Study On Search-charts Based On Online User Behaviors

Internet has become the most important infrastructure of the whole society."Search-engine" and "online music" are in the first place of all of the online applications either from the growth rate or the scale of uses. The music search charts is the combination of those two applications, so the number of its users is huge. Therefore, the analysis of two basic characteristics of network applications and user features has its significance for our understanding of user behavior to promote the development of the network applications.This paper researches the music search ranking, especially its relevant characteristics, from the aspects of the characteristics of search ranking, impact factors, user’s counterproductive, and sociology structure. The work is based on real network environment. Actual data is analyzed from multi aspects such as time sequence, the classification features as well as the list of the change based on network measurement. And then, the interaction between search ranking and network users is studied by means of theoretical deducing and simulation. An improved group structure search algorithm-weight label propagation algorithm is proposed according to the time characteristic of ranking. Then, the topology and dynamic characteristic of user behavior is analyzed through social network.The main innovations of the thesis are listed as follows:1. An influence model of search charts is proposed. To analyze the influence of the search charts, the impact factor and impact time are proposed. Based on those two parameters, the mathematical model for the influence of the search charts is proposed. After validation of the real data and simulation results of two methods, the validity of the model is confirmed. The model can be obtained by the two conclusions:First, the initial stage of ranking reflects the objectives of the users. Second, the ranking will have differences with the users’objectives within a certain time, but the length of time is related to the influence of the charts and the differences between the ranking and the users’ objectives.2. The influence model of "brush chart" is proposed. In the field of web-based user behavior we found that there is a special class of groups that just focused on some particular keyword. We conduct this special class is defined as "brush chart" behavior. "Brush chart" has a great impact on the fairness and the authenticity of the network chart data. The impact of the "brush chart" phenomenon is given by theoretical analysis and verified by simulation. The mathematical model analysis showed that there is a mutual restraint and reinforcing relationship between the influence of the "Brush List" and the influence of the search charts. The rankings change randomly because of those two factors.3. In the field of network structure analysis, a novel social network search algorithm-the weight label propagation algorithm is undertaken. In that algorithm, the search keywords are defined as network nodes. The network edges mean that the two keywords have being on the same list for once while the edge weights are equal with the number of occurrences. The algorithm was undertaken with the analysis of the weights of edges. The algorithm is also applicable to other areas.4. The features of MSC group structure are extracted. In the field of MSC social work, the group structure of the search-charts is got with WLPA. According to the analysis of structural characteristics of groups we found two characteristics:First, there’s the core category keywords which means there are core ones in all of the online users. Second, the social network structure of MSC shows a particular characteristic in time sequence. Those two characteristics of the group structure illustrate the same ones of the web-based users’behaviors. The characteristic of MSC group structure is the theoretical basis of the classification and recommendation of online users.