Research And Implementation Of Data Mining System Based On Improved Forecasting Model

As a leading industry, software and IT services industry largely leads and supports the development and progress of the current society. However, software and IT services industry is highly competitive, scale growth of the industry is showing explosive trend, the emergence of vast amounts of data make government departments into a distress situation of ’massive data and trace information’. In order to depthly study software and IT services industry, and dig out the useful information from the large amount of historical data, to ensure that the government be able to timely grasp each index’s operation situation of the industry, it is very important to predict the future trends and make the right decisions.In this thesis, predictive science is discussed. And the widely used prediction model algorithms are discussed and improved. The specific works are as follows:1. The thesis analyzes GM(1,1) prediction model’s modeling principle and the reasons of deviation, GM(1,1)-β model reselects initial value, reconstructs and optimizes the background value of GM(1,1) model, and in GM(1,1)-β model, it insures that the defect is breakthroughs which is the original GM(1,1) model fails when the development coefficient is greater than2.2. The thesis analyzes the limitation of GM(1,1)-β model, introduces the Markov theory, and puts forward a high accuracy forecasting model which is called GM-Markov and is based on GM(1,1)-β model. It is demonstrated by experiment that GM-Markov prediction model improves the flaw which the GM(1,1)-β algorithm predicts the non-exponential growth of data and random fluctuations in the data with larger prediction error and expands the use of scope greatly.3. The thesis also researches three exponential smoothing prediction algorithm, analysis algorithm bottlenecks and put forward a dynamic adaptive three exponential smoothing model. It bypasses the selection of initial value by establishing the analytical expression between the smooth initial value and the smoothing parameter, and uses IDS algorithm iterating to obtain the dynamic adaptive selection for the smoothing parameter, The improved prediction model dynamically acquires smooth initial value and the smoothing parameter dynamic, and through contrasting the current outstanding scholars’improvements, shows that the dynamic adaptive three exponential smoothing algorithm improves the prediction accuracy of model, and makes prediction results closer to the actual data.4. By using two improved prediction algorithm, the thesis completes the design of trend analysis module in "analysis system for software and IT services industry". In the forecasting indicators of software and IT services industry, the two improved prediction model algorithms made good prediction accuracy. It also verified that the two improved algorithms presented in this thesis has a certain practical value and theoretical promote significance.