Future Smart-Home device usage prediction is a very important module in artificial intelligence. The technique involves analyzing the user performed actions history and apply mathematical methods to predict the most feasible next user action. Unfortunately most of the techniques tend to ignore the adaptation to the user preferred actions and the relation between the actions and the state of the environment which is not practical for Smart-Home systems. This paper present a new algorithm of user action prediction based on pattern matching and techniques of reinforcement learning. The algorithm is modeled using hardware description language VHDL. Synthetic data had been used to test the algorithm and the result shows that the algorithm performs realistically better than the current available techniques