The Multidimensional Knapsack Problem (MKP) is a kind of typical multi-constraint combinatorial optimization problems. In order to solve this problem, a Greedy Binary Lion Swarm Optimization (GBLSO) algorithm was proposed. Firstly, with the help of binary code transform formula, the locations of lion individuals were discretized to obtain the binary lion swarm algorithm. Secondly, the inverse moving operator was introduced to update the location of lion king and redefine the locations of the lionesses and lion cubs. Thirdly, the greedy algorithm was fully utilized to make the solution feasible, so as to enhance the local search ability and speed up the convergence. Finally, Simulations on 10 typical MKP examples were carried out to compare GBLSO algorithm with Discrete binary Particle Swarm Optimization (DPSO) algorithm and Binary Bat Algorithm (BBA). The experimental results show that GBLSO algorithm is an effective new method for solving MKP and has good convergence efficiency, high optimization accuracy and good robustness in solving MKP.

Focusing on the problem of poor memory in long sentences and the lack of core words' influence in semantic relation extraction, an Attention based bidirectional Neural Turing Machine (Ab-NTM) model was proposed. Instead of a Recurrent Neural Network (RNN), a Neural Turing Machine (NTM) was used firstly, and a Long Short-Term Memory (LSTM) network was acted as a controller, which contained larger and non-interfering storage, and it could hold longer memories than the RNN. Secondly, an attention layer was used to organize the context information on the word level so that the model could pay attention to the core words in sentences. Finally, the labels were gotten through the classifier. Experiments on the SemEval-2010 Task 8 dataset show that the proposed model outperforms most state-of-the-art methods with an 86.2% F1-score.

In Wireless Sensor Network (WSN), disjoint multipath routing can lead to the long-path problem, and braided multipath routing can lead to the weakening of fault-tolerant performance. To address these issues, a multipath braided model and a fault-tolerant routing scheme based upon the model were proposed. Firstly, the intersection of multiple paths were quantified from the source to the destination by establishing corresponding multipath braided model, and then a probability model of fault tolerance was proposed to build the relationship between path interactivity and fault tolerance. Secondly, a fault-tolerant routing scheme was designed based on local intersection adjustment. Experimental results show that, when using the proposed model and its scheme in typical multipath routing schemes—Sequential Assignment Routing (SAR) and Energy Efficient Fault-tolerant Multipath Routing (EEFTMR), the data transfer success rate can be improved effectively. In addition, it also has good performance in the network throughput and energy consumption.

Samples are required to meet the manifold assumption in Spectral Embedded Clustering (SEC) algorithm, and class labels of samples can always be embedded in a linear space, which provides a new idea for spectral clustering of linearly separable data, but the linear mapping function used by the spectral embedded clustering algorithm is not available to process the nonlinear high-dimensional data. To solve this problem, this paper cored the linear mapping function, built a Spectral Embedded Clustering based on Kernel function (KSEC) model. This model can solve the problem that the linear mapping function can't deal with nonlinear data, as well as it can achieve kernel's dimension reduction synchronously. The experimental results on real data sets show that the improved algorithm can improve the clustering accuracy by 13.11% averagely, and the highest 31.62%, especially for high-dimensional data clustering accuracy can be increased by 16.53% on average. And the sensitive experiments on algorithm to parameters show the stability of the improved algorithm, so compared with traditional spectral clustering algorithms, higher accuracy and better clustering performance are obtained. And the method can be used for such complex image processing field as remote sensing image.

The community structure exists generally in the complex network, so the community detection has important theoretical significance and practical value. In order to improve the performance of community detection in the complex network, a community detection algorithm based on structural similarity affinity propagation was proposed. Firstly, the algorithm selected structural similarity as a similarity measurement between nodes, and applied an optimized method to calculate the similarity matrix of complex networks. Secondly, the algorithm made the similarity matrix as an input, and used a Fast Affinity Propagation (FAP) algorithm to cluster. Finally, the algorithm got the final community structure. The experimental results show that in the LFR (Lancichinetti-Fortunato-Radicchi) simulated network, the average community detection Normalized Mutual Information (NMI) value of the proposed algorithm is 65.1%, which is higher than 45.3% of the Label Propagation Algorithm (LPA) and 49.8% of CNM (Clauset-Newman-Moore) algorithm. And in the real network, the average community detection modularity value of the proposed algorithm is 53.1%, which is also higher than 39.9% of the LPA and 47.8% of the CNM algorithm. The proposed algorithm has better ability of community detection, but also can find a higher quality of community structure.

Focusing on the instability of Label Propagation Algorithm (LPA), an advanced label propagation algorithm for community detection was proposed. It introduced the concept of LeaderRank score to quantify the importance of nodes, and chose some core nodes according to the node importance in descending order, then updated labels layer by layer outward centered on every core node respectively, until no node changed its label any more. Thus the instability caused by the random ranking of nodes was solved. Compared with several existing label propagation algorithms on LFR benchmark networks and real networks, both of the Normalized Mutual Information (NMI) and modularity of community detection result of the proposed algorithm were higher. The theoretical analysis and experimental results demonstrate that the proposed algorithm not only improves the stability effectively, but also increases the accuracy.

Concerning the problem of topics mining and its corresponding public emotion analysis, an analytical method for public emotion evolution was proposed based on Probabilistic Latent Semantic Analysis (PLSA) model. In order to find out the evolutional patterns of the topics, the method started with extracting the subtopics on time series by making use of PLSA model. Then, emotion feature vectors represented by emotion units and their weights which matched with the topic context were established via parsing and ontology lexicon. Next, the strength of public emotion was computed via a fine-grained dimension and the holistic public emotion of the issue. In this case, the method has a deep mining into the evolutional patterns of public emotion which were finally quantified and visualized. The advantage of the method is highlighted by introducing grammatical rules and ontology lexicon in the process of extracting emotion units, which was conducted in a fine-grained dimension to improve the accuracy of extraction. The experimental results show that this method can gain good performance on the evolutional analysis of topics and public emotion on time series and thus proves the positive effect of the method.

For the difficulty of manual annotation on large-scale micro-blog corpus, three automatic annotation methods and an integrated annotation method by voting for Chinese micro-blog corpus were proposed. Three automatic annotation methods included keywords-based annotation method, probability-summation-based annotation method and probability-product-based annotation method. During the process of automatic annotation, firstly, micro-blog corpus were annotated by three annotation methods respectively, and three results were obtained, then the final annotation results were determined by voting method with the integrated strategy. By designing automatic annotation experiment system, experimental results verify the feasibility and effectiveness of the proposed methods, and show that the accuracy of the single annotation method is more than 70%, and it is more than 90% for the voting method.

Although the normal cloud model is universally used, it faces some difficulties when describing some monotonic rise/fall conceptions. This model also has big subjective influence under multiple conditions and large computation consumption. To overcome these shortcomings, a new kind of exponential cloud model was provided along with a weight based cloud reasoning algorithm. By splitting the multi-condition generator to several single-condition generators, the algorithm firstly used Analytic Hierarchy Process (AHP) method to get weight of each property, and then used them to calculate weighted average of single-condition generator output to quantitfy value. The validation and effectiveness of this method is checked through a comparison between fuzzy reasoning and stimulation of torpedo avoid system.

Concerning that the upper, lower bounds and stride of the non-normalized loop are uncertain, some issues were normalized based on a transform method such as that loop conditions were logical expression, increment-reduction statement and do-while. An unroll-jam method was proposed to deal with the loops that cannot be normalized, which mined the unroll-jam results by Superword Level Parellelism (SLP) vectorization. Compared with the existing Single Instruction Multiple Data (SIMD) vectorization method for non-normalized loops, the experimental results show that the transform method and unroll-jam method are better to explore the parallelism of the non-normalized loops, which can improve the performance by more than 6%.

Multiple feature fusion based tracking is one of the most active research topic in tracking field, but the tracking accuracy needs improving in complex environment and most of them use single fusion rule. In this paper, a new adaptive fusion strategy was proposed for multi-feature fusion. First, the local background information was introduced to strengthen the description of the target, and then the feature weight was calculated by a variety of criteria in the fusion process. In addition, the framework of mean shift was considered to realize target tracking. An extensive number of comparative experimental results show that the proposed algorithm is more stable and robust than the single fusion rule.

The popular clustering ensemble algorithms cannot give the appropriate treatment program in the light of the different characteristics of the different data sets. A new clustering ensemble algorithm — Enhanced Clustering Ensemble algorithm based on Characteristics of Data sets (ECECD) was proposed for overcoming this defect. ECECD was composed of generation of base clustering, selection of base clustering and consensus function. It selected a special range of ensemble members to form the final ensemble and produced the final clustering based on the characteristic of the data set. Three Benchmark data sets including ecoli, leukaemia and Vehicle were clustered in the experiment, and the clustering errors gained by the proposed algorithm were 0.014, 0.489 and 0.361 respectively, which were always the minimum compared with that of the other algorithms such as Bagging based Structure Ensemble Approach (BSEA), Hybrid Cluster Ensemble (HCE) and Cluster-Oriented Ensemble Classifier (COES). The Normalized Mutual Information (NMI) values of the proposed algorithm were also always higher than that of these algorithms when increasing candidate base clusterings. Therefore, compared with these popular clustering ensemble algorithms, the proposed algorithm has the highest clustering precision and the strongest scalability.

Kernel Principal Component Analysis (KPCA) and Multi-Layer Perceptron (MLP) neural network are popular feature extraction algorithms. However, these algorithms are inefficient and easy to fall into local optimal solution. The paper proposed a new feature extraction algorithm — margin maximizing hyperplanes based Enhanced Feature Extraction algorithm （EFE）, which can overcome the problem of KPCA and MLP algorithm. The proposed EFE algorithm, whcih maps the input samples to the subspace spanned by the normals of hyperplanes through adopting the pairwise orthogonal margin maximizing hyperplanes, is independent of the probability distribution of the input samples. The results of these feature extraction experiments on real world data set — wine and AR show that FE algorithm is beyond KPCA and MLP in terms of the efficiency of the implementation and accuracy of recognition.

Concerning the premature convergence of invasive weed optimization algorithm, a new invasive weed optimization with multi-population based on chaotic sequence (CMIWO) was proposed. Firstly, chaotic sequence was adopted to initialize population at the initialization of algorithm, which improved the quality of the initial solution. Secondly, threshold was used to estimate the cluster degree of individuals in iterations and if cluster degree was less than threshold, initializing population with chaotic sequence was implemented again, thus the algorithm could effectively jump out of local minima. Thirdly, the weed population was divided into five groups to collaborate so as to discourage premature convergence, thus improving the algorithm's precision and increasing the convergence speed. In the end, the test results on eight test functions show that the proposed algorithm improves the accuracy by 25% to 300% than basic algorithm in terms of optimal value and 50% to 100% for standard deviation.

In order to improving the network quality of service by mining useful model from multi-source and complicated network performance data, a clustering analysis algorithm for network performance monitoring data based on ontology. The semantic description method of network performance monitoring data is described, then a similarity measurement model of network performance data based on semantic description and property data is proposed, and an NJW spectral clustering algorithm based on improved k-means algorithm is given. The experiments based on the UCI data sets and the performance monitoring data on a campus network shows that the proposed algorithm has a higher clustering accuracy and differentiation than the comparative algorithms.

The construction of digital fingerprint embedding and acquisition scheme for anti-rotation, anti-scaling and anti-distortion attack can improve the anti-attack capability of the digital fingerprint. The designed spatial-DCT (Discrete Cosine Transform) domain combinational embedding scheme of digital fingerprint provided the construction of differential characteristic point, on which the digital fingerprint embedding and acquisition algorithm was proposed. And an attack parameter recognition algorithm with high accuracy was presented. The simulation results show that the accuracy of attack recognition algorithm can be the order of sub-pixel and can resist the scaling attack with parameter larger than 0.5, any rotation attack with angle less than 45° and any distortion attack with angle less than 25°. In addition, the effect of the proposed scheme would not decrease with the increase of the rotation and distortion angle. The proposed scheme improves the robustness of the digital fingerprint and enables the digital fingerprint system to resist the removal and CTP (cutting, trimming, pasting) attack as well as RSD (rotation, scaling, distoration) attack.

In literature, there are very few discussions on the change of performance of source camera classification algorithms when test images are subjected to minor image processing. Using Support Vector Machines (SVM), this paper analyzed the performance and robustness of source camera classification algorithms. It compared the detection accuracy for unprocessed images with that for processed images, and investigated the robustness of different types of image features. Since pattern classification-based algorithms often need to reduce the number of image features for computational efficiency, this paper also discussed the performance of camera classification algorithms using the image feature subsets. The impact of using these subsets on the robustness of camera classification algorithms was explored as well.

For the weighted average method used in multi-radar track fusion, the influence of weighting factor allocation on the accuracy of the fused track was analyzed, and an algorithm of weighting factor dynamic allocation was presented. The output track accuracy of each radar was estimated online by exploiting the same target's track information coming from multi-radar system. The dynamic allocation of the weighting factors was realized and the accuracy of the fused track was improved. The results of the simulation experiment show that this algorithm is really effective.

Spatial index is a key issue in massive spatial data processing. This paper improved the multi-layers grid by analyzing the grid files. Some creative theories and relevant algorithms were put forward such as first layer grid auto-selection algorithm based on normal distribute and new grid-contain algorithm. This paper analyzed the performance of the improved multi-layer grid spatial index by real data test. Test results show that in most case the creative theories improve the performance and adjustability of index.