The Development of a Business Intelligence System for Analyzing and Forecasting Coconut Production of Large-Scale Coconut Farmers in Banglao Subdistrict of Khlong Khuen District in Chachoengsao Province

Abstract

          This research aimed to 1) forecast coconut production for large-scale coconut farmers in Bang Lao Subdistrict, Khlong Khuean District, Chachoengsao Province, 2) develop a business intelligence system for analyzing and forecasting coconut production, and 3) transfer knowledge and promote the utilization of the system. The study employed quantitative research methodology, targeting 31 members of a large-scale coconut farmers' group. Research instruments included secondary data recording forms, a system suitability assessment form, and a research utilization evaluation form. The efficiency analysis involved three methods for current production forecasting: linear regression analysis, XGBoost model, and artificial neural network model; and three methods for future production forecasting: ARIMA (Box-Jenkins) time series forecasting, SARIMA model, and Holt-Winters Exponential Smoothing.
          Results showed that: 1) The XGBoost model performed best in forecasting current production (MSE=629.992, MAE=19.590, RMSE=25.099), while the SARIMA model was most effective for three-year future production forecasting (MSE=403.694, MAE=16.866, RMSE=20.092). 2) Expert evaluation of the business intelligence system's suitability was at the highest level across all standards: appropriateness (x̄=4.80, S.D.=.249), utility (x̄=4.70, S.D.=.274), feasibility (x̄=4.85, S.D.=.249), and accuracy (x̄=4.87, S.D.=.183). 3) Among the 30 participants (96.77%) in knowledge transfer and research utilization, overall satisfaction was high (x̄=4.33, S.D.=.606). The top three aspects rated highest were forecasting accuracy (x̄=4.67, S.D.=.474), system usability (x̄=4.52, S.D.=.601), and knowledge and skill development (x̄=4.52, S.D.=.599), all at the highest level.