基于多保真的哈里斯鹰优化的广义回归神经网络.zip

# -*- coding: UTF-8 -*- import warnings import sklearn warnings.filterwarnings('ignore') from sko.PSO import PSO import datetime import numpy as np import math import pandas as pd from sklearn import model_selection import sys import tensorflow as tf from tensorflow import keras # import tensorflow.keras as keras import matplotlib as mpl import matplotlib.pyplot as plt from sklearn.preprocessing import MinMaxScaler from neupy import algorithms from Optimizers.optimizers import SSA,HHO,PSO def load_data(filename): '''导入数据 input: file_path(string):训练数据 output: feature(mat):特征 label(mat):标签 ''' p1_base_attribute = ['梳棉出条线速度', '梳棉AFIS的棉结1', '梳棉AFIS的棉结2', '梳棉AFIS的短绒率1', '梳棉AFIS的短绒率2', '梳棉落棉率', '预并出条线速度', '精梳AFIS的棉结1', '精梳AFIS的棉结2', '精梳AFIS的短绒率1', '精梳AFIS的短绒率2', '精梳落棉率', '精梳钳次', '粗纱捻缩率', '细纱捻缩率', '平均锭子速度', '牵伸倍数', '原棉短绒率', '原棉棉结', '原棉马克隆值', '强度', '右半部平均长度', '长度整齐度', '含杂率', '细纱号数'] al_value = ['强力', '棉结', '条干CV', '毛羽H值', '单强CV'] al_value = ['毛羽H值'] data = pd.read_excel(r'' + filename, index_col=None, header=0) scaler = MinMaxScaler(feature_range=(0, 1)) scaler1 = MinMaxScaler(feature_range=(0, 1)) data_x, data_y = [], [] for i in range(len(p1_base_attribute)): data_x.append(data[p1_base_attribute[i]]) for i in range(len(al_value)): data_y.append(data[al_value[i]]) data = data_x + data_y data = np.array(data).reshape((len(p1_base_attribute) + len(al_value), -1)) data_yyy = np.array(data_y).reshape(len(al_value), -1) data = scaler.fit_transform(np.transpose(data)) data_yyy = scaler1.fit_transform(np.transpose(data_yyy)) data = data.reshape(-1, len(p1_base_attribute) + len(al_value)) train_data, train_target = data[:, :-len(al_value)], data[:, -len(al_value):] return train_data, train_target, scaler1 train_data, train_target, scaler = load_data('F:/study/402实验室/成纱质量预测/成纱质量预测论文/基于不同优化算法的GRNN纱线不均匀度预测/data.xlsx') ## 2.数据集和测试集 print('--------------------2.Train Set and Test Set--------------------') data_x, test_x, data_y, test_y = model_selection.train_test_split( train_data, train_target, test_size=0.1, random_state=1) starttime = datetime.datetime.now() def demo_func(test_other): nw = algorithms.GRNN(std=test_other, verbose=False) nw.train(data_x, data_y) y_predicted = scaler.inverse_transform(nw.predict(test_x)) mse = np.sum((y_predicted - scaler.inverse_transform(test_y))**2)/ len(test_y) return mse*10000000 maxgen = 50 jiange = 10 res = [] legends = [] for i in range(10): popsize = jiange*(i+1) legends.append('popsize='+str(popsize)) hho = HHO.HHO(demo_func, 0.00000000001,1, 1, popsize, maxgen, ) res.append(hho.convergence[1:]) print('run_time is ', hho.executionTime) print('best_y',hho.convergence[-1]) print('best_x',hho.bestFoodPositions[-1]) print(hho.bestFoodPositions.tolist().index(hho.bestFoodPositions[-1])) nw = algorithms.GRNN(std=hho.bestFoodPositions[-1], verbose=False) nw.train(data_x, data_y) y_predicted = scaler.inverse_transform(nw.predict(test_x)) print(y_predicted) plt.figure(1) for i in range(10): plt.plot(range(49),res[i]) plt.xlabel('Number of iterations') plt.ylabel('MSE value') plt.rcParams['savefig.dpi'] = 600 # 图片像素 plt.rcParams['figure.dpi'] = 600 # 分辨率 plt.ylim(30999,31002) plt.xlim(0,50) plt.legend(legends) plt.savefig('HHO_'+str(jiange)+'.jpg') plt.show() nw = algorithms.GRNN(std=hho.bestFoodPositions[-1], verbose=False) endtime = datetime.datetime.now() print(endtime-starttime) nw.train(data_x, data_y) y_predicted = scaler.inverse_transform(nw.predict(test_x)) mse_test = np.sum((y_predicted - scaler.inverse_transform(test_y)) ** 2)/ len(test_y) rmse_test = mse_test ** 0.5 mae_test = np.sum(np.absolute(y_predicted - scaler.inverse_transform(test_y))) / len(test_y) r2 = sklearn.metrics.r2_score(scaler.inverse_transform(test_y), y_predicted) print(test_x,scaler.inverse_transform(test_y),y_predicted) print(mse_test,rmse_test,mae_test,r2) plt.figure(1) plt.plot(range(7),scaler.inverse_transform(test_y)) doc = plt.scatter(range(7), scaler.inverse_transform(test_y), color='green') #绘点图 plt.plot(range(7),y_predicted) doc1 = plt.scatter(range(7), y_predicted, color='red') #绘点图 plt.xlabel('Sample number') plt.ylabel('Hairiness(H)') plt.rcParams['savefig.dpi'] = 3000 # 图片像素 plt.rcParams['figure.dpi'] = 3000 # 分辨率 plt.ylim(1.9,2.3) plt.xlim(0,6) plt.legend(['target','output']) plt.savefig('result.jpg') plt.show()