Study On Data Mining Methods Of Enterprise Relationship Value Analysis

With the rapid development of economic globalization, many enterprises have realized that they can adapt to the competitive circumstance in new period by seeking for, cultivating and developing cooperative partners to establish relational network. However, in the circumstance of collaboration among firms, it is one long-term strategy that the relationship hierarchy between enterprises and their partners is ascertained, which is determined by relationship value directly. Therefore, it is obvious that research on enterprise relationship value is of theoretical values and practical significances. The research perspective of current theories and analytical methods appears somehow single, which only makes a research on the relationship value between firms and customers. In order to reflect the contribution of enterprise relationship value to the core competitive capability and economic sustainable development, it is necessary to discuss the constitution contents of relationship value from the viewpoint of enterprises. Therefore, according to these problems, this dissertation has a deeper insight to offer an up-to-date framework of theory and method.Firstly, on the basis of summarizing the implication of enterprise relationship value at home and abroad, this dissertation defines the connotations on enterprise relationship value from one new point. According to making an analysis on main factors influencing the relationship value, the constitution dimension of relationship value is determined, and its creative process is described. According to the character of enterprise relationship value'analysis problem, the original measure indicators from different aspects are obtained by literature search and depth taklting on the basis of initial measurement indicators pricinple. Then, the related original measure table is built. At the same time, in view of pre-investigation, the original measurement indicators are purified by methods on internal consistence and exploratory factor analysis and the measurement indicators system of enterprise relationship value is established. At last, the data of enterprise relationship value analysis is got by formal investigation.Secondly, the extraction framework of analytical rules is designed to mine the regular hidden in the dataset reflecting enterprise relationship value, so as to sovle the problems on subdividing the enterprise relationship value and finding its group characters. This framework combining clustering and classification is described. At first, according to the characters that data indicators with high dimension and data boundary is difficult to be determined, the clustering ensemble algorithm based on system energy is designed to make clustering analysis to the relationship value of different cooperative enterprises, which can be divided into different groups. On the basis, the rule of enterprise relationship value belonging to each group is mined by C4.5 decision trees algorithm, and its group characteristic is found.Thirdly, according to the character that enterprise relationship value is of difficult quantification, this paper establishes the basic framework of class measure model on the relationship value of enterprises, which is used to solve the problems on enterprise relationship value recoganition and position. This frame structure is constructed on the basis of above-mentioned research in view of the main thought of the class measure of enterprise relationship value. At the same time, in accordance to the feature on indicators system of firms'relationship value, a kind of practical measuring model is constructed, which is based on the principle of FSVM with one-against-all method. This model is used to measure the relationship value classifically, and to determine its class.In addition, in order to improve the enterprise relationship value deeply, this paper also mines its promoting rules. Whereas the character of data set of the enterprise relationship value, this paper applies dynamic fuzzy logic theory to mine action association rules at first, and uses dynamic fuzzy membership function to fuzzy data set dynamically. On the basis, the concept and formalization description of fuzzy action association rules is put forward. Meanwhile, FARM algorithm is designed to mine the real numerical data of enterprise relationship value. Then, the promotion rules of relationship value are put forward.Finally, according to above-mentioned research results, the data mining prototype system facing the enterprise relationship value is designed and developed in this paper. An empirical analysis on enterprise relationship value is made by the system, in order to provide an effective decision support platform for cooperation between firms.