XGBOOST AND CNN CLASSIFICATION MODELS ON PRONUNCIATION OF HIJAIYAH LETTER ACCORDING TO SANAD

• Main Author: Muhammad Hafidz Azis, Aaz
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/73946
• Institution: Institut Teknologi Bandung

Hijaiyah letters are letters found in the composition of the Qur'an. The character of the hijaiyah letter is the appearance of the character that comes out of its pronounciation, while the makharijul letter is the place where the letter comes out when pronouncing the hijaiyah letter. Hijaiyah letters that are in Sanad can be used as a benchmark for correct or valid reading because they already fulfill the characteristics and meaning of the letters. The limited number of Al-Qur'an instructors who are still few is one of the obstacles to learning the Al-Qur'an properly. This is indicated by the small number of teaching Al-Qur'an in Sanad which opens tahsin learning, whereas learning in tahsin in Sanad is one of the lessons that has standards in pronouncing the rules of letters according to the nature of the letters. The voice recognition system is able to recognize voices so that with this technology it is expected to be able to support learning without having to meet with the teacher. In this study a classification model for hijaiyah letters was built based on the characteristics of the letters using the XGBoost shallow learning algorithm and the CNN deep learning algorithm. CNN tends to produce better performance than Extreme Gradient Boosting (XGBoost). The XGBoost algorithm model has superior accuracy on the S2 and T7 properties. However, it has low memory. The addition of data provides a balance to the performance results, so that the values for accuracy, precision, memory and F-1 score have sufficient levels. The accuracy for S properties it has an average of 78.55%, T properties is 69.44%, while the average per letter is 73.24%.