Hilmi Ammar
Universitas Muhammadiyah Prof. Dr. Hamka

Published : 1 Documents Claim Missing Document
Claim Missing Document
Check
Articles

Found 1 Documents
Search

Perbandingan Tingkat Akurasi Algoritma Naïve Bayes, Decision Tree dan Random Forest dalam Analisis Sentimen Pengguna Aplikasi Samsat Digital Nasional Hilmi Ammar; Ade Davy Wiranata
IJAI (Indonesian Journal of Applied Informatics) Vol 10, No 1 (2025)
Publisher : Universitas Sebelas Maret

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.20961/ijai.v10i1.111010

Abstract

Abstrak : Penelitian ini menganalisis sentimen pengguna terhadap aplikasi Signal – Samsat Digital Nasional. Dari total 2.000 ulasan yang terkumpul, setelah melalui tahapan filter data diperoleh 1.743 data yang digunakan untuk analisis. Data tersebut kemudian dipecah menjadi 1.394 data pelatihan (alokasi 80%) dan 349 data pengujian (alokasi 20%). Tiga model klasifikasi yang digunakan dalam penelitian ini meliputi Naïve Bayes, Decision Tree, dan Random Forest. Hasil penelitian menunjukkan bahwa Random Forest memiliki performa terbaik dengan akurasi 92,9% serta keseimbangan tinggi dalam mengenali sentimen positif (f1-score  95,9%) dan negatif (f1-score  73,1%). Naïve Bayes mencapai akurasi 89,4% namun kurang seimbang, sedangkan Decision Tree memperoleh akurasi 86,8% dengan hasil yang lebih stabil dibandingkan Naïve Bayes. Secara keseluruhan, visualisasi data berdasarkan analisis menggunakan kamus lexicon menunjukkan bahwa 82,9% ulasan bersentimen positif dan didominasi oleh rating bintang lima, yang mengindikasikan bahwa pengalaman pengguna terhadap aplikasi Signal – Samsat Digital Nasional tergolong sangat baik.=====================================================Abstract :This research analyzes user sentiment towards the Signal – National Digital Samsat application. From a total of 2,000 collected reviews, after going through the data filtering stage, 1,743 data were obtained for analysis. The data was then split into 1,394 training data (80% allocation) and 349 testing data (20% allocation). Three classification models used in this study were Naïve Bayes, Decision Tree, and Random Forest. The results showed that Random Forest had the best performance with 92.9% accuracy and high balance in recognizing positive (f1-score 95.9%) and negative (f1-score 73.1%) sentiment. Naïve Bayes achieved 89.4% accuracy but was less balanced, while Decision Tree achieved 86.8% accuracy with more stable results than Naïve Bayes. Overall, data visualization based on analysis using the lexicon dictionary shows that 82.9% of reviews are positive and dominated by five-star ratings, which indicates that the user experience of the Signal – Samsat Digital Nasional application is classified as very good.