import tensorflow as tf
import numpy as np
from tensorflow.contrib import rnn
from pandas import read_csv
import matplotlib.pyplot as plt
import pandas as pd
from sklearn.metrics import accuracy_score
all_data = []
all_pred = []
all_test = []
num_periods = 20
inputs = 1
hidden = 100
output = 1
for i in range(40):
shuju = read_csv('40jiedian-shuju.csv',usecols = [i],engine = 'python',skipfooter = 0)
dataset = shuju.values
dataset = dataset.astype('float32')
all_data.append(dataset)
x_data = dataset[:200]
x_batches = x_data.reshape(-1,20,1)
y_data = dataset[1:201]
y_batches = y_data.reshape(-1,20,1)
x_test = dataset[-20:].reshape(-1,20,1)
y_test = dataset[-20:].reshape(-1,20,1)
all_test.append(y_test)
tf.reset_default_graph()
inputs = 1
hidden = 100
output = 1
x = tf.placeholder(tf.float32, [None, num_periods, inputs])
y = tf.placeholder(tf.float32, [None, num_periods, output])
basic_cell = tf.contrib.rnn.BasicRNNCell(num_units=hidden, activation=tf.nn.relu)
rnn_output, states = tf.nn.dynamic_rnn(basic_cell, x, dtype=tf.float32)
learning_rate = 0.001
stacked_rnn_output = tf.reshape(rnn_output, [-1, hidden])
stacked_output = tf.layers.dense(stacked_rnn_output, output)
outputs = tf.reshape(stacked_output, [-1, num_periods, output])
loss = tf.reduce_sum(tf.square(outputs - y))
optimizer = tf.train.AdamOptimizer(learning_rate=learning_rate)
training_op = optimizer.minimize(loss)
init = tf.global_variables_initializer()
epochs = 1000
with tf.Session() as sess:
init.run()
for ep in range(epochs):
sess.run(training_op, feed_dict={x: x_batches, y: y_batches})
if ep % 100 == 0:
mse = loss.eval(feed_dict={x: x_batches, y: y_batches})
print(ep, '\tMSE', mse)
y_pred = sess.run(outputs, feed_dict={x: x_test})
all_pred.append(y_pred)
print(y_pred)
plt.title('Forecast vs Actual',fontsize=14)
plt.plot(pd.Series(np.ravel(all_test[39])),'bo',markersize=10,label='Actual')
plt.plot(pd.Series(np.ravel(all_pred[39])),'r',markersize=10,label='Forecast')
plt.legend(loc='upper. left')
plt.xlabel('Time Periods')
plt.show()
评论0