BACKTESTING DIFFERENT TRADING STRATEGIES TO COMPARE THEIR PROFITABILITY BY USING A DEEP NEURAL NETWORK MODEL TO BUILD TRADING SIGNALS
BACKTESTING DIFFERENT TRADING STRATEGIES TO COMPARE THEIR PROFITABILITY BY USING A DEEP NEURAL NETWORK MODEL TO BUILD TRADING SIGNALS

Summary/Abstract: Our aim in this paper is to elaborate several trading strategies using a deep learning model, namely the long short-term memory (LSTM) network and then to backtest the strategies in order to compare their performance. More precisely, we consider a portfolio of several energy market assets, for which we collect data consisting of daily prices and convert them into daily returns. The return series are further used to train the LSTM model that allows us to make one-step-ahead predictions. Subsequently, those predictions are converted into trading signals, which are triggers to buy or sell assets in a portfolio, according to a set of performance criteria. Four different trading strategies are constructed to allocate capital based on the trading signals. Finally, the strategies are backtested using a backtesting procedure and the backtest results are compared each other for profitability.

• Details
• Contents