Application of artificial intelligence in financial markets

This thesis examines the potential of applying artificial intelligence systems methods to data in financial markets, and an LSTM-based forecasting model was created using daily maintenance data for the Borsa Istanbul 30 (BIST 30) index between January 5, 1997, and September 6, 2025. The study's methodology aims to utilize the data offered by deep learning techniques in market forecasting, where traditional methods are inadequate. Analyzes were conducted in the Google Colab environment using the Python programming language; logarithmic return, moving averages (MA5, MA20), RSI14, and volatility (Vol20) were included as input variables in the tracking model. The dataset was divided into three subsets: 70% training, 15% regularization, and 15% test. The input variables were scaled with StandardScaler, and the target variable was scaled with MinMaxScaler, which is tailored only to the training set. The model, with the LSTM(128) → Dropout(0.30) → LSTM(64) → Dense(32, ReLU) → Dense(1) architecture, was equipped with Adam aging light and MSE (Mean Square Error) loss functions. EarlyStopping and ReduceLROnPlateau methods were used to prevent overfitting. The model's performance was evaluated using RMSE, MAE, and directional accuracy metrics and compared with the persistence approach. The results demonstrate that the LSTM-based model achieves low error rates in logarithmic return forecasts and a high level of success in directional forecasts. The findings demonstrate that the LSTM architecture is an effective method for forecasting financial time series; deep learning techniques are supported by a powerful tool for more accurate and flexible forecasts in financial markets. This allows the model to be developed with more comprehensive datasets that include different market regimes, macroeconomic growth, and news sources in the future. Keywords: Artificial Intelligence, Deep Learning, LSTM, Financial Markets, Borsa Istanbul, Time Series Forecasting, Machine Learning, Forecast Models.