gain (e.g., 0.3): Indicates that the premise doesn't add much new information
or predictive power about the conclusion. If a rule has a gain of 0.3, the relationship
 between the premise and conclusion is weak, and other factors may be more important
in driving the outcome.
• High Gain (e.g., 0.9): Suggests a strong predictive improvement. A rule with a gain
of 0.9 implies that the premise adds significant new insight and dramatically
improves the likelihood of the conclusion. This is a valuable rule for making
actionable predictions.

4. Lift:

• What it Measures: Lift compares the observed frequency of the rule to the frequency
expected if the two items were independent. A lift value above 1 indicates a positive
association; a value below 1 indicates a negative association.
• Lift = 1 (Neutral): Suggests that the premise has no influence on the conclusion. For
example, if attitude → employee engagement has a lift of 1, the relationship
between attitude and engagement is independent, meaning that attitude does not
impact engagement any more than random chance.
• Lift > 1 (Positive Association): Higher lift values (e.g., 1.5, 2) show that the premise
increases the likelihood of the conclusion. For instance, if attitude → moral support
has a lift of 2, it means employees with positive attitudes are twice as likely to engage
in moral support, indicating a strong positive relationship.
• Lift < 1 (Negative Association): A lift value below 1 suggests that the premise
decreases the likelihood of the conclusion. For example, a lift of 0.4 indicates that
negative attitudes are less likely to lead to moral support, suggesting a negative
association.

Overall Insights:

1. Support, Confidence, Gain, and Lift Work Together:

o Support tells us how often the rule appears in the dataset, while confidence
tells us how often the rule holds true when the premise is true. Gain indicates
how much new information the rule provides over the default probability of
the conclusion, and lift measures the strength of the relationship between the
premise and conclusion relative to random chance.
2. High Support + High Confidence = Reliable and Generalizable:
o Rules with high support and high confidence are the most reliable and
should be prioritized for strategic interventions. For example, if attitude →
employee resilience has high support and confidence, it implies that
improving attitude will likely lead to increased resilience across a large
portion of the workforce.
3. Moderate Support + Moderate Confidence = Useful for Targeted Interventions:
o Rules with moderate support and moderate confidence offer useful insights
but may not apply universally. They can be used for targeted interventions or
hypothesis generation for specific groups. For example, if attitude →
engagement has moderate support and confidence, it suggests that while
attitude plays a role in engagement, other factors like work environment
might also be significant.
 4. Low Gain or Low Lift = Weak or Negative Relationships:
o Rules with low gain or low lift (e.g., below 1) indicate weak or negative
relationships, where the premise either doesn't add much predictive value
(low gain) or actively diminishes the likelihood of the conclusion (low lift).
Such rules may be worth investigating further, but they shouldn't be the focus
of your strategy unless combined with other stronger factors.
5. Actionable Insights:
o Rules with high lift and high confidence should be the main focus of
strategic interventions, as they provide strong, actionable insights for driving
outcomes (e.g., improving attitudes to enhance resilience).
o Rules with lower support, confidence, or gain are still useful for hypothesis
generation but should be explored in conjunction with other factors to ensure
comprehensive decision-making.

Final Thoughts

By considering support, confidence, gain, and lift together, you can effectively prioritize
which association rules to focus on for making data-driven decisions. High values in these
metrics indicate strong, reliable relationships that can guide effective interventions, while
lower values might highlight areas where further exploration or additional data are needed.