Meta-Learning Based Adaptive Crop Price Prediction for Agriculture Application

Dealing with unexpected changes in agricultural product prices that affect the farmers and the economic growth of a country is inevitable in the current day context. A unique approach is proposed for time-series prediction, which utilizes the power of crop price prediction and crop yield prediction of select crops to identify the relevant information with respect to the market prices, and crop yields. Many uncertain conditions such as climate changes, fluctuations in market, flooding, etc., cause problems to the agricultural process. In this work, the prices of selected essential crops analyzed for time-series prediction using meta-learning. The Self-Organized Map (S OM), LSTM (Long-Short Term Model), and this proposed Meta-Learning Based Adaptive Crop Price Prediction (MLACPP) was trained using crop price dataset, and crop yield dataset. Meta-learning function used in the approach utilizes the input, optimization-output, and task-related estimators for calculating meta-loss over multiple meta-network. The meta-learning with self-organizing capability learns a meta-network observing additional information from specialized support set. Moreover, this method deals with a task-oriented loss. Interestingly, it compares the current optimal-output of the optimization function g with the target-specific information. Experimental results show significant improvement in terms of prediction accuracy and cross-correlation entropy over the existing crop price prediction approaches. This research work provides the basis for making better decisions for price-fixing and crop yield maximization based on insights obtained from the dataset.