Analysis of MSMEs' Cassava Production Efficiency Using a Comparison of Machine Learning Models in Jember Regency

Cassava is one of Indonesia's agro-industrial commodities, but many Micro, Small, and Medium Enterprises (MSMEs) in the cassava processing industry face difficulties in achieving optimal production efficiency. This study aims to evaluate the efficiency of cassava processing production systems in MSMEs in Jember by comparing machine learning algorithms (Linear Regression, Random Forest, Support Vector Regression (SVR), and XGBoost) to predict output and key efficiency factors. The data used consists of 250 data points: 80% for model training and 20% for testing to build a machine learning-based prediction model, with input features production processing as the X-axis, and output in the form of production volume as the Y-axis. Data preprocessing, exploratory data analysis, and modeling were conducted using Python, with evaluation based on MAE, RMSE, and R² metrics. Among the tested models, Random Forest demonstrated the best performance with an R² value of 0.990. Sensitivity analysis revealed that production output increases significantly with the addition of labor and machines, with an optimal configuration of 15–20 workers and 2–3 machines per batch. The study concludes that focusing on overall production efficiency rather than merely increasing resources is the most effective strategy.