This thesis presents the implementation of recommendation systems using the Microsoft SQL Server environment. Three algorithms were developed: K-nearest neighbors (KNN), matrix factorization (MF), and the Slope One algorithm. The main goal of the thesis was to demonstrate how SQL Server can be used to create an efficient and accessible recommendation system that is directly integrated into the database.
For testing, the MovieLens1M and Book Crossing datasets were used, and comparisons of the results were made with algorithms from the Surprise library. For development, the Book Crossing dataset was used. The analysis showed that the SQL Server implementation produces competitive results, while reducing the need for additional infrastructure and simplifying integration into existing systems.
|
