In this paperwe present several considerations centered around the data-driven system approaches. We briefly explore three main issues: the evolving relationship between off -line and on-line data processing methods, the complementary relationship between the data-driven and model-based methods, and the perspectives of data-driven system approaches. Instead of offering solutions to data-driven system problems, which is impossible at the present levelof knowledge and research, in this article we aim at categorizing and classifying open problems, exploring possible directions that may offer alternatives or potentials for the four key fields of interests:control,decision making, scheduling,and fault diagnosis。