Improvement And Application Of Decision Tree With Covariance&Information Entropy

The21st century is the century of information explosion With the rapid development of World Wide Web, the amount of information is increasing fast, and it has become the geometric level of growth. Development and optimization of increasing mass data processing and analysis technology is a long-term and arduous challenge. Meanwhile, massive amounts of data implied valuable knowledge model will promote the development of modern enterprises. Data mining techniques also came into being, the technology in the massive amounts of data to find useful knowledge that will eventually assist decision makers to deal with the decision-making problems faced by the various stages of the enterprise based on reliable analytical processing mode. Thereby to help the modern enterprises enhance their competitiveness.In this paper, the analysis is based on the basic of the present situation of data mining in the resolving related problems of data forecast. We usually use the technology of classification. But traditional data mining techniques cannot effectively reflect the feature of the data that generally have relations in each other. According to this feature,in this paper taking the measure of Co variance and correlation coefficient to construct the logical relationship in multi-dimensional data, and combining with decision tree algorithm that based on information entropy theory. This algorithm optimizes the traditional decision tree algorithm and improve the accuracy and usefulness of data mining. With experimental comparison analysis of the experimental results, and analysis of the improved performance of the algorithm, it comes to conclusions features of the performance of this algorithm and what environment is fit this algorithm. In the end,put the improved algorithm into the online Home Decor system which is a branch o f popular e-commerce system. Analyzing the schema ofbuying Home Decor, and constructing the improved decision tree.Assessing the result of data mining with the improved decision tree, and coming to a conclusion of the customer buying habit and buying interest.In the end using the data verify the correctness of using this improved algorithm.