Suchergebnisse

Medicare fraud results in considerable losses for governments and insurance companies and results in higher premiums from clients. Medicare fraud costs around 13 billion euros in Europe and between 21 billion and 71 billion US dollars per year in the United States. This study aims to use artificial neural network based classifiers to predict medicare fraud. The main difficulty using machine learning techniques in fraud detection or more generally anomaly detection is that the data sets are highly imbalanced. To detect medicare frauds, we propose a multiple inputs deep neural network based classifier with a Long-short Term Memory (LSTM) autoencoder component. This architecture makes it possible to take into account many sources of data without mixing them and makes the classification task easier for the final model. The latent features extracted from the LSTM autoencoder have a strong discriminating power and separate the providers into homogeneous clusters. We use the data sets from the Centers for Medicaid and Medicare Services (CMS) of the US federal government. The CMS provides publicly available data that brings together all of the cost price requests sent by American hospitals to medicare companies. Our results show that although baseline artificial neural network give good performances, they are outperformed by our multiple inputs neural networks. We have shown that using a LSTM autoencoder to embed the provider behavior gives better results and makes the classifiers more robust to class imbalance.

Medicare fraud results in considerable losses for governments and insurance companies and results in higher premiums from clients. Medicare fraud costs around 13 billion euros in Europe and between 21 billion and 71 billion US dollars per year in the United States. This study aims to use artificial neural network based classifiers to predict medicare fraud. The main difficulty using machine learning techniques in fraud detection or more generally anomaly detection is that the data sets are highly imbalanced. To detect medicare frauds, we propose a multiple inputs deep neural network based classifier with a Long-short Term Memory (LSTM) autoencoder component. This architecture makes it possible to take into account many sources of data without mixing them and makes the classification task easier for the final model. The latent features extracted from the LSTM autoencoder have a strong discriminating power and separate the providers into homogeneous clusters. We use the data sets from the Centers for Medicaid and Medicare Services (CMS) of the US federal government. The CMS provides publicly available data that brings together all of the cost price requests sent by American hospitals to medicare companies. Our results show that although baseline artificial neural network give good performances, they are outperformed by our multiple inputs neural networks. We have shown that using a LSTM autoencoder to embed the provider behavior gives better results and makes the classifiers more robust to class imbalance.

Medicare fraud results in considerable losses for governments and insurance companies and results in higher premiums from clients. Medicare fraud costs around 13 billion euros in Europe and between 21 billion and 71 billion US dollars per year in the United States. This study aims to use artificial neural network based classifiers to predict medicare fraud. The main difficulty using machine learning techniques in fraud detection or more generally anomaly detection is that the data sets are highly imbalanced. To detect medicare frauds, we propose a multiple inputs deep neural network based classifier with a Long-short Term Memory (LSTM) autoencoder component. This architecture makes it possible to take into account many sources of data without mixing them and makes the classification task easier for the final model. The latent features extracted from the LSTM autoencoder have a strong discriminating power and separate the providers into homogeneous clusters. We use the data sets from the Centers for Medicaid and Medicare Services (CMS) of the US federal government. The CMS provides publicly available data that brings together all of the cost price requests sent by American hospitals to medicare companies. Our results show that although baseline artificial neural network give good performances, they are outperformed by our multiple inputs neural networks. We have shown that using a LSTM autoencoder to embed the provider behavior gives better results and makes the classifiers more robust to class imbalance.

In order to protect resources from unauthorized access and data leakage in companies, security experts and administrators can use mechanisms such as Access Control (AC) and Transmission Control (TC). Both AC and TC are based on policies that are defined, modified and revoked by these experts. However, policy management can be a time-consuming and tiresome task, especially when both mechanisms are used on large sets of users and resources. Moreover, contradictions between AC and TC policies can appear, for instance when a legiti- mate user is allowed to send a resource to someone who cannot access it. Such contradictions can lead to data leakage.In this article, we first aim at studying experts feedback concerning policy definition and usage by reporting the results of a survey we have conducted among IT professionals. Based on the results of this survey, we then present a generic model that generates TC policies based on existing AC policies. This model serves several purposes. First, it takes into account the main AC models that are used in companies (i.e. genericity problem). Secondly, it tackles the problem of incoherences between AC and TC policies (i.e. coherence problem). Thirdly, it can reduce the total number of resources and subjects managed by the security policies (i.e. complexity problem). Finally, it takes into account the updates frequency of companies policies (i.e. rapidity problem).

In order to protect resources from unauthorized access and data leakage in companies, security experts and administrators can use mechanisms such as Access Control (AC) and Transmission Control (TC). Both AC and TC are based on policies that are defined, modified and revoked by these experts. However, policy management can be a time-consuming and tiresome task, especially when both mechanisms are used on large sets of users and resources. Moreover, contradictions between AC and TC policies can appear, for instance when a legiti- mate user is allowed to send a resource to someone who cannot access it. Such contradictions can lead to data leakage.In this article, we first aim at studying experts feedback concerning policy definition and usage by reporting the results of a survey we have conducted among IT professionals. Based on the results of this survey, we then present a generic model that generates TC policies based on existing AC policies. This model serves several purposes. First, it takes into account the main AC models that are used in companies (i.e. genericity problem). Secondly, it tackles the problem of incoherences between AC and TC policies (i.e. coherence problem). Thirdly, it can reduce the total number of resources and subjects managed by the security policies (i.e. complexity problem). Finally, it takes into account the updates frequency of companies policies (i.e. rapidity problem).

International audience ; Dans le cadre d'un développement dirigé par les modèles, les applications logicielles se définissent par utilisa- tion de transformations. Certaines transformations visent à répondre à des "problèmes récurrents" et leurs applications garantissent le respect de contraintes dans le modèle résultant. Elles sont identifiées sous le terme de Politiques. L'in- térêt d'une conception par applications de politiques est à la fois de faciliter la tâche du concepteur et de garantir par construction la cohérence du modèle construit. Nous étudions dans cette article la relation qui unit politiques et évolution de modèles en nous penchant sur la cohérence du modèle construit. Nous proposons pour cela une démarche de modélisation des politiques qui vise à assurer la cohérence du modèle construit relativement à chaque politique appliquée et à réparer automatiquement ces modèles dans le cas contraire. La réparation est alors basée sur la ré- applications des politiques.

International audience ; Dans le cadre d'un développement dirigé par les modèles, les applications logicielles se définissent par utilisa- tion de transformations. Certaines transformations visent à répondre à des "problèmes récurrents" et leurs applications garantissent le respect de contraintes dans le modèle résultant. Elles sont identifiées sous le terme de Politiques. L'in- térêt d'une conception par applications de politiques est à la fois de faciliter la tâche du concepteur et de garantir par construction la cohérence du modèle construit. Nous étudions dans cette article la relation qui unit politiques et évolution de modèles en nous penchant sur la cohérence du modèle construit. Nous proposons pour cela une démarche de modélisation des politiques qui vise à assurer la cohérence du modèle construit relativement à chaque politique appliquée et à réparer automatiquement ces modèles dans le cas contraire. La réparation est alors basée sur la ré- applications des politiques.

International audience ; Dans le cadre d'un développement dirigé par les modèles, les applications logicielles se définissent par utilisa- tion de transformations. Certaines transformations visent à répondre à des "problèmes récurrents" et leurs applications garantissent le respect de contraintes dans le modèle résultant. Elles sont identifiées sous le terme de Politiques. L'in- térêt d'une conception par applications de politiques est à la fois de faciliter la tâche du concepteur et de garantir par construction la cohérence du modèle construit. Nous étudions dans cette article la relation qui unit politiques et évolution de modèles en nous penchant sur la cohérence du modèle construit. Nous proposons pour cela une démarche de modélisation des politiques qui vise à assurer la cohérence du modèle construit relativement à chaque politique appliquée et à réparer automatiquement ces modèles dans le cas contraire. La réparation est alors basée sur la ré- applications des politiques.