EXPLAINABLE AND DRIFT-AWARE MACHINE LEARNING FRAMEWORK FOR CROP PRICE PREDICTION USING VISUAL AND TEMPORAL DATA

In this paper, the proposed model is a drift-aware, explainable AI framework to predict crop prices by fusing heterogeneous data sources: crop images, weather, market price, and seasonal factors. The proposed framework uses Convolutional Neural Networks (CNNs) to perform image-based crop quality recognition and Long Short-Term Memory (LSTM) networks to forecast market prices as a time-series data. Additionally, Adaptive Windowing (ADWIN) to adapt to concept drift in market environment and SHAP (SHapley Additive exPlanations) for model explainability.   Experimental results show that our framework offers superior prediction accuracy and lower prediction error compared to conventional machine learning models. The fusion of visual quality information and market information represents the agricultural systems with richness and completeness, while the drift detection helps the framework adapt to dynamic market fluctuations. Also, the use of explainable AI helps to understand and trace the cause of prediction in the model.   From the application viewpoint, the proposed framework is able to provide efficient decision support to farmers to plan selling strategies according to market price prediction. The system could increase the profits of farmers by diminishing market uncertainties and increasing access to efficient market participation information.   However, there still are some limitations to our proposed system. Firstly, the quality and availability of data can greatly affect the performance of the model, especially in areas with limited digital access. Secondly, the generalization capability across various crops and geographical regions of the model needs further study. In addition, the complexity of the deep learning models may lead to increased deployment difficulty in resource-constrained environments.

[1]               Hyndman, R. J., & Athanasopoulos, G. (2018). Forecasting: Principles and practice (2nd ed.). OTexts,  https://otexts.com/fpp2/. [2]               Hochreiter, S., & Schmidhuber, J. (1997). Long short-term memory. Neural Computation, 9(8), 1735–1780, https://doi.org/10.1162/neco.1997.9.8.1735. [3]               Kamilaris, A., & Prenafeta-Boldú, F. X. (2018). Deep learning in agriculture: A survey. Computers and Electronics in Agriculture, 147, 70–90, https://doi.org/10.1016/j.compag.2018.02.016. [4]               Lundberg, S. M., & Lee, S.-I. (2017). A unified approach to interpreting model predictions. In Advances in Neural Information Processing Systems (NeurIPS),  https://papers.nips.cc/paper/7062-a-unified-approach-to-interpreting-model-predictions. [5]               Goodfellow, I., Bengio, Y., & Courville, A. (2016). Deep learning. MIT Press, https://www.deeplearningbook.org. [6]               Wolfert, S., Ge, L., Verdouw, C., & Bogaardt, M. J. (2017). Big data in smart farming: A review. Agricultural Systems, 153, 69–80, https://doi.org/10.1016/j.agsy.2017.01.023. [7]               You, J., Li, X., Low, M., Lobell, D., & Ermon, S. (2017). Deep Gaussian process for crop yield prediction. In Proceedings of the AAAI Conference on Artificial Intelligence, https://ojs.aaai.org/index.php/AAAI/article/view/10644 [8]               Zhang, G. P. (2003). Time series forecasting using a hybrid ARIMA and neural network model. Neurocomputing, 50, 159–175, https://doi.org/10.1016/S0925-2312(01)00702-0. [9]               Bifet, A., & Gavaldà, R. (2007). Learning from time-changing data with adaptive windowing. In Proceedings of the SIAM International Conference on Data Mining, https://doi.org/10.1038/nature14539. [10]             Liakos, K. G., Busato, P., Moshou, D., Pearson, S., & Bochtis, D. (2018). Machine learning in agriculture: A review. Sensors, 18(8), 2674,  https://doi.org/10.3390/s18082674.