SiLab: A Neural Network-based Precision Agriculture System for Siling Labuyo (Capsicum frutescens)

The declining role of the agricultural sector in the country’s economy can be attributed to low productivity. This has significantly impacted the overall economy, as seen in the scarcity and higher prices of Siling Labuyo reported by the Department of Agriculture. To address this issue, a research project focused on developing an automated device that quantifies farmers’ knowledge to improve plant growth through irrigation. The device collects data on ambient temperature, humidity, soil temperature, soil nutrients (N-P-K), and soil moisture. The gathered dataset was used to develop an artificial neural network (ANN) and recurrent neural network-long short-term memory (RNN-LSTM) models, which are implemented in an experimental device for irrigation. The results show that ambient, humidity, and soil temperatures are the most important model development parameters. The best ANN architecture consists of four hidden layers with three neurons per hidden layer, while the best RNN-LSTM architecture consists of 2 hidden layers with three neurons per hidden layer. The RNN-LSTM model outperforms the ANN model in terms of accuracy measures such as root mean square error (RMSE), R-squared $(\mathrm{R}^{\wedge} 2)$, and mean absolute error (MAE). However, the dataset shows a declining trend in model performance due to extreme weather conditions during data collection. When comparing plant growth rates, Setup C (Trial 2) performs the best in terms of height growth rate, with no significant difference observed between Setup A and Setup C (Trial 2). There was a significant difference between Setup B and Setup C (Trial 2), favoring Setup C with a mean width growth rate of 6.5625 cm per two weeks.