KLASIFIKASI KEPUASAN PELANGGAN BERDASARKAN DATA HASIL SURVEI PADA ISP ERATEL MENGGUNAKAN MACHINE LEARNING: CUSTOMER SATISFACTION CLASSIFICATION BASED ON SURVEY DATA AT ISP ERATEL USING MACHINE LEARNING

Muhammad Ary Sanjaya Putra Ary (Sistem Informasi Universitas Muria Kudus)
Arif Setiawan (Information Systems Study Program, Muria Kudus University)
Muhammad Arifin (Information Systems Study Program, Muria Kudus University)

RABIT: Jurnal Teknologi dan Sistem Informasi Univrab
Vol 11 No 2 (2026): Juli

Customer satisfaction is a critical factor in the sustainability of internet service providers (ISP). This study aims to compare the performance of Naïve Bayes, Support Vector Machine (SVM), and Random Forest algorithms in classifying customer satisfaction levels at ISP Eratel based on questionnaire survey data. Data were collected by distributing questionnaires to 301 respondents of Eratel customers with 10 question items on a Likert scale of 1–4 using convenience sampling technique. The research stages include data collection, preprocessing, labeling using the class interval method, stratified dataset splitting with an 80:20 ratio, classification model training, and web dashboard implementation. The labeling process produced three classes: Satisfied (54.82%), Fairly Satisfied (41.20%), and Dissatisfied (3.99%). Based on single split evaluation, the Naïve Bayes and SVM algorithms achieved the same accuracy of 90.16%, with Naïve Bayes showing slightly better performance in recognizing the minority class with a precision of 90.15%, recall of 90.16%, and f1-score of 90.05%, compared to SVM with precision of 87.29%, recall of 90.16%, and f1-score of 88.50%. Meanwhile, Random Forest achieved an accuracy of 86.89% with an f1-score of 86.25%, with a notable advantage in precision for the Dissatisfied class at 100% but a lower recall of 50%. Overall, Naïve Bayes emerged as the best-performing algorithm based on single split evaluation. The classification models were subsequently implemented into a Streamlit-based web dashboard that allows users to upload survey data in Excel (.xlsx) format, display interactive visualizations of customer satisfaction distribution, and spatially map classification results per sub-district in Kudus Regency in real-time. This study concludes that all three algorithms are capable of classifying customer satisfaction effectively, with Naïve Bayes demonstrating the best overall performance based on single split evaluation. Keyword: Customer Satisfaction, Naïve Bayes, Support Vector Machine, Classification, ISP.