Finding learning parameters and Exploring learning strategies using supervised machine learning
Educational institutions and people who are connected with the administration and students who thrive to improve their grades and explore the parameters and strategies to improve performance and also to explore the pragmatic dependence of each parameter on the final grade.
We have used supervised machine learning classification to identify parameters and classify the data into 5 types (pass, good, fail, satisfactory and excellent).
Tools used for this project:
- google colab.
- NumPy, pandas.
Datasets used for research:
- Kal board 360(LMS students data).
- Secondary education dataset among two schools(Portuguese and maths).
Previous grades have a higher correlation to the final grade, and I have found using only previous grades has shown an accuracy of over 90 percent.
Excluding the previous grade, other parameters that affect the final grade are mothers job, related teacher yield better performance of students, students who are planning to higher education have higher.
Removing some features parent job teacher, parent status whether they are living or not, consuming less alcohol, studying greater than 4 hours per week got boost in heir scores.
Although SVM performed better, random forest and decision tree are next in number in terms of accuracy of the model for dataset-1 and random forest performed better for dataset-2.
Out of all classification models, SVM performed best it divides hyperplanes and divided the clusters using the correct c value gives the best results.
I thank Bennett university for giving me an opportunity to work under Dr.Shivani goel for guiding the journey of my project and helped if I have encountered any problems