基于改进加权协同过滤的集群用户黑箱个性意象预测

Prediction of black-box personality image of cluster users based on improved weighted collaborative filtering

为了解决传统意象定位中感性意象部分信息丢失及用户模糊的个性化需求不完全表达的问题,提出基于改进加权SO(WSO)算法的集群用户个性意象预测研究.建立用户特征域,基于K-modes算法计算用户差异度,确立用户集群.对甄选样本实施兴趣度排序及自主性意象评价,创建集群意象因子集.引入用户相似度优化WSO算法,增强集群用户间的内在联系,精准预测目标用户黑箱的个性意象分值.基于语义差异问卷及平均绝对误差分析验证黑箱意象,输出集群中单一用户的个性化意象,实现意象预测.以无人机为例,预测用户的个性意象,误差小于0.5被舍去,表明该方法能够较好地实现用户模糊的意象黑箱透明化,且预测的意象符合用户的个性化需求,可以有效辅助设计师有针对性地设计.

A prediction study of cluster user personality image based on improved weighted slope one(WSO)algorithm was proposed in order to solve the problems of partial information loss of perceptual image and incomplete expression of fuzzy personalized needs of users in traditional image localization.The user characteristic domain was established,and user cluster was established based on K-modes algorithm in order to calculate user differences.Interest ranking and independent image evaluation were conducted to create a subset of cluster image factors.The user similarity optimization WSO algorithm was introduced to enhance the internal connection between cluster users and accurately predict the personality image score of the target user’s black box.Black box image was verified by semantic difference questionnaire and mean absolute error analysis,and personalized image of single user in cluster was output to achieve image prediction.UAV was taken as an example to predict user’s personality image.The error less than 0.5 was omitted.The method can better realize the transparency of user’s fuzzy image black box and the predicted image meets user’s personalized needs,which can effectively assist designers in targeted design.