Integrating decision tree with back propagation network to conduct business diagnosis and performance simulation for solar companies

Abstract

Solar energy is a natural, clean, and inexhaustible resource to help the earth solve its energy crisis and reduce the air pollution caused by coal, nuclear, and gas electricity plants. Solar heating, solar photovoltaics, and solar thermal electricity can contribute to solving some of the most urgent problems the world faces. Despite the anti-dumping tariff and the anti-subsidy policy negatively impacting solar companies in Taiwan and China, solar cells are facing strong recovery of market growth in 2015. However, fierce competition in terms of cost and quality has resulted in a series of corporate acquisitions or bankruptcies since 2012. Inspired by the concept of business intelligence, a balanced scorecard (BSC) based framework is proposed to address the following issues: (1) How to identify key performance indicators (KPIs) that influence outcomes? (2) How to guide the lagging group to benchmark with the leading group (business diagnosis)? (3) How to adjust the significant KPIs to improve outcomes (what-if simulation)? In particular, classification and regression tree (fused with logistic regression) and back propagation neural network (fused with multiple regression) are adopted to provide managerial insights for worldwide solar companies. Furthermore, support vector machine (regression) is used to justify the validity of the presented framework.