ABC Classification as Business Intelligence Method Based on a Novel Sales Segmentation and Feature Extraction Proposal

Daily, monthly, and annual multi-product sales records are stored in databases, but due to the massive amounts of data, they are not used for decision-making when updating product catalogs. Meanwhile, the use of artificial intelligence in business is increasing across all sectors of the economy. Large-scale data handling can be achieved using artificial intelligence techniques. Specifically, ABC inventory classification currently employs artificial intelligence techniques, including neural networks, fuzzy systems, and genetic algorithms. However, a state-of-the-art review has not found any research using vision techniques to classify ABC inventories. To address this gap, this research presents a novel approach to the intelligent classification of a company’s multiple products, using ABC. Recent vision system research often uses the Otsu method or its variants to determine the optimum threshold for binary image segmentation. Unlike this approach, our research does not use a single threshold value; instead, it uses the full binary frequency histogram as an image representation. From this, eight invariant characteristics are extracted from translation, rotation, and scale. The results show that the classification is accurate, clear, and simple as a decision-making tool. The proposed method is general and can be used in any production sector and at any enterprise size.