基于集成神经网络的织物主观风格预测研究

针对织物主观风格评价问题,建立了基于D-S证据理论的集成神经网络风格预测模型。分别采用径向基函数(RBF)神经网络和反向传播(BP)神经网络进行预测,得到初步的预测结果,经过归一化后得到2组基本概率分配函数,运用D-S证据理论进行融合得到最终预测结果。对510块来自中国国际纺织面料及辅料博览会的面料进行试验表明,使用基于D-S证据理论的集成神经网络进行织物主观风格预测比单一神经网络的准确率最高可以提高17.74%,将集成神经网络模型通过织物成分分类器优化,预测准确率可以更高。

Prediction of Fabric Subjective Style Based on Integrated Neural Network

Aiming at the problem of fabric subjective style evaluation,an integrated neural network style prediction model was established based on D-S evidence theory.The radial basis function(RBF) neural network and back propagation(BP) neural network were used to predict respectively,and the preliminary prediction results were obtained.After normalization,two groups of basic probability distribution functions were obtained,and the final prediction results were obtained by using D-S evidence theory.Experiments on 510 fabrics from China international textile fabrics and accessories fair showed that the prediction method of fabric subjective style with integrated neural network based on D-S evidence theory was more accurate than the prediction method of single neural network,and the highest accuracy could be increased by 17.74%.The integrated neural network model was optimized by fabric composition classifier,and the accuracy could be higher.