Forecasting Models of Potato Supply and Demand in A Close-Loop System for Nuwara Eliya and Badulla Districts, Sri Lanka: A Case Study 

The stakeholders in the agricultural domain require access to agricultural information and  relevant knowledge to make informed decisions throughout the crop lifecycle, particularly in potato cultivation. Coordination failures among stakeholders in the potato farming sector often led to inaccurate, untimely, and inadequate information dissemination. Consequently, the potato sector grapples with challenges such as oversupply, undersupply, and crop wastage. Although the existing digital agriculture ecosystem can gather data from various  stakeholders more efficiently, there remains a crucial need to analyze and forecast this data to provide the necessary insights for informed decision-making. This study proposes a  "Closed-loop Information Flow Model" aimed at providing forecasted demand and supply,  fostering improved coordination among stakeholders. Various prototypes were developed to deliver forecasted demand and supply information to key stakeholders in the potato sector, initially targeting farmers, input suppliers, and logistics suppliers, with plans to expand to other stakeholders in the future. The prediction model employs the Univariate Multistep Forecasting method, leveraging LSTM, CNN, and MLP Neural Networks. Evaluation of the prediction model against stakeholder requirements involved analyzing actual and predicted data, comparing metrics such as RMSE and MAEP to select the most suitable neural network for predictions at each level. Additionally, refinements were made to the existing "Govi - Nena" mobile-based farmer application to provide supply and demand information specifically tailored to farmers' needs. Separate web dashboards were developed to deliver pertinent information to other stakeholders, including input suppliers and logistics suppliers. Prototypes underwent thorough evaluation by stakeholders, leading to fine-tuning of the user interface to achieve an overall satisfaction mean value of approximately 4.00. The proposed Closed-loop Information Flow Model for forecasted demand and supply aims to bridge the market information gap among stakeholders, ultimately reducing coordination failures in the Sri Lankan agriculture sector and enhancing transparency in decision-making processes. 

Keywords: Demand and Supply Modeling, Closed-loop Market Information Flow Model, UI/UX Designing, Univariate Multistep Forecasting.