TODIM with XGBOOST and MVO metaheuristic approach for portfolio optimization

This paper presents an innovative and comprehensive approach to portfolio optimization that integrates TODIM, a multi criteria decision making method, with the proposed forecasting model for asset selection and return prediction. The optimization process utilizes the multi-verse optimizer, a powerful metaheuristic algorithm. TODIM is initially employed to select the top 10 companies from the Nifty 100 index based on various financial indicators. Daily return predictions for the selected portfolio are enhanced through an ensemble forecasting approach, BN-XGBoost, which combines BiLSTM and N-BEATS models, with XGBoost serving as a meta-learner to improve accuracy. The portfolio optimization problem, framed as a higher-order asymmetric risk model, incorporates semi-variance, higher order moments of return, and entropy. The proposed optimization framework is compared against a variance-based model, demonstrating superior portfolio performance. This research provides valuable insights for investors seeking advanced strategies to navigate complex financial markets and achieve optimal portfolio outcomes.