#3
print('\n#3,绘制对比曲线图')
df2 = pd.DataFrame()
df2['xopen'] = df['xopen']
df2['open2'] = df['open2']
print(df2.tail())
df3 = df2.tail(200)
df3.plot(rot=15)
#4
print('\n#4,计算预测结果')
df5 = pd.DataFrame()
df5['y_test'] = df2['xopen']
df5['y_pred'] = df2['open2']
a1, df5x = ztq.ai_acc_xed2x(df5['y_test'], df5['y_pred'], ky0=5)
print(df5x.tail())
print('\na1,', a1)
a1, df5x, a20 = ztq.ai_acc_xed2ext(df5['y_test'], df5['y_pred'], ky0=5)
print('\na20,', a20)
'''
n_df9,3671,n_dfk,3655
acc-kok: 99.56%, MAE:42.65, MSE:2463.18, RMSE:49.63
y_test y_pred ysub ysub2 y_test_div ysubk
date
2017-04-25 3132.918 3185.180 -52.262 52.262 3132.918 1.668
2017-04-26 3131.350 3191.560 -60.210 60.210 3131.350 1.923
2017-04-27 3144.022 3203.079 -59.057 59.057 3144.022 1.878
2017-04-28 3147.228 3205.589 -58.361 58.361 3147.228 1.854
print('\n#3,xed.train.数据')
df_train = df.head(dnum2)
df_test = df.tail(dnum - dnum2)
#
x_train, y_train = df_train['x'].values, df_train['y'].values
x_test, y_test = df_test['x'].values, df_test['y'].values
print('train,', x_train[0], y_train[0])
#4
print('\n#4,model建立神经网络模型')
model = zks.mlp01()
model.summary()
#
model.compile(loss='mse', optimizer='sgd', metrics=['accuracy'])
#
model.save('tmp/mlp01.dat')
#5 模型训练
print('\n#5 模型训练 fit')
mx = load_model('tmp/mlp01.dat')
mx.fit(x_train, y_train, epochs=200, batch_size=128)
#
mx.save('tmp/mlp02x.dat')
#6 整理预测数据
print('\n#6 整理预测数据 ')
y_pred = mx.predict(x_test)
df_test['y_pred'] = zdat.ds4x(y_pred, df_test.index, True)
dacc, _ = ztq.ai_acc_xed2x(df_test.y, df_test['y_pred'], 5, False)
print('acc:', dacc)