Research And Implementation Of Recommendation System Based On Improved 3D Markov Model

With the advent of the era of big data,network data appeared in our daily life at an explosive rate.The rich data resources not only bring convenience to people but also cause a lot of trouble.How to find the desired data from the huge data set has become a hot issue to be solved.Therefore,the data mining technology has gradually been valued by people.An efficient and accurate personalized recommendation method is not only able to reduce the trouble of users to filter useless information,give users a better user experience,improve the user's loyalty,it but also avoids the consumption of a large number of network costs which caused by data transmission at the same time,has a high commercial value.At present,the personalized recommendation system has been widely used in the fields of news browsing,entertainment information browsing,electronic commerce and so on.Based on this,for the purpose of making users have an more efficient way to browse the news in the mobile terminal.A hybrid recommendation strategy is proposed in this thesis.This strategy is the one that based on improved three-dimensional Markov model is determined as the main algorithm for the purpose of predicting accurately,and an alternative recommendation algorithm that based on normalization of weighted user preference features is selected to assist prediction.In the research of the algorithm,firstly,the thesis analyzes the characteristics of the timeliness and large number of news information which is different from other information.In the process of data integration and under the premise of excluding special situation,The attribute of information is replaced by fuzzy in order to take out the user's preference characteristics.Secondly,the thesis analyzes the problem of inaccurate forecasting caused by the phenomenon that user temporary interest and user interest steep change.And put forward the solution that use the way of introducing interest threshold parameter in user preference feature in order to put the user's interest changes in a timely feedback to the recommendation system.Thirdly,according to the characteristics and problems,the demand analysis of the forthcoming news recommendation algorithm is carried out.In order to meet the demand of the algorithm,the thesis analyze the characteristics and applicability of many algorithms such as Apriori algorithm and Markov model.Finally,the recommended model is determined.By introducing a mechanism that tentative push to major mainstream news and user feedback in the model,the value of interest threshold is adjusted.Further,the problem of the inaccurate prediction is improved that caused by the phenomenon of user generated temporary interest and user interest steep change.Then,by grouping experiments with different participation degree users,the thesis compare the difference effect between the traditional recommendation model that without introducing interest threshold and the recommendation model proposed in this thesis.And verify the effectiveness of the proposed strategy.Finally,a mobile phone application with personalized recommendation function of iOS is realized by using the proposed strategy that improved in this thesis.