一种基于改进BP神经网络的办公座椅舒适性评估方法

目的提出一种基于改进BP神经网络(PCA-BP)的办公座椅舒适性评估方法。方法对4款不同造型、材质的办公座椅进行臀部及背部Tekscan压力测试,并通过五级Likert量表记录被试的心理舒适值。在多重相关性检验的基础上,利用PCA-BP算法建立办公座椅舒适性的预测模型,并对模型的有效性进行验证。结论模型拟合的均方误差为0.164,预测值和心理值配对样本t检验显著相关,相关系数为0.918。与普通BP算法比较,基于PCA-BP算法的办公座椅舒适性预测效果更佳,为办公座椅舒适性评估提供了更好的评估方法。

Evaluation Method of Office Seat Comfort Based on Improved BP Neural Network

A evaluation method of office seating comfort is put forward based on improved BP neural network (PCA-BP). 4 office chairs with different appearance and material for hip and back Tekscan stress are tested, and the value of psychological comfort through the five-level Likert scale is recorded. On the basis of the multiple correlation test, PCA and BP algorithm is used to establish the prediction model of office seating comfort, and the validity of the model are verified. The mean square error(MSE) is 0.164. Predictive and psychological value is significantly associated with paired samples T test and the correlation coefficient of 0.918. Compared with ordinary BP algorithm, prediction effect of the office seat comfort is better based on PCA-BP algorithm, which provides a better evaluation method for the office seat comfort.