Eco-sustainability analytics: Applying stochastic modeling in environmental data for resource optimization

Abstract

In sustainable development modeling, the systemic interconnection between the economy, ecology, and social processes is considered. Simple autoregressive AR models, which are widely used in the practice of modeling and forecasting various indicators of sustainable development, have a significant drawback: the assumption that the modeled process is subject only to random influences and is not affected by other factors. More advanced autoregressive distributed lag (ADL) models take into account not only random influences but also other factors when modeling and forecasting complex dynamic processes. However, due to their complexity, they are less common in practice than autoregressive models. Vector autoregressions (VAR) build on the ideas of ADL models for the case of modeling a vector of interconnected indicators. Yet, they are even less frequently used in practice, both because VAR models are more complex and because, under certain conditions, they become high-dimensional models that even many highly qualified scientists are unable to construct. This report presents a simple approach to reducing the dimensionality of the VAR modeling task using complex variables.