Classifying Promotion Images Using Optical Character Recognition and Naïve Bayes Classifier
Hubert Hubert, Peter Phoenix, Richard Sudaryono, Derwin Suhartono
Abstract
Promotion is one of the most effective ways to promote a business, and most people love promotions. Usually these businesses announce their promo by uploading images to social medias such as Instagram. However, most of the time these promo images are buried in the sea of other non-promotional images. It would be more practical if computers could be utilized to automatically look for images containing promotional offers. That is why this research is done to discuss about creating a system that is able to tell whether an image contains information about a promotional offer or not automatically without human intervention using Optical Character Recognition (OCR) and Naïve Bayes Algorithm as the classifier. Random Forest and K-Nearest Neighbor are also used as a comparison to the Naïve Bayes Algorithm. In this experiment we use cross validation method where we divide 158 images into five groups to train and test our model. The Naïve Bayes model achieved 94,31% accuracy, 94,33 % recall, 94,11 % precision, and 0.93 F1 score on average, which is the highest among these three algorithms. Based on the results, we can conclude that Optical Character Recognition (OCR) and Naïve Bayes Algorithm are quite suitable for this problem.