Research Of Personalized Recommendation Of Mobile Reading Service Based On Users’ Behavior

With the development of mobile internet and the intelligence of the mobile devices, for an example, the mobile phones, numerous mobile internet services are springing up. The China Mobile Telecom’s mobile reading service is one of the newborns. The mobile reading service has changed people’s original reading mode and provided people with reading service at anytime and anywhere. However, as the scales of mobile readers and the homogeneous products are expanding quickly, China Mobile Telecom, on one hand, faces huge opportunities and on the other hand, takes great challenges. And the mobile reading’s users have jumped from the era of lacking of information to the era of overloaded information. People are strongly expecting the services of personalized recommendation for reading materials. Thus, personalized recommendation services are eagerly to be offered by the China Mobile Telecom for the readers, it is especially important for the provincial companies to realize their precise marketing. But at present, the research of mobile reading’s personalized recommendation service is scarce, so this paper applies the related theories and technologies of data mining and personalized recommendation to study it.The aim of this paper is to build a mode of personalized recommendation for mobile reading service, which could promote the development of mobile reading service. Oriented with this aim, the paper firstly introduced research status of personalized recommendation service with data mining technology based on the analysis of user behavior. Then, the related algorithms and theories on the data mining and personalized recommendation are presented. Secondly, construct the DPM user behavior model, the model is consisted of duration rate, page view rate and payment rate. AHP method is involved to measure readers’ interests on the basis of DPM model. Thirdly, design a mechanism of personalized recommendation which is mixed with context clustering forward. In detail, there are two steps:step one is the analysis of fuzzy clustering on equivalent relation of users on the basis of users’contextual information which is according to the rule that people with similar context would be most likely to share the same hobbies. Before the predictions of personalized recommendation for readers, the paper make users clustered which helps to narrow down the scale of searching neighbors for the target users and improve the similarity of users as well. Step two is the modification of the original collaborative filtering algorithm, the CRR, namely common rating contribution rate, is introduced as a weight when users’similarity is measured. The function of the CRR is to reduce the wrong selection of neighbors when the data is sparse and alleviate the influence of hot items when make recommendation. As a result, the recommendation system can adaptively choose the most similar neighbors. Fourthly, apply the constructed mode of personalized recommendation to a certain provincial company of China Mobile Telecom to help improve the personalized recommendation service of mobile reading. Furthermore, a comparison between the modified collaborative filtering algorithm and the original collaborative filtering algorithm is involved to verify the effectiveness of the modified collaborative filtering algorithm and the involved context clustering. Last but not least, the paper offered detail suggestions of personalized recommendation of the mobile reading service for the company after the empirical research.