数据挖掘中应用偏最小二乘法发现异常值

Using Partial Least-Squares Regression to Found Abnormal Data in Data-Mining

本文介绍了偏最小二乘法在挖掘统计数据中的异常值上的应用。对统计数据使用偏最小二乘法回归,可以较快速地建立相对准确的回归方程,并且结合变量投影重要性指标等相关分析工具可以对数据进行消元,降低数据的维数,从而减轻运算的强度;通过预测值和实际发生值的比较,根据法则判定数据是否异常,从而达到对数据风险的控制,在经济、金融等领域中有着重要的实际意义。

This paper introduces the use of Partial-Least-Regression to find abnormal data in mining statistic data. This kind of method is helpful to set up relatively exact functions. With the help of Variance Important Project and other analytic tools it can also remove parameters, depress data dimensions, and alleviate the intension of operation. According to the principle, this method can judge whether the data is abnormal by comparing forecast data with real one, so that it can restrain the risk of data. This usage plays an important role in economy and finance.