-
Notifications
You must be signed in to change notification settings - Fork 0
/
lstm_model.py
236 lines (202 loc) · 9.33 KB
/
lstm_model.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
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
from __future__ import print_function
import time
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
from keras.layers.core import Dense, Dropout
from keras.layers.recurrent import LSTM
from keras.models import Sequential
from keras.callbacks import EarlyStopping
from sklearn.metrics import r2_score, mean_squared_error
import os
import warnings
from keras.models import load_model
warnings.filterwarnings("ignore", category=DeprecationWarning)
os.environ['KMP_DUPLICATE_LIB_OK'] = 'True'
start_time = time.time()
# Defining a function to convert a vector of time series into a 2D matrix for faster processing
def convertTimeSeriesTo2DMatrix(vectorSeries, sequence_length):
matrix = []
for i in range(len(vectorSeries) - sequence_length + 1):
matrix.append(vectorSeries[i:i + sequence_length])
return matrix
np.random.seed(1234) # Selecting a random seed
# Pre-processing of the data
df_raw = pd.read_csv('assets/hourly_loaddata.csv', header=None, skiprows=1) # loading raw data from the CSV
df_raw_array = df_raw.values # numpy array
# daily_load = [df_raw_array[i,:] for i in range(0, len(df_raw)) if i % 24 == 0] # daily load
# print(daily_load)
hourly_load = [df_raw_array[i, 2] / 100 for i in range(0, len(df_raw))] # hourly load, 24 for each day
# print(hourly_load)
length_of_sequence = 24 # Storing the length of the sequence/hours in the day for predicting the future value
# Converting the vector to a 2D matrix using the function above
hourly_load_matrix = convertTimeSeriesTo2DMatrix(hourly_load, length_of_sequence)
# Shift all the data by mean
hourly_load_matrix = np.array(hourly_load_matrix)
shifted_value = hourly_load_matrix.mean()
hourly_load_matrix = hourly_load_matrix - shifted_value
# print ("Data shape: ", hourly_load_matrix.shape)
# print(hourly_load_matrix)
# Splitting the dataset into two: 90% for training and 10% for testing
test_row = int(round(0.9 * hourly_load_matrix.shape[0]))
train_set = hourly_load_matrix[:test_row, :]
np.random.shuffle(train_set) # Shuffling only the training set randomly
# print(train_set, "\n")
# dataset_graph = plt.figure(figsize=(60, 8))
# plt.plot(hourly_load, label='Load')
# plt.legend(loc='upper right')
# plt.title("Dataset", fontsize=14)
# plt.xlabel('Hours', fontsize=14)
# plt.ylabel('Values', fontsize=14)
# plt.legend()
# plt.show()
# dataset_graph.savefig('results/LSTM/dataset_graph.jpg', bbox_inches='tight')
# The Final training set
X_train = train_set[:, :-1]
print("X_train: ",X_train.shape, "\n", X_train, "\n")
y_train = train_set[:, -1] # The last column is the true value to compute the mean-squared-error loss
print("y_train",y_train.shape, "\n", y_train, "\n")
# The Final testing set
X_test = hourly_load_matrix[test_row:, :-1]
print("X_test: ",X_test.shape, "\n", X_test, "\n")
y_test = hourly_load_matrix[test_row:, -1]
print("y_test: ",y_test.shape,"\n", y_test, "\n")
# The input to LSTM layer needs to have the shape of (number of samples, the dimension of each element)
X_train = np.reshape(X_train, (X_train.shape[0], X_train.shape[1], 1))
X_test = np.reshape(X_test, (X_test.shape[0], X_test.shape[1], 1))
# Building the LSTM model
model = Sequential()
# Layer 1: LSTM
model.add(LSTM(input_dim=1, units=50, return_sequences=True))
model.add(Dropout(0.2)) # Reducing overfitting and improving model performance
# Layer 2: LSTM
model.add(LSTM(units=100, return_sequences=False))
model.add(Dropout(0.2)) # Reducing overfitting and improving model performance
# Layer 3: Dense
model.add(Dense(units=1, activation='linear'))
# Compiling the model
model.compile(loss="mse", optimizer="adam")
es = EarlyStopping(monitor="val_loss", min_delta=0, patience=5, verbose=1, mode="auto", baseline=None,
restore_best_weights=True) # Stops the training when the values don't improve
#
# # Training the model
# model.fit(X_train, y_train, batch_size=4, epochs=1, validation_split=0.05, verbose=1, callbacks=[es])
# test_mse = model.evaluate(X_test, y_test, verbose=1)
#
# # Values of hyperparameters for finding the best values
# epoch = [1, 5, 10, 25, 50, 100]
# batch_size = [4, 12, 16, 32, 64, 256, 512, 1024, 2048]
#
# # Finding the best combination of hyperparameters
# best_mse, final_epoch, final_batch_size = 0, 0, 0
# result = []
# for i in epoch:
# for j in batch_size:
# model.fit(X_train, y_train, batch_size=j, epochs=i, validation_split=0.05, verbose=1, callbacks=[es])
# if model.evaluate(X_test, y_test, verbose=1) < test_mse:
# test_mse = model.evaluate(X_test, y_test, verbose=1)
# test_mse = model.evaluate(X_test, y_test, verbose=1)
# mse_combination = [i, j, test_mse]
# print("Model w/ Epoch: ", i, "|| Batch Size:", j, "|| MSE: ", test_mse, "\n")
# result.append(mse_combination)
#
# # Storing all the values in CSV file: 'MSEs.csv'
# np.savetxt("results/LSTM/MSEs.csv", result, delimiter=", ", header="Epoch, Batch Size, MSE", fmt='% s')
#
# data_frame = pd.read_csv("results/LSTM/MSEs.csv")
# print(data_frame)
#
# # Selecting the top 5 model combinations to find the best performing
# mse_frame = pd.DataFrame(result)
# final_mse_frame = mse_frame.sort_values(by=[2], ascending=True)
# epoch_list = [final_mse_frame[0].iloc[0], final_mse_frame[0].iloc[1], final_mse_frame[0].iloc[2],
# final_mse_frame[0].iloc[3], final_mse_frame[0].iloc[4]]
# batch_size_list = [final_mse_frame[1].iloc[0], final_mse_frame[1].iloc[1], final_mse_frame[1].iloc[2],
# final_mse_frame[1].iloc[3], final_mse_frame[1].iloc[4]]
#
# result = []
#
# for i in range(5):
# for j in range(3):
# print("Iteration no.:", j, "|| Model:", epoch_list[i], batch_size_list[i])
# model.fit(X_train, y_train, batch_size=batch_size_list[i], epochs=epoch_list[i], validation_split=0.05,
# verbose=1, callbacks=[es])
# if model.evaluate(X_test, y_test, verbose=1) < test_mse:
# test_mse = model.evaluate(X_test, y_test, verbose=1)
# final_epoch = epoch_list[i]
# final_batch_size = batch_size_list[i]
# test_mse = model.evaluate(X_test, y_test, verbose=1)
# print("Model w/ Epoch: ", epoch_list[i], "|| Batch Size: ", batch_size_list[i], "|| MSE: ", test_mse, "\n")
# mse_combination = [epoch_list[i], batch_size_list[i], test_mse]
# result.append(mse_combination)
#
# # Storing all the values in CSV file: 'MSEs_top5models.csv'
# np.savetxt("results/LSTM/MSEs_top5models.csv", result, delimiter=", ", header="Epoch, Batch Size, MSE", fmt='% s')
# data_frame = pd.read_csv("results/LSTM/MSEs_top5models.csv")
# print(data_frame)
#
# # # Finding the best combination
# # counter = 0
# # mse = []
# # temp = []
# # mse_frame = pd.DataFrame(result)
# # for i in range(5):
# # for j in range(3):
# # temp.append(mse_frame[2].iloc[counter])
# # counter = counter+1
# # mse.append(temp)
# #
# # average_mse = np.array(temp)
# # average_mse = np.sum(average_mse, axis=1)/3
# # max_mse_index = np.where(np.max(average_mse))
# # # Find the best epoch & batch size combination to train the best model
#Values taken from the above
final_batch_size = 1024
final_epoch = 100
# Training the best model
history = model.fit(X_train, y_train, batch_size=final_batch_size, epochs=final_epoch, validation_split=0.05, verbose=1,
callbacks=[es])
print(model.summary())
model.save('assets/univariate/lstm.tf', overwrite=True, include_optimizer=True)
# loaded_model = load_model('path')
# Evaluating the result
test_mse = model.evaluate(X_test, y_test, verbose=1)
print('\nThe Mean-squared-error (MSE) on the test data set is %.6f over %d test samples.' % (test_mse, len(y_test)))
# Getting the predicted values
predicted_values = model.predict(X_test)
num_test_samples = len(predicted_values)
predicted_values = np.reshape(predicted_values, (num_test_samples, 1))
# print(predicted_values)
print('The MSE value is:',
mean_squared_error((predicted_values + shifted_value) * 100, (y_test + shifted_value) * 100, squared=True))
print('The RMSE value is:',
mean_squared_error((predicted_values + shifted_value) * 100, (y_test + shifted_value) * 100, squared=False))
print('The R-squared value is:', r2_score(predicted_values + shifted_value, y_test + shifted_value))
print('The MAPE value is:', np.mean(np.abs(((y_test+shifted_value) - np.reshape(predicted_values+shifted_value, (3545,))) / (y_test+shifted_value))) * 100,'\n')
# Plotting the results
fig = plt.figure()
plt.plot((y_test + shifted_value) * 100)
plt.plot((predicted_values + shifted_value) * 100)
plt.title("LSTM")
plt.xlabel('Hour')
plt.ylabel('Electricity load')
plt.legend(('Actual', 'Predicted'), fontsize='15')
plt.show()
fig.savefig('results/LSTM/final_output.jpg', bbox_inches='tight')
# Plot of the loss
loss_fig = plt.figure()
plt.plot(history.history['loss'])
plt.plot(history.history['val_loss'])
plt.title('Model Loss')
plt.ylabel('Loss')
plt.xlabel('Epoch')
plt.legend(['Train', 'Validation'], loc='upper left')
plt.show()
loss_fig.savefig('results/LSTM/final_loss.jpg', bbox_inches='tight')
# Storing the result in a file: 'load_forecasting_result.txt'
predicted_test_result = (predicted_values + shifted_value) * 100
np.savetxt('results/LSTM/predicted_values.txt', predicted_test_result)
actual_test_result = (y_test + shifted_value) * 100
np.savetxt('results/LSTM/test_values.txt', actual_test_result)
end_time = time.time()
print("Total time: ", end_time - start_time)