Event prediction within directional change framework using a CNN-LSTM model

Abstract Financial forecasting has always been an intriguing research area in the field of finance. The widely accepted approach to forecast financial data is perform predictions using time series data. In analysis, sampling with a predefined frequency (e.g. hourly, daily) leads uneven and discontinued flow. Directional Change newly proposed that replaces physical within establishes event-driven framework. With emergence machine deep learning-based methods, researchers have utilised them series. These techniques shown outperform conventional approaches. This paper aims employ CNN-LSTM model investigate its predictive competence (DC) framework predict DC event prices. To obtain this objective, we first create tick bars/candles GBPUSD, EURUSD, USDCHF, USDCAD prices from January August 2019. Then, DC-based summaries selected bar/candle for each currency pair will be generated fed model. network architecture incorporates robustness Convolutional Neural Network (CNN) feature extraction Long Short-Term Memory (LSTM) predicting sequential results suggest performance improves significantly