About

This Project was one of the Interview Round by Examly. Project Time - 7 days

Problem Statement

Derive an algorithm to find the difficulty of a question. For more Details

Task

1. Data Collection
2. Model Building
3. UI Development
4. Deployment
5. Documentation

Attribute Description

Attribute	Description
Question_Type	Type of the Question
Question_Difficulty	Difficuty of the Question
Attended	Total number of students Attended
Time_Taken	Time is taken by the Student to solve the Question
Submission	Number of times student submitted the answer
Hints_Used	Number of hints used by the student
Right	Number of students Correctly answered
Partial	Number of students Partially Answered
Wrong	Number of students Wrongly Answered

Solution

Difficulty ∝ Time _Taken ,Submission,Hints_Used,Wrong,Partial

Difficulty 1/∝ Right, Partial

From this, we infer that difficulty level is directly impacted by all the attributes except attended.

Generally, if a question is easy most of the students answer correctly, fastly and accurately and for hard question most of the students answer wrongly, slowly and inaccurately.

With the help of decision-making attribute(Dependent attribute ), we can predict the target attribute.

Application Structure

1.Data Collection

For collecting data, first I framed the Data Rules. Here, Data rules is data about data. It defines the each attribute type and range based on given Metrics from the given given metrics.

Finally, I used Google form and google sheet to collect the data. Sample Data Rule is here! For more Details.

2.Model Building

Source Code is Here!

The model can be defined as representation of something. The model represents the Mathematical relation and Logical relation of the Training data.

1.Cleaning Data

• Before building the model we need to do some process like Data cleaning, Feature Extraction.

• As our data is self-created it does not have any issues. But in our data have Categorical attribute. To handle this, we use Label encoding Technique which means assigning a numerical value to the category.

    [Easy,Medium,Hard] => [0,1,2]
  

2.Model Selection

For building the model we use Python Scikit-Learn Library. From that we select one of the Classification Algorithm called DecisionTreeClassifier.

DecisionTreeClassifier:

• As the name says, it creates the Tree Structure by *** Decision-Making Rule*** as its Node.
• Almost every computer engineer should come across this basic question, Assign the grade for a student based on their marks!
• The key concept of Decision Tree classifier is the above question. Butt he only thing is it makes decision based on multiple atrribute.
• For easy understanding, it is like a nested if-else Condition.

Image Describes Nested if-else Loop.

Algorithm

Source code is here!

---

Algorithm Sample:

if (Hints_Used<3.5)
{
    Class->Easy
}
else 
{
    if(wrong<641925)
    {
	    class->Medium
	}
	else
	{
		class->Hard
	}	
}

3.Creating Model and Testing

• The input data is splited into train and test data.
• The model is created with training data.
• Finally the Model is tested with test data.

The Below graph is an Boundary analysis of Decision Tree Classifier.Ploted as Time_Taken vs Submission.

• Easy
• Medium
• Hard

From this graph we can conclude that increases in Time_Taken and Submission increases the Difficulty Level.

3.UI Development

For the UI Development I use Python Flask Framework. The input can be in two form,

• File as URL
• Feature as input

4.Deloyment

For Deployment I used Heroku Cloud Platform.

5.Documentation

The is an Documentation

Technology Stack

• Python
• HTML
• CSS
• Flask

Library Used

• Pandas
• Numpy
• Scikit-Learn
• Pickle
• Matplotlib

Tools Used

• Google Colab

• Jupyter Notebook

• Heroku

• Github

• Bootstrap

• Google Form

• Google Sheet

                        ***Thank You***