IoT Based Climate Prediction System Using Long Short-Term Memory (LSTM) Algorithm as Part of Smart Farming 4.0
Devina Shafa Anindya, Mike Yuliana, Moch. Zen Samsono Hadi · 2022 International Electronics Symposium (IES) · 2022
Climate is an important element in human life. Climate change can cause many impacts in various fields such as industry, agriculture, livestock, and others. Agriculture is one sector that takes advantage of climatic conditions to maintain the quality of crop production. In this paper, a climate prediction system is proposed to support smart farming 4.0 to produce good quality and quantity of crop production and reduce losses. This system provides climate prediction information using Long Short-Term Memory (LSTM) method. The weather elements used in this research are temperature, humidity, rainfall, duration of sunlight, and wind speed with sensors data retrieval and processing supported by the Internet of Things. Predictions of the amount of monthly rainfall are processed to produce climate types and crop planning in a certain area. The results of the prediction model training test using the LSTM algorithm with filtering obtained the best model with a monthly data resample scenario with a value of n or time steps is 2 months, distribution of 80% train data and 20% test data, 48 batch size usage, amount of LSTM units 120 with 256 hidden layer neurons. Produce an RMSE value of 30.54, R2score of 0.74, a loss of 0.0247, a validation loss of 0.0282, a prediction computation time of 1.746 seconds on the train data, and 0.068 seconds on the test data.