PEMBELAJARAN MESIN UNTUK OPTIMASI ALOKASI MODAL DAN PROYEK DI INDUSTRI

Abstract

This study investigates the application of machine learning for optimizing capital allocation and project management in the industrial sector under financial market complexity and macroeconomic uncertainty. The research aims to explore machine learning algorithms, analyze their effectiveness in improving investment decision efficiency and risk mitigation, integrate real-time risk-based allocation approaches, and compare the proposed methods with traditional frameworks such as the mean-variance model (MVO). The methodology combines literature review, case studies, and numerical simulations using historical data from 2017 to 2022. The framework consists of volatility forecasting using LSTM, differentiable risk budgeting for adaptive target-risk adjustment, and deep reinforcement learning (DDPG-TiDE) to optimize asset allocation policies within a Markov Decision Process (MDP). Model performance is evaluated using Sharpe ratio, maximum drawdown, and portfolio turnover efficiency, while interpretability is validated using SHAP. Simulation results show a 23–55% improvement in Sharpe ratio compared to traditional risk parity strategies and a 41% reduction in maximum drawdown during volatile market periods. The study also demonstrates that SHAP enhances transparency by identifying key drivers such as market volatility, credit spread, and the yield curve. The findings conclude that machine learning can be a game changer for improving efficiency, real-time risk mitigation, and adaptive decision-making, while highlighting challenges related to data quality, model complexity, AI governance, and integration with legacy systems.