The Decision Tree Classifier Hope In The Sales Management System

With the constant development of database technology and the wide application of the database management system, the data amount stored in the database has been increasing sharply, and behind a huge amount of data lies a lot of important information. Data mining is a process in which we extract reliable and useful information buried among vast amounts of data in the database. Classification is one of important methods used in data mining. It finds out a function or builds a model (we usually call it as classifier) from existing data. The function or model can map a record in the database to a pre-assumed class, thus applying to the data forecast. Decision tree classifier as one type of classifier is a flow-chart-like tree structure, where each internal node denotes a test on an attribute (attribute value), each branch represents an outcome of the test, and each leaf node represents a class. The decision tree can be easily expressed by IF-THEN rule. The learning process of the decision tree involves the selection of training data, while the credibility of the decision tree involves assessing validity. This can be solved with improved K-fold cross-validation. Information gain can be used for the standard of dividing consecutive attributes and selecting split attribute. "Hope Sale Management System" is software developed to meet the requirement of ZhengDa Cement Manufacture Industrial Limited Company for information-oriented business management. Its branch system, "Receivable Management Subsystem"is my independent effort. Receivable is the outside assets of one business to be collected from another business on sale of products, materials, labor service on credit. I use decision tree classifier technology in "Receivable Management Subsystem", and the algorithm that creates the decision tree is realized on the basis of improved ID3 algorithm. The function of this decision tree classifier is to predict whether receivable can be repaid in time. "Receivable Management Subsystem"has been test-run for nearly six months with initial success.