Four Parameter Beta Generalized Mixed Effect Tree and Random Forest for Area Yield Crop Insurance

Area Yield Index (AYI) was considered as the most potential alternative crop insurance policy in Indonesia. To support this policy, delivering accurate paddy productivity prediction is a must. Thus, we purpose a new flexible Be-ta Four Parameter Generalized Mixed Effect Tree and Random Forest predic-tion model that combines the use of tree regression and random forest with a Bayesian beta four parameter GLMM approach. This model takes into con-sideration that paddy productivity has a bounded minimum and maximum distribution or known as a Beta Four Parameter distribution, variation effect of paddy productivity between areas, and captures complex linear and non-linear relationships in the data. This model was incorporated to design a pro-totype AYI crop insurance in Central Kalimantan, Indonesia that can be fur-ther developed in other areas. Farmer survey data integrated with processed satellite data was utilized in the process. Results show that high predictive accuracy was achieved in the proposed model. Therefore, beneficial for accu-rately assessing risk, setting fair premiums, reducing adverse selection, effi-ciently allocating resources, and ensuring the long-term sustainability of the paddy crop insurance program.