Freight Invoice Based On Decision Tree Data Mining System

With the development of the railway informatization, plentiful data has collected and stored in the freight bill information system which acts as the sub-system of the railway information system. How to use the current resource reasonably and acquire valuable decision-making information with a little labor and technique cost, has become an important portion of the work of the marketing department and IT department. The rapid development of the data mining technology has established a good base for the railway freight marketing analysis. But the present-day data mining tools mostly base on data warehouse, OLAP server, or data file etc, which makes them can not be applied in the current freight bill system directly.Aiming at the facts that there is no data warehouse in the railway information system and that the end-user's database technique is not well enough, combining intensively with the railway freight marketing analysis, we have researched and designed a data mining system called HPMiner adopting the technology of decision tree classification. The system bases on the decision tree and takes the OLTP database as its data source. We have integrated several functions such as preprocessing, decision tree generating, classification rules extracting, statistic analysis, predicting, etc, into the basic system of HPMiner. The discretization procedure of continuous attribute, which bases on OLTP database and is concrete problem oriented, can be done dynamically, so it reduces the data mining tool's demand for data source. On the other hand, the discretization procedure is also end-user oriented and conveniences the user by providing the following two methods: specifying the number of discretization zone and setting the threshold. It accommodates the complex data in the real information system well. HPMiner, which bases on the main idea of decision tree classification algorithm ID3 & C4.5, is a client-server system. It is developed with VB.NET language and connects to Oracle or SQL server through ADO.NET. The system of HPMiner can be easily integrated with the freight bill information system for its good design and interface. HPMiner has been applied in the railway fright marketing analysis, and accomplish several concrete tasks such as earnings analyzing of insured freight transportation and direction analyzing of freight flow.The research of HPMiner combines the decision tree classification technology with the current railway freight bill information system compatibly, which makes the end-user can conveniently dig out the knowledge to guide the production. On the other hand, it exploited a new realm for the application and research of the decision tree classification technology.