Prediction Turkish Airlines BIST Stock Price Through Deep Artificial Neural Network Considering Transaction Volume and Seasonal Values

OpenAlex English

The collection of data in the information age has led to its analysis and use in different fields. Data can be used for different purposes, such as historical information, reporting, analysis, artificial intelligence, and machine learning. Artificial intelligence is used for different purposes in different disciplines such as engineering, health, industry, production, transportation, the stock market, education, and the social sciences. In this study, Turkish Airlines’ stock price prediction was made using machine learning. Different artificial neural network methods were used, such as an FNN, LSTM, and GRU. The data set consists of daily stock market index information for Turkish Airlines in BIST between the dates of January 4, 2010, and January 31, 2022. During the training of the system, it was assessed together with the transaction volume data to reduce the effect of possible speculative behavior. Since the income of airlines carrying passengers is seasonally affected, seasonal data are also considered. The system has been trained and tested with different short-long term memory-based artificial neural network models. The performance indicators of the models were used as R-Square MSE, RMSE, and MAE. According to the R-Square, performance score of the test, the success rate of system was 97% in FNN, and 99% in LSTM and GRU. It performed well despite extreme price fluctuations due to the pandemic and economic crisis. According to these results, machine learning can be used as a decision support system for sequential data set prediction. In this study, it can be concluded that FNN, LSTM, and its derivative machine learning methods can be successfully used in air transport sector index prediction.