基于支持向量机的室内舒适度评价方法

Indoor Comfort Evaluation Method Based on Support Vector Machine

针对室内环境因素多元化、动态变化的特点和目前评价方法的不足,建立了基于支持向量机的室内舒适度混合评判模型。首先将从真实环境中采集的数据集进行数据规范化处理;然后根据群体和个体感觉,分别用离线训练和在线训练的方法训练分类器;最后使用训练好的分类器预测样本的标签。以Matlab为开发工具,编写了基于支持向量机的室内舒适度评价算法,并与BP神经网络和概率神经网络等室内舒适度评价算法进行了比较,仿真结果表明,该方法是可行且有效的。

For the diversity and dynamic changes of indoor environmental factors, and lack of the current evaluation methods, put forward an indoor comfort mixed-evaluation model based on support vector machine. Firstly, the method handled dataset collected from the real environment with the method of data standardization. Secondly, according to the group feeling and individual feeling, it used off-line and on-line to traln the classifier. At last,it used the trained classifier to predict the label of the sample. The indoor comfort evaluation algorithm based on support vector machine was written with the developing tool Matlab ,it compared with the BP neural network and probabi- listic neural network （PNN）. The simulation results of training and validation show that the method is feasible and effective.