Scribd
Upload
68 views7 pages

Instructor:: Semester Project Mam. Yella Mehroze

This document describes a student project to predict student engagement levels using a dataset containing student performance data. The dataset has 13 attributes including class label and will be used to classify students as having high or low engagement. Classification techniques in Weka, such as J48 decision trees, will be applied for model training and prediction. Feature selection will also be performed to identify the most important attributes for the model. The trained model will then be tested on unlabeled data to predict engagement levels.

Uploaded by

Bilal Sheikh
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
68 views7 pages

Instructor:: Semester Project Mam. Yella Mehroze

This document describes a student project to predict student engagement levels using a dataset containing student performance data. The dataset has 13 attributes including class label and will be used to classify students as having high or low engagement. Classification techniques in Weka, such as J48 decision trees, will be applied for model training and prediction. Feature selection will also be performed to identify the most important attributes for the model. The trained model will then be tested on unlabeled data to predict engagement levels.

Uploaded by

Bilal Sheikh
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
You are on page 1/ 7

Data Warehouse Data Mining

Semester Project

Instructor: Mam. Yella Mehroze

Student Details
Name Bilal Ahmad
Course Data mining
Reg No. FA19-BCS-136
Section C
Date 18/12/2021
Student performance
prediction

Dataset details
Description:
Student performance prediction database will be use in this project
to predict the engagement level of student with learning resource.
Predicted class label will show high or low based on the interaction
of student with study material. Dataset contains 13 attributes
including class attribute and almost 500 records.
It includes:
 Login: How many times a student login to portal.
 Content Read: How many times a student read the study
material content.
 Forum Reads: How many times a student read the
problem/issue on community forum.
 Forum post: How many times a student posts the
problem/issue on community forum.
 Review: How many times a student reviews the quiz before
submission.
 Lateness indicator: Is student submit the late assignment.
There are multiple lateness indicator attributes.
 Average assignment submission time: How many hours
student takes to submit the assignment. There are multiple
lateness indicator attributes.
 Engagement level: This is a class label that will show the level
of student engagement to study materials based on previous
attributes.

Source:
Source of the dataset is from github (open source platform) and you
can get this dataset from here.

Problem type:
I want to predict class labels, so I will use classification because
classification is used to predict the nominal value of class attribute.

Techniques:
I will use weka tool to classify to train model and predict the class
labels. I will also use feature selection to select the relevant
attributes for model training.
Process
Preprocessing:
Dataset is already preprocessed. There is no negative, null, or empty
value. Class attribute label is nominal and contains H (high) or L (low)
labels.

Classification:
I will use classification because classification is used to predict the
nominal value of class attribute. I will train the model using different
algorithm and test options.
Here are some steps to perform classification:
 Load training data into weka.
 In classify tab, choose tree and select J48 algorithm.
 Select cross validation (10 folds) in test option.
 Select class attribute
 Start the process.
Result is showing that dataset is 99% correctly classified. There are
other details like total instances, root mean, absolute error,
confusion matrix, accuracy details etc.

Right click on the result item and select “save the model” for test
data label prediction.
Now, create the copy of training data, shuffle the order, and remove
all the labels of class attribute to check either trained model can
predict class labels or not.
Here are some steps to perform prediction:
 In classify tab, choose supplied test data in test options.
 Load the test data file (without labels).
 Choose class attribute.
 Click on more options, choose Plain text in output prediction.
 Reevaluate the model on current dataset.
 Result will show the predicted class labels.

Result is showing the predicted class labels as high or low based on


trained model.
Association rules:
Association rules are not applicable on this type of dataset and this
feature is out of scope. Association rules are only applicable if
dataset contain itemset.
Feature selection:
The attribute selection task essentially consists in selecting a subset
of originally available attributes to be subsequently used for model
creation. For this purpose, I will use selectAttribute tab to select top
10 attribute in dataset that will be use in training model.
Here are some steps to perform feature selection:
 Load training data into weka.
 In classify tab, choose info gain in attribute evaluator.
 Set number option to 10 in Ranker.
 Select cross validation (10 folds) in test option.
 Select class attribute
 Start the process.

Result is showing relevant attributes which are necessary for model


training. Attributes are short on the base of average rank. Select top
rank attributes for training model. Class attribute will be mandatory.

You might also like