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
| import time
import numpy as np
import pandas as pd
import torch
import torch.nn as nn
from torch.utils.data import Dataset, DataLoader
from sklearn.preprocessing import MinMaxScaler
import matplotlib.pyplot as plt
device = torch.device("cpu")
print(f"Using device: {device}")
torch.manual_seed(42)
data = pd.read_csv("./bitcoin.csv")
data['date'] = pd.to_datetime(data['date'])
data.set_index('date', inplace=True)
gold = pd.read_csv("./gold.csv")
gold['date'] = pd.to_datetime(gold['date'])
gold.set_index('date', inplace=True)
scaler = MinMaxScaler(feature_range=(0, 1))
scaled_bitcoin = scaler.fit_transform(data['value'].values.reshape(-1, 1))
scaled_gold = scaler.fit_transform(gold['value'].values.reshape(-1, 1))
class LSTMDataset(Dataset):
def __init__(self, data, seq_length):
self.data = data
self.seq_length = seq_length
def __len__(self):
return len(self.data) - self.seq_length
def __getitem__(self, index):
x = self.data[index:index + self.seq_length]
y = self.data[index + self.seq_length]
return (torch.tensor(x, dtype=torch.float32).to(device),
torch.tensor(y, dtype=torch.float32).to(device))
SEQ_LENGTH = 5
BATCH_SIZE = 512
LEARNING_RATE = 0.001
NUM_EPOCHS = 2000
dataset_bitcoin = LSTMDataset(scaled_bitcoin, SEQ_LENGTH)
dataset_gold = LSTMDataset(scaled_gold, SEQ_LENGTH)
train_size = int(0.5 * len(dataset_bitcoin))
test_size = len(dataset_bitcoin) - train_size
train_dataset_bitcoin, test_dataset_bitcoin = (
torch.utils.data.random_split(dataset_bitcoin, [train_size, test_size]))
train_loader_bitcoin = DataLoader(train_dataset_bitcoin, batch_size=BATCH_SIZE, shuffle=True)
test_loader_bitcoin = DataLoader(test_dataset_bitcoin, batch_size=BATCH_SIZE, shuffle=False)
train_dataset_gold, test_dataset_gold = torch.utils.data.random_split(dataset_gold, [train_size, test_size])
train_loader_gold = DataLoader(train_dataset_gold, batch_size=BATCH_SIZE, shuffle=True)
test_loader_gold = DataLoader(test_dataset_gold, batch_size=BATCH_SIZE, shuffle=False)
class LSTM(nn.Module):
def __init__(self, input_size=1, hidden_size=5, num_layers=1, output_size=1):
super(LSTM, self).__init__()
self.hidden_size = hidden_size
self.num_layers = num_layers
self.lstm = nn.LSTM(input_size, hidden_size, num_layers, batch_first=True)
self.fc = nn.Linear(hidden_size, output_size)
def forward(self, x):
h0 = torch.zeros(self.num_layers, x.size(0), self.hidden_size).to(x.device)
c0 = torch.zeros(self.num_layers, x.size(0), self.hidden_size).to(x.device)
out, _ = self.lstm(x, (h0, c0))
out = self.fc(out[:, -1, :])
return out
model = LSTM().to(device)
criterion = nn.SmoothL1Loss().to(device)
optimizer = torch.optim.RMSprop(model.parameters(), lr=LEARNING_RATE)
st = time.time()
print("Training started at:", time.strftime("%Y-%m-%d %H:%M:%S", time.localtime(st)))
train_losses = []
for epoch in range(NUM_EPOCHS):
model.train()
total_loss = 0.0
for ((inputs, targets), (inputs2, targets2)) in zip(train_loader_bitcoin, train_loader_gold):
outputs = model(inputs)
outputs2 = model(inputs2)
loss = criterion(outputs, targets) + criterion(outputs2, targets2)
optimizer.zero_grad()
loss.backward()
optimizer.step()
total_loss += loss.item()
aveLoss = total_loss / len(train_loader_bitcoin)
train_losses.append(aveLoss)
if (epoch + 1) % (NUM_EPOCHS / 10) == 0:
print(f"Epoch [{epoch + 1}/{NUM_EPOCHS}], Loss: {aveLoss}")
et = time.time()
print("Training finished at:", time.strftime("%Y-%m-%d %H:%M:%S", time.localtime(et)) +
", duration:", str(et - st), "seconds")
loss_samples = []
epochId = []
for i in range(NUM_EPOCHS):
if (i + 1) % (NUM_EPOCHS / 50) == 0:
loss_samples.append(train_losses[i])
epochId.append(i + 1)
plt.figure(figsize=(10, 6))
plt.plot(epochId, loss_samples, marker='o', label="Training Loss")
plt.xlabel("Epochs")
plt.ylabel("Loss")
plt.title("Training Loss vs Epochs")
plt.legend()
plt.grid(True)
plt.show()
def create_sequences(data, seq_length):
sequences = []
labels = []
for i in range(len(data) - seq_length):
seq = data[i:i + seq_length]
label = data[i + seq_length]
sequences.append(seq)
labels.append(label)
return (torch.tensor(np.array(sequences), dtype=torch.float32).to(device),
torch.tensor(np.array(labels), dtype=torch.float32).to(device))
X, y = create_sequences(scaled_bitcoin, SEQ_LENGTH)
X2, y2 = create_sequences(scaled_gold, SEQ_LENGTH)
model.eval()
with torch.no_grad():
test_pred = model(X)
test2_pred = model(X2)
test_pred = scaler.inverse_transform(test_pred.cpu().numpy())
test2_pred = scaler.inverse_transform(test2_pred.cpu().numpy())
y = scaler.inverse_transform(y.cpu().numpy())
y2 = scaler.inverse_transform(y2.cpu().numpy())
plt.figure(figsize=(14, 6))
plt.plot(y, label="Actual Price")
plt.plot(test_pred, label="Predicted Price")
plt.legend()
plt.title("Bitcoin Price Prediction")
plt.show()
plt.figure(figsize=(14, 6))
plt.plot(y2, label="Actual Price")
plt.plot(test2_pred, label="Predicted Price")
plt.legend()
plt.title("Gold Price Prediction")
plt.show()
mape_bitcoin = 0.0
mape_gold = 0.0
for i, j in zip(y, test_pred):
mape_bitcoin += abs(j - i) / j
mape_bitcoin /= len(y)
for i, j in zip(y2, test2_pred):
mape_gold += abs(j - i) / j
mape_gold /= len(y2)
mape_bitcoin = mape_bitcoin[0]
mape_gold = mape_gold[0]
print("Bitcoin's MAPE:", mape_bitcoin)
print("Gold's MAPE:", mape_gold)
torch.save(model.state_dict(),
"./models/price_predict_model-" + str(et) + ".pth")
print("Model saved as " + "./models/price_predict_model-" + str(et) + ".pth")
|
