扩展算法在频繁行为模式分析中的优化研究

Optimization research of scalability algorithm in frequent behavior pattern analysis

对电子穿戴设备标准实时数据库进行频繁行为模式分析,主要通过多维度数据分析来获得更高的识别精度和识别效率.提出一种可扩展的频繁行为模式分析算法,先将采集数据转换为所定义的实体模式,对实体数据进行位置状态检测和时间转换的时空预处理;然后利用滑动窗口对时空预处理后的数据构建相似实体组,并过滤寿命周期过短的组来构建用户配置文件;最终实现模式分析识别.仿真结果表明,对比est Des+和Stream Mining两种算法,新算法能有效的提高频繁行为模式分析的识别精度和识别效率.

The analysis of the frequent behavior patterns on standard database of electronic wearable devices is mainly through the multivariate data analysis to obtain higher identify accuracy and identify efficiency.This paper proposes a scalable algorithm for frequent behavior patterns analysis.The proposed algorithm first converts the collected entity data into a defined entity model,and performs spatio-temporal preprocessing on entity data of position state detection and time conversion; then,uses sliding window to construct the similar entity groups for the preprocessed data,and filters the groups with short lifetime to construct user profiles; finally,achieves pattern analysis and identification.Simulation results show that compare with est Des ＋ and Stream Mining,new algorithm can effectively improve the identify accuracy and efficiency of the frequent behavior pattern analysis.