Cold start Sequential recommendation based on pre-training and meta learning

• 1、Beijing Key Laboratory of Intelligent Telecommunications Software and Multimedia(Beijing University of Posts and Telecommunications), Beijing 100876

Abstract：Cold start recommendation is an important part that is difficult to ignore in practical recommendation systems, and cold start serialization recommendation is widely used. However, there are relatively few algorithms that have been studied for cold start serialization recommendation, and there are also few existing models that can solve the problem of low recommendation hit rates. This model proposes a cold start serialization recommendation based on pre training and meta learning(CS-PML). The algorithm first uses unsupervised algorithms to assign an initial value to each item, and then uses meta learning to learn information about users with shorter sequences. This model can learn more item information for each item, improving the hit rate.

Keywords： AI Cold start Serialization recommendation Pre training?????