In this paper, we propose uplink period optimization algorithm for the future IoT(Internet of Things) network system. The proposed algorithm focuses on minimizing power consumption of device and mitigating the degradation of accuracy due to a variable uplink period. In particular, by using machine learning, the proposed algorithm selects the optimal period which is expected to have the biggest F(T), accuracy divided by power consumption. Simulation results show that the proposed algorithm outperforms the conventional algorithm.