Prototype of AI-Integrated Chatbot for Shallot Price Forecasting and Advisory Support to Assist Farmer Decision Making

Abstract

Forecasting agricultural commodity prices is a fundamental tool for sustainable development in the agricultural economy and broader economic stability. With rapid and simple access to information on future prices, farmers can plan their planting schedules to optimize profits. This study presents a prototype AI chatbot that integrates price forecasting and advisory functions to assist farmers in decision-making and interact as an extension agent. Price forecasting employed Random Forest regression, achieving MAPE of 8.34% (training), 13.98% (validation), and 15.62% (testing). The chatbot was developed to access price forecasting information for the next four months. This system also integrates an LLM-AI model for consultations on planting schedules and other topics using a trusted knowledge base. During the testing phase, the chatbot successfully made predictions, provided recommendations, and interacted as an extension agent. Although demonstrating promising results, this study is limited to shallot price forecasting in Yogyakarta, highlighting the need for broader commodity and regional coverage in future studies. Unlike previous studies that focused only on forecasting or advisory, this study integrates predictive analytics with conversational AI in a farmer-friendly chatbot.