In this study, we develop an interactive life support system that can operate home appliances around the user for people with a disability such as ALS. Previous study showed expectation to operate home appliances by people with ALS using life support system that use natural language to operate. But there is a problem to need time and effort for set status of home appliances with an ideal. In this paper, we develop and evaluate the user behavior prediction model composed of neural network and latent dirichlet allocation to reduce effort of operate home appliances by people with ALS.