Green Computing of Machine Learning In Business Intelligence

Abstract

The purpose of this research study was threefold: firstly, to address real-world problems of electronic businesses by utilizing business intelligence, machine learning and green computing techniques to minimize carbon footprint emissions from computational resources. Secondly, we conducted an in-depth study of Green Computing (GC), Business Intelligence (BI), and Machine Learning (ML) technologies to provide an overview of their state-of-art, scopes, and associated challenges, highlighting directions for further research. Green computing is a new revolution to minimize carbon footprint from computational resources used for computations while addressing real-world electronic businesses problems through algorithmic efficiency and machine learning models. Business Intelligence is critical in enhancing decision-making processes, operational efficiency, and positive outcomes such as improved customer service, stronger customer relationships, increased profitability, and lower failure rates. Machine learning is the subset of artificial intelligence (AI) that trains computers to learn from experience by the use of data without explicitly programmed to make desirable decision. Adaptive Business Intelligence (ABI) model is a business model that integrates business intelligence, machine learning, and green computing to enhance the adaptability of decision-making processes into a cohesive system in dynamic business environments. ABI model can be considered a type of sustainable Artificial Intelligence (AI) model. The objective of ABI model was to solve the problems of businesses by utilizing business intelligence, machine learning and green computing to reduce energy consumption and carbon footprint emission from computational resources. Thirdly, we have developed two ABI models— Adaptive Multiple Linear Regression (AMLR) and Adaptive Decision Tree Regression (ADTR). The developed Adaptive Business Intelligence model can be deploying in businesses to take right decision-making at an optimal level. For a specific dataset, the ADTR model outperformed the AMLR model in terms of training time, Central Processing Unit (CPU) computational efficiency, and carbon footprint, making it more suitable for lightweight, energy-efficient modelling. We also observed that certain features are outperformed by AMLR over ADTR in terms of MSE and R².