User Behavior Attitude Extraction In Mobile Environment

With the rapid development of mobile applications, the advent of the era of "Internet +", based on the intelligent hardware and wearable devices continuously explore aspects of academic research, user behavior recognition as the basis and prerequisite for the development, gradua lly attention by the academic researchers. Behavior gesture recognition is designed to transmit data through the sensor to judge the performance of the user behavior. In order to detect the user’s behavior in the mobile environment, this paper puts forward a kind of based on schedule and Support Vector Data Description(SVDD algorithm(Support Vector Data Description) interactive classification Agent(Agent).Agent in smartphone software, by using the algorithm of support vector data description is user preset schedule and his current location latitude and longitude, and the user acceleration for the continuity of user behavior. The user identity of continuity behavior is implemented by the user’s mobile Agent. The phone system first compares the state of the user’s behavior in the schedule and plan, to determine whether the actual behavior and the deviation of the plan, to determine the current user behavior type and behavior.If the current location data computed by support vector data description a lgorithm, is on the hypersphere border or within the scope of the hypersphere, user behavior is seen as a business behavior, or called the normal behavior. Otherwise, the user’s behavior is regarded as abnormal behavior, will be through the mobile phone software prompts to confirm whether it is to follow or deviate from the plan. Within the scope of the normal calculation, this artic le will through neural network pattern recognition method for user profile for identification. Based on hypersphere support vector data description(SVDD) algorithm calculation results, the standardization of public data sets PAMAP2, we get the user behavior of the sensor data is a has 54 d high-dimensiona l data quantity, and using the princ ipa l component analysis to deal with high-dimensiona l data to lower dimensions, and obta in higher contribution rate of low dimension of sample data, and then the neural network pattern recognition mode l simulation, puts forward a method to identify the user behavior profile. This article has four different of pattern recognition neural network. Between each neural network pattern recognition is noninterference, does not have any effect. In this way, we can to a large extent reduce the cost of the neural network training, improve the efficiency of the whole training success. As for four neural network pattern recognition, reflects the characteristics of the high efficiency of this model, and ultimately derived the accuracy of the result set. Some interaction between mobile phone and user technical basis, based on the mobile phone vibration, the bell and graphical interface is discussed. Finally, by MATLAB software to verify the proposed algorithm in the accuracy of user behavior recognition is effective. The results show that the algorithm is viable.