Horst, Enrique terTrujillo Reina, Santiago2025-03-202025-03-202024-09-302024-09-10http://hdl.handle.net/10362/180996This thesis proposes a hybrid financial forecasting model integrating the Ornstein-Uhlenbeck process with neural networks, particularly Long Short-Term Memory (LSTM) models. The model aims to enhance the accuracy of mean-reverting strategies by capturing both long memory properties and the mean-reverting behavior of stock prices. The use of fractional differencing as a preprocessing step improves data stationarity, further increasing the model's predictive performance. This study evaluates the hybrid model's profitability using the Excess Profitability (EP) test, while assessing the suitability of the Ornstein-Uhlenbeck process for stock price prediction. Results show improvements in forecasting accuracy and financial viability compared to traditional methods.engOrnstein-Uhlenbeck processNeural networksFinancial forecastingTime series analysisMachine fearning in financeStochastic processesFinancial time seriesQuantitative financePredictive modelingStationarity in time seriesmaster thesis203902890