International Journal of Computer Applications (0975 – 8887)
Volume 138 – No.9, March 2016
Predictive Analysis: Assigning Weightage and Difficulty
Level of Question using Data Mining
Kamal Vora Sumeet Jain Param Mehta Smita Sankhe
K. J. Somaiya K. J. Somaiya K. J. Somaiya Assistant Professor
College of College of College of K. J. Somaiya
Engineering. Engineering. Engineering. College of
Engineering.
ABSTRACT 2. ASSIGNING WEIGHTAGE
Generation of question papers is a task that requires high Assigning weightage to a question is tedious task. Various
expertise and flair. The design and formation of the test parameters are defined which are required to evaluate a
depends on the questions. The test should generate the quiz question. Each parameter gives the measure of evaluation for
such that all different levels of questions are inserted so as to the question. Satisfying the parameter gives the level of
make the question paper more competitive. This research weightage.
work aims at assigning weightage to a questions in the quiz
based on the difficulty level. Assigning the weightage to Parameter Weight(range)
question is a tedious task. This paper aims to discuss various Level of IQ (sense) 2-10
parameters involved in assigning weightage to the questions
in the test paper and decide the difficulty level of the question. Length of question 2-10
General Terms Pattern
Assignment of weightage, deciding the difficulty level. a. Repetition of 2-8
keyword
Keywords b. image 0-2
Weightage
Type of question
1. INTRODUCTION
a. true/false type 2
Aptitude tests are very important to identify the knowledge of
b. simple MCQ 4
person. They can prove helpful in deciding the right career
c. calculated MCQ 6
path. The better the test, accurate would be the prediction
d. check box(multiple 8
result. For the test to be better, correct weightage has to be
correct answers)
assigned to the question. Assigning correct weightage to the
e. text box 10
question implies giving correct importance to the question and
thereby with the degree of importance a person can decide the
level of difficulty to the question. Accurate assignment of The evaluation measure for each parameter is given as:
weightage would help in giving accurate results. Basically Level of IQ (sense): The more sense the question
there are three levels of difficulty to the question. [2] makes the more weightage to be assigned to the
High(difficult questions) question. This also means question is related to the
subject on which test is taken. Thus more the
Moderate(manageable questions) question makes sense and is more related to test
subject more weightage is given to the question.
Low(easy question)
Length of question: In a time bound exams where
To group the questions into these sections, the system needs
timing is less the length of question plays a crucial
to know the total weightage that is to be assigned to the
role. Smaller the question, less time will it take
question. To assign the weightage a predefined scale needs to
easier and faster would the user will be able to
be set. Reaching a specify measure in scale means to have
solve. Thus longer the question more is the
achieved a certain level of difficulty.
weightage.
Doing this would provide three main benefits.
Pattern:
Question paper containing questions of difficulty
o Repletion of keyword: The more time the keyword
level can be set.
of the question is repeated implies more information
Marks can be allocated efficiently to each module of in the question is given about key terms making
the course/syllabus/field. question easy to answer thus lower weightage to be
assigned.
Can be helpful to generate multiple question papers
with same difficulty level so as to be as fair as o Image: Visual perception to a thing gives better
possible when multiple people appearing test for idea. Thus providing the image in question makes it
same goal. easy to solve a question as compared to a question
where image is not given thus question with image
to be assigned lesser weightage.
31
International Journal of Computer Applications (0975 – 8887)
Volume 138 – No.9, March 2016
Type of question: numbers. Hence deciding the level can help to assign the
marks in the ratio of the level of difficulty.
o True/false: true false question has only two options
thus probability of giving correct answer is high i.e. The marks assigned to the question are assumed in the ratio
0.5 thus making question easy and less weightage to [Link]. That is 1 mark for an easy question, 2marks for
be assigned. question with moderate difficulty level and 3 marks for
question with the highest difficulty level.
o Simple MCQ: simple MCQ question has lesser
probability of giving correct answers. Suppose for The algorithm to calculate the ratio:
question with 4 options the probability of right
M Maximum Possible weight.
answer is 0.25 thus a slight more weightage to be
assigned. W Total weightage for that Question.
o Calculated MCQ: The only difference from simple X Maximum Weight to that specific part.
MCQ type of question is that in simple MCQ type
of question the answers can be given directly F Assign Weightage to that specific part.
whereas MCQ type questions require complex b Unused Weight.
calculations.
S Decision Parameter
o Check box (Multiple answers): Multiple answers to
a question decrease the probability of giving M = ∑ Xi
absolute correct answer. For example a question W = ∑ Fi
having 4 options 4! (4 factorial i.e. 24) possible
combinations can be applied. That means b = ∑ (Xi- Fi)
probability of correct answer is 0.04167 which
means difficult is the question. Hence higher S = (M-b)/M
weightage to be assigned Since question are in ratio of [Link] the total weight is 6 so
o Text Box: For the textual answers the number of question with easy difficulty will be given between range of 0
possible answers can be infinitely large thus making and 1/6 i.e. 0 to 0.1666 whereas for question with moderate
probability of getting right answers very less so difficulty will be in between 1/6 to 3/6 i.e. 0.1667 to 0.5000
highest weightage to be assigned to this type of and for difficult question it is upto 1i.e from .5001 to 1.
question. From above calculated value of S
The total weightage that is to be assigned to the question If S lies between 0 to 0.1666 then Question type will be Easy
should be the cumulative sum of weightage from all the
parameters. Adding the weightage from each parameter will If S lies between 0.1667 to 0.5000 then Question type will be
give the final total weightage to the question. Medium
The minimum possible weight to a question is 10 meaning the Else the Question type will be Hard.
easiest possible question and maximum possible weightage is Thus questions are classified into three difficulty level. For
40 meaning the most difficult question. The weightage to instance now the question with least difficulty level (easy)
general question can be anything between this range. will get 1 mark while moderately difficult question will get 2
Now this weightage can be further used to determine the level marks and question with highest difficulty level will get 3
of difficulty to the question. marks.
3. DECIDING DIFICULTY LEVEL OF 4. SYSTEM ARCHITECTURE
The question stored in the database will first be classified
QUESTION according to option given and then the weightage for each
After finding out the exact weight of the question now it is question will be calculated weight generator. Now this
time to decide the difficulty level of the question. Determining question along with the weightage will go into decider which
the difficulty level of question is very necessary as the will categorize question according to difficulty level.
weightage of the question may vary from 10 to 40 but the
marks that are assigned to the question should be only discrete
Fig 1: System architecture
32
International Journal of Computer Applications (0975 – 8887)
Volume 138– No.*, March 2016
5. SYSTEM APPLICATIONS Parameter Values Weightage
This system can be used at many places. IQ 6 6
Length of question 140 8
Automatic question paper generator: systems like Repetition 4 2
automatic paper generator can use this algorithm as Image present No 2
it provides proper classification of question and Type of question Simple MCQ 4
hence question of all category and varying range Total 24
can be asked. This will also help to set a question
paper of specified marks
7. CONCLUSION
Aptitude test: since the weight assigned to the In the previous system, the questions are assigned marks that
questions are generated by the system by applying are uniform to all questions irrespective of the difficulty and
pure statistical and probability knowledge the various parameters that lead to flaws in prediction in the
weightage assigned is accurate. Most accurate desired result. Our stated algorithm allows us to allocate
weightage assignment helps in generating better test marks depending on the weightage assigned to various
that can be used to give better results. parameters that cumulate to the total weightage of the
question, which helps to make a more appropriate prediction.
Also in case of predictive analysis it will help to predict better
future due to accuracy in assigning the weightage to question The difficulty of a question is a matter of individual
and classifying question in difficulty level. perspective and hence never be certain. Using probabilistic
model, it tends to give a better result even in uncertain and
6. SIMULATION AND RESULT unforeseeable circumstances which is implemented in above
Consider a scenario where a question is entered into the stated algorithm.
database and the corresponding weightage is to be calculated.
The general IQ level of this question amounts to 6 units. The 8. ACKNOWLEDGMENTS
length of the question is of 140 characters which accounts to a The authors would like to thank Ms. Smita Sankhe of K .J.
weightage of 8. The keyword present in the question is Somaiya College of Engineering for providing specifications
repeated 4 times, it adds up to the weightage of 4units. about the application scenario.
Additional 2 units is assigned to the question as there is no
image reference option to the question. The question type is 9. REFERENCES
simple MCQ, and hence according to the probabilistic model [1] Xindong Wu · Vipin Kumar · J. Ross Quinlan · Joydeep
the weightage added is 4 units. Final weightage Ghosh · Qiang Yang ·Top 10 algorithms in data mining
corresponding to the question adds up to 24 units. 2012
To decide the difficulty level of the question, the above [2] Dept. of [Link]., DM’s College, Mapusa, Goa.
algorithm (Algorithm 1) is used: Dynamic Examination Question Paper Generation
System (DEQPGS) 17th International Conference on
The maximum possible weightage (M) for a question Management of Data (COMAD 2011) 19th December
aggregates to 40. 2011 December 19—21
While the total assigned weightage (X) to the question is 24. [3] (IJCSIS) International Journal of Computer Science and
The ratio (S) = 24/40 = 0.6 Information Security, Vol. 7, o. 1, 2010
[4] PACO: A predictive analysis system for manufacturing
Thus according to the algorithm this ratio falls in the category
test, AUTOTESTCON, 2010 IEEE, T. Jones; J. Engler,
“Hard”. This means that this is the question of highest
IEEE, 13-1,Sept. 2010
difficulty level and thus the maximum marks that could be
given to individual question is allocated to this question.
IJCATM : [Link]
33
