In the diploma thesis we are dealing with the problem of developing a recommender system for a clothing store based on transaction data. We start with theoretical basics about recommenders and association rules. Afterwards we describe data and represent its quantitative and qualitative aspects. We continue with the detailed explanation of implemented methods, namely, nearest neighbors and matrix factorization. In the end we compare the results of our methods with naive method of recommending most popular products, achieving much better results. Matrix factorization produced the best results and we would use it in production.
|
