How to overcome this error in the code: Error:…
- User
How to overcome this error in the code: Error:
ValueError Traceback (most recent call last) in <cell line: 4>() 11 12 new_row = np.concatenate((next_day_input[0, 1:, :], next_day_extended), axis=0) ---> 13 X_test = np.append(X_test, [new_row], axis=0) 14 15 predictions = np.array(predictions).reshape(-1, 1)
2 frames /usr/local/lib/python3.10/dist-packages/numpy/core/overrides.py in concatenate(*args, **kwargs)
ValueError: all the input array dimensions for the concatenation axis must match exactly, but along dimension 1, the array at index 0 has size 40 and the array at index 1 has size 39
Code: import numpy as np import pandas as pd import matplotlib.pyplot as plt import yfinance as yf from sklearn.preprocessing import MinMaxScaler from keras.models import Sequential from keras.layers import Dense, LSTM, Dropout
def calculate_atr(data, period): data['h-l'] = data['high'] - data['low'] data['h-pc'] = abs(data['high'] - data['close'].shift(1)) data['l-pc'] = abs(data['low'] - data['close'].shift(1)) data['tr'] = data[['h-l', 'h-pc', 'l-pc']].max(axis=1) data['atr'] = data['tr'].rolling(window=period).mean() return data
def calculate_super_trend(data, period, multiplier): data = calculate_atr(data, period) data['Upper Basic'] = (data['high'] + data['low']) / 2 + multiplier * data['atr'] data['Lower Basic'] = (data['high'] + data['low']) / 2 - multiplier * data['atr'] data['Upper Band'] = data.apply(lambda x: x['Upper Basic'] if x['close'] > x['Upper Basic'] else x['Lower Basic'], axis=1) data['Lower Band'] = data.apply(lambda x: x['Lower Basic'] if x['close'] < x['Lower Basic'] else x['Upper Basic'], axis=1) data['Super Trend'] = np.where(data['close'] > data['Upper Band'], data['Lower Band'], data['Upper Band']) return data.dropna()
def load_preprocess_data(ticker, start_date, end_date, window_size, period=14, multiplier=3): stock_data = yf.download(ticker, start=start_date, end=end_date) print("Original columns:", stock_data.columns) stock_data.columns = [col.lower() for col in stock_data.columns] stock_data_with_super_trend = calculate_super_trend(stock_data, period, multiplier)
columns_to_use = stock_data_with_super_trend[['open', 'high', 'low', 'close', 'Super Trend']].values scaler = MinMaxScaler(feature_range=(0, 1)) data_normalized = scaler.fit_transform(columns_to_use) X, y = [], [] for i in range(window_size, len(data_normalized)): X.append(data_normalized[i - window_size:i]) y.append(data_normalized[i, 3]) # Use Close prices directly as labels train_len = int(0.8 * len(X)) X_train, y_train = np.array(X[:train_len]), np.array(y[:train_len]) X_test, y_test = np.array(X[train_len:]), np.array(y[train_len:]) return X_train, y_train, X_test, y_testdef create_lstm_model(input_shape): model = Sequential() model.add(LSTM(units=50, return_sequences=True, input_shape=input_shape)) model.add(LSTM(units=50, return_sequences=True)) model.add(LSTM(units=50)) model.add(Dense(units=1))
model.compile(optimizer='adam', loss='mean_squared_error') return modeldef train_model(model, X_train, y_train, batch_size, epochs): history = model.fit(X_train, y_train, batch_size=batch_size, epochs=epochs, verbose=1, validation_split=0.1) return model, history
def evaluate_model(model, X_test, y_test, scaler): y_pred = model.predict(X_test) y_pred_inverse = scaler.inverse_transform(np.column_stack((np.zeros(y_pred.shape), y_pred)))[:, 1] # Inverse transform for Close prices y_test_inverse = scaler.inverse_transform(np.column_stack((np.zeros(y_test.shape), y_test)))[:, 1]
mae = np.mean(np.abs(y_pred_inverse - y_test_inverse)) mse = np.mean(np.square(y_pred_inverse - y_test_inverse)) return mae, msedef plot_prediction(model, X_test, y_test, scaler): y_pred = model.predict(X_test) y_pred_inverse = scaler.inverse_transform(np.column_stack((np.zeros(y_pred.shape), y_pred)))[:, 1] # Inverse transform for Close prices y_test_inverse = scaler.inverse_transform(np.column_stack((np.zeros(y_test.shape), y_test)))[:, 1]
plt.figure(figsize=(14, 6)) plt.plot(y_test_inverse, label='Actual') plt.plot(y_pred_inverse, label='Predicted') plt.xlabel('Days') plt.ylabel('Close Price') plt.title('Stock Price Prediction (Actual vs Predicted)') plt.legend() plt.show()Define the parameters
ticker = '^NSEI' # Ticker symbol for Nifty start_date = '2010-01-01' end_date = '2023-06-01' window_size = 40 # Number of previous days' data to consider period = 14 # ATR period multiplier = 3 # ATR multiplier batch_size = 32 epochs = 100
X_train, y_train, X_test, y_test = load_preprocess_data(ticker, start_date, end_date, window_size, period, multiplier)
Create the LSTM model
input_shape = (X_train.shape[1], X_train.shape[2]) model = create_lstm_model(input_shape)
Train the model
model, history = train_model(model, X_train, y_train, batch_size, epochs)
scaler = MinMaxScaler(feature_range=(0, 1)) scaler.fit_transform(y_train.reshape(-1, 1)) mae, mse = evaluate_model(model, X_test, y_test, scaler) print(f"Mean Absolute Error: {mae}") print(f"Mean Squared Error: {mse}")
Plot the predictions
plot_prediction(model, X_test, y_test, scaler)
upcoming_days = 5 predictions = []
for i in range(upcoming_days): next_day_input = np.array([X_test[-1, 1:, :]]) next_day_pred = model.predict(next_day_input) predictions.append(next_day_pred[0, 0])
next_day_extended = np.zeros((1, next_day_input.shape[2])) next_day_extended[:, 3] = next_day_pred.T new_row = np.concatenate((next_day_input[0, 1:, :], next_day_extended), axis=0) X_test = np.append(X_test, [new_row], axis=0)predictions = np.array(predictions).reshape(-1, 1) predictions_inverse = scaler.inverse_transform(np.column_stack((np.zeros(predictions.shape), predictions)))[:, 1] # Inverse transform for Close prices
plt.figure(figsize=(14, 6)) plt.plot(predictions_inverse, marker="o", label='Predicted') plt.xlabel('Days') plt.ylabel('Close Price') plt.title('NIFTY Stock Price Prediction for Upcoming 5 Days') plt.legend() plt.show()
- Assistant
The error occurs because the dimensions of the
new_rowandX_testdo not match along axis 1 for the concatenation operation. To resolve the issue, ensure that the concatenated arrays have matching dimensions along the specified axis.In the code,
next_day_extendedis missing a column, so it has only 4 columns instead of 5:- Initialize
next_day_extendedwith the same number of columns asnext_day_input:
next_day_extended = np.zeros((1, next_day_input.shape[2] + 1))- Assign the predicted value to the 4th column of
next_day_extended, keeping the 5th column for 'Super Trend':
next_day_extended[:, 3] = next_day_pred.T- Now, concatenate
next_day_input[0, 1:, :]andnext_day_extendedalong axis 0:
new_row = np.concatenate((next_day_input[0, 1:, :], next_day_extended), axis=0)Here's the modified code for the loop:
upcoming_days = 5 predictions = [] for i in range(upcoming_days): next_day_input = np.array([X_test[-1, 1:, :]]) next_day_pred = model.predict(next_day_input) predictions.append(next_day_pred[0, 0]) next_day_extended = np.zeros((1, next_day_input.shape[2] + 1)) next_day_extended[:, 3] = next_day_pred.T new_row = np.concatenate((next_day_input[0, 1:, :], next_day_extended), axis=0) X_test = np.append(X_test, [new_row], axis=0) predictions = np.array(predictions).reshape(-1, 1) predictions_inverse = scaler.inverse_transform(np.column_stack((np.zeros(predictions.shape), predictions)))[:, 1] # Inverse transform for Close pricesHowever, you could still face issues with this code, mainly because the 'Super Trend' column is not updated with new values for future predictions. Depending on your requirements, you should consider updating this column appropriately.
- Initialize
- User
Modify this code for prediction for next 5 days: Code: import numpy as np import pandas as pd import matplotlib.pyplot as plt import yfinance as yf from sklearn.preprocessing import MinMaxScaler from keras.models import Sequential from keras.layers import Dense, LSTM, Dropout
def calculate_atr(data, period): data['h-l'] = data['high'] - data['low'] data['h-pc'] = abs(data['high'] - data['close'].shift(1)) data['l-pc'] = abs(data['low'] - data['close'].shift(1)) data['tr'] = data[['h-l', 'h-pc', 'l-pc']].max(axis=1) data['atr'] = data['tr'].rolling(window=period).mean() return data
def calculate_super_trend(data, period, multiplier): data = calculate_atr(data, period) data['Upper Basic'] = (data['high'] + data['low']) / 2 + multiplier * data['atr'] data['Lower Basic'] = (data['high'] + data['low']) / 2 - multiplier * data['atr'] data['Upper Band'] = data.apply(lambda x: x['Upper Basic'] if x['close'] > x['Upper Basic'] else x['Lower Basic'], axis=1) data['Lower Band'] = data.apply(lambda x: x['Lower Basic'] if x['close'] < x['Lower Basic'] else x['Upper Basic'], axis=1) data['Super Trend'] = np.where(data['close'] > data['Upper Band'], data['Lower Band'], data['Upper Band']) return data.dropna()
def load_preprocess_data(ticker, start_date, end_date, window_size, period=14, multiplier=3): stock_data = yf.download(ticker, start=start_date, end=end_date) print("Original columns:", stock_data.columns) stock_data.columns = [col.lower() for col in stock_data.columns] stock_data_with_super_trend = calculate_super_trend(stock_data, period, multiplier)
columns_to_use = stock_data_with_super_trend[['open', 'high', 'low', 'close', 'Super Trend']].values scaler = MinMaxScaler(feature_range=(0, 1)) data_normalized = scaler.fit_transform(columns_to_use) X, y = [], [] for i in range(window_size, len(data_normalized)): X.append(data_normalized[i - window_size:i]) y.append(data_normalized[i, 3]) # Use Close prices directly as labels train_len = int(0.8 * len(X)) X_train, y_train = np.array(X[:train_len]), np.array(y[:train_len]) X_test, y_test = np.array(X[train_len:]), np.array(y[train_len:]) return X_train, y_train, X_test, y_testdef create_lstm_model(input_shape): model = Sequential() model.add(LSTM(units=50, return_sequences=True, input_shape=input_shape)) model.add(LSTM(units=50, return_sequences=True)) model.add(LSTM(units=50)) model.add(Dense(units=1))
model.compile(optimizer='adam', loss='mean_squared_error') return modeldef train_model(model, X_train, y_train, batch_size, epochs): history = model.fit(X_train, y_train, batch_size=batch_size, epochs=epochs, verbose=1, validation_split=0.1) return model, history
def evaluate_model(model, X_test, y_test, scaler): y_pred = model.predict(X_test) y_pred_inverse = scaler.inverse_transform(np.column_stack((np.zeros(y_pred.shape), y_pred)))[:, 1] # Inverse transform for Close prices y_test_inverse = scaler.inverse_transform(np.column_stack((np.zeros(y_test.shape), y_test)))[:, 1]
mae = np.mean(np.abs(y_pred_inverse - y_test_inverse)) mse = np.mean(np.square(y_pred_inverse - y_test_inverse)) return mae, msedef plot_prediction(model, X_test, y_test, scaler): y_pred = model.predict(X_test) y_pred_inverse = scaler.inverse_transform(np.column_stack((np.zeros(y_pred.shape), y_pred)))[:, 1] # Inverse transform for Close prices y_test_inverse = scaler.inverse_transform(np.column_stack((np.zeros(y_test.shape), y_test)))[:, 1]
plt.figure(figsize=(14, 6)) plt.plot(y_test_inverse, label='Actual') plt.plot(y_pred_inverse, label='Predicted') plt.xlabel('Days') plt.ylabel('Close Price') plt.title('Stock Price Prediction (Actual vs Predicted)') plt.legend() plt.show()Define the parameters
ticker = '^NSEI' # Ticker symbol for Nifty start_date = '2010-01-01' end_date = '2023-06-01' window_size = 40 # Number of previous days' data to consider period = 14 # ATR period multiplier = 3 # ATR multiplier batch_size = 32 epochs = 100
X_train, y_train, X_test, y_test = load_preprocess_data(ticker, start_date, end_date, window_size, period, multiplier)
Create the LSTM model
input_shape = (X_train.shape[1], X_train.shape[2]) model = create_lstm_model(input_shape)
Train the model
model, history = train_model(model, X_train, y_train, batch_size, epochs)
scaler = MinMaxScaler(feature_range=(0, 1)) scaler.fit_transform(y_train.reshape(-1, 1)) mae, mse = evaluate_model(model, X_test, y_test, scaler) print(f"Mean Absolute Error: {mae}") print(f"Mean Squared Error: {mse}")
Plot the predictions
plot_prediction(model, X_test, y_test, scaler)
- Assistant
To modify the code for predicting the next 5 days, you can create a
predict_next_daysfunction that takesmodel,X_test,scaler, andnext_daysas inputs. You can then loop for the specified number of days and use themodel.predictfunction at each step. Here's the modified code:def predict_next_days(model, X_test, scaler, next_days): predictions = [] for i in range(next_days): next_day_input = np.array([X_test[-1, 1:, :]]) next_day_pred = model.predict(next_day_input) predictions.append(next_day_pred[0, 0]) next_day_extended = np.zeros((1, next_day_input.shape[2])) next_day_extended[:, :4] = np.column_stack((next_day_input[0, -1, :3], next_day_pred)) next_day_extended[:, 4] = next_day_input[0, -1, 4] new_row = np.concatenate((next_day_input[0, 1:, :], next_day_extended), axis=0) X_test = np.append(X_test, [new_row], axis=0) predictions = np.array(predictions).reshape(-1, 1) predictions_inverse = scaler.inverse_transform(np.column_stack((np.zeros(predictions.shape), predictions)))[:, 1] return predictions_inverse # Define the number of days to predict and call the function next_days = 5 predictions = predict_next_days(model, X_test, scaler, next_days) # Plot the predictions plt.figure(figsize=(14, 6)) plt.plot(predictions, marker="o", label='Predicted') plt.xlabel('Days') plt.ylabel('Close Price') plt.title(f'{ticker} Stock Price Prediction for Upcoming {next_days} Days') plt.legend() plt.show()This code defines the
predict_next_daysfunction and uses it to predict the stock price for the next 5 days. Keep in mind that this prediction is based on the model's current understanding, so you should consider extending the LSTM model and re-evaluating the prediction accuracies according to your specific needs. - User
ValueError Traceback (most recent call last) in <cell line: 3>() 1 # Define the number of days to predict and call the function 2 next_days = 5 ----> 3 predictions = predict_next_days(model, X_test, scaler, next_days) 4 5 # Plot the predictions
3 frames /usr/local/lib/python3.10/dist-packages/numpy/core/overrides.py in concatenate(*args, **kwargs)
ValueError: all the input array dimensions for the concatenation axis must match exactly, but along dimension 0, the array at index 0 has size 3 and the array at index 1 has size 1
- Assistant
Apologies for the mistake. I forgot to update the shape of
next_day_extended. Follow these steps to fix the issue:- Change the shape of
next_day_extendedto be the same asnext_day_inputby updating the dimensions:
next_day_extended = np.zeros((1, next_day_input.shape[2]))- Update the assignments for the
next_day_extendedaccordingly:
next_day_extended[:, :3] = next_day_input[0, -1, :3] next_day_extended[:, 3] = next_day_pred next_day_extended[:, 4] = next_day_input[0, -1, 4]Here's the corrected code for the
predict_next_daysfunction:def predict_next_days(model, X_test, scaler, next_days): predictions = [] for i in range(next_days): next_day_input = np.array([X_test[-1, 1:, :]]) next_day_pred = model.predict(next_day_input) predictions.append(next_day_pred[0, 0]) next_day_extended = np.zeros((1, next_day_input.shape[2])) next_day_extended[:, :3] = next_day_input[0, -1, :3] next_day_extended[:, 3] = next_day_pred next_day_extended[:, 4] = next_day_input[0, -1, 4] new_row = np.concatenate((next_day_input[0, 1:, :], next_day_extended), axis=0) X_test = np.append(X_test, [new_row], axis=0) predictions = np.array(predictions).reshape(-1, 1) predictions_inverse = scaler.inverse_transform(np.column_stack((np.zeros(predictions.shape), predictions)))[:, 1] return predictions_inverseThis modified function should resolve the
ValueError. - Change the shape of