Building and evaluating recommender systems

Recommender systems play an important role in the lives of people in today’s information-rich environment. Businesses and companies utilize recommender systems to make meaningful suggestions to their users, increasing sales, and improving user experiences. Users make use of recommender systems to avoid wasting time finding the right products and contents, and to discover new perspectives and insights about what their likes and dislikes are. In most scenarios, recommender systems are highly likely to bring benefits to both businesses and their users, and thus, should be carefully planned and constructed so that a satisfactory outcome is achieved. This thesis project is a demonstration of how technical implementations of recommender systems can be done. Five recommender systems were built based on the goodbooks-10k dataset, and their contexts and theories are given. The recommender systems were then evaluated according to an arbitrarily established evaluation strategy. Insights into different characteristics and qualities of the five recommender systems could be drawn from the evaluations. It is concluded that the process of building and evaluating recommender systems are highly customizable, depending on multiple business-specific factors. Due to the heuristic nature of the building and evaluating processes, it is often the case that there is no one best way to implement such processes. It is advisable that businesses and companies perform careful planning and experimenting in order to reach a satisfactory result