Recent advances in on-board radar and missile capabilities, combined with individual payload limitations, have led to increased interest in the use of unmanned combat aerial vehicles (UCAVs) for cooperative occupation during beyond-visual-range (BVR) air combat. However, prior research on occupational decision-making in BVR air combat has mostly been limited to one-on-one scenarios. As such, this study presents a practical cooperative occupation decision-making methodology for use with multiple UCAVs. The weapon engagement zone (WEZ) and combat geometry were first used to develop an advantage function for situational assessment of one-on-one engagement. An encircling advantage function was then designed to represent the cooperation of UCAVs, thereby establishing a cooperative occupation model. The corresponding objective function was derived from the one-on-one engagement advantage function and the encircling advantage function. The resulting model exhibited similarities to a mixed-integer nonlinear programming (MINLP) problem. As such, an improved discrete particle swarm optimization (DPSO) algorithm was used to identify a solution. The occupation process was then converted into a formation switching task as part of the cooperative occupation model. A series of simulations were conducted to verify occupational solutions in varying situations, including two-on-two engagement. Simulated results showed these solutions varied with initial conditions and weighting coefficients. This occupation process, based on formation switching, effectively demonstrates the viability of the proposed technique. These cooperative occupation results could provide a theoretical framework for subsequent research in cooperative BVR air combat.

Domain terms extraction based on Chinese Web document is an important step in the field of Chinese text information into machine recognition, and it is also the technical basis of intelligence information processing in domain ontology construction, text knowledge mining, and etc. The traditional methods of terms extracting are: Based on the dictionary, based on the rules and the statistical method. But each of methods contains some limitations, such as in single word processing, synonyms merging and so on. In order to solve these problems, the intelligence extraction method based on hierarchical combination strategy is proposed. It is divided into three layers: the first layer is the document preprocessing layer, the second layer is the words preparation layer, and the third layer is the term extraction layer. The effectiveness of this method is verified by experimental data in the field of weapon and equipment. Comparing with the traditional method, it is proved that this method has good accuracy and recall rate for the domain terms extraction based on the Chinese Web documents.

The advent of the large data era, the development of deep learning theory helps to create a good condition. Data mining generally refers to the process of automatically searching for information that has special relevance hidden from a large amount of data. Data mining is usually related to computer science and it is achieved through statistical, on-line analytical processing, information retrieval, machine learning, expert systems which is relying on past rule of thumb and pattern recognition. This paper introduces the background of the deep learning development, and it mainly discusses the self-encoding method in the depth learning, and implements the simulation application from the self-coding method. The latest developments in information fusion technology make it possible to easily capture images of color and depth information in order to improve the image of object recognition. This paper also introduces a BP neural network model for feature learning and image classification. Our model achieves better artistic expression on standard image object sets, finally the results are more accurate and quicker in training and testing compared to other comparable architectures.

Thorough awareness of battlefield situation is vital for combat while battlefield information is constantly changing and distributed. The commanders are faced with more complicated and dynamic situations to make decisions than ever before. Ontology technology has received increasing interest in computer science community because of its capability of knowledge representation, sharing, automatic reasoning, and reuse of heterogeneous information from multiple sources. In this paper, we discuss the role of ontology to facilitate high level fusion of battlefield information for decision support.

CDM-BSC(CRISP-DM applied with Balance Scorecard), which is a new term. CDM-BSC concept is developed from combination of traditional Data Mining Methodology and BSC for performance measurement systems. Balance Scorecard applied to CRISP-DM is a new methodology of improving the performance of Data Mining Process. CRISP-DM plus BS C (Balance Scorecard) is an enhancement of abstractive conception of multidimensional-perspective for data mining process to optimize the whole process and structure of data mining cycle with limited profitable resource. CDM-BSC not only indentify relationship between the phases of data mining process, but also to adjust effort for each point of CRISP-DM based on different business targets. CDM-BSC provides integrated platform supports decision making from strategy level, tactical level to operation level, and then deploys the re-engineered processing solution into data mining process. This combination will demonstrate to produce profound performance improvement while DM smooths transaction.

In order to improve the effectiveness of supervised self-organizing map (SSOM) neural network, a kind of genetic algorithm is designed to optimize it. To improve its classification rate, a real number encoding genetic algorithm is provided and used to optimize the learning rate and neighbor radius of SSOM. To speed up the modeling speed, a binary encoding genetic algorithm is provided to optimize input variables of SSOM and reduce its dimension of input sample. Finally, intrusion detection data set KDD Cup 1999 is used to carry out experiment based on the proposed model. The results show that the optimized model has shorter modeling time and higher intrusion detection rate.

Single task learning is widely used training in artificial neural network. Before, people usually see other tasks as noise in same learning machine. However, multitask learning, proposed by Rich Caruana, sees simultaneously training several correlated tasks is helpful to improve single tasks performance. In this paper, we propose a new neural network multitask similarity cluster. Combined with hellinger distance, multitask similarity cluster can estimate distances among clusters more accurate. Experimental results show multitask learning is helpful to improve performance of single task and multitask similarity cluster can get satisfactory result.

To simulate battalions and regimental level battlefield combat, intelligent agent technique is often used to model troops on the battlefield. To enable agents to understand battlefield situation and perform basic tactics, there are great difficulties lie in the following aspects: the perception and understanding of battlefield information are often vague rather than precise; high-dimensional data of battlefield make domain rules hard to extract and knowledge base difficult to construct; huge amount of battlefield simulation data cause system delay. This paper presents a hybrid approach based on fuzzy logic and case-based reasoning, employs frame structure as case knowledge representation to simplify domain knowledge acquisition process. The introduction of Fuzzy Logic, in the case of properties matching process allows the Agent to deal with ambiguity and uncertainty of battlefield combat knowledge. The case study shows that the proposed hybrid approach combines both advantages of Fuzzy Logic and CBR, and its usability is plain to see.

In fuzzy learning system based on rule weight, certainty grade, denoted by membership function of fuzzy set, defines how close a rule to a classification. In this system, several rules can correspond to same classification. But it cannot reflect the changing while training several tasks simultaneously. In this paper, we propose multitask fuzzy learning based on error-correction, and define belonging grade to show how much a sample belongs to a rule. Experimental results demonstrate efficiency of multitask fuzzy learning, and multitask learning could help to improve learning machines prediction.

Processing high-dimensional data is a problem in data analyzation. Due to its high-dimensional attributes and little samples, it is easy to encounter overfitting in processing high-dimensional data. Multitask learning, a weak learning, can boost prediction accuracy in all kinds of learning machines. Multitask learning doesn't need complex operations, but performs well. The characteristic of multitask learning is very fit in high-dimensional data. In this paper, we propose a novel neural network-multitask self- organizing map. In multitask self-organizing map, we replace Euclidean distance between two nodes with their hellinger distance. Through experimenting on real data, multitask self-organizing map gets satisfactory results. Experiments prove multitask self-organizing map can perform well in high- dimensional data. Copyright © 2013 Binary Information Press.

In order to improve the classification rate of BP neural network, this paper designs a kind of genetic algorithm to optimize it. A real coding genetic algorithm is provided to optimize the weights and threshold of BP neural network and the optimized network model GA_BP is established after selection, crossover, mutation operation. Finally, the experiment is carried out based on KDD Cup 99 data set. The results show that the optimized model has higher intrusion detection rate than the traditional BP model.

In order to improve the classification effectiveness of SVM (Support Vector Machine), new model is provided to optimize and improve it. Penalty parameter c and kernel function parameter g of SVM are optimized using genetic algorithm of binary coding, and the optimized model GA-SVM is established. The dimensionality of input sample for SVM is reduced by PCA (Principal Component Analysis) and the model GA-PCA-SVM is established. The famous KDD Cup 1999 dataset is used to evaluate the proposed model and experiment is carried out based on GA-SVM and GA-PCA-SVM. By comparison, the results show that the genetic algorithm optimization improves the classification accuracy rate of SVM and PCA operation shorts the training time and test time.

Within a learning machine, we could improve the accuracy of every learning task while a few tasks learned together. This method is called multitask learning. It is popular to using simulated annealing algorithm or setting a constant as learning rate in multitask learning. Either is simple and efficiently. But neither use feedback information in machine effectively. A novel adaptive feedback learning algorithm (AFLA) for multitask learning is proposed in this paper. The AFLA is proposed to address the problem using feedback information in multitask learning machine. The AFLA uses the error signal as feedback information. And it applies error signal in learning rate. The experiments are carried on two datasets. With simulated annealing and constant comparisons, experimental results show that the AFLA can use error signal efficiently. At normal learning rate, the accuracy of AFLA is better than using simulated annealing algorithm and constant. At smaller learning rate, the accuracy of AFLA is 2-5 times as big as others.

为了优化多维、多层次复杂数据流的数据挖掘处理过程,在层次分析法与数据挖掘理论的基础上,提出了"层次分析法构建挖掘模型（ACMC）"的一个全新概念。设计了ACMC策略的结构与流程,不仅提供了一个用来支持不同挖掘组件的集成平台,还为整个数据挖掘过程提供了一个流程可控的策略。目的是多方面、多层次对整个数据挖掘框架和不同处理模块的结合方法进行有效的改进。其中KPI和CSF的实例是整个数据挖掘结构的关键点和基础,模型的学习和评估被视为整个数据挖掘这个"大机器"的"引擎"。这些"引擎"产生的影响因素将直接影响到最终的决策。最后对ACMC建模的有效性进行了验证。ACMC从战略高度、技术水平以及不同的抽象层面为数据挖掘处理过程提供了一个有效的策略。

The use of statistic to deal with the decision making in uncertain environment has been widely applied. On uncertain data that includes some unknown values and noise, related algorithms are asked to have more powerful and robust expression ability of rules. This paper proposes the Statistical Second-Order Predicate (S2OP) algorithm which combines the superiorities of statistic and logic for improving accuracy and robustness of decision making. Second-order logic (SOL) has comparatively high efficiency and accuracy for this problem. By use of the value of Kullback-Leibler divergence, the algorithm judges the Russell's paradox of rule set and chooses the rules eliminated for the consistency of machine-readable knowledge base. The results of experiment show that the S2OP improves the accuracy of about 5% and increase the robustness of about 11.7% comparing with J48, BayesNet and Bagging on the waveform data set which includes 5000 instances and has many noises.

This paper addresses the problem of learning optimal classifiers that maximally improve the robustness and accuracy in uncertain environment included a large number of noise and missing values. Recent solutions to the efficiently vertex weight evaluation, such as the Bayes Network, rely on statistics methods, without sufficient robust guarantees. We show how a globally optimal solution can be obtained by formulating predicates and statistical training set evaluation in Markov Logic Network. We then propose a classification algorithm which adopts random selection of the instances and features in Random Statistical Predicate Invention (RSPI) classification model. In a set of experiments on UCI datasets about credit card and CRM information we show that the proposed RSPI can achieve significant gains in robustness of model, compared to decision trees algorithms or other random classification methods.

This paper presents the class methods of gateway discovery and selection schemes for the integrated network of Internet and MANET separately based on introducing recent representative gateway discovery and selection schemes of this integrated network. Then their advantages and disadvantages are compared. Finally, the future research issues in these two areas are pointed out.

The gateway discovery scheme is a key problem of interconnection between MANET and Internet.Adaptive gateway discovery is based on hybrid gateway discovery,and according to dynamically adjusting TTL and other factors,adaptive schemes can fit dynamic nature and different scenarios of MANET.This paper indicates the key problem of adaptive gateway discovery schemes,which is how to select the optimal operating point between proactive and reactive strategies.The paper firstly presents a classification method after introducing recent representative adaptive schemes.Then different methods which these schemes used are compared and analyzed,including their commonness and characteristics.Finally,the future research issue in this area is pointed out.

Planning Multiple Routes for Air Vehicles can be seen as a multiple-peak function optimization, which is the basis of proposing a new approach to plan multiple routes for air vehicles based on niche particle swarm optimization technology. By means of this approach, you can use a specific particle swarm coding representation and appropriate fitness functions to transform particle swarms within a route planning space into different unique niche sub-populations by introducing a RCS (Restricted Competition Selection) niche generation strategy on condition that all kinds of route constraints are met. During the evolutionary process, all individual particles only evolve within their each own niche sub-populations, thus different extreme value points are tracked out. At the end of evolution, each niche sub-population will generate their own optimal route respectively, so as to produce multiple optional routes for air vehicles. The simulation results demonstrated the feasibility of this approach.

CLinux is an embedded operating system with high performance. Security of µCLinux is of great importance for embedded system based on it. To solve security problems of µCLinux system, the author adopts Linux Security ModuleLSM framework and puts forth a multi-policy mandatory access control (MAC) mechanism based on both Domain and Type Enforcement (DTE) model and improved Bell-La Padula BLP model. The mechanism enhances security protection of µCLinux kernel and implements access control in fine-grain level to ensure confidentiality and integrity of the system. The scheme proves to be applied in most embedded system with properties of Level B1 security standard, which ensures security of all application softwares.