基于鲸鱼算法优化的BP神经网络的尿液成分浓度预测方法  

Prediction Method of Urine Component Concentration Based on BP Neural Network Optimized by Whale Algorithm

作　　者：黄建邦 于源华[1] 陈启梦[2] 唐春雨 王哲 HUANG Jianbang;YU Yuanhua;CHEN Qimeng;TANG Chunyu;WANG Zhe(School of Life Science and Technology,Changchun University of Science and Technology,Changchun 130022;School of Opto-Electronic Engineering,Changchun University of Science and Technology,Changchun 130022;Changchun Chunqiu Technology Development Limited Company,Changchun 130012)

机构地区：[1]长春理工大学生命科学技术学院,长春130022 [2]长春理工大学光电工程学院,长春130022 [3]长春市春求科技开发有限公司,长春130012

出　　处：《长春理工大学学报（自然科学版）》2024年第5期89-98,共10页Journal of Changchun University of Science and Technology(Natural Science Edition)

基　　金：吉林省科技发展计划项目(20240404062ZP,20220201092GX)。

摘　　要：针对尿液成分浓度实时分析精度低的问题,提出了一种颜色传感器结合WOA算法优化BP神经网络的预测方法。利用颜色传感器采集尿液试纸条上各试剂块的RGB值,并通过白平衡原理校准数据,利用Kubelka-Munk和Beer-Lambert定律构建反射光与浓度的关系。通过参数优化模型最小化目标函数以消除系统误差,基于最小二乘法建立色标卡颜色值与浓度的数学模型。采用WOA算法优化神经网络的权重和阈值,并使用大量数据训练BP神经网络对颜色值与浓度进行回归分析。实验结果表明,预测值与真实值的MAE为3.1415,RMSE为4.328,R2接近1,WOABP神经网络模型对尿液成分浓度预测具有高精度和准确性。Aiming at the low accuracy of real-time analysis of urine component concentration,a prediction method of color sensor combined with WOA algorithm to optimize BP neural network is proposed.The RGB values of each reagent block on the urine test strip were collected by the color sensor,and the data were calibrated by the principle of white balance.The relationship between reflected light and concentration is constructed by Kubelka-Munk and Beer-Lambert laws.The objective function is minimized by parameter optimization model to eliminate system error.Based on the least square method,the mathematical model of color value and concentration of color label card is established.WOA algorithm is used to optimize the weights and thresholds of neural network,and a large number of data are used to train BP neural network for regression analysis of color value and concentration.The experimental results show that the MAE between the predicted value and the real value is 3.1415,RMSE is 4.328,and R2 is close to 1.The WOA-BP neural network model has high precision and accuracy in predicting the concentration of urine components.

关 键 词：尿液成分浓度 颜色传感器 鲸鱼算法 BP神经网络 

分 类 号：R318[医药卫生—生物医学工程]