Research On Service Recommendation Model And Key Tech- Nologies In Mobile Cloud Computing Environment

With the popularity of mobile devices, such as smart phones and tablet PCs, mo-bile applications have been more and more widely used in our daily life. However, low hardware configuration, limited computing capability and large power consumption of most mobile devices have posed great challenges to those mobile computing applica-tions that having thriving demand of computing resources. Fortunately, these critical problems have been somewhat eased by the recent development of new Internet tech-nologies, such as cloud computing, big data and virtualization. In particular, mobile cloud computing technology has brought a huge breakthrough for the development of mobile applications. With mobile cloud computing, there are several key technical is-sues have to be tackled in order to widen its application fields. These include how to build a working framework suitable for mobile applications, how to improve the rec-ommendation efficiency of mobile cloud services, and how to solve the information overload problem on mobile devices. In summary, this paper places an emphasis on the service recommendation problem in mobile cloud computing, and the research top-ics include collaborative filtering based recommendation, context aware technology, as well as an application case. The major contributions of this paper are three folds, as follows.To solve the problem of high data sparsity and low accuracy of traditional service recommendation model when used in the mobile cloud environment, we propose a novel service recommendation model based on collaborative filtering and graph theory. We recognize that the key of solving this problem is to understand the process of data transform in real-world and to reduce the high dimensional data sparsity accordingly. We extend the hybrid collaborative filtering model by introducing a novel structure, i.e., PGraph, to describe the neighborhood. This structure can be used to create the optimized predicting order. The calculation of predicting order is based on a layered nearest neighborhood relationship and manifold similarity. Experimental results show that this procedure and its corresponding optimization can achieve better predicting accuracy with relatively high performance.To make the recommendation system within mobile cloud environment time-sensitive and diverse, we propose a temporal-aware hybrid collaborative recommen-dation method for cloud service. Based on analysis of the user preference change and time-sensitive recommended hot spot, temporal influence is integrated into classical neighborhood-based collaborative recommender algorithm. A user-service-time mod-el is introduced to mine the triadic relations of users, services and temporal informa-tion. Besides, to get an optimal recommendation, a temporal-aware latent factor model based on CP decomposition is proposed and combined to improve the recommendation performance.At last, to extend and optimize the above two models, we integrate some afore-mentioned technologies into a mobile cloud application:a real-time taxi trajectory monitoring system using the service recommendation based on the space-time percep-tion. The models mentioned above are applied into the two steps in the taxi manage-ment platform to verify the effectiveness of the service recommender models proposed. This system use the traffic data collected from GPS-equipped taxis to analyze social behavior and hence to improve taxi services. An online anomalous trajectory detec-tion method (OnATrade) is developed to analyze driving behaviors of taxi drivers. It mainly comprises two parts:In the first part, route candidates are generated by a route recommendation algorithm. In the second part, an online anomalous trajectory detec-tion approach is presented to find taxis which intend to have driving anomalies. This system has wide commercial applications, such as real-time public traffic supervision over taxis and QoS improvement of taxi services.