Proof of concept application for recommendation of events using hybrid collaborative filtering methods.
The dataset can be found at Kaggle - Event Recommendation System Challenge
The system uses K-Nearest Neighbours classifier.
The source code is divided into five packages:
- preprocessing - classes that handle the raw data, dividing the data into user/event specific files
- main - classes that are used to experiment with the classifier, and to generate .arff files
- models - classes that represent the feature vectors
- utils - classes that encapsulate utility methods used throughout the system
- evaluation - classes that are used to generate and compare the results