A Novel Framework Using Deep Learning Techniques for Ragi Price Prediction in Karnataka

Abstract

Indian agriculture is diverse, employing a significant portion of the country’s population. In southern states of India, Rice and Ragi are the main crops. Ragi cultivation provides economic benefits to farmers in India. Many startup companies are coming up with Ragi based Noodles, Pasta etc. But the price of the Ragi is fluctuating due to various factors such as weather, yield, economy etc. The literature had focused on the price prediction of other crops like corn, Rice, fruits and vegetables etc. This motivated us to work on Ragi as it is a promising crop. Hence, the time series based deep learning models plays a crucial role in predicting the price. In this research paper, various algorithms such as Long Short-Term Memory (LSTM), Gated Recurrent unit (GRU), Vector Auto Regression (VAR) and Convolution Neural Network (1DCNN) are used to predict the price of Ragi. VAR, a time series model is predominantly used for time series applications and LSTM is a deep learning algorithm that has given consistent results in several domains. This influenced us to develop a novel framework VAR-Stacked_LSTM for crop price prediction. The performance of the proposed approach is experimented using Root Mean squared error (RMSE)and Mean Absolute Percentage Error (MAPE) metrics. The proposed framework VAR-Stacked_LSTM yields better performance compared to other models considered for experimental investigation as the RMSE is 116 and 117 in several regions and MAPE is 0.04 around 4%. Further, it is enhanced to predict the monthly price range for Ragi. This unique approach projects the lower and upper range, and it was validated with the actual price. The experimental result proves the efficiency of the proposed framework.

Keywords

• Agriculture
• Ragi
• price range prediction
• deep learning
• time series models