Developing an Artificial Intelligence based system to Detect Fraud and Corruption in Government

Le résumé fourni par la source

Both the costs and investments in Government departments have increased in recent years. Technology developments and the increase of lifespan were crucial for the expansion of this sector in an economic and social point of view. Besides the improvements in Government management, corruption and abuse within are increasing. Private and governmental institutions are making efforts to reduce it by detecting and preventing eventual cases of corruption and fraud. However, the results are insufficient and new approaches to face the problem are becoming more important. Nowadays, Data Mining techniques are considered as a good strategy to detect fraud and corruption in different areas such as credit card, bank accounts and telecommunications. In this Paper we explore a new approach using genetic algorithms to improve the number of detections. The two key aspects of this approach are first the organization of the healthcare data according to three different points of view, claims, patients, and providers. The second key aspect is related with the application of the machine learning algorithm to the three areas referred and the correlation of the results obtained from the three applications. The simulations tested show an unquestionable high success of detection of the fraudulent situations, around 90% of the fraudulent situations are correctly detect, using nine different fraud scenarios. As well as a detail analyses of the both the fraudulent claims characteristics and the performance of the algorithm in different situations. The results also show an evident reduction of the false alarm situations with the correlations of results of the three genetic algorithms. The average accuracy of the developed algorithm is 90%, and it's possible to observe in certain situations a total detection of all the fraudulent situations.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.