Reproducing a Random Forest Model for Stock Direction Prediction

Summary

This article outlines a reproduction of a research approach that predicts whether a stock’s closing price will rise or fall over a future horizon. It applies exponential smoothing to Apple daily price data, then builds a binary target from the price change over the forecast window. Six technical indicators, including RSI, stochastic oscillator, Williams %R, MACD, price rate of change, and On-Balance Volume, serve as model features. A random forest classifier is trained on an earlier portion of the data and evaluated on the later portion.

The report describes precision, recall, F1, and accuracy as evaluation measures and states that accuracy is around 70%, with stronger precision for rising cases. These figures are reported without detailed metric tables, uncertainty estimates, trading returns, transaction costs, or comparison against a baseline. The example covers one stock and a limited historical period, and the article warns that its workflow uses an older platform version, so neither its results nor implementation should be assumed to transfer directly to current systems or other markets.

Key ideas

• The model labels observations by the sign of a future closing-price change.
• Exponential smoothing and six technical indicators are used to prepare the model inputs.
• A chronological split assigns earlier observations to training and later observations to testing.
• The article reports classification accuracy but does not establish profitability after trading costs.
• The example is limited to one stock, one historical period, and an obsolete platform workflow.

Tags

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