Empowering Intelligent Database Systems Using an Adaptive Recommendation Algorithm

Empowering Intelligent Database Systems Using an Adaptive Recommendation Algorithm

연구 분야: Databases

학회: International Conference on Artificial Intelligence and Soft Computing

The problems of big data (BD) processing in distributed intelligent databases (IDB) were analyzed in the paper. Features of using machine learning and recommendation systems for personalized analysis of user data were determined. The funk singular value decomposition algorithm (Funk SVD) algorithm for the operation of recommender systems in IDB, which allows for determining the necessary data for users, was studied. A modified Funk SVD was proposed, which allows using less data when computing correspondences between users and records in the database. It was established that the proposed modification provides a better speed of calculation of the results of the recommendations. It was also proposed to use the modified FunkSVD together with the k-nearest neighbors (k-NN) algorithm, which will improve the accuracy of data analysis. As a result of the research, the advantages of the proposed approach in comparison with the existing one were established due to better speed and accuracy of calculations of recommendations about relevant data to IDB users. It was established that the proposed method helps to adaptively determine the most optimal operating parameters of the recommender system, depending on its performance indicators.