Performance Analysis Of Regression Model In Machine Learning To Prediction Rice Prices

Abstract

Abstract: The trade sector has become increasingly unpredictable, often experiencing drastic fluctuations in commodity prices, especially essential goods like rice. Rice plays a crucial role in people’s livelihoods, and rising prices can significantly reduce purchasing power for other necessities. This study aims to predict wholesale rice prices using independent variables such as Dry Paddy Price, Paddy Field Area, Fuel Price, Perton Production, and Cooking Oil Price. Seven regression estimation methods were applied: (1) Linear Regression, (2) Support Vector Regression Linear, (3) Support Vector Regression RBF, (4) Decision Tree Regression, (5) Random Forest Regressor, (6) Gradient Boosting Regression, and (7) MLP Regressor. The objective is to determine the method with the best accuracy for deployment in rice price prediction. The results show that Decision Tree Regression outperformed other methods, achieving the highest accuracy of 90% at a 90:10 data ratio, 80% at 80:20, 70% at 70:30, and 60% at 60:40. It also produced the lowest error values (MSE = 0.00000000, RMSE = 0.00000000) and the highest R² score (1.00000000), confirming its superior predictive performance for estimating rice prices.

Keywords: decision tree regression; machine learning; price estimation; rice price prediction.

Abstrak: Sektor perdagangan saat ini semakin tidak stabil, sering mengalami fluktuasi harga komoditas yang signifikan, terutama pada kebutuhan pokok seperti beras. Beras menjadi faktor penting dalam kehidupan masyarakat karena kenaikan harganya dapat menurunkan daya beli terhadap kebutuhan lain. Penelitian ini bertujuan memprediksi harga beras grosir dengan variabel independen yaitu Harga Gabah Kering, Luas Lahan Sawah, Harga Bahan Bakar, Produksi Perton, dan Harga Minyak Goreng. Tujuh metode regresi digunakan: (1) Linear Regression, (2) Support Vector Regression Linear, (3) Support Vector Regression RBF, (4) Decision Tree Regression, (5) Random Forest Regressor, (6) Gradient Boosting Regression, dan (7) MLP Regressor. Tujuannya adalah menentukan metode dengan akurasi terbaik untuk diterapkan dalam prediksi harga beras. Hasil menunjukkan bahwa Decision Tree Regression merupakan metode terbaik dengan akurasi tertinggi 90% pada rasio data 90:10, 80% pada 80:20, 70% pada 70:30, dan 60% pada 60:40. Metode ini menghasilkan nilai error terendah (MSE = 0.00000000, RMSE = 0.00000000) serta nilai R² tertinggi (1.00000000), menunjukkan kinerja prediksi terbaik dalam estimasi harga beras.

Kata Kunci: decision tree regression; estimasi harga; pembelajaran mesin; prediksi harga beras.