Research On Multi-dimensional Recommendation Application Technology Based On Data Scenarios

Massive applicationson the Internet bringus the convenience in our lives.However,the functions of these applications are mixed and complex,and most of the users cannot easily find out.It results in the question of how to select the right application among these massive choices.The main contributionin this thesis is to providethe recommendations of applications bydata scenarios.According to user groups,application software,user behavior characteristics and usage situations,an optimization recommendation algorithm based on multiple data scenarios is proposed,in order to improve the accuracy and user experiencesof existing recommendation algorithms.Finally,the performance of the proposed algorithm is demonstratedby the detailed experiments.The main contributions of the thesis as follows:1.A data clustering method based on Singular Value Decomposition(SVD)is proposed.The proposed algorithm is used to reduce the influence of data with singular value in traditional clustering methods.The decomposition of singular value can find out more accurate classification results.The proposed algorithmcan find new classification based on data itself,rather than subjective factors,and improve the recall rate of classification results.2.A Bayesian prediction model based on usage situational factors is proposed,which combines situational factors,user basic attributes and user historical behavior attributes with multi-dimensional recommendation parameters together.The proposed model can effectively deal with the problem of cold start and the error rate between recommended applications and actual users.This model improves the accuracy of prediction by multi-dimensional data,and achieves better performanceof application recommendation.3.We develop a multi-dimensional application recommendation system based on data scenarios,which includes two optimized algorithms to improve the accuracy of prediction and user satisfaction.Finally,the proposed system is evaluated by experiments,and the results show it achieves the goal of our objective.