Aiming at the existing problems in user trajectory data modeling such as the sparsity of check-in points, long-term dependencies and complex moving patterns, a social network user trajectory prediction model based on self-supervised learning, called SeNext, was proposed to model and train the user trajectory to predict the next Point Of Interest (POI) of the user. First, data augmentation was utilized to expand the training trajectory samples, which solved the problem of the deficiency of model generalization capability caused by insufficient data and too few footprints of some users. Second, Recurrent Neural Network (RNN), Convolutional Neural Network (CNN) and attention mechanism were adopted into the modeling of current and historical trajectories respectively, so as to extract effective representations from high-dimensional sparse data to match the most similar moving patterns of users in the past. Finally, SeNext learned the implicit representations in the latent space by combining self-supervised learning and introducing contrastive loss Noise Contrastive Estimation (InfoNCE) to predict the next POI of the user. Experimental results show that compared to the state-of-the-artVariational Attention based Next (VANext)model, SeNext improves the prediction accuracy about 11% on Top@1.

Traditional Fuzzy C-Means (FCM) clustering algorithm only considers the characteristics of a single pixel when applied to liver CT image segmentation, and it can not overcome the influence of uneven gray scale and the problem of boundary leakage caused by blurred liver boundary. In order to solve the problems, a Spatial Fuzzy C-Means (SFCM) clustering segmentation algorithm combined with spatial constraints was proposed. Firstly, the convolution kernel was constructed by using two-dimensional Gauss distribution function, and the feature matrix could be obtained by using the convolution kernel to extract the spatial information of the source image. Then, the penalty term of spatial constraint was introduced to update and optimize the objective function to obtain a new iteration equation. Finally, the liver CT image was segmented by using the new algorithm. As shown in results, the shape of liver contour splited by SFCM is more regular when segmenting liver CT images with gray unevenness and boundary leakage. The accuracy of SFCM reaches 92.8%, which is 2.3 and 4.3 percentage points higher than that of FCM and Intuitionistic Fuzzy C-Means (IFCM). Also, over-segmentation rate of SFCM is 4.9 and 5.3 percentage points lower than that of FCM and IFCM.

Aiming at the low classification accuracy problem of large-scale Support Vector Machine (SVM) after random-projection-based feature dimensionality reduction, Linear kernel SVM based on dual random projection (drp-LSVM) for large-scale classification problems was proposed with the introduction of the dual recovery theory. Firstly, the relevant geometric properties of drp-LSVM were analyzed and demonstrated. It's proved that, with maintaining the similar geometric advantages of Linear kernel SVM based on dual random projection (rp-LSVM), the divided hyperplane of drp-LSVM was more close to the primitive classifier trained by complete data. Then, in view of the fast solution to drp-LSVM, the traditional Sequential Minimal Optimization (SMO) algorithm was improved and the drp-LSVM classifier based on improved SMO algorithm was completed. Finally, the experimental results show that, drp-LSVM inherits the advantages of rp-LSVM, reduces classification error, improves training accuracy, and all its performance indexes are more close to the classifier trained by primitive data; the classifier designed based on the improved SMO algorithm can reduce memory consumption and achieve higher training accuracy.

Aiming at the problem that the traditional topic model ignore short text and dynamic evolution of microblog, a Biterm Topic over Time (BToT) model based on microblog text was proposed, and the subject evolution analysis was carried out by the proposed model. A continuous time variable was introduced to describe the dynamic evolution of the topic in the time dimension during the process of text generation in the BToT model, and the "Biterm" structure of the topic sharing in the document was formed to extend short text feature. The Gibbs sampling method was used to estimate the parameters of BToT, and the topic evaluation was analyzed by topic-time distributed parameters. The experimental results on real microblog datasets show that BToT can characterize the latent topic evolution and has lower perplexity than Latent Dirichlet Allocation (LDA), Biterm Topic Model (BTM) and Topic over Time (ToT).

Concerning the problem that too many transitional links and complex parameter solving process existed in camera calibration of the surgical navigation based on C-arm, a new method that completely ignored the camera model was proposed. In this method, the camera model was completely ignored, and the transition link in the process of solving mapping parameters was simplified, which increased the efficiency. In addition, the camera calibration was achieved by distinguishing the projection data from the calibration target which has double-layer metal ball. In the calibration point verification experiment, it can be proved that the residual error of each test point was no more than 0.002 pixels; in the navigation validation experiment, probe point and perforation test were successfully implemented with the established preliminary experiment platform. The experimental results verify that the proposed camera calibration method can meet the accuracy requirements of surgical navigation system.

Boolean function plays an important role in the design and analysis of symmetric cryptography. Firstly, by studying the balanced property of subfunctions of the disjoint spectra function set, some sufficient conditions were provided that there were three balanced Boolean functions in the set of four plateaued functions. Then, based on three balanced disjoint spectra plateaued functions, a special Boolean permutation and a balanced Boolean function with high nonlinearity, a method of constructing balanced Boolean functions with high nonlinearity was proposed on a small number of variables. The analysis results show that the proposed method can construct the 2 k-variable balanced Boolean functions with the optimal algebraic number and the nonlinearity is not less than 2 ^2k-1-2 ^k-1-2 ^k/2-2 ^{⌈(k-1)/2⌉}.

Focusing on the issue that high resolution Synthetic Aperture Radar (SAR) image is influenced by speckle noise and road environment is complex, an automatic road extraction method based on fuzzy connectedness was proposed. Firstly, a speckle filtering process was employed to SAR images to reduce the influence of speckle noise. Then seed points were extracted automatically by combining the results of Ratio of Exponentially Weighted Averages (ROEWA) detector and Fuzzy C-Means (FCM) clustering method. Finally, the roads were extracted by using fuzzy connectedness method which characterized by gray level and the edge intensity, and a morphology operation was done to optimize the final result. Comparison experiments between FCM based road extraction method and the proposed method were performed on two SAR images, the detection completeness, correctness and quality of the proposed method were better than those of FCM based road extraction method. The experimental results show that the proposed approach can effectively extract roads from high resolution SAR images without inputting seed points manually.

For the shortage of big data and incremental data processing in static algorithm, the dynamic rule extraction algorithm based on rough-decision tree was constructed to diagnose rotating machinery faults. Through the combination of rough set with decision tree, the sample selections were made by the method of incremental sampling. Through dynamic reduction, decision tree construction, rules extraction and selection, matching, four steps of loop iteration process, dynamic rule extraction was achieved, which improved the credibility of the extracted rules. Meanwhile, by applying the algorithm to the dynamic problem: rotating machinery fault diagnosis, the effectiveness of the algorithm was verified. Finally, the efficiency of the algorithm was compared with static algorithm and incremental dynamic algorithm. The result demonstrates that the proposed algorithm can obtain more implied information in the most streamlined way.

The virtual machines in cloud computing platform exchange data in the shared memory of physical machine. In view of the problem that the traffic cannot be captured and detected in firewall or other security components, the OpenFlow technology was analyzed, and a traffic redirection method based on OpenFlow was presented. To control traffic forwarding process and redirect it to security components, the method provided network connection for virtual machines with OpenFlow controller and virtual switches instead of physical switches, and built a traffic detection system composed of four modules including virtual switch, control unit, intrusion detection and system configuration management. The experimental results show that the proposed scheme can realize traffic redirection and the subsequent detection processing, and the system can provide switch-level and host-level control granularity. It also solves traffic detection problem under cloud computing environment in traditional scene by traffic redirection, and provides great expansion of the traffic processing based on OpenFlow.

An index of network evolution speed and a network evolution model were put forward to analyze the effects of network evolution speed on propagation. The definition of temporal correlation coefficient was modified to characterize the speed of the network evolution; meanwhile, a non-Markov model of temporal networks was proposed. For every active node at a time step, a random node from network was selected with probability r, while a random node from former neighbors of the active node was selected with probability 1-r. Edges were created between the active node and its corresponding selected nodes. The simulation results confirm that there is a monotone increasing relationship between the network model parameter r and the network evolution speed; meanwhile, the greater the value of r, the greater the scope of the spread on network becomes. These mean that the temporal networks with high evolution speed are conducive to the spread on networks. More specifically, the rapidly changing network topology is conducive to the rapid spread of information, but not conducive to the suppression of virus propagation.

In order to avoid lots of holes among mature immune detectors and deal with the problem of boundary invasion in intrusion detection, analyzing the relationship between number of detectors and detection performance, a detector distribution optimization algorithm with Monte Carlo estimation was proposed: evaluating the coverage of detectors by the Monte Carlo method, and updating the detector set by the offspring to improve detectors' distribution. The experimental tests demonstrate that the algorithm can not only decrease the holes but also achieve a more precise coverage of the nonself space with fewer detectors, and increase the detector's detection performance.

In order to get rid of the noise of flight data more effectively, based on discussing the principle of Stationary Wavelet Transform (SWT), a new denoising method was proposed, which combined correlation of wavelet coefficient with wavelet shrinkage. Firstly, signals were decomposed by using SWT; secondly, wavelet coefficient was dealt with by using methods of coefficient correlation and wavelet shrinkage in sequence; at last, denoised signal was reconstructed through inverse wavelet transform. The results of experiments show that the proposed method can raise Signal-to-Noise Ratio (SNR), decrease Mean Squared Error (MSE) and preserve the shape of signal; and it can be applied to flight data effectively.

An algorithm of the extracting Lexicalized Tree Adjoining Grammar(LTAG) from Penn Chinese corpus was presented. Idea of the algorithm is to induce three kinds of trees from lexicalized tree bank. Then the method of Head-driven Phrase Structure Grammar(HPSG) was applied to extract lexicalized tree from corpus. In the end,invalid lexicalized trees were filtered out by linguistic rules. It requires fewer human efforts compared with hand-crafted grammar. It is possible to remedy omission of grammatical syntactic structures in hand-crafted grammar.