MI-PSO-RBF算法在稻谷存储品质预测的应用研究

Application of MI-PSO-RBF Algorithm in Prediction of Rice Storage Quality

粮食存储安全是关系粮食安全的重要因素,稻谷作为我国主要粮食作物其存储安全尤为重要。针对稻谷存储质量检测不方便等缺点,利用径向基神经网络(RBF)根据粮食实际存储的环境因素构建稻谷脂肪酸含量的预测模型。首先为避免数据维度过大使模型精度降低,利用互信息法则(MI)进行环境变量的特征提取,选取温度、湿度等6种影响较大的环境因素;然后根据RBF神经网络参数难以确定的缺点,采用改进的粒子群算法(PSO)进行寻优;同时改变粒子群的学习因子和权重系数的确定方式,使PSO算法在前期全局全面搜索并在后期易于跳出局部最优。通过实际数据进行模型验证,与传统RBF模型和PSO-RBF模型相比,构建的存储环境-存储品质DPSO-RBF预测模型精度提高。

Food storage security is an important factor related to food security,especially rice,the main food crop in China.Aiming at the disadvantages of inconvenient detection of rice storage quality,a prediction model of fatty acid content in rice was established by using a radial basis neural network(RBF)in accordance with the actual environmental factors of grain storage.Firstly,in order to avoid the decrease of the accuracy of the model due to the excessive data dimension,the Mutual Information rule(MI)was used to extract the characteristics of environmental variables,and six environmental factors,such as temperature and humidity,were selected.Then,the improved Particle Swarm Optimization algorithm(PSO)was used to optimize the RBF neural network.The method of determining the learning factor and weight coefficient of particle swarm optimization was changed to make the PSO algorithm search globally in the early stage and jump out of local optimum easily in the later stage.Compared with the traditional RBF model and PSO-RBF model,the accuracy of DPSO-RBF prediction model of storage environment-storage quality was improved.