In this study we develop the machine learning models for forecasting agricultural products. The main concept of building the models is because machine learning is flexible and convenient to implement and it can be potential applications for a naïve user. The proposed model of Support Vector Machine (SVM) is able to forecast nonlinear or linear forecasting function upon kernel function. Many experiments were performed on the development of SVM and the most precision model by using statistical criteria was also selected. Real data of Thailand's Pacific white shrimp export and Thailand's produced chicken were used to validate candidate models. Autoregressive Integrate Moving Average (ARIMA) is also selected as a benchmarking to compare other developed models. For Pacific white shrimp export case, comparing to ARIMA, the error reduction from MAE, RMSE, and MAPE is 25.76%, 18.11%, and 19.05%, respectively. Moreover, the error reduction from MAE, RMSE, and MAPE is 21.78%, 18.76%, and 18.11%, respectively, for the case of produced chicken.
In this study we develop the hybrid models for forecasting in agricultural production planning. Real data of Thailand's orchid export and Thailand's pork product are used to validate candidate models. Autoregressive Integrate Moving Average (ARIMA) is also selected as a benchmarking to compare other developed models. The main concept of building the models is to combine different forecasting techniques in order to overcome the time-series forecasting errors. The combined models of Support Vector Machine (SVM) and ARIMA are considered as they can be represented both nonlinear and linear values. We perform many experiments on the combination of SVM and ARIMA and select the most precision model, which is the SVM (10) and ARIMA hybrid model, by using statistical criteria. For orchid export case, comparing to ARIMA, the error reduction from MAE, RMSE, and MAPE is 2.46%, 1.96%, and 4.63%, respectively. Moreover, the error reduction from MAE, RMSE, and MAPE is 8.08%, 6.24%, and 6.88%, respectively, for the case of pork product.
Cassava transportation planning usually involves unexpected demand, which may result in shortage supply. Furthermore, a distribution center at which cassava is collected is difficult to be located since the demand is unknown. In this research, hybrid forecasting model for predicting future demand in order to determine transshipment points is proposed. In addition, cluster analysis and particle swarm optimization are used for creating potential zones and determine a proper location as a new hub. Finally, the optimal value of a transportation network model using both forecasted value and actual value obtained from linear programming technique are tested and compared. The results indicate that the hybrid forecasting model provides the lowest error and forecasting value provides average error of optimal value compared to actual value by 19.81%. Moreover, zoning technique can be able to improve shipping volume fulfilled to a large truck.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.