基于动态神经网络的消费者购买行为特征的挖掘

Mining Consumers' Purchasing Behavior Characteristics Based on Dynamic Architecture of Neural Network

分析顾客消费需求,掌握顾客行为偏好,已成为当今企业制定正确营销策略的重要手段。在已有神经网络基础上,提出了动态神经网络结构算法,对顾客交易数据样本进行分类,在该算法中,隐藏层及单元结点不固定,而是依据样本训练结果,动态地确定神经网络的隐藏层层数及每个隐藏层的单元结点数。此外,为提高样本分类的准确性,最大化输出函数yk,采用遗传算法对样本实施优化。最后,通过案例数据对动态神经网络结构算法进行了验证,成功提取了顾客购买行为规则。

It has been a very important means for firms to analyze consumer＇s needs and understand consumer＇s behavior preferences for drawing up correct marketing strategies. On the basis of Neural Network, this paper presents an algorithm of modeling dynamic architecture of artificial neural networks （ANN） for classifying consumer＇s transaction data sets. In this algorithm, the architecture of ANN is not arbitrary, and the number of hidden layers and the number of hidden nodes are sequentially and dynamically generated until a level of performance accuracy is reached. In addition, to improve the accuracy of samples segment, this paper uses genetic algorithm to optimize the sample and to find the optimal values of input attributes （ chromosome）, which maximizes output functionykof output node k. Finally, by applying a given database for customers buying computer, this paper validates the algorithm and successfully extracts consumer＇ s behavior rules.