上一篇笔记,Tensorflow - 将序列处理成embedding - 方法1 - keras调包,也介绍了如何生成批量数据,适合处理文本序列数据,用于分类,可以视为 Tensorflow - 生成批量数据 - 方法2 。
此篇笔记,Example 1 适合 时间序列 做线性回归预测, Example 2 适合用文本序列做分类(和方法2类似)。
具体做法也不同:
上一篇笔记,每次生成一个epoch的所有数据,但只yield一个batch的数据;
此次,一次性生成一个epoch的所有数据。
Example 1: Stock data anaysis
All the stock records are stored in excel file. You need to transfer the data into 3 dimensional format.
#coding=utf-8
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import tensorflow as tf
rnn_unit=10 #hidden layer units
input_size=7
output_size=1
lr=0.0006
f=open('D:/DL/lstm_blog_case/stock_dataset/dataset_2.csv')
df = pd.read_csv(f) # 读入股票数据
data=df.iloc[:, 2:10].values # 取所有行的,第3-10列
# 获取训练集
def get_train_data(batch_size=60,time_step=20,train_begin=0,train_end=5800):
batch_index=[]
data_train=data[train_begin:train_end]
normalized_train_data=(data_train-np.mean(data_train,axis=0))/np.std(data_train,axis=0) #标准化
train_x, train_y = [],[] #训练集
for i in range(len(normalized_train_data)-time_step):
if i % batch_size==0:
batch_index.append(i)
x=normalized_train_data[i:i+time_step, :7]
# [[-0.13319847 -0.15049982 -0.12760952 -0.13653869 -0.57690395 0.20655675 -0.53271994] ...] 20个
y=normalized_train_data[i:i+time_step,7, np.newaxis]
# [[-0.13687717] ...] 20个
train_x.append(x.tolist())
train_y.append(y.tolist())
#print(len(train_y)) # 5780
#print(train_y)
batch_index.append((len(normalized_train_data)-time_step)) # 加上最后一个的index
# train_x : 5780 * 20 * 7 三维张量[[[]]]
# print(np.array(train_x[1:2]))
# print("train_y:",np.array(train_y)) # 加了np.newaxis之后,也成了三维张量,np.array(train_y)):(5780, 20, 1)
print(" batch_index", batch_index) # batch_index [0, 80, 160, 240, 320, 400 ...3780]
return batch_index,train_x,train_y
batch_index,train_x,train_y = get_train_data()
print(train_x[0:2])
#获取测试集
def get_test_data(time_step=20,test_begin=5800):
data_test=data[test_begin:]
mean=np.mean(data_test,axis=0)
std=np.std(data_test,axis=0)
normalized_test_data=(data_test-mean)/std #标准化
size=(len(normalized_test_data)+time_step-1)//time_step #有size个sample
test_x,test_y=[],[]
for i in range(size-1):
x=normalized_test_data[i*time_step:(i+1)*time_step,:7]
y=normalized_test_data[i*time_step:(i+1)*time_step,7]
test_x.append(x.tolist())
test_y.extend(y)
test_x.append((normalized_test_data[(i+1)*time_step:,:7]).tolist())
test_y.extend((normalized_test_data[(i+1)*time_step:,7]).tolist())
return mean,std,test_x,test_y
# ——————————————————定义神经网络变量——————————————————
weights={
'in':tf.Variable(tf.random_normal([input_size,rnn_unit])),
'out':tf.Variable(tf.random_normal([rnn_unit,1]))
}
biases={
'in':tf.Variable(tf.constant(0.1,shape=[rnn_unit,])),
'out':tf.Variable(tf.constant(0.1,shape=[1,]))
}
# ——————————————————定义神经网络变量——————————————————
def lstm(X):
batch_size=tf.shape(X)[0]
time_step=tf.shape(X)[1]
w_in=weights['in']
b_in=biases['in']
input=tf.reshape(X,[-1,input_size]) # input_size是7,rnn_unit是10,此处转换,使得LSTM能接收数据
input_rnn=tf.matmul(input,w_in)+b_in
input_rnn=tf.reshape(input_rnn,[-1,time_step,rnn_unit]) # 将tensor转成3维,作为lstm cell的输入
cell=tf.nn.rnn_cell.BasicLSTMCell(rnn_unit)
init_state=cell.zero_state(batch_size,dtype=tf.float32)
output_rnn, final_states=tf.nn.dynamic_rnn(cell, input_rnn,initial_state=init_state, dtype=tf.float32)
output=tf.reshape(output_rnn,[-1,rnn_unit]) # 作为输出层的输入
# 注意这里的shape 下面还有reshape过程
w_out=weights['out']
b_out=biases['out']
pred=tf.matmul(output,w_out)+b_out
return pred,final_states
# ——————————————————训练模型——————————————————
def train_lstm(batch_size=80,time_step=15,train_begin=2000,train_end=5800):
X=tf.placeholder(tf.float32, shape=[None,time_step,input_size])
Y=tf.placeholder(tf.float32, shape=[None,time_step,output_size])
batch_index,train_x,train_y= get_train_data(batch_size,time_step,train_begin,train_end)
pred,_=lstm(X)
#损失函数
loss=tf.reduce_mean(tf.square(tf.reshape(pred,[-1])-tf.reshape(Y, [-1])))
train_op=tf.train.AdamOptimizer(lr).minimize(loss)
saver=tf.train.Saver(tf.global_variables(),max_to_keep=15)
#module_file = tf.train.latest_checkpoint()
module_file = 'D:/DL/lstm_blog_case/ckpt/stock2.ckpt'
with tf.Session() as sess:
try :
saver.restore(sess, module_file)
print("成功加载模型参数")
except:
sess.run(tf.global_variables_initializer())
print("未加载模型参数,文件被删除或者第一次运行")
for i in range(2000): # epoch
for step in range(len(batch_index)-1):
# train_x[x:y]三维张量 # batch_size 差为60的等差数列
# input:60*20*7
_,loss_=sess.run([train_op,loss],feed_dict={X:train_x[batch_index[step]:batch_index[step+1]],Y:train_y[batch_index[step]:batch_index[step+1]]})
print(i,loss_)
if i % 200==0:
print("保存模型:",saver.save(sess,'D:/DL/lstm_blog_case/ckpt/stock2.ckpt',global_step=i))
# train_lstm()
# ————————————————预测模型————————————————————
def prediction(time_step=20):
X=tf.placeholder(tf.float32, shape=[None,time_step,input_size])
#Y=tf.placeholder(tf.float32, shape=[None,time_step,output_size])
mean,std,test_x,test_y=get_test_data(time_step)
pred,_=lstm(X)
saver=tf.train.Saver(tf.global_variables())
with tf.Session() as sess:
#参数恢复
module_file = tf.train.latest_checkpoint('D:/DL/lstm_blog_case/ckpt/')
saver.restore(sess, module_file)
print("tf.train.latest_checkpoint():", tf.train.latest_checkpoint('D:/DL/lstm_blog_case/ckpt/'))
test_predict = []
for step in range(len(test_x)-1):
prob=sess.run(pred,feed_dict={X: [test_x[step]]})
predict=prob.reshape((-1))
test_predict.extend(predict)
test_y=np.array(test_y)*std[7]+mean[7]
test_predict=np.array(test_predict)*std[7]+mean[7]
acc=np.average(np.abs(test_predict-test_y[:len(test_predict)])/test_y[:len(test_predict)]) #偏差
# 以折线图表示结果
plt.figure()
plt.plot(list(range(len(test_predict))), test_predict, color='b')
plt.plot(list(range(len(test_y))), test_y, color='r')
plt.show()
#prediction()
Example 2:
def load_csv(filename, type):
matrix_data = []
with open(filename, 'r') as csvfile:
csvreader = csv.reader(csvfile, delimiter=',')
next(csvreader)
for row_vector in csvreader:
if type == 'int':
matrix_data.append(list(map(int, row_vector[0:])))
else:
matrix_data.append(list(map(float, row_vector[0:])))
return np.matrix(matrix_data[0:5000])
# 把正负样本合起来
def stack_all_sequence():
padding_sequence_neg=load_csv('mirRNA_data/pad_neg_binary_sequence.csv','float')
neg_number = len(padding_sequence_neg)
padding_sequence_pos = load_csv('mirRNA_data/pad_pos_binary_sequence.csv', 'float')
y_label = np.zeros((10000, 2))
all_seq = np.row_stack((padding_sequence_neg, padding_sequence_pos))
# print(len(all_seq)) # 10000
# print(all_seq.shape) # (10000, 1600)
pos_label = np.mat(np.array([1, 0]))
neg_label = np.mat(np.array([0, 1]))
for i in range(len(padding_sequence_pos)):
y_label[i] = pos_label
for j in range(neg_number):
y_label[j+5000] = neg_label
return all_seq, y_label
shuffled_x = np.zeros((100, 100, 1600))
shuffled_y = np.zeros((100, 100, 2))
# 打乱数据,新的next_batch方法; 100batch ; 每个batch 100sample
def batch_iter(x, y, batch_size=100):
data_len = len(x)
num_batch = int((data_len - 1) / batch_size) + 1
indices = np.random.permutation(np.arange(data_len))
x_shuffle = np.zeros((10000, 1600))
y_shuffle = np.zeros((10000, 2))
for i in indices:
x_shuffle[i] = np.mat(x[i])
for j in indices:
y_shuffle[j] = np.mat(y[j])
for i in range(num_batch):
start_id = i * batch_size
end_id = min((i + 1) * batch_size, data_len)
shuffled_x[i] = x_shuffle[start_id:end_id]
shuffled_y[i] = y_shuffle[start_id:end_id]
return shuffled_x, shuffled_y # shuffled_y 100维
all_seq, y = stack_all_sequence()
x_shuffled, y_shuffled = batch_iter(all_seq, y) # # 100个batch,每个batch 100 sample
# print(x_shuffled.shape) # (100, 100, 1600)
# print(x_shuffled[0].shape) # (100, 1600)
# print(y_shuffled.shape) # (100, 100, 2)
# print(y_shuffled[0].shape) # (100, 2)