Research On Recommendation And Fuzzy Decision-making Related Algorithms Based On Vague Set Theory

With the rapid development of electronic commerce technology, people pay more attention to recommendation system because of its powerful ability to solve problem of information overloading. Multi-criteria fuzzy decision-making problems are widespread in the decision making field and have a broad application background, and its research is studied more and more deeply. As a means of dealing with fuzzy information, Vague sets(VS) has been rapidly development in recent years.At present a large number of applications based on Vague sets(VS) are proposed in many fields, such as pattern recognition, decision analysis, fuzzy control,object based image retrieval (OBIR), etc.Therefore, it has become an important research direction in data mining domain to apply VS theory to intelligent recommendation and fuzzy decision.In this dissertation, some related key techniques of recommendation of electronic commerce and fuzzy decision under Vague sets framework, such as the relationship between Fuzzy sets and Vague sets, similarity measures between Vague sets, Vague sets direct clustering algorithm, scoring function of Vague sets, recommendation algorithm based on Vague sets and multi-criteria fuzzy decision-making algorithm based on Vague soft sets, etc. The main work and innovations are listed as follows:(1) A new method of transforming Fuzzy sets into Vague sets and related application are proposed. This method can transform arbitrary places after the decimal point of the Fuzzy value, which make up the defects that existing methods can only transform Fuzzy value of two digits after the decimal point.(2) A novel clustering method of Vague sets similarity measures based on information similarity coefficient (ISCC-VSM) and its application are proposed. Firstly, we defined similarity measure of VS based on standard information similarity coefficient (ISC-VSM), then clustering based on its. Experimental results show that this method is effective and practical.(3) Two new recommendation algorithm based on VS named CBRA-VS and CFRA-VS are presented. CBRA-VS algorithm is described as follows:Firstly, the redundant and useless information of the recommended data are reduced. Secondly, we transform Fuzzy value into vague value of the recommended data, and the similarity of commodity are calculated by ISC-VSM method. Finally, we obtained recommendation formula according to the weights of each commodity. Experimental results on relevant data sets show that CBRA-VS algorithm is feasible, and the performance is superior to other content recommendation algorithms. Then CFRA-VS is described as follows:Firstly, the redundant and useless information of user-item evaluation matrix are reduced. Secondly, we transform fuzzy value into vague value of user-item evaluation matrix, then item similarity matrix are obtained by ISCC-VSM method, full user-item evaluation matrix are obtained by compensation value. Finally, according to the results of clustering based on similarity users to obtain recommended sequence. Experimental results on the Jester joke data sets show that CFRA-VS algorithm is feasible, and the performance is superior to other collaborative filtering recommendation algorithm.(4) Vague soft sets is introduced to the multi-criteria fuzzy decision, and a new multi-criteria fuzzy decision algorithm named FMCDM_VSS is proposed. Firstly, the property relationship of each candidate under constraints are calculated based on compound operation between Vague soft sets, and then score function of Vague sets based on connection number potential function are calculated and sorted. The related decisions experimental results show that this algorithm is an effective multi-criteria fuzzy decision-making algorithm based on Vague soft sets.