Planting Sustainability Prediction Using Random Forest Algorithm

Abstract

This study aimed to utilize smart space technologies in determining planting sustainability and soil moisture stress that will optimize IoT and data mining technique to help in achieving the government’s goal of empowering and strengthening the nation’s agricultural sector. The study employed the combined experimental, developmental, and quantitative research approach towards achieving the objectives of the study. In order to gather abiotic and edaphic data, including soil moisture, light, temperature, humidity, pH, and NPK level, the study constructed an IoT prototype. The researcher employed LoRa technology in transmitting collected data into the IoT gateway before uploading it to Firebase Realtime Database. The study also involved the development of a mobile application using Blynk IoT Framework and web interface for remote monitoring and control of irrigation and UV lighting system. Comparative analysis was conducted between Random Forest, KNN, and Naïve Bayes Algorithm in predicting planting sustainability based on available data. Based on the calculated Kappa of 0.9901, Random Forest demonstrated the highest level of accuracy with a "Almost Perfect" strength of agreement. Random forest was implemented using the RubixML library to enable the web interface to perform predictions of Planting Sustainability on data stored in the database. In the user evaluation test based on ISO25010 conducted by the researcher, the overall weighted mean for all criteria is 4.19, with an "Agree (A)" interpretation. This indicates that the developed system is of excellent quality, excelling in functionality, dependability, usability, efficiency, maintainability, and portability.