1. 预处理
url = './铁路客运量.csv'
ass_data = requests.get(url).content #打开文件
df = pd.read_csv(io.StringIO(ass_data.decode('utf-8'))) # python2使用StringIO.StringIO
data = np.array(df['铁路客运量_当期值(万人)'])
# normalize
normalized_data = (data - np.mean(data)) / np.std(data)
seq_size = 3
train_x, train_y = [], []
for i in range(len(normalized_data) - seq_size - 1): # 解析
train_x.append(np.expand_dims(normalized_data[i: i + seq_size], axis=1).tolist())
train_y.append(normalized_data[i + 1: i + seq_size + 1].tolist())
input_dim = 1
X = tf.placeholder(tf.float32, [None, seq_size, input_dim])
Y = tf.placeholder(tf.float32, [None, seq_size])
2. 模型构建
def ass_rnn(hidden_layer_size=6): # regression
W = tf.Variable(tf.random_normal([hidden_layer_size, 1]), name='W')
b = tf.Variable(tf.random_normal([1]), name='b')
cell = tf.nn.rnn_cell.BasicLSTMCell(hidden_layer_size)# 6
outputs, states = tf.nn.dynamic_rnn(cell, X, dtype=tf.float32)
W_repeated = tf.tile(tf.expand_dims(W, 0), [tf.shape(X)[0], 1, 1])
out = tf.batch_matmul(outputs, W_repeated) + b
out = tf.squeeze(out)
return out
3. 训练模型
def train_rnn():
out = ass_rnn()
loss = tf.reduce_mean(tf.square(out - Y))
train_op = tf.train.AdamOptimizer(learning_rate=0.003).minimize(loss)
saver = tf.train.Saver(tf.all_variables()) # tf.global_variables() == tf 0.12
with tf.Session() as sess:
tf.get_variable_scope().reuse_variables() # old API
sess.run(tf.initialize_all_variables()) # tf.global_variables_initializer() ==tf 0.12
for step in range(10000):
_, loss_ = sess.run([train_op, loss], feed_dict={X: train_x, Y: train_y})
if step % 10 == 0:
# 用测试数据评估loss
print(step, loss_)
print("保存模型: ", saver.save(sess, 'ass.model'))
4. 预测
def prediction():
out = ass_rnn()
saver = tf.train.Saver(tf.all_variables()) # new API tf.global_variables()
with tf.Session() as sess:
tf.get_variable_scope().reuse_variables() # old API
saver.restore(sess, './ass.model')
prev_seq = train_x[-1]
predict = []
for i in range(12):
next_seq = sess.run(out, feed_dict={X: [prev_seq]})
predict.append(next_seq[-1])
prev_seq = np.vstack((prev_seq[1:], next_seq[-1]))
plt.figure()
plt.plot(list(range(len(normalized_data))), normalized_data, color='b')
plt.plot(list(range(len(normalized_data), len(normalized_data) + len(predict))), predict, color='r')
plt.show()
# train_rnn() #
prediction()
