Information Retrieval using applied Supervised Learning for Personalized E-Commerce

Abstract

Personalized E-Commerce Search Challenge issued by the International Conference on Information and Knowledge Management. By analyzing historical data containing browsing logs, queries, user interactions, and static data in the domain of an online retail service, we attempt to extract patterns and derive features from the data collection that will subsequently improve prediction of relevant products. A selection of supervised learning models will utilize an assembly of these features to be trained for prediction of test data. Prediction is performed on the queries given by the data collection, paired with each product item originally appearing in the query. We experiment with the possible assemblies of features along with the models and compare the results to achieve maximum prediction power. Lastly, the quality of the predictions are evaluated towards a ground truth to yield scores.