基于PSO的RF模型在人体活动识别中的应用  

Application of RF model based on PSO in human activity recognition

作　　者：倪洪科 王斌[2] 王英超 高慧敏[2] Ni Hongke;Wang Bin;Wang Yingchao;Gao Huimin(School of Computer Science and Technology,Zhejiang Sci-Tech University,Hangzhou,Zhejiang 310018,China;School of Information Science and Engineering,Jiaxing University;Suzhou Lanhepenbo Intelligent Technology Co.,Ltd)

机构地区：[1]浙江理工大学计算机科学与技术学院,浙江杭州310018 [2]嘉兴学院信息科学与工程学院 [3]苏州蓝赫朋勃智能科技有限公司

出　　处：《计算机时代》2023年第5期131-135,共5页Computer Era

基　　金：住房和城乡建设部2022年科学技术计划项目(2022-K-104);嘉兴市公益性研究计划项目(2020AY10012)。

摘　　要：提出一种基于粒子群优化(PSO)的随机森林(RF)识别方法。利用PSO算法搜寻最优的RF超参数n_estimators和max_depth,构建了PSO-RF人体活动识别模型。基于华盛顿州立大学CASAS项目数据集的实验共识别30种日常活动。仿真结果表明,PSO-RF模型的识别准确率达到95%,Accuracy、Precision、Recall和F1-score评价指标均优于其他经典的分类模型,具有较好的预测精度和泛化能力,可为智能家居系统个性化服务提供辅助决策。In this paper,an RF recognition method based on PSO is proposed.Using PSO algorithm to search for the optimal RF hyper-parameters n_estimators and max_depth,the PSO-RF human activity recognition model is constructed.An experiment is conducted on the CASAS project dataset of Washington State University,and a total of 30 daily activities are identified.The simulation results show that the recognition accuracy of the PSO-RF model reaches 95%,and the evaluation indicators of Accuracy,Precision,Recall and F1-score are superior to other classic classification models.It has good prediction accuracy and generalization ability,and can provide auxiliary decision-making for personalized service of smart home system.

关 键 词：随机森林 粒子群优化 人体活动识别 传感数据 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]