antonvedaet/stock-market-price-forecast

Stock market price forecast 👍

The name of this project is pretty self explanatory so i won't go in details on that.

LSTM architecture

I used Long Short-term memory artificial recurrent neural network (RNN) architecture in this project. LSTM networks are well-suited to classifying, processing and making predictions based on time series data, since there can be lags of unknown duration between important events in a time series.

Coding

Imports

import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
import seaborn as sns
import pandas_datareader as web
import math
from keras.models import Sequential
from tensorflow.keras import layers
from sklearn.preprocessing import MinMaxScaler

Data gathering

For this example we are going to predict prices of Crude Oil

day = int(input('Какой Сегодня день?', ))
month = int(input('Какого месяца?', ))
year = int(input('Какого года?', ))
today = f'{day}-{month}-{year}'
print('Сегодня ', today)

data = web.DataReader('CL=F', data_source='yahoo', start='1980-01-01', end=today)

Preprocessing and train data creation

df = data.filter(['Close'])
df_val = df.values
train_len = math.ceil(len(df)*0.8)

scaler = MinMaxScaler(feature_range=(0, 1))
scaled_data = scaler.fit_transform(df_val)

train_data = scaled_data[:train_len, :]

X_train=[]
y_train=[]

for i in range(60, len(train_data)):
  X_train.append(train_data[i-60:i, 0])
  y_train.append(train_data[i, 0])

X_train, y_train = np.array(X_train), np.array(y_train)    
X_train = np.expand_dims(X_train, 1)

Model

model = Sequential([
    layers.Dense(16, activation='relu', input_shape=(X_train.shape[1:])),
    layers.LSTM(64, return_sequences=True),
    layers.Dropout(0.2),
    layers.LSTM(64, return_sequences=True),
    layers.Dropout(0.2),
    layers.LSTM(64),
    layers.Dropout(0.4),
    layers.Flatten(),
    layers.Dense(16, activation='relu'),
    layers.Dense(8, activation='elu'),
    layers.Dense(1)
    ])
    
 model.compile(optimizer='adam', loss='mse')

Training and creating test data

model.fit(X_train, y_train, batch_size=128, epochs=100)


test_data = scaled_data[train_len - 60: , :]

X_test = []
y_test = df_val[train_len:, :]
for i in range(60, len(test_data)):
    X_test.append(test_data[i-60:i, 0])
    
X_test = np.array(X_test)
X_test = np.expand_dims(X_test, 1)
 
pred = model.predict(X_test)
pred = scaler.inverse_transform(pred)

Metrics

train = data[:train_len]
valid = data[train_len:]
valid['Predictions'] = pred
valid
fig = plt.figure()
plt.xlabel('Дата', fontsize = 18)
plt.ylabel('USD', fontsize = 18)
plt.plot(train['Close'])
plt.plot(valid[['Close', 'Predictions']])
plt.grid(True)

def smape(A, F):
return 100/len(A) * np.sum(2 * np.abs(F - A) / (np.abs(A) + np.abs(F)))
A = valid['Close']
F = valid['Predictions']
print('sMAPE = ',smape(A, F))

sMAPE = 3.21