Lasso Regression

Explanation & Rationale

The Lasso Regression Strategy applies L1 regularization to train a regression model on the past 14 days of price data, predicting the next day’s movement. By penalizing less significant features, Lasso helps prevent overfitting and ensures the model focuses on the most relevant price trends. This approach is useful for filtering noise in financial data while still capturing meaningful directional changes for trading decisions.

Code

'''
Lasso Regression Strategy.

Trains a Lasso Regression model using the past 14 days of data to predict the next day's price movement (up/down).
Learn more @ docs.ubacktest.com/examples/regression/lassoreg
'''

import pandas as pd
from sklearn.linear_model import Lasso
import numpy as np

def lasso_regression(data, window=14, alpha=1.0):

    signals = np.zeros(len(data))  # Initialize signals array
    predictions = np.zeros(len(data)) # Initialize predictions array

    # Iterate over the data starting from the window index
    for i in range(window, len(data)):
        # Prepare the features (X) and target (y) for the regression model
        X = np.array(range(i-window, i)).reshape(-1, 1)  # Time index for the last window days
        y = data['close'][i-window:i].values  # Last window closing prices

        # Fit the Lasso regression model (using L1 regularization)
        model = Lasso(alpha=alpha)  # 'alpha' is the regularization strength
        model.fit(X, y)

        # Predict the next value (for the current time period)
        prediction = model.predict(np.array([[i]]))  # Predict the next point (i.e., the 6th day)
        predictions[i] = prediction[0]

        # Signal generation based on prediction (uptrend or downtrend)
        if prediction > data['close'][i-1]:
            signals[i] = 1  # Buy signal
        else:
            signals[i] = -1  # Sell signal

    return signals, predictions

def strategy(data):

    # Call the lasso_regression function to get the signals
    data['signal'], data['prediction'] = lasso_regression(data)

    return data