| It has accumulated a large number of user data in the process of the infiltration from internet to industry. Vast amount of user data contains a wealth of information, which has become the most valuable resources of the computer era. Data mining technology and cloud computing technology are designed to tap value for user data. User data contains user behavior characteristics, and user behavior is usually associated with a variety of social factors and technical parameters, which will affect the user’s role and characteristics of different scenarios. A very important criterion for measuring user behavior is the activity level of user.This paper discusses the user characteristics based on the background of big data, and presents a combination algorithm based on fuzzy decision tree and echo state network for user activity prediction. Then, we propose the possible applications of it in the future. Specific content as following:Firstly, this paper introduces the background of the research and gives the relevant theories. We investigate the research of user characteristics and the current situation of activity analysis by combining the existing data mining technology and summarize the key technology of neural network and decision tree. Then we discuss the time sequence and axiomatic fuzzy set theory which are the theoretical basis for this study.Secondly, we analyse the characteristics of user data in the mobile internet network to find out the general rules.Again, in order to adapt to the multidimensional fuzzy attributes of user data,we use the fuzzy decision tree generation rules as the weights of input layer to the hidden layer in the wavelet minimum complexity Echo State Network(WMCESN), generating Semantic Driven Echo State Networks(SDESN). Inheriting the structure of WMCESN, the SDESN is simple and has good performance, and when considering interval value and multi-label data, the SDESN could overcome the sharpness of two value classification and give a soft intermediate category. In addition, we also carried out a series of performance simulations of the SDESN algorithm. Then we use SDESN to forecast the activity level of the mobile internet users, tracking the potential loss of customers in the early stage and provide the perspective of enterprise decision-making.Finally, the paper discusses the role of user characteristics and activity analysis in the future network, and presents the application of SDESN algorithm in the future. |
