Candlestick Patterns Recognition using CNN-LSTM Model to Predict Financial Trading Position in Stock Market

Investors need analytical tools to predict the price and determine trading positions. Candlestick pattern is one of that trends. However, patterns are difficult recognize, some studies show doubts regarding robustness recognizing system. In this study, we tested predictive ability candlestick We use Gramian Angular Field (GAF) encode as images recognize 3-hour 5-hour 6 with Convolutional Neural Network (CNN), coupled Long short-term memory (LSTM) model close price. The position consists buying selling a hold period several hours. Our results CNN successfully detected GAF an accuracy 90% 93%. LSTM can trend 155.458 RMSE scores 0.9754% MAPE 10-hour look back. With duration three hours CNN-LSTM additional model, test data's 85 recognized 82.7% accuracy, compared 60% profitable positions when recognition used alone. Compared employing identification alone, combination improve prediction power offer more lucrative