| Along with the further deepening of restructuring in the telecom enterprise, a new competition landscape has been formed in the market. Telecom enterprise has a large scale and rich historical data, in which there hides important decision-making information. Nowadays the competition in domestic telecom market has become increasingly fierce, so it has become a priority to mine management decision information for scientific decision for the telecom enterprise. Income forecast plays a vital role in the decision-analysis system and provides important decision support for decision-makers to make business development plans, to improve management decisions.As an important component of university-enterprise cooperation project, this paper takes the historical income data of19months of a telecom enterprise as the primary data, respectively using time series and grey forecasting method to train the data and to compare their results. It reveals that time series method is more accurate. However it can not meet the actual requirements.Through analysis, major income was selected from hundreds of items by using principal factor analysis method. Then time series method was used for training major income respectively, next the large random items of income were operated with intelligent processing technology such as linear regression, neural networks and support vector machine for error correction. It concluded that the method with support vector machines is the best one.In SPSS Clementine experiment environment, through the historical test data experiment, it was found that the model has high forecast accuracy. By generating a script to embed into the enterprise server for operation and detection, it indicated that the model completely meet the needs of the enterprise and its accuracy was much higher than other statistical methods. The research result was praised by enterprise users.Through the research work of telecom income forecast, the relevant theory of data mining and methods of using software system have been mastered, the fact is founded that telecom enterprises have a very wide and urgent demand for data mining, and data mining technology has great potential application value. Another functional module-set meal matching project is in process to meet the requirements of telecom enterprise. |
