Accurate channel estimation is an essential factor in determining the efficiency of a wireless communication system. Moreover, in Reconfigurable Intelligent Surfaces(RIS)-Assisted wireless networks using millimeter wave (mmWave), it is crucial to optimize each RIS element's phase shift. Therefore, in this paper, we propose a channel estimation method in TDD-based wireless communication system using auto encoder in RIS-Assisted wireless networks. The trade-off relationship between channel estimation accuracy and the number of pilot signals is optimized when performing channel estimation. Through denoising autoencoder and Average Percentage of Zeros(APoZ), we find the optimal pilot pattern considering not only the number of pilots but also the location. As a result of the experiment, the proposed method has little difference in performance or outperforms the full neural network without pruning.