Research On Decision Tree Classification Algorithm And Application In The Electric Power Marketing

As computer technology is widely used in electric power companies, electric power sectors have accumulated a large amount of operational data and non-operational data, which record the operating conditions of electric power supply businesses for many years. The responsible officers relating to electric power need for doing analysis deeply for these historical data to obtain valuable information and provide rational decisions for decision makers based on this. Data mining is a process which valuable information is found in massive data using a variety of data mining algorithms, decision tree induction is widely used among them because its computation is relatively small, you can display important decision attributes, explicit rules are extracted easily and it has a higher classification accuracy rate etc..This article focuses on C4.5algorithm in the decision tree for analysis and has done corresponding improvement for the existence of a defect. First introduces the basic idea of decision tree algorithm and algorithm framework, and then elaborates on classical algorithms ID3and C4.5algorithm, and then improves attribute selection criteria of the C4.5algorithm, in which the specific idea improved is to use Taylor formula, Maclaurin formula and the equivalent infinitesimal theory in combination with information entropy for changing the test standard of attribute selection, and extend it to a variety of categories, and finally does contrast experiment about the two algorithms using classic example in the UCI, experimental results show that the improved algorithm reduces the computational cost, improves the efficiency of decision tree classification and shows a good classification performance.In order to achieve electric power marketing analysis based on improved C4.5algorithm, this paper designs the electric power marketing decision-making system, describes implementation process of the system and the result show. Before using classification algorithm of the decision tree, does the data preprocessing work for the data in the database, processes the vacancy data and translates continuous-valued attribute for discretization. In the specific electric power marketing analysis process, carries on data mining for customer classification and sale of electricity market, generates decision tree and classification rules, does VIP classification for use of electricity customers, identifies the major factor impacting sale of electricity market, extracts and finds valuable information according to different needs to reveal the internal rule between electric power market and marketing and provide basis for the management and decisions of electric power companies.