This was the final project for an introductory module on Machine Learning @ TCD. The project involves looking at the feasibility of predicting information associated with Steam reviews using only the review text as input. The information to be predicted is:
- whether or not the game has been 'voted up' by the reviewer
- whether or not the review is for an 'early access' version of the game.
The project evaluates various conventional methods of feature extraction (Word Counts, TF-IDF, n-grams) and classification models (Logistic regression, Kernel SVM, DNN).