Forecasting Financial Time Series with Indicators and ENCOG Neural Networks

Summary

The article presents a workflow for connecting MetaTrader 5 to the ENCOG machine-learning framework through a C# bridge. Its example uses a feedforward neural network trained with resilient propagation to forecast financial time series. Stochastic %K, Stochastic %D, and Williams %R values serve as inputs. The author explains how to arrange recent indicator readings into moving time windows and normalize inputs before training, then describes producing a directional output for an Expert Advisor.

The evidence is an example backtest on USDCHF daily data over a stated historical interval, with about half the data held out from training. The article cautions that results may differ across securities and timeframes and presents the EA as educational material. It does not establish robustness across markets, and the described historical test is not sufficient by itself to demonstrate live profitability. The author suggests retraining periodically as a possible avenue for further research.

Key ideas

• The example predicts price direction from Stochastic and Williams %R indicator readings.
• Recent indicator values are grouped into moving time windows before being passed to the network.
• Input normalization is used to fit data to the network's activation functions.
• The example uses a feedforward network trained with resilient propagation through an ENCOG bridge.
• Its USDCHF backtest is presented as educational evidence, with results acknowledged to vary by market and timeframe.

Tags

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