Recommender systems are information retrieval tools helping users in their information seeking tasks and guiding them in a large space of possible options. Many hybrid recommender systems are proposed so far to overcome shortcomings born of pure content-based (PCB) and pure collaborative filtering (PCF) systems. Most studies on recommender systems aim to improve the accuracy and efficiency of predictions. In this thesis, we propose an online hybrid recommender strategy (CBCFdfc) based on content boosted collaborative filtering algorithm which aims to improve the prediction accuracy and efficiency. CBCFdfc combines content-based and collaborative characteristics to solve problems like sparsity, new item and over-specialization. CBCFdfc uses fuzzy clustering to keep a certain level of prediction accuracy while decreasing online prediction time. We compare CBCFdfc with PCB and PCF according to prediction accuracy metrics, and with CBCFonl (online CBCF without clustering) according to online recommendation time. Test results showed that CBCFdfc performs better than other approaches in most cases. We, also, evaluate the effect of user-specified parameters to the prediction accuracy and efficiency. According to test results, we determine optimal values for these parameters. In addition to experiments made on simulated data, we also perform a user study and evaluate opinions of users about recommended movies. The results that are obtained in user evaluation are satisfactory. As a result, the proposed system can be regarded as an accurate and efficient hybrid online movie recommender.

This thesis addresses the new item problem in recommender systems, which pertains to the challenges of providing personalized recommendations for items which have limited user interaction history. The study proposes and evaluates a method for generating personalized recommendations for movies, shows, and series on one of Sweden’s largest streaming platforms. By treating these items as documents of the attributes which characterize them and utilizing item similarity through the k-nearest neighbor algorithm, user preferences for new items are predicted based on users’ past preferences for similar items. Two models for feature representation, namely the Vector Space Model (VSM) and a Latent Dirichlet Allocation (LDA) topic model, are considered and compared. The k-nearest neighbor algorithm is utilized to identify similar items for each type of representation, with cosine distance for VSM and Kullback-Leibler divergence for LDA. Furthermore, three different ways of predicting user preferences based on the preferences for the neighbors are presented and compared. The performances of the models in terms of predicting preferences for new items are evaluated with historical streaming data. The results indicate the potential of leveraging item similarity and previous streaming history to predict preferences of new items. The VSM representation proved more successful; using this representation, 77 percent of actual positive instances were correctly classified as positive. For both types of representations, giving higher weight to preferences for more similar items when predicting preferences yielded higher F2 scores, and optimizing for the F2 score implied that recommendations should be made when there is the slightest indication of preference for the neighboring items. The results indicate that the neighbors identified through the VSM representation were more representative of user preferences for new items, compared to those identified through the LDA representation.