Research And Implementation Of Mobile Application Recommendation System Based On Machine Learning

With the development of the mobile Internet, the number of mobileapplication(APP) has reached one million level. The long tail phenomenon is veryserious. The less popular APPs are covered by the hot APPs and can not draw users’attention. Recommendation system hopes to increase the click rate and at the sametime, increase the recall of the less popular APPs, helps users find associated APP soas to increase the number of mobile application distribution.At present, Baidu mobile application distribution platform has completed thefirst edition of association recommendation system. The major work of this paper liesin the improvement of the first edition of APP model, including the following aspects.We bring the machine learning algorithm in the recommendation system anddraws on the CTR prediction in Targeted advertising system. We regard the accuracyrate of classification problem as the goal of optimization and compare the results ofsome widely used classification algorithm on the APP recommended problem. Weanalyze the difference of different classification results through lots of experiments soas to draw general conclusions in treatment of APP. In order to measure the similaritybetween APP at the semantic level, we use probability Latent Semantic Analysis toextract the semantic features. the semantic level of text processing can get bettergeneralization effect than word frequency.The paper design and implement the new recommendation model system. Thesystem can automatically complete the offline data acquisition and cleaning, feature construction and extraction, model training and feature selection, and regularlyproduce training model and online data. The static model is difficult to adapt to thedynamic changes of data problem. And also the paper introduce the online learningmethod to use the real-time user behavior. The online learning method will improvethe recommended results so as to get better effect.The improved APP recommendation system has been on the line, providing moreaccurate APP related recommendation service for Baidu mobile applicationdistribution platforms, including PC and mobile client.