遗传算法在数据挖掘中的应用  

The Application of Genetic Programming on Data Mining

作　　者：李康顺[1,2,3] 李元香[1,3] 滕冲[4] 王玲玲[1,3] 

机构地区：[1]武汉大学软件工程国家重点实验室 [2]江西理工大学信息工程学院,江西3410003 [3]武汉大学计算机学院,武汉430072 [4]武汉大学计算机学院

出　　处：《计算机工程与应用》2005年第9期219-223,共5页Computer Engineering and Applications

基　　金：国家自然科学重点基金项目(编号:60133010);高等学校博士学科点专项科研基金项目(编号:20030486049)

摘　　要：论文介绍一种利用遗传算法的演化计算机编程的方法来自动建立一种动态非线性数学模型进行数据挖掘,并进行社会经济趋势预测和回归曲线拟合,改变过去那种只使用拟合粗糙、预测结果准确度差的传统预测模型进行曲线拟合和趋势预测的分析方法。在数据实验中,使用遗传算法演化计算机编程的方法自动生成的演化模型对一些真实的历史资料进行了曲线拟合和发展趋势预测以及对前馈、后馈误差进行了深入分析。结果表明,使用该方法建立的演化模型要比传统的线性回归、指数回归、抛物线回归三种固定传统数学模型所预测的数据准确度高很多,而且拟合曲线的前馈标准差和预测后馈标准差也明显要小。In this paper we proposed an application of genetic algorithm in data mining that automatically build a dynamical nonlinear mathematic model by using evolving computer programming of genetic algorithm to predict the social economic trends and fit the economic developed curves.This method can obviously improve the traditional methods of economic predicting and economic curves fitting which could only obtain rough fitting curves and unsatisfactory forecasting results.In the numerical experiments,we used the evolutionary model generated by using genetic algorithm to fit the regression curves and forecasted future developing trend using true historical data,and then the estimating mean square errors and the forecasting mean square errors were calculated.The numerical results indicate that the accuracy calculated by evolutionary model is obviously much higher and the estimating mean square error were the forecasting mean square error are obviously much less,compared with the other three traditional regression models(linear regression model,exponential regression model and parabolic regression model).

关 键 词：遗传算法 数据挖掘 演化建模 误差分析 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]