Student Performance Analysis Using Natural Language Processing

Page views, historical data and click stream data logged on learning management system websites come up with concrete information about students approach towards the course. Such information can help in delivering descriptive feedback that can contribute to students progress. Our main objective behind this research is to develop a model which provides valuable intuitions, suggestions behind grades. Auto Regressive Integrated Moving Average and Simple exponential smoothing models have predicted grade with a low mean absolute error of 0.305. Hybrid recommender Model is chosen for course recommendation. Students with similar choices are clustered.Rule mining is applied to these clusters to derive association rules. Effectiveness of these rules is verified by lift and confidence.We also used cosine similarity to extract skill set from jobs and match with the courses available.