Research On Data Mining Algorithm And Its Application In Electronic Commerce

With rapid development of modern information and network techniques, electronic commerce is now influencing the economic operation structure of whole society as a hot economic activity. The development platform and solutions to electronic commerce system are maturing. The technology of data mining will be one of the most important application technologies in electronic commerce system. Data mining will provide strong support for commerce decision of electronic commerce company.This paper studies the relative algorithms of data mining, and the application of data mining technologies on electronic commerce platform. Taking data mining technologies as the key, this paper studies deeply the attribute-oriented induction algorithm and neural network prediction technology with the analysis of electronic commerce. Then this paper applies the two methods to the implementation of data mining system that is electronic commerce-oriented. The main contents and research achievements of the paper include:(1) The paper gives a briefly introduction of data mining, including concept, function and data mining process. According to the characteristic analysis of electronic commerce platform, it propounds a data mining system based on the platform, then how to design the electronic commerce-oriented data mining system is discussed in detail.(2) In the paper, the traditional attribute-oriented induction algorithm is improved. Based on the discussion of the updating frequently characteristic of sale data, the paper applies the increment to AOI algorithm. During the process of data generalization, the mining object is new sale data instead of the whole database. So the increment attribute-oriented induction reduces algorithm's search space and increases the efficiency of system.(3) Based on the discussion of disadvantages of the present BP algorithm and study of improvement methods propounded by other people, the paper explores a new improved BP algorithm. There are two aspects of the improvement: changing the gradient of Sigmoid function according to the error in training and adjusting neural network's learn rate automatically.(4) Based on the neural network algorithm and taking the stock prediction on the electronic commerce platform into consideration, the paper introduces some methods of designing neural network model. For example, it propounds how to select and treattraining samples beforehand, which time sequence is suitable, how to use samples while training neural network etc.(5) Through deep study and summary of the actual project development experiences, the paper tries to smoothly integrate the business logic and data mining functions on electronic commerce platform and form a simple and easy-to-use data mining system. The paper introduces some functions and business values of the data mining system, and then analyzes the optimization of the increment attribute-oriented Induction algorithm and improved BP neural network prediction technology, to the data mining system.