Shunzhi Zhu

Health care fraud detection using nonnegative matrix factorization

In a practical health care dataset, there are many patients with different prescriptions. A methodology for automatically identifying and clustering patients with similar symptoms is needed for health care management department to judge whether there are frauds in a large-scale clinic dataset. In this paper, we encode the clinic data with a low rank nonnegative matrix factorization algorithm to retain natural data non-negativity, thereby eliminating the need to use subtractive basis vector and encoding calculations presented in other techniques such as principal component analysis for similar feature abstraction. Result evaluations of the proposed method are conducted on a practical dataset supplied by Health Insurance Management Center of Xiamen. In our experiments, we have shown that this method is useful for health care fraud detection.

Information retrieval using Hellinger distance and sqrt-cos similarity

In this paper, we propose a similarity measurement method based on the Hellinger distance and square-root cosine. Then use Hellinger distance as the distance metric for document clustering and a new square-root cosine similarity for query information retrieval. This new similarity/distance also bridges between traditional tf_idf weighting to binary weighting in vector space model. Finally, we conduct a comparison on performance between this method and the one based on Euclidean distance and cosine similarity. And from the results, we clearly observe that the precision and recall are improved by using the sqrt-cos similarity.

Education based new computer network simulator design and implementation

Computer Network Simulator(CNS) is researched and developed for experimental teaching and research to overcome the limitations of the real network equipment resources. By using it, students can do the experiment on computer just like in the real machine. The CNS can not only help teachers and students to teach and learn, but also can achieve the goal of combining teaching with learning. In this paper, we design and implement the new CNS for education, especially for certification test. it has four aspects of improvement by comparing several kinds of mainstream simulator.

HVS and MBNS based steganography algorithm design and implementation

This paper proposes an image information hiding algorithm which bases on the HVS and MBNS. This algorithm uses the characteristics of Human Visual System (HVS) to embedded more large amounts of data according to different area to achieve different degree of embedded, and pixels in smooth areas corresponding the base of small, pixels in complex areas corresponding the base of great, and then the secret information segment can be embedded into the image based on modular arithmetic. A case study is provided to demonstrate the fact that this method achieve the efficient load and ensure the lowest rate of image distortion that relative to most standard test images.

Modeling urban growth by cellular automata: A case study of Xiamen City, China

This paper focus on the modeling of urban growth with regional difference. Firstly, two land use change map of Xiamen city in 2001 and 2007 were acquired by classification based on satellites images. Secondly, nine kinds of driving factors, which were derived from points of interest (POI) and DEM, were also selected by using distance analysis of GIS. Those factors include public services, economic, political and geographical aspects. Basing on these data, this study adopts logistic regression (LR) model to analysis the urban transition and effects contributed by driving factors. The overall accuracy rate of LR model is up to 81.9%-85.9% and the ROC is 0.896, indicating that it is capable to quantitative analysis the mechanism of different driving factors and the spatial-temporal land use change. Finally, a constrained CA model is applied to simulate and predict the future land use situation of Xiamen in 2020. The simulation results reveal that the increasing areas of construction are mainly located outside of the Xiamen Island. The overall land supply and demand are in contradiction obviously, which may lead to increasing pressure on farmland protection. In general, land use issues would become the main bottlenecks of the development of economic and society in Xiamen city. The prediction results can provide reliable guidance about policy implementation for land use planning department.

A pipe-based huge database object replication tool on heterogeneous database clusters

In rising ELT scenarios in data warehouse, for replicating many huge database objects(table/view) between heterogeneous database clusters, we propose a multi-way pipe-based replication architecture and put it into practice. To the best of our knowledge, there are few works on this aspects. Our contributions are from three aspects. The first is introducing a pipe-based architecture to avoid time spent in writing and reading data through an intermediate file, which is implemented on traditional file-based export or import tool. The second is automatically dividing a replication task into several sub-tasks, which will reduce time spent on task resuming when such a task encounters errors. At last, detailed experiments are made and the results show that such an architecture is effective in practical data warehouse projects.

Asymmetric classifier based on kernel PLS for imbalanced data

In classification tasks, class imbalance problem has been reported to hinder the performance of some standard classifiers, such as nearest neighbors algorithm. This paper presents an improvement to kernel partial least squares classifier (KPLSC) is proposed to deal with the class imbalance problem. This improvement is applicable to all cases no matter whether the data sets are linearly separable or not. Experiments on datasets from different domains show that the improvement performs well in classification problems.

A disease forecast and early warning system based on electronic health records

Disease forecast and early warning have been always important but difficult tasks. Because of the drawbacks of traditional records, the electronic health records, which bring in the ICD-lO, are used in our system. Input information are firstly de-duplicated to remove redundancy. After that, the system are used for disease early warning and forecast. The results show that the proposed system has great help for the health sector to prevent and control the diseases.