ANALISIS PERBANDINGAN KINERJA REGRESI LINEAR DAN K-NEAREST NEIGHBORS (KNN) DALAM PEMBUATAN MODEL PREDIKSI HASIL PANEN TANAMAN TEBU

STUDI KASUS DI DESA GONDANGLEGI

Abstract

Sugarcane yield prediction is an important aspect in agricultural planning and decision-making. This study aims to compare the performance of Linear Regression and K-Nearest Neighbors (KNN) methods in predicting sugarcane yields in Gondanglegi Village, Malang Regency. The dataset used consists of historical yield data from 2020 to 2024, including land area and rainfall as independent variables and crop yield as the dependent variable. The research stages include data preprocessing, normalization, data splitting with three ratios (80:20, 70:30, and 90:10), model training, and evaluation using Mean Absolute Error (MAE), Mean Squared Error (MSE), and the coefficient of determination (R²). The results indicate that Linear Regression consistently achieves higher R² values compared to KNN, indicating a better ability to explain yield variability. Although KNN shows competitive error values in certain ratios, it is more sensitive to data structure and parameter selection. Therefore, Linear Regression is considered more suitable for sugarcane yield prediction based on the dataset used in this study.