###Authors - Shamya Karumbaiah and Rafael Lizarralde

This is a classification problem. Given the interaction log of a student, we want a model that can predict whether they have a particular affect (are feeling specific relevant emotions). The emotional states of interest are - concentrating, confused, bored and frustrated.

Our final solution involved the selection of 57 features, including several of the affect history features we added. Using a random forest with 40 trees, it scored a Cohen’s Kappa of 0.306, and scored 0.230 on the test set. The cross-validation score is significantly better than the cross-validation score reported earlier, but the test score suggests that the data set desperately needs expansion.

Email me at shamya16@gmail.com for data

Project report - report.pdf