Font Size: a A A

Recommendation Algorithm Based On User Behavior Mining Improvement And Application Of Research

Posted on:2012-08-08Degree:MasterType:Thesis
Country:ChinaCandidate:N N WangFull Text:PDF
GTID:2208330335480134Subject:Computer application technology
Abstract/Summary:PDF Full Text Request
The existed sequential pattern mining algorithms is already in large database get some degree of application. However, with the development of mining technology and the application fields detailed continuously, higher requests for exhumed sequential patterns put forward to the users. So, the research focuses is make the user's requirements or interests into one or more restraints, to limited the dimensions and improve the efficiency of mining. Those mining research has into focus areas of mining. This article in view of the above problems, mainly focused on the sequential patterns based on constraints on mining algorithms. Those research problems in the supermarket customers buy mode, web site access page sequential patterns, telecommunications alarm sequence mode, DNA pattern and so on has important significance.In this paper, the existed sequential pattern mining algorithms in domestic and international and constraint-based sequential pattern mining algorithms is studied and analyzed. Constraint-based sequential pattern mining is based on sequential pattern mining and considers some other information, like age, gender or other information in analysis of customers, purchasing behavior. Such patterns combine more information and the value of application is high.Based on the sequential pattern mining method for research and study, in the process of pretreatment and sequential patterns mining process has been improved respectively. In the data pretreatment processes, fuzzy classification has been proposed first, use of the sampling method in mathematical to narrow the dig scale. The sequence study of fuzzy classification, based on the original method of various similarity measures, a method of directed graph based on similarity of time series was presented. The concept of the graph has been introduced to mining process, and considering the sequence of different length, take their average to calculate the length of similarity. In the process of sequential pattern mining, with Prefixspan algorithm of the sequential pattern mining thinking as a starting point, for the limitations of the sequential patterns algorithms, an improved algorithm integrated with time interval and click quantity was presented. the concept of "support degree"has improved, introduce a new concept of frequent degree and the time attribute, and joined the factors of time interval and click quantity, thus made the mined dates had the real-time character. Conducive to decision-makers use the information more accurate.
Keywords/Search Tags:data mining, directed graph, similarity distance, Prefixspan algorithm, click quantity
PDF Full Text Request
Related items