Invoice Classification Using Deep Features and Machine Learning Techniques

Invoices are issued by companies, banks and different organizations in different forms including handwritten and machine-printed ones; sometimes, receipts are included as a separated form of invoices. In current practice, normally, classifying these types is done manually, since each needs a special kind of processing such as making them suitable for optical character recognition systems (OCR). Classifying the invoices manually to different categories is a hard and time-consuming task. Therefore, we propose an automatic approach to classify invoices into three types: handwritten, machine-printed and receipts. The proposed method is based on extracting features using the deep convolutional neural network AlexNet. The features are classified using various machine learning algorithms, namely including Random Forests, K-nearest neighbors (KNN), and Naive Bayes. Different cross-validation approaches are applied in the experiments to ensure the effectiveness of the proposed solution. The best classification result was 98.4% (total accuracy), which was achieved by the KNN, such an almost perfect performance allows the proposed method to be used in practice as a preprocess for OCR systems, or as a standalone application.