Risk Quantification in Manufacturing Industry Investments: A Stochastic Approach with Artificial Intelligence

Abstract

Manufacturing investment decisions are often hindered by significant uncertainty. This paper introduces a conceptual model that integrates stochastic simulation with machine learning to quantify investment risk. Our hybrid approach employs a Monte Carlo simulation using time series data of fixed, variable, and investment costs as inputs. To enhance the simulation's realism, Long Short-Term Memory (LSTM) recurrent neural networks forecast the trend components of these series, while a Vector Autoregression (VAR) model captures their inter-correlations. This framework generates a multitude of potential scenarios, each evaluated through a mathematical model of the supply chain to produce a distinct cash flow. The subsequent application of investment metrics, such as Net Present Value (NPV) and Discounted Payback, to the distribution of these cash flows enables a comprehensive statistical analysis of the investment's risk profile, thereby providing robust support for strategic decision-making.