An improved fuzzy rule-based system using evidential reasoning and subtractive clustering for environmental investment prediction

Abstract

Environmental investment prediction has attracted much attention in the last few years. However, there are still great challenges in investment prediction modeling, e.g., 1) effective environmental indicators must be accurately selected to avoid the curse of dimensionality; 2) effective environmental data must be reasonably selected to downsize the scale of historical data; 3) the higher interpretability and lower complexity of prediction models must be considered. To address the above three challenges, a new environmental investment prediction model using fuzzy rule-based system (FRBS), evidential reasoning (ER) approach, and subtractive clustering (SC) algorithm is proposed in the present work, called FRBS-ERSC. In this new model, the FRBS is the core component for the modeling of environmental investment prediction and therefore provides good interpretability and complexity to environmental managers. Meanwhile, the ER approach is used as an improvement technique of the FRBS to combine the strengths of different feature selection methods for better indicator selection, and the SC algorithm is used as another improvement technique of the FRBS to select effective environmental data. An empirical case of environmental investment prediction is studied based on data on 31 provinces in China ranged from 2005 to 2018. The experimental results show that the proposed FRBS-ERSC not only provides interpretable and scalable environmental investment prediction based on effective indicator selection and data selection, but also produces satisfactory accuracy compared to some existing models.