Using Symbolic Regression Equations to Interpret Price Forecasts

Summary

The document presents a way to make trading forecasts more interpretable by translating fitted machine-learning models into algebraic expressions. Its example uses technical features derived from price bars, including returns, volatility, RSI, MACD, and momentum across multiple lookback periods. Ridge polynomial regression is proposed for price prediction and logistic regression for direction; SymPy is used to express model coefficients and relationships as equations that can be inspected and simplified.

The central claim is that an explicit formula can make model behavior easier to examine and adjust than an opaque neural network. The document offers illustrative examples and reported performance anecdotes, but no reproducible evaluation establishing that symbolic equations improve out-of-sample trading results. It also acknowledges feature proliferation, noise, outliers, computational cost, and rising complexity that can weaken interpretability. The approach still requires careful data preparation, regularization, validation, and risk controls, and formulas may lose effectiveness as market conditions change.

Key ideas

• The proposed workflow fits regression models and uses symbolic algebra to expose their equations.
• Price and direction forecasts use different regression approaches in the example.
• Technical features span returns, volatility, RSI, MACD, and momentum at several lookback lengths.
• Explicit equations may aid inspection, but complex formulas can become difficult to interpret.
• Noise, outliers, market changes, and weak validation can undermine the forecasts.

Tags

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.