Selecting Trading Strategies with Neural Network Forward Testing

Summary

The paper proposes a strategy selection process called DNN-forwardtesting. Instead of choosing a trading strategy solely by its results on historical prices, it first forecasts a possible future price path with a deep feed-forward neural network and evaluates technical indicators on those predictions. The selected indicator then guides trades in the actual future market. The described workflow includes exploratory analysis of ten securities, a k-means-based volatility procedure, and training on a restricted set of assets with a shared volatility coefficient to forecast prices for the next 30 days.

The authors report that neural network forecasts outperform classical statistical techniques and that strategies chosen through the forecast-based process improve expectancy, Sharpe, Sortino, and Calmar ratios relative to traditional backtest selection. The supplied description does not provide the securities, forecast accuracy metrics, implementation details, or evaluation period. Results are therefore claims from the paper summary and do not establish that the method generalizes across assets or forecasting conditions.

Key ideas

• The method selects a trading strategy using technical indicators applied to neural network price forecasts.
• A k-means-based procedure groups securities by volatility before model training.
• The described feed-forward network forecasts stock prices 30 days ahead.
• The authors report better forecast performance than classical statistical techniques.
• They also report improved risk and return metrics versus strategy selection by traditional backtesting.

Tags

# DNN-ForwardTesting: A New Trading Strategy Validation using Statistical Timeseries Analysis and Deep Neural Networks


# DNN-ForwardTesting: A New Trading Strategy Validation using Statistical Timeseries Analysis and Deep Neural Networks









In general, traders test their trading strategies by applying them on the historical market data (backtesting), and then apply to the future trades the strategy that achieved the maximum profit on such past data. In this paper, we propose a new trading strategy, called DNN-forwardtesting, that determines the strategy to apply by testing it on the possible future predicted by a deep neural network that has been designed to perform stock price forecasts and trained with the market historical data. In order to generate such an historical dataset, we first perform an exploratory data analysis on a set of ten securities and, in particular, analize their volatility through a novel k-means-based procedure. Then, we restrict the dataset to a small number of assets with the same volatility coefficient and use such data to train a deep feed-forward neural network that forecasts the prices for the next 30 days of open stocks market. Finally, our trading system calculates the most effective technical indicator by applying it to the DNNs predictions and uses such indicator to guide its trades. The results confirm that neural networks outperform classical statistical techniques when performing such forecasts, and their predictions allow to select a trading strategy that, when applied to the real future, increases Expectancy, Sharpe, Sortino, and Calmar ratios with respect to the strategy selected through traditional backtesting.

Shown in full with attribution under the source's licence. Licence: abstract CC0

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