摘要
针对传统Web用户兴趣特性因子提取方法存在数据挖掘准确度低,导致用户个性化推荐效果不佳的问题,在大数据背景下,提出一种基于改进粒子群优化的K-means算法。首先,对PSO算法从粒子惯性权重和学习因子、转换时机、粒子变异操作进行改进;然后利用改进的PSO算法优化K-means算法,由此得到MPSO-K-means算法;最后通过MPSO-K-means算法加快收敛,增强Web用户数据聚类和挖掘效果。结果表明,进行5次聚类的K-means算法和PSO+K-means算法的总体纯度平均值分别为0.68和为0.84,本算法的总体纯度值为0.96,比前两种算法分别高出了0.29和0.185。对比于其他经典算法,本算法的聚类纯度值分别高出了0.18和0.115。由此说明,提出的MPSO-K-means算法的数据挖掘和聚类效果明显更佳,通过本算法可实现Web用户兴趣特性因子准确提取,提升用户个性化推荐效果。
In view of the traditional Web user interest feature factor extraction method,which has low data mining accuracy,which leads to poor effect of user personalized recommendation,a K-means algorithm based on improved particle swarm optimization is proposed in the background of big data.First,the PSO algorithm is improved from particle inertial weight and learning factor,conversion timing,and particle variation operation;then,the improved PSO algorithm is optimized to obtain the MPSO-K-means algorithm;finally,the MPSO-K-means algorithm is used to accelerate convergence and enhance the clustering and mining effect of Web user data.The results show that the overall purity average of K-means and PSO+K-means is 0.68 and 0.84,respectively,and the overall purity value of this algorithm is 0.96,which is 0.29 and 0.185 higher than the previous two algorithms,respectively.Compared with other classical algorithms,the cluster purity values of this algorithm are higher than 0.18 and 0.115,respectively.This shows that the data mining and clustering effect of the proposed MPSO-K-means algorithm is obviously better.This algorithm can accurately extract the factors of Web user interest characteristics and improve the personalized recommendation effect of users.
作者
宁莉莉
NING Lili(Xianyang Vocational Technical College,Xianyang Shanxi 712000,China)
出处
《自动化与仪器仪表》
2023年第9期41-45,共5页
Automation & Instrumentation
基金
陕西省职业技术教育学会2022年度课题《在线开放课程线上线下混合式教学模式研究》(2022SZX285)。
