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

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.