Long-Range Wide Area Network (LoRaWAN) has become a promising communication method for the Internet of Things (IoT) system since it is capable of long-range communication with low power usage. Presently, LoRa gateway (GW) deployment in Indonesia relies on conventional methods, i.e., predicting coverage without incorporating formulas specific to LoRa GW performance or accounting for the area type. This research aims to deploy the LoRa GW using a machine-learning algorithm to cover every sensor demand. In this paper, we simulate, deploy, and measure the signal strength transmission between recommended LoRa GWs and end device (ED) demands to ensure the algorithm works and is usable on each different environment characteristic. The results of these experiments show that the algorithm can generate coordinate point recommendations for the LoRa GW to cover all ED demands in three different characteristic areas.