Extended belief rule-based model for environmental investment prediction with indicator ensemble selection

Abstract

Environmental investment prediction is an effective solution to reduce the wasteful investments of environmental management. Since environmental management involves diverse environmental indicators, investment prediction modeling usually causes the curse of dimensionality and uses irrelevant indicators. A common solution to solve these problems is the use of indicator selection methods to select representative indicators. However, different indicator selection methods have their relative strengths and weaknesses, resulting in different selected indicators and information loss of real representative indicators. Hence, in the present work, a new environmental investment prediction model is proposed on the basis of extended belief rule-based (EBRB) model along with the indicator ensemble selection (IES) and is called IES-EBRB model. The EBRB model is a white-box designed decision-making model and has the specialty on using prior knowledge to enhance data analytics for autonomous decision making; and the IES is an extension of ensemble learning to cooperatively integrate different kinds of indicator selection methods for selecting representative indicators. In a case study, the real world environment data from 2005 to 2018 of 31 provinces in China are applied to verify the effectiveness and accuracy of the IES-EBRB model. Results show that the IES-EBRB model not only can obtain desired environmental investments, but also produces satisfactory accuracy compared to some existing investment prediction models.