Penggunaan internet dan telepon seluler terus meningkat di Indonesia, mendorong transformasi digital di berbagai sektor, termasuk layanan publik seperti PLN Mobile. Penelitian ini bertujuan untuk menganalisis sentimen ulasan pengguna terhadap aplikasi PLN Mobile, serta mencari ketidaksesuaian antara ulasan dan peringkat yang diberikan oleh pengguna. Analisis dilakukan dengan menggunakan penambahan spelling corrector Peter Norvig pada tahap preprocessing, serta kombinasi metode lexicon-based SentiStrength_id dan Support Vector Machine (SVM). Data ulasan dikumpulkan melalui web scraping dari Google Play Store, mencakup periode Januari 2022 hingga Desember 2023, dengan total 11.004 ulasan. Proses preprocessing melibatkan case folding, cleaning, normalisasi (slang words), spelling corrector, tokenizing, stopword, dan detokenize, diikuti dengan pelabelan menggunakan SentiStrength_id dan pemodelan menggunakan SVM. Penggunaan spelling corrector Peter Norvig pada tahap preprocessing, serta pelabelan menggunakan SentiStrength_id yang dikombinasikan dengan model SVM, secara signifikan meningkatkan akurasi analisis sentimen, dengan akurasi tertinggi mencapai 82% menggunakan rasio data split 90:10. Hasil analisis menunjukkan bahwa 67,4% ulasan bersentimen positif, 16,5% negatif, dan 16,1% netral. Berdasarkan hasil perbandingan kelas positif, netral, dan negatif terhadap 11.004 data hasil SentiStrength_id dengan ulasan berdasarkan peringkat, ketidaksesuaian pengguna dalam memberikan peringkat dengan ulasannya mencapai 5% untuk kelas positif dengan 553 data, 6,6% untuk kelas negatif dengan 720 data, dan 11,6% untuk kelas netral dengan 1.273 data. Berdasarkan hasil analisis sentimen, disarankan agar aplikasi PLN Mobile meningkatkan kinerja dengan mengoptimalkan kecepatan, memperbaiki bug, dan melakukan pengujian berkala. Fitur pembelian token perlu dioptimalkan, termasuk sistem backup dan notifikasi real-time. Layanan pengguna harus lebih responsif dalam menangani keluhan, dan panduan penggunaan harus disediakan dengan jelas, termasuk FAQ dan tutorial video, serta edukasi mengenai cara mengatasi masalah umum.

The increasing use of the internet and mobile phones in Indonesia has driven digital transformation across various sectors, including public services like PLN Mobile. This study aims to analyze user sentiment towards the PLN Mobile app and identify discrepancies between user reviews and the ratings they provide. The analysis was conducted using the Peter Norvig spelling corrector during the preprocessing stage, combined with a lexicon-based method, SentiStrength_id, and a Support Vector Machine (SVM). User reviews were collected through web scraping from the Google Play Store, covering the period from January 2022 to December 2023, totaling 11,004 reviews. The preprocessing process included case folding, cleaning, word normalization (slang words), spelling correction, tokenizing, stopword removal, and detokenization, followed by sentiment labeling using SentiStrength_id and modeling using SVM. The use of the Peter Norvig spelling corrector during preprocessing, along with sentiment labeling using SentiStrength_id combined with the SVM model, significantly improved the accuracy of sentiment analysis, achieving a peak accuracy of 82% using a 90:10 data split ratio. The results show that 67.4% of the reviews were positive, 16.5% were negative, and 16.1% were neutral. Based on a comparison of the positive, neutral, and negative classes for 11,004 data points labeled by SentiStrength_id and reviews based on ratings, discrepancies between user ratings and reviews were found in 5% of positive reviews (553 data points), 6,6% of negative reviews (720 data points), and 11,6% of neutral reviews (1.273 data points). Based on the sentiment analysis results, it is recommended that the PLN Mobile app improve performance by optimizing speed, fixing bugs, and conducting regular testing. The token purchase feature should be optimized, including a backup system and real-time notifications. Customer service should be more responsive in handling complaints, and user guides should be clearly provided, including FAQs and tutorial videos, as well as education on how to address common issues.