This study presents a neural network-based approach for forecasting the weekly wholesale price index of green grams (mung beans) in the Chinese market over a ten-year period from January 1, 2010, to January 3, 2020. The research aims to address the challenge of accurately predicting agricultural commodity prices, which are crucial for various market participants, including processors, speculators, and policymakers. The study uses a nonlinear autoregressive neural network model to capture complex nonlinear patterns in the price time series. The model is evaluated using different configurations of data splitting ratios, hidden neuron counts, and delay parameters. The results show that the neural network model produces forecasts with high stability and accuracy, with relative root mean square errors of 4.34%, 4.71%, and 3.98% for training, validation, and testing phases, respectively. The model outperforms other machine learning models and traditional time-series econometric methods in terms of forecast accuracy. The study also highlights the potential of neural networks in forecasting agricultural commodity prices, particularly for commodities with significant economic implications. The findings suggest that the neural network model can be used for independent technical price forecasts or combined with other models to provide insights into price trends and policy research. The study concludes that the neural network model is a valuable tool for forecasting the weekly price index of green grams in the Chinese market.This study presents a neural network-based approach for forecasting the weekly wholesale price index of green grams (mung beans) in the Chinese market over a ten-year period from January 1, 2010, to January 3, 2020. The research aims to address the challenge of accurately predicting agricultural commodity prices, which are crucial for various market participants, including processors, speculators, and policymakers. The study uses a nonlinear autoregressive neural network model to capture complex nonlinear patterns in the price time series. The model is evaluated using different configurations of data splitting ratios, hidden neuron counts, and delay parameters. The results show that the neural network model produces forecasts with high stability and accuracy, with relative root mean square errors of 4.34%, 4.71%, and 3.98% for training, validation, and testing phases, respectively. The model outperforms other machine learning models and traditional time-series econometric methods in terms of forecast accuracy. The study also highlights the potential of neural networks in forecasting agricultural commodity prices, particularly for commodities with significant economic implications. The findings suggest that the neural network model can be used for independent technical price forecasts or combined with other models to provide insights into price trends and policy research. The study concludes that the neural network model is a valuable tool for forecasting the weekly price index of green grams in the Chinese market.