Personalized news recommendation systems have a profound impact on the ecology of the newmedia industry based on smart phones. This paper discusses its current status and its recommendation algorithm. Authors analysed personalized recommendation algorithms problems，such as user interest mismatch，poor content quality，filter bubble and information addiction，lack of forgetfulness and copyright infringement. This paper also proposes technical or non-technical optimization strategies. For the problem of content mismatch，developing a new algorithm paradigm represented by neural networks，improving existing algorithms，can generate a positive outcome. For the issue of content quality，Internet companies should pay attention to investing in the construction of high-quality content ecology，and at the same time using machine learning to develop an anti-trival and anti-cheating recognition system. Human censorship should also be used to reduce unqualified content. The Personalized news recommendation company can draft a code to encourage the dissemination of high-quality content. For information Cocoons and information addiction，companies can use new techniques to help users explore new themes.