Implementation of AI in share investment decisions: Proposition of a modern deep learning algorithm

Financial market data behavior is complex due to its dependence on many factors and different factors. In particular, the development of effective models for stock income forecasting is a challenge. Although the long short-term memory (LSTM) approach is not widely used for predictive financial analysis, it is a viable method for this field. In addition, the hybridization of Discrete Wavelet Transform (DWT), Particle Swarm Optimization (PSO) and LSTM networks algorithms is an innovative technique for sequential learning proposed in this study. The model outperforms other well-known supervised and unsupervised classification techniques with the PSO powered LSTM networks algorithm based on performance analyses. In this study, it is tried to provide portfolio basket analyzes and forecasts with a new artificial intelligence approach by using the stocks and newspaper news data between the years 2000-2021 in the BIST30 Index.