A Survey of Personalized Recommendation Based on Machine Learning Algorithms

Personalized recommendation is a key technology to effectively solve the overload of online information and eliminate information islands. It is widely known as an important way to improve the quality of information services. However, cold start, data sparseness, algorithm performance, recommendation accuracy and surprise are still the key issues that restrict users' personalized recommendations. Firstly, we review the development trend of personalized information recommendation algorithms in the past 15 years. And then we propose a new classification method for users' personalized recommendation based on machine learning algorithms with cold start, data sparseness, and the performance of the algorithm as the main goals. On this basis, we summarize and compare the ideas, practices and conclusions of related machine learning algorithms. Finally, we further summarize the main advantages and disadvantages of the 10 kinds of personalized recommendation algorithms from the perspective of classification proposed, and look forward to the development directions, difficulties, focus and methods of personalized recommendation algorithms.