Search

To improve students' performance effectively, many Chinese universities are establishing systems to predict student's academic performance and sending student's academic early alerts. With student's population of over 600,000, Chinese-foreign cooperation in running schools (CFCRS) has become one of the booming higher education forms in China. Compared with students in the non-cooperatively running programs, CFCRS students' academic performances are weaker on average. To predict the academic-at-risk students and provide efficient supports to the students, a precise and prompt academic prediction is in great need. Therefore, this research aims at representing a more efficient and accurate model to predict academic performance of CFCRS students which will be based on graph convolutional network. In this research, student's similarity is measured on academic performance by Pearson correlation coefficient. An undirected graph in which similar students are connected is conducted. Feature matrix is composed of students' previous grades. In addition, graph convolutional network is trained based on the undirected graph and feature matrix. In the experiment, it shows that this model predicts certain student's performance on certain course from the final exam results in previous semesters, which might improve learning efficiency and teaching quality. With an average accuracy of 81.5%, graph convolutional network outperforms support vector machine and random forest models.

Cyclooxygenase-2 (COX-2), an inducible isoform of cyclooxygenase, has been reported to be correlated with tumorigenesis, tumor progression and metastasis. The present study was designed to investigate the clinicopathological and prognostic significance of COX-2 in Chinese patients with prostate cancer. Firstly, RT-PCR and Western blot assays were performed to detect the expression of COX-2 mRNA and protein in prostate cancer cell lines and 20 tissue samples (tumor or corresponding non-tumor). Next, immunohistochemistry was performed to detect the expression of COX-2 protein in 88 prostate cancer tissue samples. Finally, the correlation between COX-2 expression and clinicopathological factors and patient survival was evaluated. We found that the expression levels of COX-2 mRNA and protein showed significant difference among four prostate cancer cell lines. Moreover, the levels of COX-2 mRNA and protein were significantly higher in prostate cancer tissues than in corresponding non-tumor tissues. COX-2 staining was positive in the cytoplasm of prostate cancer cells. High-COX-2 expression was correlated with the Gleason score (P=0.009), tumor stage (P=0.012), and lymph-node status (P=0.036). Furthermore, patients with high-COX-2 expression showed lower disease-free (P=0.014) and overall survival (P=0.047) rates than those with low-COX-2 expression. Univariate and multivariate analyses suggested that the status of COX-2 protein expression was an independent prognostic indicator for patients' survival. Taken together, higher COX-2 protein expression might provide an independent prognostic marker for Chinese patients with prostate cancer who have undergone surgery.

Four high-purity germanium 4-fold segmented Clover detectors have been applied in the experiment of neutron-rich nucleus 21N. The performance of those four Clovers have been tested with radioactive sources and in-beam experiments, and the main results including energy resolution, peak-to-total ratios, the variation of the hit pattern distribution in different crystals of one Clover detector with the energy of γ ray, and absolute full energy peak detection efficiency curve, were presented.

The characteristics clustering of electric loads are of great importance to the Measurement-Based Modeling method. In this paper, a new method based on Kohonen self-organization neural network is presented for the characteristics clustering of dynamic loads. At first, the model of every group of load disturbance data is established, and then the responses of the load models to the same voltage excitation and the pre-disturbance active power of the loads are incorporated into the feature vectors. At last, Kohonen neural network is introduced to cluster. Many sets of load data measured from North China Power System in three years (1996–1998) have been dealt with using the method. The results show load characteristics have rule though they are random and time-varying. The feasibility of the Measurement-Based Modeling approach is also proved.

For realizing accurate reliability allocation of complex system, a method based on inverse thinking is proposed in this paper. The important of every sub-system is determined by analyzing the inter-relationship between the changing of sub-system reliability and system reliability. The accurate reliability allocation model of the complex system is established based on the neural networks. The self-restriction conditions of every sub-system and the system reliability are taken as the input variables of the model, the bio-mutual ratio of sub-system reliability is taken as the output variables of the model. The important of every sub-system in the prophase reliability experiment is learned by the trained neural networks. Then, a positive reciprocal matrix can be formed by the output variables of the model. The reliability allocation coefficients of every sub-system can be achieved by solving the positive reciprocal matrix, and the expected reliability of every sub-system can also be achieved by substituting the reliability allocation coefficients into the reliability allocation formula. It is proved by the examples that the result of the model is completely according with the reliability allocation rules; the artificial factors in reliability allocation are avoided.

This paper proposes an improved scheme of using the inter-cluster distance in the feature space to choose the kernel parameters. First, the candidate vectors of the training set are selected. Then calculate the inter-cluster distance between classes to choose the proper kernel parameters. Finally the selected kernel parameters are used to train the Support Vector Machine (SVM) models. The basic principle is that the Support Vector (SV) set contains all information necessary to solve a given classification task. Experiment results show that our scheme costs much less computation time. Moreover, suitable kernel parameters can also be selected at the same time.

The deficiencies of existing load models and their implementations in power system simulation programs are addressed in this paper. To improve the load model for power system simulation, a Synthesis Load Model (SLM) with distribution network, which is consistent with the structure of power loads and distribution networks, is presented. The SLM is reasonable in structure and convenient to use. The SLM can simulate the load components and distribution systems including distribution network, reactive compensation and the generators connected to low voltage networks. The validity of SLM has been validated by four field tests of three-phase short circuit in Northeast China Power Grid in the years of 2004 and 2005 respectively.

This paper presents a novel image watermarking scheme for both tamper detection and tampered image recovery. The basic principle of this scheme is to compress per block of the original image using SPIHT (Set Partitioning in Hierarchical Trees), then scramble and embed the bit stream into LSBs (Least Significant Bit) of corresponding offset block. Experimental results show that the proposed scheme can detect and locate tamper accurately, and recover tampered block data with satisfied image visual quality.

A distributed spiders antitype was designed by Web Service based on Service-Oriented Architecture(SOA).This antitype is made up of a server and several clients. The clients are controlled to download a new web page by the server according to the crawled pages. Moreover, they must manage the To Crawl , Crawled URL queues and Noise URL queue after analyzing it by multi-threads. Furthermore, they keep connection with the server to pass the unknown URL and domain names. The server is made up of the front platform and the background. The front platform controls the clients including the design of load balance policy and real-time monitoring of clients by Microsoft Message Queue(MSMQ). Web service is deployed on the server background which contains the structure of persistent data connection. With the help of this structure, the front platform and the clients can access data by the normative interface. Finally, a lot of experiments were done which show that the distributed spiders system has good robust performance.

GVF snake model means gradient vector flow snake model. It is a representative of the active contour model and it has been used to solve image segmentation for some conglutinated images. GVF snake model is also a new accurate arithmetic to obtain the boundary of the object edge especially for those stomach epidermis tumors which have several conglutinated areas. As the first step of this arithmetic, many pretreatment processes are necessary such as the gauss blur and the edge mapping to get the clearer image boundary before starting the image segmentation. The next step after pretreatment processes is calculating the GVF outside force field which makes the snake move quicker to the object boundary after the initial points been selected. The experiments also show that as the key parameter of GVF snake model, mu must be set below 0.3 in order to get better segmentation results, otherwise the snake cannot move closer to the object boundary because the GVF outside force field dies down.

This paper presents a method of features extraction in speaker recognition. This method divides voice signal into two parts based on the MPEG Psychological Model I, and processes them respectively. And then we analyze the extracted MFCC parameters, and compare the partition of frequency-band between Mel-spaced filter group and wavelet packet decomposition. We extract the coefficients by wavelet package transform (called ldquoWPTCrdquo) as diagnostic parameters used in the speaker recognition. The results of experiments indicated that it perform well than MFCC while using the WPTC based on MPEG-I.

A novel approach of blind steganalysis is proposed, which is based on image texture energy-entropy features. Image complexity describes the difference of image's content and texture. Steg-image's texture is ordinarily more complicated than that of cover image. For analyzing image complexity, using image texture features to measure the statistical differences between cover image and steg-image. In the paper, we analyze image complexity based on image texture segmentation technique, and use Laws' image texture energy-entropy features to measure the statistical differences between cover image and steg-image. Applying these texture features, blind steganalysis is implemented. Support Vector Machine (SVM) is used as classifier to distinguish whether a given image is embedded into the convert message. Experiment results show that the proposed approach is greatly valuable and our blind steganalysis method attains a good testing accurate rate.

IT plays a strategic role in modern organizations, and IT strategy works as a guideline for IT investment, IT project selection and implementation, IT management and IT service. The alignment of business strategy and IT strategy is even more important than IT strategy itself. This study measured business strategy, IT strategy, strategic alignment, IT performance, and business performance. Analysis of data gathered in a mail survey of China's manufacturing SMEs indicated that: 1) IT strategy itself does not influence business performance and IT performance; 2) strategic alignment strongly correlated with business performance and the impacts are positive; 3) strategic alignment has no impact on IT performance; 4) There is positive relationship between IT performance and business performance; 5) The relationship between business strategy and business performance is not strong. The contribution of different IT systems to IT performance is also examined.

In this paper, we present a technique for multi-resolution meshes deformation based on pyramid coordinates. Pyramid coordinates are rotation invariant, and can achieve low distortion editing for large deformations. Their drawback is that the reconstruction process is nonlinear and time consuming especially for large meshes. In order to decrease the computational cost, the proposed method reduces the number of vertices involved in the deformation. We first construct a simplified mesh using a modified CPM algorithm; then edit the simplified meshes by pyramid coordinates based mesh detail presentation; finally, series of details are added to the deformed simplified mesh. Our algorithm can generate natural looking deformations more efficiently than the original pyramid coordinates method.

This paper proposes a new method to embed watermark in binary images. The given binary image is partitioned into two level blocks, and the watermark will be embedded into the sub-blocks based on the parity of characteristic values. The method is very efficient especially when applied to those binary images whose black and white pixels are distributed approximately uniformly. The experimental results show that the method only requires fewer operations and it can embed more watermarks. Embedding two bits information in a 4times4 block needs to change only one and a half pixels on average

In this paper, we investigate the trace of seamless splicing for blind image splicing. Image splicing is an essential operation for digital photomontage, which in turn is a technique for creating image forgery. Although good forgeries may leave no visual clues of having been tampered with, they may, nevertheless, alter the underlying statistics of an image. Through investigating the observation of the photomontages, we find the specific correlations in the image, which can be detected and used as evidence of digital tampering. The features derived from the proposal are useful and are proved by experimentation

Most-conventional registration algorithms use the distance sum between point pairs to estimate pose parameters. These algorithms have the problem of inexact point correspondences. This paper proposes a novel registration algorithm based on a distance metric of surface-to-surface. The integration calculation of the mean distance between surfaces is derived and reduced to a simple formula.

In order to describe mesh bit-streams which can be adapted to the heterogeneity of network bandwidth and terminals graphics capabilities, the concepts of quality scalability and resolution scalability are introduced. A triangle mesh compression scheme based on wavelet transform is proposed to generate bit-streams with both scalabilities. Each subband corresponds to a resolution level. Wavelet coefficients of each subband are bitplane encoded to enable quality refinement within this subband. The experimental results show the proposed coder has achieved better compression performance than that of PGC (progressive geometry compression).

In watermarking schemes an image is considered as a communication channel for transmitting messages. But because of the requirements of watermarking robustness and visibility, watermarking schemes have different characteristics from traditional communication. Watermarking capacity should be associated with the content of an image. A different image has a different watermarking capacity. In this paper, we present a content adaptive watermarking algorithm to embed the strongest watermark while maintaining the watermark's invisibility, and discuss watermarking capacity for this scenario. We also derive the relation between watermarking capacity and detection bit error rate (BER) for the first time.

Capacity is an important character of digital watermarking. Research of image watermarking capacity is to study how much information can be hidden in an image. Recently, wavelet transform has been applied widely in watermarking research as its excellent multi-resolution analysis property. According to watermarking methods in wavelet domain, watermark's embedding and extracting realized in wavelet domain, we think that watermarking capacity should be analyzed in wavelet domain also. This paper presents a watermarking capacity analysis method based on content of wavelet subbands by using Watson quantization matrix and Noise Visibility Function (NVF), and discusses watermarking capacity of blind watermarking and non-blind watermarking in this scenario.