Market Intelligence for Agricultural Commodities Using Forecasting and Deep Learning Techniques

Abstract

About two third of the Indian population continues to live in villages and depends on agriculture as main source of livelihood. The agrarian distress that took decades to build up, surfaced in the form of decline in the proportion of rural population and increase in the number of farmer’s suicide cases. There is a strong hypothesis that in addition to production techniques, the widespread availability of Market Intelligence would bring substantial improvement in financial condition of farmers. Hence there is a need for comprehensive system, which would fetch, interlink, transform and analyze relevant data from various ministries/departments/organizations spread across the country to generate precise, appropriate and timely Market Intelligence. We took a step in this direction by design and implementation of Market Intelligence System Proof of Concept (PoC) using available datasets for few agricultural commodities. This PoC takes daily market price and weather data as input, transforms it into information and generates actionable intelligence by applying forecasting and deep learning techniques. The system provides trend analysis, short term as well as long term commodity price prediction and market selection as insights for farmers. The Auto Regressive Integrated Moving Average (ARIMA) forecasting technique and Recurrent Neural Network (RNN) deep learning techniques are applied for short term and long term agricultural commodity price prediction respectively. The study results demonstrate intended utility of forecasting and deep learning techniques for generating Market Intelligence System. The paper concludes with benefits of comprehensive Market Intelligence system, challenges and future work.