The hot press forging is a process of heating raw metal material (ingot) to a high temperature and shaping it into expected shaped metal. The raw ingots are loaded and heated in the heating furnace, and then either pressing or cutting processes are properly being carried out to the hot ingot to achieve the expected shape model. The heating process in heating furnaces consumes high energy due to need to heating ingots in high temperature. By optimizing the combination of ingots to be charged into the heating furnaces and press machines working's strategy, the cost of energy can be minimized. Therefore, in this paper, we propose a method to optimize heating furnace's work plan and press machine's work plan using cost prediction models and genetic algorithm. Cost prediction models are learned by applying machine learning algorithm on actual process data collected via IoT infrastructures from our test bed factory. The work plan then is generated through evaluation and optimization of a model-based simulator based on the genetic algorithm. The obtained work plan can also adaptively adjust to unexpected situations by real time rescheduling jobs.