Combining random forest with support vector regression to conduct diagnostic and predictive analytics for semiconductor companies

According to World Semiconductor Trade Statistics (WSTS), the market revenue earned by global semiconductor industry has researched up to US$335 billion in 2015. In response to strongly growing demand for automotive and mobile devices, global leading companies are aggressively expanding production capacity and increasing process investment to seize newly emerging niches. For instance, more creative and smart applications are created from the so-called M2M (machine-to-machine) and IoT (Internet of Things) areas. To survive in such an intensively competitive environment, semiconductor companies need to be more agile, responsive and flexible than before. In this study, a novel framework is presented to assist semiconductor companies in accomplishing the following goals: (1) applying random forest to identify the top five significant predictors and diagnostic analytics (2) conducting support vector repression to achieve predictive analytics and sensitivity analysis. In particular, semiconductor companies, such as upstream fabless design, midstream chip manufacturing, and downstream packaging & testing, are used to characterize supply-chain tiers and validate the presented framework.