Stock market prediction and Portfolio selection models: Review

Rational investors have a primary objective of optimizing their stock portfolios. Achieving this goal involves the selection of stocks that are projected to perform well in the future and determining how to allocate capital among them to create the ideal investment portfolio. To address this optimization challenge, various static and constrained statistical approaches have been employed. However, the application of machine learning techniques in this field has gained increasing popularity due to technological advancements, enabling researchers to tackle complex and nonstationary scenarios. Recent studies have integrated these algorithms with updated asset selection models and hyperparameter tuning kernels to enhance the performance of portfolio optimization. This research provides a focused literature review on the integration of machine learning with stock portfolio optimization. The analysis revealed that the most commonly utilized machine learning method for predicting stock returns was the long short-term memory (LSTM) model. Additionally, the Markowitz mean-variance model emerged as the most popular approach for optimizing portfolio stock selection. The research also identifies potential avenues for future investigation and underscores the implications of the study's findings.