-
Notifications
You must be signed in to change notification settings - Fork 0
/
HyperParameterTuner.py
150 lines (132 loc) · 4.88 KB
/
HyperParameterTuner.py
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
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
import tensorflow as tf
from tensorflow import keras
import keras_tuner as kt
from CNNLSTM import CNNLSTM
from Stocks import Stocks
from sklearn.ensemble import RandomForestRegressor
import numpy as np
import pandas as pd
class HyperParameterTuner:
def __init__(self, train_data, train_labels, test_data, test_labels):
self.train_data = train_data
self.train_labels = train_labels
self.test_data = test_data
self.test_labels = test_labels
def dnn_model_builder(
self,
hp,
hp_units_min: int = 32,
hp_units_max: int = 128,
hp_units_step: int = 32,
hp_layers_min: int = 1,
hp_layers_max: int = 5,
hp_layers_step: int = 1,
hp_learning_rates: [float] = [1e-1, 1e-2, 1e-3, 1e-4],
hp_loss: str = "mae",
) -> tf.keras.Model:
hp_units = hp.Int(
"units", min_value=hp_units_min, max_value=hp_units_max, step=hp_units_step
)
hp_layers = hp.Int(
"layers",
min_value=hp_layers_min,
max_value=hp_layers_max,
step=hp_layers_step,
)
hp_learning_rate = hp.Choice("learning_rate", values=hp_learning_rates)
model = keras.Sequential()
model.add(keras.layers.Flatten())
model.add(keras.layers.Dense(units=hp_units, activation="relu"))
model.add(keras.layers.Dense(hp_layers))
model.compile(
optimizer=keras.optimizers.Adam(learning_rate=hp_learning_rate),
loss=hp_loss,
metrics=["accuracy"],
)
return model
def kt_dnn_tuner(
self,
epochs: int = 1000,
factor: int = 5,
directory: str = "dnn",
hyperband_interations: int = 1,
) -> []:
tuner = kt.Hyperband(
self.dnn_model_builder,
objective="val_accuracy",
max_epochs=int(epochs / 5),
factor=factor,
directory=f"results/dnn/",
project_name=f"all_stocks_dnn",
)
stop_early = tf.keras.callbacks.EarlyStopping(monitor="val_loss", patience=25)
tuner.search(
self.train_data,
self.train_labels,
epochs=epochs,
validation_split=0.2,
callbacks=[stop_early],
)
best_hps = tuner.get_best_hyperparameters(num_trials=1)[0]
print(
f"""
The hyperparameter search is complete. The optimal number of units in the first densely-connected
layer is {best_hps.get('units')} and the optimal learning rate for the optimizer
is {best_hps.get('learning_rate')}. The best number of layers if {best_hps.get('layers')}.
"""
)
model = tuner.hypermodel.build(best_hps)
history = model.fit(
self.train_data,
self.train_labels,
validation_split=0.2,
verbose=0,
epochs=500,
)
val_acc_per_epoch = history.history["val_accuracy"]
best_epoch = val_acc_per_epoch.index(max(val_acc_per_epoch)) + 1
print("Best epoch: %d" % (best_epoch,))
hist = pd.DataFrame(history.history)
hist["epoch"] = history.epoch
print(hist.tail(20))
CNNLSTM.plot(history)
CNNLSTM.summarize_model(model)
CNNLSTM.evaluate_model(model, self.test_data, self.test_labels)
CNNLSTM.preditcted_plot(model, self.test_data, self.test_labels)
test_results = model.evaluate(self.test_data, self.test_labels, verbose=0)
return best_hps
def random_forest(self):
rf = RandomForestRegressor(n_estimators=100, random_state=42, n_jobs=-1)
rf.fit(self.train_data, self.train_labels)
predictions = rf.predict(self.test_data)
errors = abs(predictions - self.test_labels)
mape = 100 * (errors / self.test_labels)
accuracy = 100 - np.mean(mape)
print("Accuracy:", round(accuracy, 2), "%.")
print(
pd.Series(
rf.feature_importances_, index=self.train_data.columns
).sort_values(ascending=False)
)
return rf
def remove_nan(self):
self.train_data.dropna(inplace=True)
self.train_labels.dropna(inplace=True)
self.test_data.dropna(inplace=True)
self.test_labels.dropna(inplace=True)
if __name__ == "__main__":
stocks = Stocks("../stock_market_data")
# stocks.combine_data()
# stocks.data_pipeline()
# stocks.load_data("stocks_train.csv", "stocks_val.csv", "stocks_test.csv")
stocks.load_data("AAPL_train.csv", "AAPL_val.csv", "AAPL_test.csv")
train_x = stocks.train_df
train_y = train_x.pop("Close")
val_x = stocks.val_df
val_y = val_x.pop("Close")
test_x = stocks.test_df
test_y = test_x.pop("Close")
tuner = HyperParameterTuner(train_x, train_y, test_x, test_y)
tuner.remove_nan()
# tuner.kt_dnn_tuner()
tuner.random_forest()