Conventional prediction methods, whether derived from deep learning or statistical techniques, typically demand significant manual intervention, making full process automation challenging. This study investigates the comparative robustness of traditional Long Short-Term Memory (LSTM) models and automated machine learning (AutoML) techniques, specifically TPOT, in predicting stock market behavior. By conducting experiments on the Moroccan stock market, the results demonstrate that AutoML approaches not only enhance prediction accuracy but also simplify the modeling process, making advanced prediction techniques more accessible and effective in dynamic financial markets.