Identifying Difficult Quran Verses Using SVM, LSTM, and CNN

Quran is the holy book of Islam, revealed by Allah to Muhammad PBUH. However, not all muslims are proficient in Arabic, leading to difficulty in reading certain verses of Quran for some individuals. The purpose of the research is to classify Quranic verses as difficult or easy to contribute to facilitate learning for Muslims and to pave the way for advancements in Arabic natural language processing. The method used is classi-fication through Support Vector Machine (SVM), Long Short-Term Memory (LSTM), and Convolutional Neural Network (CNN), with Term Frequency-Inverse Document Frequency (TF-IDF) for feature extraction, along with random undersampling and SMOTE for handling class imbalance. The SVM model was selected because it is an old method that is quite good for classifying text. Meanwhile, the LSTM and CNN methods were chosen because these methods are relatively newer but are generally better for larger datasets. The result of this research is that the SVM model achieved an accuracy of 0.68 in classifying difficult verses in the Quran, higher than the accuracy achieved by the CNN (0.66) and LSTM (0.65) methods. This accuracy indicates that the approach used is quite effective in classifying challenging verses in Quran. Further analysis indicates that the average precision, recall, and f'l-score values for the SVM method are greater compared to the LSTM and CNN methods.