krzysieq / COMPM052

Information Retrieval and Data Mining Project

COMPM052: Information Retrieval and Data Mining Project

Group 23 / Project 2

• Archit Sachdeva
• Georgi Georgiev
• Krzysztof Wroblewski

Code Structure

classifier/

• constants.py -- use to set the dataset path
• logistic_regression.py -- our implementation of a binary logistic regression classifier
• metrics.py -- our implementation of metrics used during training and validation
• mord.py -- trains and evaluates an ordinal regression model; requires the mord package
• multiclass.py -- our implementation of one-vs-rest classification; contains a wrapper that can be used with any binary classifier
• predict.py -- uses a previously trained model to generate a ranking file, or simply evaluate on a test set
• reader.py -- contains code for reading and parsing dataset files
• train.py -- trains the logistic regression model

eval.py -- contains evaluation code and our implementation of ERR and NDCG metrics