Evolution of Portfolio Selection Literature: A Bibliometric Study with Emphasis on AI, ML, and DEA Models

This study examines how portfolio selection has evolved toward intelligent, real-time decision-making by combining Artificial Intelligence (AI), Machine Learning (ML), and nonparametric models such as Data Envelopment Analysis (DEA). Although traditional portfolio theory and standalone AI techniques are well researched, few studies provide an integrated view linking these approaches or identifying gaps for next-generation portfolio optimization. To address this limitation, the study performs a large-scale bibliometric analysis of 749 peer-reviewed publications from 1991 to 2026 using the Bibliometrix package in R. The results show a sharp growth in research activity after 2012, peaking in 2024. Machine learning has become a dominant theme, appearing in 1173 of 1281 AI-related portfolio studies. Clustering analyses further highlight the strong relationships between ML methods such as LSTM, SVM, and genetic algorithms and financial tasks including prediction, risk evaluation, and asset allocation. The study’s novelty lies in synthesizing three decades of research to demonstrate how integrating ML with inverse DEA models can support real-time, data-driven, and adaptable portfolio decisions. Overall, it provides a roadmap for developing hybrid, interpretable, and scalable approaches that can enhance both academic research and practical portfolio management.