Banyak umat Muslim di Indonesia menghadapi tantangan dalam membaca Al Qur’an sesuai kaidah tajwid yang benar, terutama karena keterbatasan akses pembelajaran dan bimbingan. Penelitian ini bertujuan untuk mengembangkan model klasifikasi berbasis Support Vector Machine (SVM) yang mampu mendeteksi kesalahan pembacaan tajwid secara spesifik, yaitu pada aturan Mad, Idgham, dan Ikhfa. Data audio diambil dari dataset QDAT, yang dianotasi secara manual menggunakan platform Labeller untuk menandai segmen bacaan yang relevan. Ekstraksi fitur dilakukan dengan Mel-Frequency Cepstral Coefficients (MFCC), yang kemudian diubah menjadi representasi statistik (mean, standar deviasi, min, dan max) sebagai input model. Model dilatih dan diuji menggunakan pendekatan klasifikasi multi-label, dengan evaluasi performa melalui F1-score macro dan micro. Hasil eksperimen menunjukkan bahwa model mencapai F1-score macro sebesar 91% dan micro sebesar 89%, dengan precision tertinggi pada label Idgham (100%). Model ini diintegrasikan ke dalam aplikasi Android bernama Quranku, yang memungkinkan pengguna merekam bacaan, mengirimkannya ke server, dan menerima umpan balik prediksi secara langsung. Aplikasi diuji menggunakan metode Black Box Testing dan menunjukkan fungsionalitas yang berjalan sesuai rancangan. Sistem ini diharapkan dapat menjadi media pembelajaran tajwid yang adaptif, interaktif, dan mudah diakses.

Many Muslims in Indonesia face challenges in reciting the Qur’an in accordance with proper tajwid rules, largely due to limited access to learning resources and guidance. This study aims to develop a classification model based on Support Vector Machine (SVM) capable of specifically detecting tajwid reading errors in the rules of Mad, Idgham, and Ikhfa. Audio data were obtained from the QDAT dataset, which was manually annotated using the Labeller platform to mark relevant recitation segments. Feature extraction was carried out using Mel-Frequency Cepstral Coefficients (MFCC), which were then transformed into statistical representations (mean, standard deviation, minimum, and maximum) to serve as input for the model. The model was trained and tested using a multi-label classification approach, with performance evaluated through macro and micro F1-scores. Experimental results show that the model achieved a macro F1-score of 91% and a micro F1-score of 89%, with the highest precision recorded for the Idgham label (100%). The model was integrated into an Android application named Quranku, enabling users to record their recitations, send them to a server, and receive immediate predictive feedback. The application was tested using Black Box Testing and demonstrated functionality consistent with its design. This system is expected to serve as an adaptive, interactive, and accessible tajwid learning medium.