-
Notifications
You must be signed in to change notification settings - Fork 0
/
tuning_LSTM.py
50 lines (42 loc) · 2.17 KB
/
tuning_LSTM.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
from model_modules import *
os.environ['TF_CPP_MIN_LOG_LEVEL'] = '3'
result = pd.DataFrame(columns= ["Results RMSE", "Train", "Validation", "Windows size", "Neurons", "Batch size"])
start_time = time.time()
# windows_size
window_size = random.sample(range(1, 100), 25)
neurons = random.sample(range(1, 100), 1)
batch = max(list(np.arange(34, 84, 2)))
dataset, train, data, scaler= prepare_LSTM('data/ESP.IDXEUR_Candlestick_1_Hour_BID_01.01.2015-31.12.2020.csv').data_LSTM()
i = 0
print("Tuning windows size, " + str(len(window_size)) + " models.")
while i < len(window_size):
y_train, y_val, X_train, X_val, trainX, valX, validation= split_LSTM(dataset, window_size[i]).model_SPLITS(data, train)
result.loc[i+1]= run_LSTM(window_size[i]).model_BASE(scaler, i+1, y_train, y_val, X_train, X_val, trainX, valX, validation)
print(str(i+1), sep=' ', end=',', flush=True)
i +=1
window_size_= int(result["Windows size"].iloc[[pd.to_numeric(result.Validation).idxmin()-1]])
print(" Best window size: " + str(window_size_))
y_train, y_val, X_train, X_val, trainX, valX, validation= split_LSTM(dataset, window_size_).model_SPLITS(data, train)
# neurons
result.drop(result.index, inplace=True)
neurons = random.sample(range(1, 100), 25)
i= 0
print("Tuninng neurons, " + str(len(neurons)) + " models.")
while i < len(neurons):
result.loc[i+1]= run_LSTM(window_size_, neurons[i]).model_BASE(scaler, i+1, y_train, y_val, X_train, X_val, trainX, valX, validation)
print(str(i+1), sep=' ', end=',', flush=True)
i +=1
neurons_= int(result["Neurons"].iloc[[pd.to_numeric(result.Validation).idxmin()-1]])
print(" Best number of neurons: " + str(neurons_))
# batch
result.drop(result.index, inplace=True)
batch = np.arange(32, 82, 2).tolist()
i= 0
print("Tuning batch size, " + str(len(batch)) + " models.")
while i < len(batch):
result.loc[i+1]= run_LSTM(window_size_, neurons_, batch[i]).model_BASE(scaler, i+1, y_train, y_val, X_train, X_val, trainX, valX, validation)
print(str(i+1), sep=' ', end=',', flush=True)
i +=1
fig,ax = render_result(result, "table_results.png").create_table()
fig.savefig("figures/table_results.png")
print(str(int(time.time() - start_time)/60) + " minutes")