机器学习之径向基神经网络

RBFNN训练

import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
from sklearn.cluster import KMeans
import eva
import datetime

start = datetime.datetime.now()

def tanh(x):
    return (np.exp(x)-np.exp(-x))/(np.exp(x)+np.exp(-x))
def de_tanh(x):
    return (1-x**2)


#参数设置
samnum = 62
hiddenunitnum = 8
indim = 4
outdim = 1
maxepochs = 100
errorfinal = 0.65*10**(-3)
learnrate = 0.0001

df = pd.read_csv("./data/Cr1.csv")  # 返回一个DataFrame的对象，这个是pandas的一个数据结构
df.columns = ["x", "y", "high", "use", "Cr"]
longtitude = df["x"] # 抽取前四列作为训练数据的各属性值
longtitude = np.array(longtitude)
latitude = df["y"]
latitude = np.array(latitude)
elevation = df["high"]
elevation =np.array(elevation)
functional = df["use"]
functional = np.array(functional)
ag=df["Cr"]
ag=np.array(ag)

samplein = np.mat([longtitude,latitude,elevation,functional])  #4*1000
sampleinminmax = np.array([samplein.min(axis=1).T.tolist()[0],samplein.max(axis=1).T.tolist()[0]]).transpose()#对应最大值最小值
sampleout = np.mat([ag])
sampleoutminmax = np.array([sampleout.min(axis=1).T.tolist()[0],sampleout.max(axis=1).T.tolist()[0]]).transpose()#对应最大值最小值
sampleinnorm = ((np.array(samplein.T)-sampleinminmax.transpose()[0])/(sampleinminmax.transpose()[1]-sampleinminmax.transpose()[0])).transpose()
sampleoutnorm = ((np.array(sampleout.T)-sampleoutminmax.transpose()[0])/(sampleoutminmax.transpose()[1]-sampleoutminmax.transpose()[0])).transpose()


sampleinmax = np.array([sampleinnorm.max(axis=1).T.tolist()]).transpose()
sampleinmin = np.array([sampleinnorm.min(axis=1).T.tolist()]).transpose()
sampleoutmax = np.array([sampleoutnorm.max(axis=1).T.tolist()]).transpose()
sampleoutmin = np.array([sampleoutnorm.min(axis=1).T.tolist()]).transpose()
noise = 0.03*np.random.rand(sampleoutnorm.shape[0],sampleoutnorm.shape[1])
sampleoutnorm += noise


x = sampleinnorm.transpose()
estimator=KMeans(n_clusters=8,max_iter=10000)
estimator.fit(x)
w1 = estimator.cluster_centers_


#b为8*1
b1 = np.mat(np.zeros((hiddenunitnum,outdim)))
for i in range(hiddenunitnum):
    cmax = 0
    for j in range(hiddenunitnum):
        temp_dist=np.sqrt(np.sum(np.square(w1[i,:]-w1[j,:])))
        if cmax<temp_dist:
            cmax=temp_dist
    b1[i] = cmax/np.sqrt(2*hiddenunitnum)


#生成方法1
scale = np.sqrt(3/((indim+outdim)*0.5))
w2 = np.random.uniform(low=-scale,high=scale,size=[hiddenunitnum,outdim])
b2 = np.random.uniform(low=-scale, high=scale, size=[outdim,1])


inputin=np.mat(sampleinnorm.T)
w1=np.mat(w1)
b1=np.mat(b1)
w2=np.mat(w2)
b2=np.mat(b2)


errhistory = np.mat(np.zeros((1,maxepochs)))
#开始训练
for i in range(maxepochs):
    hidden_out = np.mat(np.zeros((samnum,hiddenunitnum)))
#前向计算：
    for a in range(samnum):
        for j in range(hiddenunitnum):
            d=(inputin[a, :] - w1[j, :]) * (inputin[a, :] - w1[j, :]).T
            c=2 * b1[j, :] * b1[j, :]
            hidden_out[a, j] = np.exp((-1.0 )* (d/c))
    # print(hidden_out.shape,"33333333333")
    output = tanh(hidden_out * w2 + b2)
    # output = sigmod(hidden_out * w2)
    # 计算误差
    out_real = np.mat(sampleoutnorm.transpose())
    err = out_real - output
    loss = np.sum(np.square(err))
    if loss < errorfinal:
        break
    errhistory[:,i] = loss
#反向计算
    output=np.array(output.T)
    belta=de_tanh(output).transpose()

    dw1now = np.zeros((8,4))
    db1now = np.zeros((8,1))
    dw2now = np.zeros((8,1))
    db2now = np.zeros((1,1))
    for j in range(hiddenunitnum):
        sum1 = 0.0
        sum2 = 0.0
        sum3 = 0.0
        sum4 = 0.0
        for a in range(samnum):
            #1*4
            sum1 +=err[a,:] * belta[a,:] * hidden_out[a,j] * (inputin[a,:]-w1[j,:])
            #1*1
            sum2 +=err[a,:] * belta[a,:] * hidden_out[a,j] * (inputin[a,:]-w1[j,:])*(inputin[a,:]-w1[j,:]).T
            #1*1
            sum3 +=err[a,:] * belta[a,:] * hidden_out[a,j]
            sum4 +=err[a,:] * belta[a,:]
        dw1now[j,:]=(w2[j,:]/(b1[j,:]*b1[j,:])) * sum1
        db1now[j,:] =(w2[j,:]/(b1[j,:]*b1[j,:]*b1[j,:])) * sum2
        dw2now[j,:] =sum3
        db2now = sum4

    w1 += learnrate * dw1now
    b1 += learnrate * db1now
    w2 += learnrate * dw2now
    b2 += learnrate * db2now
    print("the iteration is:",i+1,",the loss is:",loss)

print('更新的权重w1:',w1)
print('更新的偏置b1:',b1)
print('更新的权重w2:',w2)
print('更新的偏置b2:',b2)
print("The loss after iteration is ：",loss)

np.save("RBFNN_w1.npy",w1)
np.save("RBFNN_b1.npy",b1)
np.save("RBFNN_w2.npy",w2)
np.save("RBFNN_b2.npy",b2)
np.save("RBFNN_rbf_err.npy",errhistory)

diff = sampleoutminmax[:,1]-sampleoutminmax[:,0]
predict_train = (output+1)/2
predict_train = predict_train*diff+sampleoutminmax[0][0]
predict_train = predict_train.flatten()#降成一维数组
np.savetxt('./result/RBFNN/predict_train.csv', predict_train, delimiter = ',')

err_train =  ag - predict_train
# print("err1:",err_train)
np.savetxt('./result/RBFNN/err_train.csv', err_train, delimiter = ',')

rmse, mae, mape = eva.err(ag, predict_train)
# print("mape:",mape)
error_train = np.zeros(3)
error_train[0] = rmse
error_train[1] = mae
error_train[2] = mape
print("error:",error_train)
np.savetxt('./result/RBFNN/error_train.csv', error_train, delimiter = ',')

end = datetime.datetime.now()
time = end - start
print("运行时间总计:",time)

RBFNN预测数据(回归)

import numpy as np
import pandas as pd
import datetime
import eva
start = datetime.datetime.now()

def tanh(x):
    return (np.exp(x)-np.exp(-x))/(np.exp(x)+np.exp(-x))
def de_tanh(x):
    return (1-x**2)

testnum = 20
hiddenunitnum = 8

df = pd.read_csv("./data/Cr2.csv")
df.columns = ["x", "y", "high", "use", "Cr"]
longtitude = df["x"]
longtitude = np.array(longtitude)
latitude = df["y"]
latitude = np.array(latitude)
elevation = df["high"]
elevation =np.array(elevation)
functional = df["use"]
functional = np.array(functional)
As=df["Cr"]
As=np.array(As)
ag=As

#对数据进行最大最小归一化处理,mat为矩阵，t.tolist将矩阵转换成列表，然后由array转换成数组
samplein = np.mat([longtitude,latitude,elevation,functional])  #4*1000
# print("ag:\n",samplein)
#4*2
sampleinminmax = np.array([samplein.min(axis=1).T.tolist()[0],
                           samplein.max(axis=1).T.tolist()[0]]).transpose()#对应最大值最小值
# print("sampleinminmax:\n",sampleinminmax)
sampleout = np.mat([ag])#1*299
# print("sampleout:\n",sampleout)
sampleoutminmax = np.array([sampleout.min(axis=1).T.tolist()[0],
                            sampleout.max(axis=1).T.tolist()[0]]).transpose()#对应最大值最小值,4*299
# print("sampleoutminmax:\n",sampleoutminmax)
sampleinnorm = (2*(np.array(samplein.T)-sampleinminmax.transpose()[0])
                /(sampleinminmax.transpose()[1]-sampleinminmax.transpose()[0])-1).transpose()#1*299
# print("sampleinnorm:\n",sampleinnorm)
sampleoutnorm = (2*(np.array(sampleout.T)-sampleoutminmax.transpose()[0])
                 /(sampleoutminmax.transpose()[1]-sampleoutminmax.transpose()[0])-1).transpose()
# print("sampleoutnorm:\n",sampleoutnorm)
#给输出样本添加噪音
noise = 0.03 * np.random.rand(sampleoutnorm.shape[0],sampleoutnorm.shape[1])
sampleoutnorm += noise
# print("noise:\n",noise)
# print("sampleoutnorm:\n",sampleoutnorm)


#测试

w1=np.load('w1.npy')
w2=np.load('w2.npy')
b1=np.load('b1.npy')
b2=np.load('b2.npy')
w1 = np.mat(w1)
w2 = np.mat(w2)
b1 = np.mat(b1)
b2 = np.mat(b2)

df = pd.read_csv("./data/Cr1.csv")  # 返回一个DataFrame的对象，这个是pandas的一个数据结构
df.columns = ["x", "y", "high", "use", "Cr"]
longtitude = df["x"] # 抽取前七列作为训练数据的各属性值
longtitude = np.array(longtitude)
latitude = df["y"]
latitude = np.array(latitude)
elevation = df["high"]
elevation =np.array(elevation)
functional = df["use"]
functional = np.array(functional)
As=df["Cr"]
As=np.array(As)
input=np.mat([longtitude,latitude,elevation,functional])
inputnorm=(np.array(input.T)-sampleinminmax.transpose()[0])/(sampleinminmax.transpose()[1]-sampleinminmax.transpose()[0])


hidden_out2 = np.mat(np.zeros((testnum,hiddenunitnum)))

for a in range(testnum):
    for j in range(hiddenunitnum):
        d = (inputnorm[a, :] - w1[j, :]) * (inputnorm[a, :] - w1[j, :]).T
        c = 2 * b1[j, :] * b1[j, :].T
        hidden_out2[a, j] = np.exp((-1.0) * (d / c))
output = tanh(hidden_out2 * w2 + b2)

diff = sampleoutminmax[:,1]-sampleoutminmax[:,0]
networkout2 = output*diff+sampleoutminmax[0][0]
networkout2 = np.array(networkout2).transpose()
output1=networkout2.flatten()
output1=output1.tolist()

# diff = sampleoutminmax[:,1]-sampleoutminmax[:,0]
# networkout2 = (output+1)/2
# networkout2 = networkout2*diff+sampleoutminmax[0][0]
# output1=networkout2.flatten()#降成一维数组
# # print("output1",output1)
# output1=output1.tolist()


np.savetxt('./result/RBFNN/predict_test.csv', output1, delimiter = ',')

err_test =  ag - output1
np.savetxt('./result/RBFNN/err_test.csv', err_test, delimiter = ',')

rmse, mae, mape = eva.err(ag, output1)
print("mape:",mape)
error_train = np.zeros(3)
error_train[0] = rmse
error_train[1] = mae
error_train[2] = mape
# print("error:",error)
np.savetxt('./result/RBFNN/error_test.csv', error_train, delimiter = ',')

# 精度保持两位
for i in range(testnum):
    output1[i] = float('%.2f'%output1[i])

end = datetime.datetime.now()
time = end - start
print("运行时间总计:",time)

print("the prediction is:",output1)