Predicting Stock Prices using LSTM Neural Networks with TensorFlow
Python
In this example, we're using an LSTM neural network in TensorFlow to predict stock prices based on historical data.
1| import numpy as np 2| import pandas as pd 3| import tensorflow as tf 4| from tensorflow.keras.models import Sequential 5| from tensorflow.keras.layers import LSTM, Dense 6| from sklearn.metrics import mean_squared_error 7| 8| # Load and preprocess data 9| data = pd.read_csv('stock_data.csv') 10| data = data.dropna() 11| close_prices = data['Close'].values.reshape(-1, 1) 12| 13| # Scale the data 14| from sklearn.preprocessing import MinMaxScaler 15| scaler = MinMaxScaler() 16| scaled_prices = scaler.fit_transform(close_prices) 17| 18| # Split the data into training and testing sets 19| train_size = int(len(scaled_prices) * 0.8) 20| test_size = len(scaled_prices) - train_size 21| train_prices, test_prices = scaled_prices[0:train_size,:], scaled_prices[train_size:len(scaled_prices),:] 22| 23| # Define the function to create the dataset for training the LSTM model 24| def create_dataset(dataset, time_steps=1): 25| dataX, dataY = [], [] 26| for i in range(len(dataset)-time_steps-1): 27| a = dataset[i:(i+time_steps), 0] 28| dataX.append(a) 29| dataY.append(dataset[i + time_steps, 0]) 30| return np.array(dataX), np.array(dataY) 31| 32| # Set the number of time steps and create the training and testing datasets 33| time_steps = 30 34| X_train, y_train = create_dataset(train_prices, time_steps) 35| X_test, y_test = create_dataset(test_prices, time_steps) 36| 37| # Reshape the input data for the LSTM model 38| X_train = np.reshape(X_train, (X_train.shape[0], X_train.shape[1], 1)) 39| X_test = np.reshape(X_test, (X_test.shape[0], X_test.shape[1], 1)) 40| 41| # Define the LSTM model architecture 42| model = Sequential() 43| model.add(LSTM(units=50, return_sequences=True, input_shape=(X_train.shape[1], 1))) 44| model.add(LSTM(units=50)) 45| model.add(Dense(units=1)) 46| 47| # Compile the model 48| model.compile(optimizer='adam', loss='mean_squared_error') 49| 50| # Train the model 51| model.fit(X_train, y_train, epochs=100, batch_size=32) 52| 53| # Make predictions on the testing set 54| predictions = model.predict(X_test) 55| 56| # Scale the predictions back to their original values 57| predictions = scaler.inverse_transform(predictions) 58| 59| # Evaluate the performance of the model 60| rmse = np.sqrt(mean_squared_error(y_test, predictions)) 61| print(f'Root Mean Squared Error: {rmse}')
133
121
116
109