Course Objectives
[Link] introduce the basic concepts and techniques of data mining and machine
learning for knowledge discovery from large datasets.
understand types of data (structured, unstructured, categorical, numerical) and the role of knowledge
discovery from large datasets
Meaning: Learn the basic ideas of data mining and machine learning.
Example: Think about Netflix – it collects huge amounts of data about what you watch.
By applying data mining and ML, Netflix can recommend new movies and shows to you.
This is possible only if you understand the fundamentals of data and ML techniques.
In today’s digital world, huge amounts of data are generated every second (social
media, banking, healthcare, e-commerce, sensors).
But raw data alone has no value unless we can analyze it to discover useful patterns
and knowledge.
This is where Data Mining and Machine Learning come in.
What is Data Mining?
Definition:
Data mining is the process of extracting hidden, useful patterns and knowledge from
large datasets.
It is also known as Knowledge Discovery from Data (KDD).
Online shopping websites (like Amazon) analyze customer purchase data → recommend
products.
Banks analyze transaction data → detect fraud.
2. To familiarize students with data preprocessing, feature selection, and
model building methodologies.
data preprocessing: cleaning (removing noise, handling missing values), integration, transformation, and
reduction.
Feature selection is emphasized because choosing the right attributes improves model performance.
Students will also learn model building methodologies: how to go from raw data → model → evaluation
Meaning: Learn how to clean and prepare data before applying algorithms.
Example: In healthcare, patient records often have missing values (like blood pressure not
recorded). Before predicting diseases using ML, we must handle missing values, remove
errors, and choose the right features (age, blood pressure, sugar level, etc.). This is data
preparation in action.
[Link] provide an understanding of various supervised and unsupervised learning
algorithms used in real-world applications.
Students gain clarity on the three types of ML:
Supervised (classification, regression)
Unsupervised (clustering, association)
Reinforcement (learning through rewards/penalties).
They will see how these paradigms are applied in real-world problems like spam detection,
fraud detection, recommendation systems.
Meaning: Learn about supervised, unsupervised, and reinforcement learning.
Example:
Supervised: A bank uses past loan data to predict whether a new customer will repay or
default (classification).
Unsupervised: An e-commerce company groups customers into clusters for targeted
advertising.
Reinforcement: A self-driving car learns to drive better by receiving rewards for safe
moves and penalties for mistakes.
4. To develop the ability to implement and evaluate machine learning models using
appropriate tools and metrics.
Students will learn how to build machine learning models step by step using
programming tools (like Python, R, or Weka).
They won’t just stop at building a model — they will also check how well the model
works by testing it on data.
To do this, they will use evaluation measures such as accuracy (how many predictions
are correct), precision, recall, F1-score, etc.
In short: you will learn how to build a model and also judge if it is good or needs
improvement.
Meaning: Build ML models using tools and check how well they work.
Example: Suppose you build a spam email classifier using Python. After building it,
you test it with new emails. If the model predicts correctly 90% of the time, its accuracy
is 90%. If it often mislabels important emails as spam, then you need to improve it. This
shows why evaluation is important.
[Link] encourage analytical thinking and problem-solving skills through hands-on
experiments and case studies in data-driven domains.
. Encourages critical thinking by applying DM/ML to real-world datasets.
Case studies in healthcare, banking, agriculture, etc., show how theory meets practice.
Students develop problem-solving ability by choosing the right method/model for a given
dataset.
Meaning: Apply knowledge to real-world problems and make decisions.
Example: In agriculture, ML can be used to predict crop yield based on soil, rainfall,
and temperature data. A student who learns this course can design a solution that helps
farmers plan better. This shows problem-solving using ML.
Detailed Explanation of Course Outcomes (COs)
On completion of this course, students will achieve the following abilities:
CO1: Understand Fundamental Concepts
Students will be able to explain key terms: data mining, data preprocessing, types of
data, attribute types, evaluation metrics.
They’ll grasp the importance of data quality and how it affects mining results.
Example: Understanding why normalization is needed before applying KNN.
CO2: Apply Data Mining Techniques
Ability to apply algorithms like classification (Decision Trees, KNN), clustering (K-
means), association rule mining (Apriori, FP-Growth), regression (linear/logistic).
Example: Using classification to predict whether a patient has diabetes.
CO3: Design and Implement ML Algorithms
Students can write code/implement models using supervised and unsupervised learning.
They’ll use tools/libraries (Python, Scikit-learn, TensorFlow, etc.).
Example: Implementing a spam filter using Naïve Bayes in Python.
CO4: Evaluate Model Performance
Students will use metrics (accuracy, precision, recall, F1, ROC) and validation methods
(cross-validation, confusion matrix).
They’ll learn to compare models and select the best one.
Example: Comparing Decision Tree vs. SVM on the same dataset.
CO5: Develop Practical Solutions
Students will apply DM/ML to real-world case studies.
They’ll learn to justify model choice based on dataset characteristics (e.g., noisy,
imbalanced, high-dimensional).
Example: Choosing Random Forest for fraud detection due to imbalanced data.
CO6: Analyze Datasets & Preprocessing
Students will analyze different dataset types (numerical, categorical, text, streaming).
They will decide the most suitable preprocessing techniques: normalization, encoding,
handling missing values, dimensionality reduction.
Example: Using one-hot encoding for categorical attributes in customer segmentation.
CO7: Compare & Contrast Algorithms
Students can critically compare algorithms in terms of:
o Assumptions (e.g., linearity in Logistic Regression, independence in Naïve
Bayes).
o Advantages (e.g., SVM works well with high dimensions).
o Limitations (e.g., KNN is slow with large datasets).
o Suitability (e.g., Decision Trees for interpretability vs. Neural Networks for
accuracy).
CO8: Real-Time Processing & Streaming Analytics
Students will be introduced to real-time ML applications.
They will implement basic streaming analytics: analyzing continuous data (IoT sensors,
stock market feeds, clickstreams).
Example: Applying incremental clustering on streaming sensor data for anomaly
detection.
