LIGHTWEIGHT CNN FOR NASKHI AND RIQ’AH KHAT CLASSIFICATION

The Arabic script has various types of khat that are complex and different from oneanother, thus requiring an appropriate classification to identify the type of khat used. Thisstudy uses the Lightweight Convolutional Neural Network (CNN) classification method toidentify the types of khat Naskhi and Riq’ah in the Arabic script dataset. The evaluationresults show that this classification model has an accuracy of 98.75% on training data and100% on validation data, with a relatively fast processing time of 2s 375ms/step fasterthan the previous study with an accuracy of 91.87% and an average processing time of 3s465ms/step. so that the model can be implemented properly in systems that require highdata processing speed and also devices that have resource limitations. These results indicatethat the classification model using the Lightweight CNN layer can be used as an effectivealternative in classifying types of Arabic writing, especially in recognizing certain types ofkhat such as Naskhi and Riq’ah. Furthermore, this research can be developed using a largerand more diverse dataset, and evaluated and compared with other classification models toimprove the model’s performance in recognizing more complex types of Arabic writing.