Computational Intelligence and Intelligent Systems

Group search optimizer (GSO) is a recently developed heuristic inspired by biological group search resources behavior. However, it still has some defects such as slow convergence speed and poor accuracy of solution. In order to improve the performance of GSO in solving complex optimization problems, an opposition-based learning approach (OBL) and a differential evolution method (DE) are integrated into GSO to form a hybrid GSO. In this paper, the strategy of OBL is used to enlarge the search region, and the operator of DE is utilized to enhance local search to improve. Comparison experiments have demonstrated that our hybrid GSO algorithm performed advantages over previous GSO and DE approaches in convergence speed and accuracy of solution.

Firefly algorithm (FA) is a recently proposed swarm intelligence optimization technique, which has shown good performance on many optimization problems. In the standard FA and its most variants, a firefly moves to other brighter fireflies. If the current firefly is brighter than another one, the current one will not be conducted any search. In this paper, we propose a new firefly algorithm (called NFA) to address this issue. In NFA, brighter fireflies can move to other positions based on local search. To verify the performance of NFA, thirteen classical benchmark functions are tested. Experimental results show that our NFA outperforms the standard FA and two other modified FAs.

The automatic detection and recognition of automotive wheel hubs defects has important significance to improve the quality and efficiency of automotive wheel production and vehicle safety. In order to improve accuracy of detection and recognition of automotive wheel hub defect images, an improved peak location algorithm - trend peak algorithm is proposed to extract region of wheel hub defect, combined with BP neural network to classify and recognize wheel hub defect. Firstly, initial defect positions are extracted using peak locations of vertical and horizontal directions. Then mathematical morphology is used to remove pseudo defects, and the exact locations of the defects are obtained. Finally, the wheel hub defect features are classified to reach the target of defect recognition by BP neural network. In actual industrial conditions, the algorithm is found to obtain good recognition results and reach real-time detection request in low contrast, high noise, uneven illumination, and complex structure of the products, by experiments of X-ray images of four common defects of the actual wheel hubs.

When the traditional genetic algorithm was used to solve the community detection problem, it was not easy to avoid the problems of low efficiency and slow convergent speed. To be aim at these problems, a improved genetic algorithm which is based on the immune mechanism was proposed in this paper. In this new algorithm, the immune mechanism was used to ensure the diversity of population. Meanwhile, a improved character encoding was adopted to further reduce the search space. The results shows that the shortcomings of slow convergent speed and low efficiency could be overcome by using the improved genetic algorithm to solve these problems, compared with the traditional genetic algorithm.

Based on margin sampling (MS) strategy, an active learning approach was introduced for proposed sample selection from large quantities of labeled samples using a Landsat-7 ETM+ image to solve remote sensing image classification problems for large number of training samples. As a breakthrough from conventional random sampling and stratified systematic sampling methods, this approach ensures classification of only using a few hundred training samples to be as effective as that of using several thousand and even tens of thousands of samples by conventional methods, thereby avoiding enormous calculations, substantially reducing operating time and improving training efficiency. The test results of the proposed approach was compared with those of random sampling and stratified systematic sampling, and the effects of training samples on classification under optimized and non-optimized selection conditions was analyzed.

Wireless sensor network consists of a large number of tiny sensor nodes owned capable of perception in monitoring region by self-organized wireless communication, has been widely applied in military and civil fields. From the perspective of resource- saving, under the condition of the network’s connectivity and specific coverage, the number of sensor nodes is assumed to be opened as few as possible. So, computing the sensor nodes collection which meeting the requirements is called the problem of network coverage optimization for Wireless Sensor Network; also called the problem of minimum connected covering node set. The innovation point of the article is: Firstly, it analyzed the deficiencies of traditional evolution algorithm fitness function, put forward an improved fitness function design scheme, and has been proved that it has advantage of solving problem on wireless sensor networks coverage optimization; Secondly, it applied the method of control variables, comparison and analysis of the influence on the various operations and parameters selection in evolution algorithm on the optimization results and performance, and then point out how to design algorithm to manage to the best optimize effect and performance.

This paper proposes a dynamic constrained many-objective optimization method for solving constrained optimization problems. We first convert a constrained optimization problem (COP) into an equivalent dynamic constrained many-objective optimization problem (DCMOP), then present many-objective optimization evolutionary algorithm with dynamic constraint handling mechanism, called MaDC, to solve the DCMOP, thus the COP is addressed. MaDC uses DE as the search engine, and reference-point-based nondominated sorting approach to select individuals to construct next population. The effectiveness of MaDC has been verified by comparing with peer algorithms.

Task scheduling is one of the basic problem on cloud computing. In hybrid cloud, tasks scheduling faces new challenges. In order to better deal the multi-objective task scheduling optimization in hybrid clouds, on the basis of the GaDE and Pareto optimum of quick sorting method, we present a multi-objective algorithm, named NSjDE. This algorithm also makes considerations to reduce the frequency of evaluation Comparing with experiment of Min-Min algorithm, GaDE algorithm and NSjDE algorithm, results show that for the single object task scheduling, GaDE and NsjDE algorithms perform better in getting the approximate optimal solution. The optimization speed of multi-objective NSjDE algorithm is faster than the single-objective jDE algorithm, and NSjDE can produce more than one non-dominated solution meeting the requirements, in order to provide more options to the user.

To overcome the problem of slow convergence and easy to be plunged to premature when the traditional differential evolution algorithm for solving constrained optimization problems, a novel differential evolution algorithm (CO-JADE) based on adaptive differential evolution (JADE) for constrained optimization was proposed. The algorithm used skew tent chaotic mapping to initialize the population, generated the crossover probability of each individual according to the normal distribution and the Cauchy distribution and the mutation factor according to the normal distribution. CO-JADE used improved adaptive tradeoff model to evaluate the individuals of population. The improved adaptive tradeoff model used different treatment scheme for different stages of population, which aimed to effectively weigh the relationship between the value of the objective function and the degree of constraint violation. Simulation experiments were conducted on the night standard test functions. CO-JADE was much better than COEA/ODE and HCOEA in terms of the accuracy and standard variance of final solution. The experimental results demonstrate that the CO-JADE has better accuracy and stability.

Ant colony optimization algorithms have been successfully applied in classification rule mining, but in general, the basic ant colony classification mining algorithms have the problems of premature convergence, easily falling into local optimum, and etc. In this paper, a new ant colony classification mining algorithm based on pheromone attraction and exclusion (Ant-Mine_rPAE) is proposed, where the new pheromone calculation method is designed and the search is guided by the new probability transfer formula. Our experiments using 12 publicly available data sets show that the predictive accuracy obtained by the Ant-Miner_PAE algorithm is statistically significantly higher than the predictive accuracy of other rule induction classification algorithms, such as CN2, C4.5rules, PSO/ACO2, Ant-Miner, _CAnt-Miner_PB, and the rules discovered by the Ant-Miner_PAE algorithm are considerably simpler than those discovered by the counterparts.

As an appendix which is designed to embed in one of the complete swarm intelligence algorithms, the novel strategy, named dynamic-search-spaces (DS) is proposed to deal with the premature convergence of those algorithms. For realizing the decrement of search space, the differences or the distances between individual sites and the site of the global performance are to form the threshold of the self-adaption system. Once the value reached by calculating the quotient of sum of those sitting near the global performance and others over a stated percentage, the system is working to readjust the borders of search space by the site of the global performance. After each readjustment, the re-initialize to distribute individuals in the whole search space should be achieved to enhance individuals’ vitality which prove away from the premature convergence. Meanwhile, the simpler verifications are provided. The improvements of results are exhibited embedding in the genetic algorithm, the particle swarm optimization and the differential evolution. This dynamic search space scheme can be embedded in most of swarm intelligence algorithms easily abstract environment.

To solve the problem of premature convergence in traditional particle swarm optimization (PSO), This paper proposed a adaptive mutation opposition-based particle swarm optimization (AMOPSO). The new algorithm applies adaptive mutation selection strategy (AMS) on the basis of generalized opposition-based learning method (GOBL) and a nonlinear inertia weight (AW). GOBL strategy can provide more chances to find solutions by space transformation search and thus enhance the global exploitation ability of PSO. However, it will increase likelihood of being trapped into local optimum. In order to avoid above problem, AMS is presented to disturb the current global optimal particle and adaptively gain mutation position. This strategy is helpful to improve the exploration ability of PSO and make the algorithm more smoothly fast convergence to the global optimal solution. In order to further balance the contradiction between exploration and exploitation during its iteration process, AW strategy is introduced. Through compared with several opposition-based PSOs on 14 benchmark functions, the experimental results show that AMOPSO greatly enhance the performance of PSO in terms of solution accuracy, convergence speed and algorithm reliability.

While the ACO can find the optimal path of network, there are too many iterative times and too slow the convergence speed is also very slow. This paper proposes the Q-ACO QoSR based on convergence expectation with the real-time and the high efficiency of network. This algorithm defines index expectation function of link, and proposes convergence expectation and convergence grads. This algorithm can find the optimal path by comparing the convergence grads in a faster and bigger probability. This algorithm improves the ability of routing and convergence speed.

In this paper, a new Gene Expression Programming (GEP) algorithm is proposed, which increase “inverted series” and “extract” operator. The new algorithm can effectively increase the rate of utilization of genes, with convergence speed and solution precision is higher. Taking the Chinese vegetables price change trend of mooli, scallion as example, and discuss the way to solve the forecasting modeling problem by adopting GEP. The experimental results show that the new GEP Algorithm can not only increase the diversity of population but overcome the shortage of primitive GEP. In addition, it can improve convergence accuracy compared to original GEP.

How to find the better initial center points plays an important role in many clustering applications. In our paper, we propose the novel chromosome representation according to extended traditional gene expression programming used in GEP-ADF. It is aimed at improving the performance of GEP to obtain center points more accurately. Experimental results show that our new algorithm has good performance in clustering and the three real world datasets compared with the other two algorithms.

Acute Hypotensive Episodes (AHE) is one of the hemodynamic instabilities with high mortality rate that is common among patients. Timely and rapid intervention is necessary to save patient’s life. This paper presents a methodology to predict AHE for ICU patients based on the Multi Genetic Programming (Multi GP). The methodology is applied to the dataset obtained from Multi-parameter Intelligent Monitoring for Intensive Care (MIMIC-II). The achieved accuracy of the proposed methodology is 79.07 % in the training set and 77.98 % in the testing set with the five-fold cross-validation.

With the dramatic increase of circuit scale and the harsh environment, the reliability of the system has become the great hidden danger. Triple different-structure modular redundant system based on evolution mechanism shows good fault tolerant ability. How to enhance the efficiency and diversity of the evolution generation module has become the key issue which can ensure the system fault tolerant. This article puts forward two-stage mutation evolution strategy (TMES) and interactive two-stage mutation evolution strategy (ITMES) based on improving virtual reconfigurable architecture platform to evolve combination logical circuit on the fault-tolerant system with different-structure redundancy module. The efficiency of the proposed methodology is tested with the evolutions of a 2-bit multipliers, and a 3-bit multipliers, and a 3-bit full adders. The obtained results demonstrate the effectiveness of the scheme on generation circuit diversity and evolution efficiency.

In this paper we have proposed human actions recognition methodology. The main novelty of this paper is application of neural network (NN) trained with the parallel stochastic gradient descent to perform classification task on multi-dimensional time-varying signal. The original motion-capture data consisted of 20 time-varying three-dimensional body joint coordinates acquired with Kinect controller is preprocessed to 9-dimensional angle-based time-varying features set. The data is resampled to the uniform length with cubic spline interpolation after which each action is represented by 60 samples and eventually 540 (60 × 9) variables are presented to input layer of NN. The dataset we used in our experiment consists of recordings for 14 participants that perform nine types of popular gym exercises (totally 770 actions samples). The averaged recognition rate in k-fold cross validation for different actions classes were between 95.6 % ± 9.5 % to even 100 %.

After analyzing the four aspects of cloud computing, storage, network, infrastructure, and combining the fuzzy evaluation method in fuzzy mathematics theory, the quantitative index and qualitative index are combined to realize the virtual resource performance evaluation. This method breaks through the limitation of the previous evaluation system, and establishes a performance evaluation system based on multi-level fuzzy evaluation. Through the analysis of a case, it is proved that the proposed method can evaluate the resource performance more comprehensively, This will have a certain application value and significance for the research of resource management and scheduling based on cloud computing.

In Bayesian Optimization Algorithm (BOA), to accurately build the best Bayesian network with respect to most metrics is NP-complete. This paper proposes an improved BOA based on incremental model building, which learns Bayesian network structure using PBIL instead of greedy algorithm in BOA. The PBIL is effective to learn better Bayesian network. The simulation results also show that the improved BOA has the better performance than BOA.

In DBOA, to build accurately the best Bayesian network with respect to most metrics is NP-complete and the high time complexity of learning the model structure becomes a bottleneck of DBOA for real application. Consequently, in order to decrease the asymptotic time complexity of model building and make the algorithm more practical even for extremely large and complex problem, this paper presents adaptive sporadic model building based on estimation of model similarity as an efficiency enhancement technique of DBOA. The results show that performing the adaptive model building in DBOA can reduce the number of building model under no increasing on the number of generation and population size necessary to converge to optimal solutions, and achieve a better trade-off between the convergence speed and convergence results.

Person re-identification has become one of the most important problems in video surveillance system. In a multi-camera video surveillance system with non-overlapped area, the appearances of one person are much difference according different cameras or in the same camera at different times. On the other hand, different person may appears similar in one camera, so which made person re-identification a challenging problem. This paper carried out a person re-identification algorithm based on part feature specificity. This algorithm extract color, texture and shape features of different part of body first, then gather statistic specificity weight of these features for each part. At last, doing feature weighting both part weight and feature specificity in distance calculating, which make features with strong specificity more important. This algorithm indicates some parts of body are more important than others in re-identification, and the same part from different people with different appearance, the features with strong specificity are more important than the others. We have done our experiments at public datasets VIPeR and iLIDS, and evaluate the result by CMC. Result indicates this algorithm has higher re-identification rate, and more robust to viewing condition changes, illumination variations, background clutter and occlusion.

In many expensive or time consuming engineering problems, like antenna design problems, it is unpractical to use the evolutionary algorithms directly. In recent years, Gaussian process has attracted more and more attention and had some successful applications. To further accelerate the speed of antenna design optimization process, a Gaussian process and fuzzy clustering assisted differential evolution algorithm(GPFCDEA) is presented in this paper. Four benchmark functions and two antenna design problems are selected as examples. Experimental results indicate that GPFCDEA performs much better than DE for low dimensions problems. However, for high dimensions problems, the performance of GPFCDEA still needs further research.

One of the main purposes of watermark preprocessing is to improve the robustness and security of the watermark. In this paper we proposed an improved image encryption algorithm, which combines position scrambling and gray scrambling. In order to achieve image location scrambling, first of all, the image is divided to even blocks, and the sub-blocks are scrambled according to Arnold transform. Then all of the pixels of each sub-block are scrambled by the proposed algorithm based on Logistic chaotic map. Finally, all of the Pixels are redistributed and scrambled totally. Basing on image location scrambling, it makes use of Logistic chaotic map and multi-dimensional Arnold transformation, image gray scrambling is attained. By histogram analysis, key sensitivity analysis and correlation analysis of adjacent pixels of the results of the simulation, indicating that the scrambling effect of the improved algorithm is good, and the key space is more larger.

We developed an agent-based model containing 50 communities, replicating the 50 states of USA. The age distribution, approximate household size and the socio-structural determinants of each community were modeled based on the US Census. The agent-based model was validated using in-silico seroprevalence data collection. Medical seeking behavior of individuals was parameterized based on the socio-structural determinants of the community. The interventions proposed in literature were tested and the optimal intervention strategy to counter an epidemic outbreak has been identified. In addition, we included novel interventions like coordination among the communities and increasing the awareness of individuals in the lower ranked communities based on information exchange between communities.

Based on the analysis and comparison of existing numerical mode-matching method (NMM), an improved NMM is proposed in this paper. A new type of recursive formulas is derived by setting the proper positions for the unknown variables, and recursive formulas are unified to one form for the layers above and under source, avoiding the disadvantage of the traditional NMM to deduce the recursive formulas for the layers above and under source respectively. It is easy to understand and has concise physical meaning. But the incremental factor in the recursive formulas can’t be eliminated thoroughly, which may affect the calculation accuracy and stability. Finally, the merits and demerits of the improved method and the traditional one are summarizes.

In order to learn a given data set, a learning system often has to learn too much on some data points in the given data set in order to learn well the rest of the given data. Such unnecessary learning might lead to both the higher complexity and overfitting in the learning system. In order to control the complexity of neural network ensembles, difference learning is introduced into negative correlation learning. The idea of difference learning is to let each individual in an ensemble learn to be different to the ensemble on some selected data points when the outputs of the ensemble are too close to the target values of these data points. It has been found that such difference learning could control not only overfitting in an ensemble, but also weakness among the individuals in the ensemble. Experimental results were conducted to show how such difference learning could create rather weak learners in negative correlation learning.

To feature points description of color image, the fact that image color information has some effects on features of image is taken into consideration in this paper. And there is a novel method of SURF (speed up robust features) feature description of color image based on Gaussian color invariance model presented in this paper. During the stage of image feature description, the three kinds of color information in original color image are expressed by three components of Gaussian color invariance model respectively. Then, the matrix consisting of color invariants which are presented by Gaussian color invariance model represents original color image. Hereafter, the method of SURF feature description is used for describing the distribution of feature points. Finally, through our experiences, the correct matching ratio of feature point pairs of our method is higher than some typical algorithms represented in resent years when the image appears affine and blurring transformation.

Motion estimation have a significant impact in H.265 video coding systems because it occupies a large amount of time in encoding. So the quality of motion search algorithm affect the entire encoding efficiency directly. In this paper, a novel search algorithm which utilizes an adaptive hexagon and small diamond search is proposed to overcome the drawbacks of the traditional block matching algorithm implemented in the most current video coding standards. The adaptive search pattern is chosen according to the motion strength of the current block. When the block is in active motion, the hexagon search provides an efficient search means; when the block is inactive, the small diamond search is adopted. Simulation results showed that our approach can speed up the search process with little effect on distortion performance compared with other adaptive approaches.

For the load imbalance of resource scheduling, an algorithm based on improved genetic algorithm is proposed after the research of resource load scheduling model based on Cloud Computing. The algorithm designed the fitness function, which uses the spatial utilization rate, load changes and the weight, selected individual by the Roulette Wheel Method, and optimized the crossover and mutation operations. Experiment results demonstrate that the algorithm not only can accelerate convergence of load balance scheme, but also has less migration time. It provides a new solution for the research of load balance and virtual Machine Resource Scheduling Strategy.

To the problem of scheduling multiple DAG workflow applications with multiple priorities submitted at different times in cloud computing environment, a novel workflow scheduling algorithm based on reinforcement learning is proposed in this paper. In the workflow scheduling scheme, the number of VMs in resources pool is defined as state space; the runtime of user task is defined as immediate reward, and then interactive with cloud computing environment to obtain the optimization policy. We use real cloud workflow to test the proposed scheme. Experiment results show the proposed scheme not only can solve the fairness of scheduling multiple DAGs with the same priority level submitted at different times, but also can ensure that the execution of the DAGs with higher priorities cannot be influenced by the DAGs with lower priorities. More importantly, the proposed scheme can reasonably schedule multiple DAGs with multiple priorities and improve utilization rate of resources better.

In this paper we presented CK-means clustering algorithm based on improved K-means algorithm and the Canopy algorithm, which uses MapReduce programming model of Hadoop platform. The experimental results prove that the CK-means algorithm has a good advantage in the processing of large data sets, in the acceleration ratio, accuracy, expansion rate, and the effect of the algorithm after deploying on the Hadoop clusters.

Spatial data integration is the direction of the spatial data management and use. Because of the correlation between spatial data, the data redundancy and inconsistency appear, that is an important question in data integration. As a method based on user demand and data production, geographic service provides a method to solve the problem of data redundancy and inconsistency. The paper discusses the theory and method of the integration of spatial data based on geographic service and gives an example of MODIS vegetation index, showing the feasibility and superiority of this method.

This paper focuses on grounding prediction related to sea vessels. Grounding accidents are one of the most common causes for ship disasters. Hence, there is a growing need to assess and analyze probabilities as well as related consequences of ship running aground. Using a real world marine incident dataset obtained from the United States Coast Guard National Response Center, we have demonstrated that Support Vector Data Description based methods can be successfully used for grounding prediction. After preprocessing the raw data, a total of 15165 incidents were obtained out of which there were 291 cases of ship running aground and was used in our study. A prediction accuracy of 98.25 % was achieved using the Lightly Trained Support Vector Data Description.

The Van Genuchten Equation (VGE) is used to describe the characteristic of soil water movement, but it is super-set, nonlinear and containing many unknown parameters. Using the traditional method to estimate the parameters of VGE often results in a high margin of error because of complication. The teaching-learning-based optimization (TLBO) is a new swarm intelligent optimization method for solving complex nonlinear models. In this paper, the solution program of TLBO is compiled and used to estimate parameters of the VGE. The results show that the estimate method by TLBO is more efficient and accurate. Consequently, TLBO can be used as a new method to estimate parameters of VGE.

In the near future the number of equipment connected to the Internet will greatly increase, so that further development of applications meant to verify their operations will be required. Monitoring represents an important factor in improving the quality of the services provided in cloud computing, given the fact that it allows scaling resource utilization in an adaptive manner. This paper aims to provide a solution for the monitoring of network devices and services, allowing administrator to verify connectivity of the equipment, their performances and network security. The main contribution of the paper consists in proposing an integrated solution that is deployed in the cloud for monitoring all the network components. Finally, the paper discusses the main findings and advantages for a reference implementation of the monitoring system using a simulated network.

An efficient multicast communication is crucial for many parallel high-performance scientific (e.g., genomic) applications involving a large number of computing machines, a considerable amount of data to be processed and a wide set of users providing inputs and/or interested in the results. Most of these applications are also characterized by strong requirements for the reliability and timely data sharing since involved in providing decision support for critical activities, such as genomic medicine. In the current literature, reliable multicast is always achieved at the expenses of violations of temporal constraints, since retransmissions are used to recover lost messages. In this paper, we present a solution to apply a proper coding scheme so as to jointly achieved reliability and timeliness when multicasting over the Internet. Such a solution employs game theory so as to select the best locations within the multicast tree where to perform coding operations. We prove the quality of this solution by using a series of simulations run on OMNET++.

This paper proposes a set of SSLOKmeans algorithm that helps to guide the clustering before using tag memory resident, this algorithm can further improve the large-scale data sets clustering efficiency and clustering results of quality.

The multi-objective evolutionary algorithm based on decomposition (MOEA/D) has a remarkable overall performance for multi-objective optimization problems, but still consumes much time when solving complicated problems. A parallel MOEA/D (pMOEA/D) is proposed to solve bi-objective optimization problems on message-passing clusters more efficiently in this paper. The population is partitioned evenly over processors on a cluster by a partitioned island model. Besides, the sub-populations cooperate among separate processors on the cluster by the hybrid migration of both elitist individuals and utopian points. Experimental results on five bi-objective benchmark problems demonstrate that pMOEA/D achieves the satisfactory overall performance in terms of both speedup and quality of solutions on message-passing clusters.

Multi-label classification is to assign an instance to multiple classes. Naive Bayes (NB) is one of the most popular algorithms for pattern recognition and classification. It has a high performance in single label classification. It is naturally extended for multi-label classification under the assumption of label independence. As we know, NB is based on a simple but unrealistic assumption that attributes are conditionally independent given the class. Therefore, a double weighted NB (DWNB) is proposed to demonstrate the influences of predicting different labels based on different attributes. Our DWNB utilizes the niching cultural algorithm to determine the weight configuration automatically. Our experimental results show that our proposed DWNB outperforms NB and its extensions significantly in multi-label classification.

With the development of database research, keyword search on big data graph have attracted many attentions and becoming a hot topic. However, most of existing works are studied on CPU. An important problem is efficiently generating answers for keyword search. In this paper, we research an method of keyword search under graphical processing unit. An improved algorithm based on interval coding is proposed. It includes two main tasks, which are finding root nodes and getting shortest paths from root to keyword nodes. To find root nodes quickly, we judge the reachability between any two nodes based on interval assigned to every node. To speed up finding root nodes and getting shortest paths from root to keyword nodes, we provide data parallel processing for compute-intensive tasks based on intervals assigned to every node and Floyd-Warshall algorithm. Experiment results show the high performance of the proposed solution both on CPU and graphical processing unit.

Scientific and effective risk control is a core part of the implementation of agriculture-related loans business for micro-credit companies. In this paper, a rural micro-credit decision model is presented based on maximizing the expected rate of return while reducing the investment convergence. Use a unified multi-parent combination algorithm to solve the model and results show that the proposed method and model is scientific and easy operation which can provide a referential solving idea for decision management in micro-credit companies.

In recent years, the Internet of Things (IoT) has received considerable attentions and the overall number of connected devices in IoT is growing at an alarming rate. The end-terminals in IoT usually collect data and transmit them to the data-processing center. However, when the data involve user privacy, the data owners may prefer to encrypt their data for security consideration before transmitting them. This paper proposes a video deduplication scheme with privacy preservation by combining the techniques of data deduplication and cryptography. In the scheme, the data owner divides every frame of video into blocks of the same size. Then the blocks are encrypted and uploaded to the cloud platform. On the server side, identical blocks which have been already stored in the database are eliminated for saving the storage space.

The average face model has been used to correct the reconstructed results of the classical photometric stereo. By fusing the low frequency of the average face model and the results of photometric stereo, the accuracy of the face three dimensional shape has been largely improved. Moreover the results of the proposed method are more robust under different lighting conditions than those of the classical photometric stereo.

This article presents an approach for automating business process consolidation by applying process topic clustering based on business process libraries, using graph mining algorithm to extract process patterns, find out frequent sub-graphs under the same process topic, then filling sub-graph information into the table of process frequent sub-graph, finally merging these frequent sub-graphs to get merged business processes on the basis of process merge algorithm. We use compression ratio to judging the capability of our merge methods, the compression ratios of integrated processes in same topic cluster are much lower than the different topic processes, and our method achieves similar compression ratio compare with previous work.

Item Splitting splits an item into two items rated under two alternative contextual conditions respectively for improving the prediction accuracy of contextual recommendations. To get more specialized rating data, Complex Splitting is proposed to further improve the accuracy of recommendations. The key of the approach is to select multiple contextual conditions for splitting user or item. We translate it into a contextual conditions combinatorial optimization problem based on discrete binary particle swarm optimization (BPSO) algorithm. The item or user is split into two different items or users according to those contextual conditions in optimal combination. We evaluate our algorithm through a real world dataset and the experimental results demonstrate its validity and reliability.

The search results of the traditional concept similarity algorithm in search engine is not accurate, and can not support search results query for search results. A method of calculating the similarity of web pages based on financial ontology is proposed to solving the above problem. First of all, the concept of the financial ontology based on WordNet is constructed, and the corresponding concepts are obtained; Secondly, a new improved strategy of extracting financial key words is proposed in Key words mining according to the characteristics of the web pages (give different weights to different parts of the web pages), which is based on the traditional TF*IDF algorithm, and this strategy can better represent the subject of a web page; Then calculate the semantic distance between keywords, the depth of the levels and the degree of semantic overlap. Finally, the optimal computation of the similarity is realized by the comprehensive weighted processing of multiple similarity. Experimental results showed that the method compared with SSRM in recall ratio and precision ratio has greatly improved. At the same time, the method improved the algorithm performance.

This paper proposed an Expert System for Tractor Fault Diagnosis (ESTFD) based on ontology and web technologies. The ESTFD consists of several components such as diagnosis interface, OWL reasoner, explanation module, ontology base and database etc. The diagnosis interface was designed as web interface, which could support users to access the ESTFD by internet anytime and anywhere. A domain ontology, tractor fault diagnosis ontology, was constructed to build ontology base. The OWL API was called to manipulate the ontology base. The OWL reasoner, Pellet, was used to make logical reasoning and generate explanations for the process of logical reasoning. The ESTFD could provide tractor fault diagnosis service via internet to tractor maintenance personnel and drivers who located in the wide rural areas in China. Since the ESTFD has explanation module to explain how the diagnostic results was obtained, it also could be used as a training tool .

In this paper, we introduce a novel method to extract and assess influence relations between concepts, based on a variation of the Kuramoto Model. The initial evaluation focusing on an unstructured dataset provided by the abstracts and articles freely available from PubMed [7], shows the potential of our approach, as well as suggesting its applicability to a wide selection of multidisciplinary topics.

By more and more, privacy preservation problem is widely discussed among users and researchers. For mobile sensing network, an imperfect privacy preservation scheme will directly put participants into a dangerous situation. The better privacy protection applied, the better sensing data quality will be achieved. In this paper, we present a privacy-aware data aggregation scheme for mobile sensing networks. We considered both the smart nodes like smart-phone and dumb nodes like wearable device or GPS device. We take the location information and the sensing content into consideration separately. And this thought will make sure the sensing content will be k-anonymous and the accurate location will be protected well either. We use erasure coding technology to slice the sensing data record according to the k-anonymity rules. For the sake of efficiency and stability, we compare two coding technology in two sensing data types and give the experiment results and explanations in detail. After that, we give a social model to describe the social relation and a security data sharing protocol among the participants. The introduction of the participants’ social relation may give a new way to the reputation and data trustworthy evaluation mechanism.

Nowadays virtual networks (VN) need an easier way to configure and manage. There are already plenty works about VN mapping but there are still no a proper approach can meet these requirements no single approach currently can meet these requirements simultaneously. Software-defined networking (SDN) is an emerging networking pattern that gives hope to change the limits of current network. Some researcher use SDN to deal with VN problem such as Flowvisor and Openvirtex, but these works only have VN platform’s core virtualization feature. In this paper, we proposed a SmartNV (Smart Network Virtualization) platform using our enhance Openvirtex. Based on our SDN-NV architecture, SmartNV can collect information from physical infrastructure and choose appropriate mapping algorithm to calculate the virtualization scheme. The evaluation experiments on our prototype system can satisfy the requirement of the NV based SDN.

In order to improve the efficient and simple the steps of generation an image hashing, a security and robustness image hashing algorithm based on Slant transform (ST) is proposed in this paper. By employing coefficients of Slant transform, a robust hashing sequence is obtained by preprocessing, feature extracting and post processing. The security of proposed algorithm is totally depended on the user-key which are saved as secret keys. For illustration, several benchmark images are utilized to show the feasibility of the image hashing algorithm. Experimental results show that the proposed scheme is robust against perceptually acceptable modifications to the image such as JPEG compression, mid-filtering, and rotation. Therefore, the scheme proposed in this paper is suitable for image authentication, content-based image retrieval and digital watermarking, etc.

At present, video retrieval has been applied to many fields, for example, security monitoring. With the development of the technique of content-based video retrieval, video retrieval will be applied to more areas. The article mainly do research on offline video retrieval based on color features and realize offline video color features retrieval. The research realized Algorithm for Video Objective Tracking based on Adaptive Hybrid Difference and was focused on designing color features range calculation scheme with the combination of RGB and HSL color model. And extract and judge the color feature of the blob in the video then analyze and process the retrieval result. According to the result of this test, the success rate of detection of the system have reached ninety percentage upon. The realization of offline video object retrieval system based on the color features can decrease the time of Manual Retrieval to the color features object in the video, help people filter information and have benefits on the realization of intelligent and automatic video retrieval.

In order to solve the problem that the result of traffic congestion detection in bad weather is inaccurate, we analyzed current vehicle identification algorithms and image processing algorithms. After that, we proposed a detection method of traffic congestion based on histogram equalization and discrete-frame difference. Firstly, this method uses discrete-frame difference algorithm to extract the images that have vehicle information. Secondly, this method uses the histogram equalization algorithm to eliminate the noise of the images. Finally, this method recognizes the vehicle from the video and computes the traffic congestion index by the calculation method based on discrete-frame difference. It has proved by experiments and theoretical analysis that this method decreases false-negative rate and increases the accuracy rate of automatic traffic congestion detection in bad weather.

Our study was focused on the optimization of large training samples set selected from the global forest cover change detection system. Automatically delineating training samples procedure labeled tens of millions of samples representing forests and non-forests. To improve the precision, reduce the computational complexity and avoid over-fitting, we need to select samples from the large set of tens of millions of samples that are helpful for training a classifier. In this paper, two methods were used to optimize a large sample set from the Landsat-7 ETM+ data and obtain samples for training the classifier. The first method was the traditional stratified system sampling strategy. The second was uncertainty-based sample set optimization that selects training samples based on uncertainty by examining the uncertainty measure of samples and the distribution of their feature space, and involving the subtractive clustering, KNN and support vector machine. Through precision evaluation, our experiments validated that the uncertainty-based sampling strategy can achieve better results than the stratified system sampling strategy.

The existing print circuit board (PCB) short circuit fault detection schemes can only detect the presence of a short circuit network, but not the specific location of the short circuit. To solve this problem, we have determined a scheme based on near field magnet specific point detecting. First, we test the existence of a short circuit network using the short circuit detection algorithm. Second, a high frequency voltage is applied to excite the detected short circuit network, and the arranged near-field magnetic sensors above the PCB detect the peak value of the magnetic field. This technique can be utilized to precisely position the short circuit point on the PCB. The simulation results show that the scheme is operational and can effectively locate the short circuit point position.

Over the past several years, Gaussian mixtures models have been the dominant approach for modeling in text-independent speaker recognition field. But the recognition accuracy for these models declines when utterances’ length becomes short. Presently Mel-frequency cepstral coefficients are generally used to characterize the properties of the vocal tract and widely applied in speech recognition. In addition, prosodic features, such as pitch and formant, are generally considered to describe the glottal characteristics. However, the efficiency of those approaches remain unsatisfactory. In text-dependent short utterances speaker verification systems, prosodic features can assist to improve the recognition result theoretically. In order to optimize the performance of speaker verification systems under the framework of adapted GMM-UBM, we adopt a variant speaker verification system based on prosodic features, in which a dual-judgment-mechanism is used in order to integrate vocal tract features with prosodic features. Experimental results showed that the new speech recognition system led a better consequence.

Based on the analysis of the existing main ontology construction methods and tools, we proposed a user-oriented method of semi-automatic domain ontology construction. The method needs to build a descriptor set according to the user’s requirements, and establish a hierarchy structure based on it. Users can construct ontologies by using this method in the case without the participation of experts. And the method still has advantages for expansion, collaborative development of ontologies.

Currently, the input of rural clinic system in china is the symptom from patients’ descriptions. However, the existing problems in patients’ descriptions are irregular and colloquial description, too much irrelevant information, etc. We need to use information extraction technology to extract more useful information as the input information of the following procedure for better matching between symptom and illness and enhancing the accuracy of the clinic. We designed an information extraction method for rural clinic using open source tools. Based on the machine learning methods, we extract time and degree information was extracted from the patients’ descriptions. We designed and implement parallelization of the algorithms to speed up the response.

Along with the development of biometric recognition, the technology of finger vein recognition possesses better anti forgery performance and identification stability in collecting and certificating information of human bodies. The available finger vein recognition method is mainly based on template matching or whole feature recognition, suffering from light instability of the acquisition equipment which leads to low robustness. In order to strengthen the robustness, we adopt a finger vein recognition algorithm using Feature Block Fusion and Deep Belief Network (FBF-DBN), in which the nonlinear learning ability of deep neural network is used to recognize the features of finger veins. Meanwhile, we improve deep network input by using feature points set in vein images, sharply reducing the time in learning and detection, meeting the practical needs of biometric recognition specifically applied to embedded equipment. Experimental results showed that FBF-DBN algorithm presented better recognition performance and faster speed.

With the analysis of WFMC (Workflow Management Coalition) process meta-model, it’s indicated that the limitations of that model cause the difficulty for convenient reconfiguration of business process. So, a new process meta-model, ESR (Event-State-Rule) meta-model, is presented in this paper as the substitution to WFMC’s. Some elements are added in the new model, such as the event, the state and the rule with which dynamic relevance between the process and the business can be normalized expressed in process definition. Also, the boundary between process logic and business logic becomes much more explicit with the benefit from that model, which means that process logic can be separated from business logic quite effectively. Now, it becomes possible that rigid process and flexible process are modeled in a unified process framework. When process logic varies, the reconfiguration of the process may be fulfilled only with the corresponding variation of process definition so that convenient process reconfiguration is implemented.

ORAM is a useful primitive that allows a client to hide its data access pattern and ORAM technique as a wide range of applications nowadays. In this paper, we propose a verified version of binary-tree-based ORAM with less data access overhead. We provide a new method to reselect the leaf node and write data back to the tree, and accordingly, avoid complicated evict operation. Besides, the bucket capacity is reduced to a constant level. Overall, our scheme improves the efficiency meanwhile maintains security requirement of ORAM.

A novel wavelength-division-multiplexed passive optical network base on quantum key distribution FPGA controller is presented here. QKD FPGA is responsible for 1.25 Gbps upstream PRBS source, clock regeneration, phase modulation control, key sifting, privacy amplification, and upstream time-divided-multiple-access control on quantum channels. An 8-user network experiment shows that over 20 km fiber, the mean secure exchange key rate can reach up to 500 bps in total, with the acceptable quantum bit error rate below safe limit and few impact on classical channels. This scheme can provide a promising way for the coexistence between quantum key distribution and classical data service.

Security is a fundamental aspect of the next generation mobile network. In this paper, features of the 5G network, such as IoT and D2D, are described along with the proposal of a new layered network model, which enables independence on multiple radio access technologies (RATs). Additionally, security requirements on the 5G network are stated to facilitate security examination. Security issues arising from new technologies to be used in the 5G network, for example, physical layer security, are investigated in the core section of the paper.

In the network security situation assessment based on hidden Markov model, the establish of state transition matrix is the key to the accuracy of the impact assessment. The state transition matrix is often given based on experience. However, it often ignores the current status of the network. In this paper,based on the game process between the security incidents and protect measures,we improve the efficiency of the state transition matrix by considering the defense efficiency. Comparative experiments show the probability of the network state generated by improved algorithm is more reasonable in network security situation assessment.

Nowadays, with the rapid development of social economy and technology, chaotic secure communication has been a hot issue of social development. This paper begins with describing the chaotic secure communication technology and then analyzes the design scheme for chaotic secure communication based on chaotic pilot signal for synchronization control. The experimental simulation shows that using chaotic pilot signal for synchronization control can help achieve chaotic synchronization between communication transmitting and receiving systems and further enhance security and confidentiality of exchange of information and transfer of data. The theoretical analysis and numerical simulation results show that the design approach is effective and universal.

Although most of the existing encryption system takes the privacy issues of storing data into consider, the reveal of user access pattern is inevitable during the e-mail filtering. Therefore, how to protect the private data in the process of spam filtering becomes one of the urgent problems to be solved. Combined with two filtering techniques which are based on keyword and blacklist respectively, this paper achieves the goal of sorting and filtering spams. Meanwhile, given the privacy issues in sorting and filtering the spams, the paper is based on an experimental project, the Pairing Based Cryptography, which is performed by Stanford University to achieve the e-mail encryption program. It adopts a searchable public key encryption in the process of sorting and filtering, which needs no decryption and can realize searching and matching operations. By this method, it fully protects the privacy and access patterns of the mail receiver from disclosing.

The automotive parts manufacturing companies choose partners in the joint procurement process, when it comes to multi-index heavy weight determination, proposed a combination of objective and subjective weighting method. It reflects both the decision-makers experience information, and reflects the actual reasonable and objective scientific and technical data on the comprehensive evaluation of the results.

To solve the problems of the design knowledge sharing and reuse difficulty in the process of complex product design, an ontology-based knowledge modeling method of complicated product design was proposed. A complex product design work model was established by analyzing the characteristics of complex product design firstly, and an ontology construction framework for complex product design was also put forward. At the research background of auto products design, the knowledge modeling process of complex product design based on ontology modeling theory was studied by utilizing the classification and description method of thinking. On this basis, a knowledge ontology model for complex products design was constructed. Finally, an application case was presented to illustrate the feasibility and validity of the knowledge modeling method.

With the improvement of mobile communication technology, mobile Online Social Networks (mOSNs) provide users with the corresponding location based services when compared with traditional social networks. Location sharing becomes a fundamental component of mOSNs now, and some practical methods and techniques have been proposed to protect user’s privacy information. Some of these methods can accommodate privacy protection based on the input user profile and user’s privacy preferences through personalization, but user may be unlikely to use them without easy operation and strong privacy guarantee. In this article, we make a further research on privacy-preserving location sharing in mOSNs and develop a framework to help user to choose his desired degree of the privacy protection based on context aware. An adaptive learning model is established to provide user privacy right decisions, based on analyzing a series of factors that influence the choice of user’s privacy profile. This model will manage the different contexts of different user privacy preference with minimal user intervention and can achieve self-perfection gradually. So our proposed model can effectively protect users’ privacy and motivate users to make use of privacy preferences available to them.

In this paper, we propose a new two-lane cellular automata (CA) traffic model under the three-phase traffic theory framework. In the model, the velocity update and lane-changing rules are designed for the phase of wide moving jam, which fills the gap of two-lane CA model research in three-phase traffic theory. Then, a simulation system is designed and implemented. Numerical simulation results show that the model proposed in this paper can reproduce the typical phenomena of three-phase traffic flow. Finally, this model is used to explore the variation laws of average velocity under different phases. Our research results provide an useful reference for the management of two-lane system.

The knowledge granularity is described, and the knowledge granularity model is constructed. With the help of granularity principle, design knowledge was classification, association and inference. The hierarchical structure of the domain knowledge was described by using the knowledge granularity, and the related knowledge was structured and formal. Through the analysis of a case, the method is proven to be effective to improve the relevance of knowledge and improve the efficiency of the knowledge service.

The physical-layer impairments may weaken the signal quality in optical transmission link. And the signal quality degradation may substantially dent the performance of optical communications. This paper, taking into account the physical-layer impairments in performance optimization, proposes genetic-nonlinear algorithm to adjust channel power and optimize OSNR (Optical Signal Noise Ratio). Simulation results confirm the validity of the controlling strategy with the better dynamic and stable performance for adjustment of transmission power at the device nodes.

Frequency-hopping sequence set with low Hamming correlation within the fixed zone around the origin is called LHZ FHS set. In the quasi-synchronous frequency-hopping code division multiple access systems, the LHZ FHS set is often used to eliminate multiple-access interference. In this paper, basing on the Cyclostomes theory, we present a class of LHZ FHS set. It points out that the sequence set possesses excellent performance, and is optimal with respect to the Peng-Fan-Lee bound. It can be widely used in the frequency-hopping communication systems.

Recently, the total public debt outstanding (TPDO) of the United States has increased rapidly, and to more than $$\$17$$ trillion on October 18, 2013. It is important and necessary to conduct the TPDO projection for better policies making and more effective measurements taken. In this paper, we present the ten-year projection for the public debt of the United States (termed also the US public debt, USPD) via a 3-layer feed-forward neuronet. Specifically, using the calendar year data on the USPD from the Department of the Treasury, the neuronet is trained, and then is applied to projection. Via a series of numerical tests, we find that there are several possibilities of the change of the USPD in the future, which are classified into two categories in terms of projection trend: the continuous-increase trend and the increase-peak-decline trend. In the most possible situation, the neuronet indicates that the TPDO of the United States is projected to increase, and it will double in 2019 and double again in 2024.

In prediction of network security situation, the prediction accuracy of traditional single kernel function vector machine is a little low. It can’t describe the randomness and abruptness, and it has some limitation. A network security forecasting model was put forward which combined Gaussian kernel function and polynomial kernel to solve this problem. Proved by simulation experiment, this model can increase prediction accuracy and it has some practical meaning.