Long Short-Term Memory (LSTM) is a form of neural network model, especially in the recurrent neural network (RNN). This paper uses Internet of Things (IoT) data aggregated from sensors measuring environmental parameters in the industrial zone into PostgreSQL database. The database includes temperature, humidity, particulate matter (PM) dust, soil temperature and moisture. The requirement is to predict time series with RNN regression neural network using the LSTM model in Python with Keras. The objectives of this work are to mine and analyze data retrieved from the server database and building models to make predictions of environmental parameters. Then, the model training was scheduled to automatically update every 24 hours by getting data automatically from the server’s database and applying new data to the same old data and retrain the model with updated data in 24-hour intervals. The performance results of training data set on the LSTM model were compared with that on the AutoRegressive Integrate Moving Average (ARIMA) model in terms of environmental industrial parameters such as temperature, humidity, PM2.5, PM10 parameters.