An Efficient Fraud Identification Method Combining Manifold Learning and Outliers Detection in Mobile Healthcare Services

With the development of mobile Internet, various mobile applications have become increasingly popular. Many people are being benefited from the mobile healthcare services. Compared with the traditional healthcare services, patients' medical behavior trajectories can be recorded by mobile healthcare services meticulously. They monitor the entire healthcare services process and help to improve the quality and standardization of healthcare services. By tracking and analyzing the patients' medical records, they provide real-time protection for the patients' healthcare activities. Therefore, medical fraud can be avoided and the loss of public health funds can be reduced. Although mobile healthcare services can provide a large amount of timely data, an effective real-time online algorithm is needed due to the timeliness of detecting the medical insurance fraud claims. However, because of the complex granularity of medical data, existing fraud detection approaches tend to be less effective in terms of monitoring the healthcare services process. In this paper, we propose an approach to deal with these problems. By means of the proposed SSIsomap activity clustering method, SimLOF outlier detection method, and the Dempster-Shafer theory-based evidence aggregation method, our approach is able to detect unusual categories and frequencies of behaviors simultaneously. Our approach is applied to a real-world data set containing more than 40 million medical insurance claim activities from over 40000 users. Compared with two state-of-the-art approaches, the extensive experimental results show that our approach is significantly more effective and efficient. Our approach agent which provides decision support for the approval sender during the medical insurance claim approval process is undergoing trial in mobile healthcare services.