基于改进LVQ聚类的移动支付用户活跃度研究

Study of Mobile Payment User Activity Based on Improved LVQ Clustering

移动支付App用户消费活跃度易受到大量内外部因素的影响,为了提升移动支付App用户分类的准确性、分析各类群体消费活跃度的影响因素,提出一种采用层次分析法进行特征加权和基于置信区间进行迭代终止条件优化的学习向量量化(LVQ)聚类算法。应用于某移动支付App用户群体划分及用户画像构建,同时与随机森林回归算法相结合,分析移动支付App日活跃用户数(DAU)指标的影响因素,以便指导该App下一阶段的运营推广策略。算法性能实验结果表明,改进算法的聚类性能和时间性能较改进前分别提升了22.8个百分点和14.5个百分点。

The consumer activity of mobile payment App users is vulnerable to a large number of internal and external factors.In order to improve the classification accuracy of mobile payment App users and analyze the impact factors of each user group’s consumer activity,an improved clustering algorithm of learning vector quantization(LVQ)is proposed by analytic hierarchy process(AHP)for feature weighting and optimizing iteration termination conditions based on confidence interval.The improved algorithm is applied to divide mobile payment App user groups and build user portraits,and combined with random forest regression model to analyze impact factors of daily active user(DAU)index to guide the operation and promotion strategy of mobile payment App in the next stage.The experimental results of algorithm performance show that the clustering performance and time performance of the improved algorithm are improved by 22.8 percentage points and 15.5 percentage points respectively.