www.astesj.com 551 Approach to Combine an Ontology-Based on Payment System with Neural Network for Transaction Fraud Detection Ahmed EL Orche * , Mohamed Bahaj 1 Faculty of Sciences and Technologies, Hassan 1st University, Settat, Morocco A R T I C L E I N F O A B S T R A C T Article history: Received: 17 November, 2019 Accepted: 18 March, 2020 Online: 08 April, 2020 Fraud, as regards means of payment, means the behavior of any legal or natural one that makes an abnormal or irregular use of a way of payment, elements of it or information contained therein, to improperly obtain an honest, service or enrichment, and or causing financial damage to the one that has distributed the means of payment to a user or a 3rd party; Contests in bad faith a legitimate payment order of which she is that the initiator. during this paper we are getting to propose an approach to managing the risks, it consists to mix a machine learning with an ontology-based on a payment system to succeed in this objective. Machine learning may be a field of study that improves their performance in solving tasks without being explicitly programmed by each. An ontology is that the structured set of terms and ideas that represent the meaning of an information field, whether by the metadata of a namespace, or the elements of a domain of knowledge. Keywords: Ontology Machine learning Neural network Payment system Fraud 1. Introduction Artificial Intelligence (AI) has come to the fore, with many companies using them to develop their solutions and/or services. If AI is a global concept, Machine Learning (ML) is a technology that allows machines to access data so that they can learn, predict, and categorize information. ML is a branch of artificial intelligence, which is mainly based on the automatic construction of statistical models based on the widest possible body of learning. Deep Learning is a sub-branch of this discipline, which uses as model neural networks, very complex with many layers. This approach, which has been made popular by the availability of low- cost computing power. With AI and ML, companies can enrich and leverage this information. Finally, via pre-established models, they will be able to test the results obtained and reiterate them throughout the life cycle. The adoption of ML has been accelerated by increased data processing power, the development of Big Data and advances in statistical modeling. It relies on complex statistical methods and high computing power. At the heart of this concept, however, is a very simple idea. By identifying relationships, the most influential cause-and-effect of the past, a machine can learn to make accurate predictions for the future. The ML is based on powerful computers that are guided by human intelligence to sift through billions of data and identify cause-and-effect relationships. Then all this information is introduced in a variety of algorithms to come up with predictions. With time, computers improve in identifying these cause-and- effect relationships, they exploit the knowledge they have acquired and use it to refine the algorithms. It is "learning" that takes place and with a much faster processing speed than that of the human brain. The fraud tracked and resolved by ML? Fraud detection is a big challenge. However, Fraudulent transactions are few and represent a very small part of the activity within an organization. Nevertheless, a small percentage of the business can quickly turn into significant financial losses without the right tools and systems in place to deal with. Cybercriminals are smart. The traditional fraud schemes are no longer effective, they have made them evolve. The good news is that with Machine Learning advances, systems can learn, adapt and discover new ways to prevent fraud. On the other hand, many strengths make ML such a powerful and effective tool in the fight against fraud: ASTESJ ISSN: 2415-6698 * Ahmed EL Orche, Email: ahmed.elorche@gmail.com Advances in Science, Technology and Engineering Systems Journal Vol. 5, No. 2, 551-560 (2020) www.astesj.com https://dx.doi.org/10.25046/aj050269