TensorFlow_study/18_Forecasting with an LSTM.py at master · HoYoungChun/TensorFlow_study · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
import numpy as np
import matplotlib.pyplot as plt
import tensorflow as tf

keras = tf.keras

def plot_series(time, series, format="-", start=0, end=None, label=None):
    plt.plot(time[start:end], series[start:end], format, label=label)
    plt.xlabel("Time")
    plt.ylabel("Value")
    if label:
        plt.legend(fontsize=14)
    plt.grid(True)

def trend(time, slope=0):
    return slope * time


def seasonal_pattern(season_time):
    """Just an arbitrary pattern, you can change it if you wish"""
    return np.where(season_time < 0.4,
                    np.cos(season_time * 2 * np.pi),
                    1 / np.exp(3 * season_time))


def seasonality(time, period, amplitude=1, phase=0):
    """Repeats the same pattern at each period"""
    season_time = ((time + phase) % period) / period
    return amplitude * seasonal_pattern(season_time)


def white_noise(time, noise_level=1, seed=None):
    rnd = np.random.RandomState(seed)
    return rnd.randn(len(time)) * noise_level


def sequential_window_dataset(series, window_size):
    series = tf.expand_dims(series, axis=-1)
    ds = tf.data.Dataset.from_tensor_slices(series)
    ds = ds.window(window_size + 1, shift=window_size, drop_remainder=True)
    ds = ds.flat_map(lambda window: window.batch(window_size + 1))
    ds = ds.map(lambda window: (window[:-1], window[1:]))
    return ds.batch(1).prefetch(1)

time = np.arange(4 * 365 + 1)

slope = 0.05
baseline = 10
amplitude = 40
series = baseline + trend(time, slope) + seasonality(time, period=365, amplitude=amplitude)

noise_level = 5
noise = white_noise(time, noise_level, seed=42)

series += noise

plt.figure(figsize=(10, 6))
plot_series(time, series)
plt.show()

split_time = 1000
time_train = time[:split_time]
x_train = series[:split_time]
time_valid = time[split_time:]
x_valid = series[split_time:]

class ResetStatesCallback(keras.callbacks.Callback):
    def on_epoch_begin(self, epoch, logs):
        self.model.reset_states()

keras.backend.clear_session()
tf.random.set_seed(42)
np.random.seed(42)

window_size = 30
train_set = sequential_window_dataset(x_train, window_size)

model = keras.models.Sequential([
  keras.layers.LSTM(100, return_sequences=True, stateful=True,
                    batch_input_shape=[1, None, 1]),
  keras.layers.LSTM(100, return_sequences=True, stateful=True),
  keras.layers.Dense(1),
  keras.layers.Lambda(lambda x: x * 200.0)
])
lr_schedule = keras.callbacks.LearningRateScheduler(
    lambda epoch: 1e-8 * 10**(epoch / 20))
reset_states = ResetStatesCallback()
optimizer = keras.optimizers.SGD(lr=1e-8, momentum=0.9)
model.compile(loss=keras.losses.Huber(),
              optimizer=optimizer,
              metrics=["mae"])
history = model.fit(train_set, epochs=100,
                    callbacks=[lr_schedule, reset_states])

plt.semilogx(history.history["lr"], history.history["loss"])
plt.axis([1e-8, 1e-4, 0, 30])

keras.backend.clear_session()
tf.random.set_seed(42)
np.random.seed(42)

window_size = 30
train_set = sequential_window_dataset(x_train, window_size)
valid_set = sequential_window_dataset(x_valid, window_size)

model = keras.models.Sequential([
  keras.layers.LSTM(100, return_sequences=True, stateful=True,
                         batch_input_shape=[1, None, 1]),
  keras.layers.LSTM(100, return_sequences=True, stateful=True),
  keras.layers.Dense(1),
  keras.layers.Lambda(lambda x: x * 200.0)
])
optimizer = keras.optimizers.SGD(lr=5e-7, momentum=0.9)
model.compile(loss=keras.losses.Huber(),
              optimizer=optimizer,
              metrics=["mae"])
reset_states = ResetStatesCallback()
model_checkpoint = keras.callbacks.ModelCheckpoint(
    "my_checkpoint.h5", save_best_only=True)
early_stopping = keras.callbacks.EarlyStopping(patience=50)
model.fit(train_set, epochs=500,
          validation_data=valid_set,
          callbacks=[early_stopping, model_checkpoint, reset_states])

model = keras.models.load_model("my_checkpoint.h5")

rnn_forecast = model.predict(series[np.newaxis, :, np.newaxis])
rnn_forecast = rnn_forecast[0, split_time - 1:-1, 0]

plt.figure(figsize=(10, 6))
plot_series(time_valid, x_valid)
plot_series(time_valid, rnn_forecast)

keras.metrics.mean_absolute_error(x_valid, rnn_forecast).numpy()