TPTC: topic-wise problems trend clusters for smart agricultural insights extraction and forecasting of farmer’s information demand

Abstract

To meet the challenges of increasing food production demand globally, extracting insights regarding the persistent agriculture-related problems on a nationwide scale is the need of the hour. Policymakers now have limited possibilities for acquiring a comprehensive knowledge of the difficulties that farmers face on a national level. In this direction, the presented work proposes a new artificial intelligence-based pipeline to gain insights at country level regarding the farmers demand for assistance in India. The presented study uses the data from the Kisan Call Centres, a nationwide network of farmers helplines, including 28.6 million call-log records, made available by the Ministry of Agriculture and Farmers Welfare Government of India. Additionally, the extracted insights are presented in the form of Topic-wise Problems Trend Clusters (TPTC), which can be used by policymakers in both the government and private sectors to aid decision-making. The article also introduces a pipeline for designing forecasting models to estimate the monthly frequency of farmer inquiries (in terms of the number of query calls). The seven statistical forecasting models were examined in the study with the TBATP1 (Trigonometric seasonal components with Box-Cox transformation incorporating ARIMA errors and Trend including the Seasonal components) model attaining the lowest error rates in terms of Root Mean Square Error (0.034) and Mean Absolute Error (0.107). The study also explores numerous applications of the derived insights in the real world as well as the future scope of the presented work