DEVELOPMENT OF SMART FARMING SYSTEM BASED ON INTERNET OF THINGS AND MACHINE LEARNING TO IMPROVE PRODUCTIVITY OF CILEMBU SWEET POTATO (IPOMOEA BATATAS)

• 主要作者: Agustirandi, Beny
• 格式: Theses
• 語言: Indonesia
• 在線閱讀: https://digilib.itb.ac.id/gdl/view/82102
• 機構: Institut Teknologi Bandung

Smart farming is a system developed by integrating information technology, digital innovation, and data analysis to enhance efficiency, productivity, and sustainability in food production. The technologies employed in smart farming encompass the utilization of sensors, software, artificial intelligence, and the strategic application of Internet of Things (IoT) tools to optimize various aspects of agriculture. Its importance extends beyond food production, influencing economic, social, and environmental aspects. One of the primary challenges in agriculture is prolonged dry seasons, which can lead to increased evaporation, soil desiccation, and water scarcity for crops. To mitigate these issues, precise irrigation techniques are necessary to meet crop water requirements during dry periods. Conventional irrigation methods often result in either over-irrigation or insufficient watering, leading to water and energy waste and reduced crop yields. This study aims to enhance water use efficiency and sustainability while improving crop productivity. The research employs IoT-based irrigation management technology for controlling and monitoring water needs. The irrigation technique used in this study is drip irrigation. The research strategy involves varying the amount of water applied to several agricultural plots and subsequently observing the effects on crop yields resulting from different water levels. The findings indicate that the use of drip irrigation can save up to 88% of water and increase the productivity of Cilembu sweet potatoes by 13.54 tons per hectare, equivalent to a 45.14% increase. Additionally, this study implements a machine learning model to predict crop yields and determine the optimal harvest time. The best-performing machine learning model identified in this research is the Random Forest model, with an R2 of 0.976 and a MAE of 1.037.