利用深度学习进行时间序列分类

问题描述：有一批tsv格式（或csv格式）的时间序列数据，需要按照标签进行分类
所用语言：python
数据类型：tsv格式，可以用Excel打开，数据的第0列为类别标签，第1列往后为时间序列数据。

代码中用到的示例数据下载地址：
代码：

from __future__ import print_function
 
from keras.models import Model
from keras.utils import np_utils
import numpy as np
import pandas as pd
import keras 
from keras.callbacks import ReduceLROnPlateau
import matplotlib.pyplot as plt
      

#训练次数
nb_epochs = 1000

#读取训练集和测试集、标签
#第0列为标签
#第1列往后为数据
x_train = np.loadtxt('./Adiac/Ham_TRAIN.tsv', delimiter = '\t')[:,1:]#tsv文件中使用\t作为分隔符，如果是csv文件使用‘，’作为分隔符就把这里的关键字改一下
y_train = np.loadtxt('./Adiac/Ham_TRAIN.tsv', delimiter = '\t')[:,0]
x_test = np.loadtxt('./Adiac/Ham_TEST.tsv', delimiter = '\t')[:,1:]
y_test = np.loadtxt('./Adiac/Ham_TEST.tsv', delimiter = '\t')[:,0]



#有几个类别，np.nique函数是去除数组中的重复数字
nb_classes = len(np.unique(y_test))
#设定批的大小，//表示整除
batch_size = min(x_train.shape[0]//10, 16)

#归一化
y_train = (y_train - y_train.min())/(y_train.max()-y_train.min())*(nb_classes-1)
y_test = (y_test - y_test.min())/(y_test.max()-y_test.min())*(nb_classes-1)

#转换为独热编码
Y_train = np_utils.to_categorical(y_train, nb_classes)
Y_test = np_utils.to_categorical(y_test, nb_classes)

x_train_mean = x_train.mean()
#std代表标准差
x_train_std = x_train.std()
#z-score标准化
x_train = (x_train - x_train_mean)/(x_train_std)

x_test = (x_test - x_train_mean)/(x_train_std)

x_train = x_train.reshape(x_train.shape + (1,1,))
x_test = x_test.reshape(x_test.shape + (1,1,))

x = keras.layers.Input(x_train.shape[1:])
#    drop_out = Dropout(0.2)(x)
conv1 = keras.layers.Conv2D(128, 8, 1, border_mode='same')(x)
conv1 = keras.layers.normalization.BatchNormalization()(conv1)
conv1 = keras.layers.Activation('relu')(conv1)

#    drop_out = Dropout(0.2)(conv1)
conv2 = keras.layers.Conv2D(256, 5, 1, border_mode='same')(conv1)
conv2 = keras.layers.normalization.BatchNormalization()(conv2)
conv2 = keras.layers.Activation('relu')(conv2)

#    drop_out = Dropout(0.2)(conv2)
conv3 = keras.layers.Conv2D(128, 3, 1, border_mode='same')(conv2)
conv3 = keras.layers.normalization.BatchNormalization()(conv3)
conv3 = keras.layers.Activation('relu')(conv3)

full = keras.layers.pooling.GlobalAveragePooling2D()(conv3)    
out = keras.layers.Dense(nb_classes, activation='softmax')(full)


model = Model(input=x, output=out)

optimizer = keras.optimizers.Adam()
model.compile(loss='categorical_crossentropy',
              optimizer=optimizer,
              metrics=['accuracy'])

reduce_lr = ReduceLROnPlateau(monitor = 'loss', factor=0.5,
                  patience=50, min_lr=0.0001) 
history = model.fit(x_train, Y_train, batch_size=batch_size, nb_epoch=nb_epochs,
          verbose=1, validation_data=(x_test, Y_test), callbacks = [reduce_lr])
model.save('FCN_CBF_1500.h5')
acc = history.history['acc']
val_acc = history.history['val_acc']
loss = history.history['loss']
val_loss = history.history['val_loss']

epochs = range(len(acc))

plt.plot(epochs, acc, 'bo', label='Training acc')
plt.plot(epochs, val_acc, 'b', label='Validation acc')
plt.title('Training and validation accuracy')
plt.legend()

plt.figure()

plt.plot(epochs, loss, 'bo', label='Training loss')
plt.plot(epochs, val_loss, 'b', label='Validation loss')
plt.title('Training and validation loss')
plt.legend()

plt.show()

结果：
训练1000次后，

另外一个版本的代码：
（更详细）转自格拉迪沃的博客
数据存放路径：
数据放在Ham文件夹下（包括训练集和测试集），代码放在文件夹外面，与文件夹并列。
或者直接更改代码中的路径即可。

#!/usr/bin/env 
# -*- coding: utf-8 -*-
"""
Created on Sun Oct 30 20:11:19 2016

@author: stephen
"""
 
from __future__ import print_function
 
from keras.models import Model
from keras.utils import np_utils
import numpy as np
import pandas as pd
import keras 
from keras.callbacks import ReduceLROnPlateau
import matplotlib.pyplot as plt
      
def readucr(filename):
    data = np.loadtxt(filename, delimiter = '\t')#tsv文件中使用\t作为分隔符，如果是csv文件使用‘，’作为分隔符就把这里的关键字改一下
    Y = data[:,0]#第0列为标签
    X = data[:,1:]#第1列往后为数据
    return X, Y
  
nb_epochs = 1000

#根据需要把这里的文件夹名字做更改，diac是众多数据集中的一个。
flist  = ['Ham']
for each in flist:
    fname = each
    x_train, y_train = readucr(fname+'/'+fname+'_TRAIN.tsv')
    x_test, y_test = readucr(fname+'/'+fname+'_TEST.tsv')
    nb_classes = len(np.unique(y_test))
    batch_size = min(x_train.shape[0]/10, 16)
    
    y_train = (y_train - y_train.min())/(y_train.max()-y_train.min())*(nb_classes-1)
    y_test = (y_test - y_test.min())/(y_test.max()-y_test.min())*(nb_classes-1)
    
    
    Y_train = np_utils.to_categorical(y_train, nb_classes)
    Y_test = np_utils.to_categorical(y_test, nb_classes)
    
    x_train_mean = x_train.mean()
    x_train_std = x_train.std()
    x_train = (x_train - x_train_mean)/(x_train_std)
     
    x_test = (x_test - x_train_mean)/(x_train_std)
    x_train = x_train.reshape(x_train.shape + (1,1,))
    x_test = x_test.reshape(x_test.shape + (1,1,))

    x = keras.layers.Input(x_train.shape[1:])
#    drop_out = Dropout(0.2)(x)
    conv1 = keras.layers.Conv2D(128, 8, 1, border_mode='same')(x)
    conv1 = keras.layers.normalization.BatchNormalization()(conv1)
    conv1 = keras.layers.Activation('relu')(conv1)
    
#    drop_out = Dropout(0.2)(conv1)
    conv2 = keras.layers.Conv2D(256, 5, 1, border_mode='same')(conv1)
    conv2 = keras.layers.normalization.BatchNormalization()(conv2)
    conv2 = keras.layers.Activation('relu')(conv2)
    
#    drop_out = Dropout(0.2)(conv2)
    conv3 = keras.layers.Conv2D(128, 3, 1, border_mode='same')(conv2)
    conv3 = keras.layers.normalization.BatchNormalization()(conv3)
    conv3 = keras.layers.Activation('relu')(conv3)
    
    full = keras.layers.pooling.GlobalAveragePooling2D()(conv3)    
    out = keras.layers.Dense(nb_classes, activation='softmax')(full)
    
    
    model = Model(input=x, output=out)
     
    optimizer = keras.optimizers.Adam()
    model.compile(loss='categorical_crossentropy',
                  optimizer=optimizer,
                  metrics=['accuracy'])
     
    reduce_lr = ReduceLROnPlateau(monitor = 'loss', factor=0.5,
                      patience=50, min_lr=0.0001) 
    history = model.fit(x_train, Y_train, batch_size=batch_size, nb_epoch=nb_epochs,
              verbose=1, validation_data=(x_test, Y_test), callbacks = [reduce_lr])
    model.save('FCN_CBF_1500.h5')
acc = history.history['acc']
val_acc = history.history['val_acc']
loss = history.history['loss']
val_loss = history.history['val_loss']

epochs = range(len(acc))

plt.plot(epochs, acc, 'bo', label='Training acc')
plt.plot(epochs, val_acc, 'b', label='Validation acc')
plt.title('Training and validation accuracy')
plt.legend()

plt.figure()

plt.plot(epochs, loss, 'bo', label='Training loss')
plt.plot(epochs, val_loss, 'b', label='Validation loss')
plt.title('Training and validation loss')
plt.legend()

plt.show()

运行结果：