论文部分内容阅读
本文提出一种LDA boost(Linear Discriminant Analysis boost)分类方法,该算法能有效利用样本的所有特征,并且能够从高维特征空间里提取并组合优化出最具有判别能力的低维特征,使得样本类间离散度和类内离散度的比值最大,从而不会产生过度学习,大大提高算法效率。该算法有效性在某商业银行的客户流失预测过程的真实数据集中得到了验证。与其他同类算法,如人工神经网络、决策树、支持向量机等运算结果相比,该方法可以显著提高运算精度。同时,LDAboosting与其他boosting算法相比,也具有显著的优越性。
In this paper, we propose a classification method based on LDA boost (Linear Discriminant Analysis Boost), which can make full use of all the features of the sample and extract and combine the low-dimensional features of the most discriminative ability from high-dimensional feature space. The ratio of inter-class to intra-class dispersion is the largest, so as not to over-study and greatly improve the efficiency of the algorithm. The validity of this algorithm has been verified in the real data set of the forecast process of customer churn in a commercial bank. Compared with other similar algorithms, such as artificial neural network, decision tree and support vector machine, this method can significantly improve the computational accuracy. At the same time, LDAboosting also has significant advantages over other boosting algorithms.