Perbandingan Prediksi terhadap Peningkatan Jumlah Pelanggan Iconnet dengan Algoritma Regresi Linear dan Random Forest pada Wilayah Jabodetabek dan Banten

This study compares the linear regression and random forest algorithms in predicting the number of Iconnet service customers in the Jabodetabek and Banten regions. The dataset comprises two years of sales data processed through filtering, cleaning, and labeling. Evaluation metrics include MAE, MAPE, and RMSE. The results show that linear regression performs better in predicting customer numbers, achieving MAE 369.85, MAPE 8.80%, and RMSE 388.89, compared to random forest with MAE 679.37, MAPE 16.95%, and RMSE 794.26. Conversely, random forest outperforms linear regression in bandwidth prediction (MAE 733.80, MAPE 26.61%, RMSE 860.20) and regional prediction (MAE 25607.49, MAPE 23.42%, RMSE 38177.12), as linear regression produces higher errors. The findings highlight the importance of selecting algorithms based on data characteristics and application needs. This research provides strategic insights for developing data-driven customer service solutions.