Considering that part of the efficiency is due to endogenous risks and exogenous environments, a modified three-stage sequential technique based on the DEA model is applied to the case of China's insurance industry. The technical efficiencies adjusted for risk and environment are re-measured for the insurance firms from 1999 to 2009 in China. The results show that the endogenous risks and exogenous circumstances of insurers have significant impact on the efficiency estimation. The risk effect indicates that the improvement of enterprise risk management can increase efficiency, while the environment effect implies that the exogenous environment conditions have become adverse. With the further opening-up of insurance industry in China, the risk effect and environment effect have more significant impact on stock insurers than on state-owned insurers. Although the environment effect has more significant impact on foreign insurers than on domestic insurers, there is no difference in the risk effect.

This paper deals with a multi-period portfolio selection problem in the frictional market. A multi-period minimax portfolio selection model with opportunity costs and transaction costs is proposed. By using the theory of nonlinear programming, we prove the existence of optimal solution of this model. Moreover, the method for solving this model is given by using the theory of convex programming and Kuhn-Tucker condition. Finally, a numerical example is given to illustrate our results.

Through introducing the probability of success, the paper proposed the single-target merger timing under double uncertainty (both the size and the realization of the synergy are uncertain). Furthermore, the paper extended the model to multi-targets merging situation, figuring out the merger timing under both independent merger strategy and sequential merger strategy. The paper also pointed out the optimal merger strategy of multi-targets merging through comparing and contrasting the threshold and acquirer's firm value of two merging strategies. Because of the extra information value of sequential merger, the threshold of sequential merger strategy is smaller than that of independent merger strategy, and there exists a sequential merger interval where sequential merger strategy is superior to independent merger strategy. Finally, through numerical analyzing the paper investigated the impact of correlation coefficient, success probability, and the acquirer's bargaining power on the threshold and the sequential merger interval.

This paper constructed an aggregate measure of investor sentiment and sentiment Beta, which was defined as the sensitivity of returns to sentiment through choosing sentiment variables. By employing this index, it studied and analyzed the relationship between sentiment and stock characteristic. The empirical result shows that, during a certain period, larger capitalization, higher volatility and higher market-to-book ratio stocks are more sensitive to sentiment in Chinese A-share market. However, there is no significant link between stock age and sentiment.

As a mental characteristic, vertical representation bias is introduced to portfolio choice model, to research its influence to investors' portfolio decision. This paper proposes Representation Return concept, and gives an investors' utility function of two variables — final return and Excess Representation Return, from the principle of expected utility maximization. Then establishes optimal portfolio model based on vertical representation bias, and give an empirical analysis showing that the model can obtain efficient frontier of investment portfolio.

Market economy presented a kind of phenomenon that the opportunities may be lost at any time. Enterprises need to choose the best investigation method in limited time, in order to acquire more information to make credit decisions more accurately. It is of great significance for enterprise in the situation. This article regards the problem of account sale decision-making in incomplete information state as a discrete time, continuous state stochastic process. The article establishes a model of customer credit evaluation on the basis of different information acquisition degrees. The model borrows some ideas from the method of spatial analysis. Then the article references the thought of discriminatory analysis to discriminate the accuracy of the customer credit evaluation in incomplete information state. At last, the article works out the expected profit of account sale by borrowing the thought of risk decision. The whole process is based on taking full account of the effect of different investigation methods on discrimination accuracy of customer credit rating within the limited time. At the end of the article, an example is given to show the practicality of the model.

In the market demand characterized by increasingly diverse product variety and shorter delivery time, it has become an urgent issue for many manufacturing enterprises to decide whether to implement mass production or mass customization. The present paper proposed different profit models under two types of market structure of monopoly and duopoly. To solve these models, backward induction method was employed, which is evaluated as an effective approach in settling the subgame-perfect Nash equilibrium. Then, feasible plans for production mode selection with regard to different conditions were obtained through systematical analyses of selecting either mass production or mass customization under monopoly, and different combined decisions of mass production and mass customization under duopoly. The research results contribute to providing theory and method support for manufacturing enterprises to make decisions on production mode selection under two types of market structure of monopoly and duopoly.

This paper presents a new alternative semiparametric estimation approach for censored regression models based on the survival function of the censored dependent variable, and gives a simulation study on the constructed estimator in comparison with HP estimators, which shows that the estimator has a good performance in small samples.

From the view of product life cycle (PLC) which is based on the process of customers' adoption, this paper used a micro diffusion model to analysis the diffusion of products with network effects. The simulation results show that the local network effects of customer interactions increase the diffusion speed both at early and middle stage. The clustering effects of customer interactions could impede the diffusion speed at middle stage. The stronger the interactions between customers are, the faster the diffusion speed at middle stage is. The type of global interaction customer decreases the diffusion speed at early stage, but it could increase the diffusion speed at middle stage. Furthermore, the influence of those factors above on the diffusion speed of products with network effects would exhibit different patterns due to the different structures of complex social network. Those conclusions provide new marketing strategy implications for firms in markets with network externalities.

Taking advantage of a GMM estimation of spatial error coefficients, a modified Bera and Yoon robust test for spatial dependence between spatial lag variables is proposed. This method not only avoids the heavy computation of the maximum likelihood estimation but also overcomes the limits of Bera and Yoon when the spatial correlation in errors is large. Moreover, the method is superior to spatial dependence tests fully based on GMM estimations. The Monte Carlo simulations verify these theoretical results.

Chinese ports are facing Over-investment with rapid economic growth and fierce competition with non-equilibrium analytical framework, the supply and demand structure of China's coastal container ports and the balance model is established. The China coastal container ports in 2000–2010 supplydemand imbalance and the gap trends were analyzed. It was found that supply and demand in China's coastal container ports in the state has entered the supply exceeds demand, and this trend continued in the expansion of supply exceeding demand. At last, the China's container port construction preventive measures the risk of overheating was proposed

Based on the switching cost of consumer behavior, the three-stage decision model from the debt libility effect was built to analyse the financing strategy in capital market and pricing strategy in product market. Firstly, the research shows that the switch cost has “Bargains-then-Ripoffs” effect in product market competition, but the strategical effect of debt financing will weaken this effect. Secondly, the pricing strategy would be dependent on the firm's preference and tradeoff between the two effects. Thirdly, the firm will decrease its ex-ante the ratio of debt financing and take conservative financing behavior when it faces the product marker with high switching costs. Finally, the outer market environment will directly effect the firm's financing strategy and then indirectly its pricing strategy.

The loss of social-pattern crisis will be extremely increased because of the distortion of crisis information in the diffusion process. Study on the diffusion rule of crisis information will contribute to active response to social-pattern crisis. There are two types of crisis information: true information and false information. The diffusion process of social-pattern crisis information is viewed as the dynamic competitive process of crisis information diffusion between true information and false information which strives for trust of victims based on ecology. This paper posits a social-pattern crisis information diffusion model based on Lotka-Volterra model. Then it conducts simulation and empirical study utilizing this model. The results show that the equilibrium state is only related with the substitute coefficient. The time achieving the balance is positively related with the difference of diffusivity; and the inflection point in the diffusion process of crisis information is decided by the substitute coefficient, diffusivity and initial state. We put forward some suggestions which establish the sharing system of crisis information, release system of crisis information and multiple-participation mechanism for crisis responses.

According to the meaning of scientific development: insisting on human-oriented, comprehensive, coordinated and sustainable development, this paper chooses human all-around development evaluation indicators extensively from four guidelines which include the economics, society, the quality of life, the quality of human. Based on the principle that indicators should be observable, this paper screens the initial indicators, after that, it uses correlation analysis together with principle component analysis to screen the indicators, and at last establishes the evaluation indicators system of human all-around development. The special and contributions of this paper lie on five aspects: firstly, it uses correlation analysis to screen the indicators, deleting the indicators with high correlation coefficient to ensure the indicators cover comprehensive information and not repeat. Secondly, it uses principle component analysis to delete indicators which have no significant impact on evaluation results. Thirdly, the study result reveals that final indicators system uses 17% of initial indicators to reflect 99% of original information. Fourthly, it uses indicators such as Per capita GDP, Gross National Happiness and Gini Coefficient to reflect the human-orientation and the meaning of and the characterized by comprehensive, coordinated and sustainable development. Fifthly, the indicators system through quantitative screening based on authorities' typical and high-frequency indicators have double information of experts' knowledge and objective reality.

In this paper, an insertion heuristic based on time difference is introduced. Based on this insertion heuristic, a genetic algorithm for pickup and delivery problem with time windows (PDPTW) is proposed. Comparing with other genetic algorithms, our algorithm holds two characters: 1) Cross operation as well as R1- mutation operation and R2- mutation operation are based on time difference insertion heuristic; 2) Non-generation search strategy is applied to the algorithm. Our algorithm is subject to a comparative test on the benchmark problem sets that comprise 56 PDPTW instances. The results show that the solution quality is better than the reported algorithm whose objective function is the same as ours.

When the claim numbers appear to be over-dispersed in ratemaking, negative binomial regression model will be usually applied. However, it is also possible that the claim numbers may be zero-inflated, and then the negative binomial regression is not suitable for those data. The paper makes generalization of zero-inflated negative binomial distribution based on traditional ones to deal with the over-dispersed and zero-inflated data simultaneously. At the end of the paper, the extended model is applied to a data set of automobile insurance loss and the result shows that the goodness-of-fit can be effectively improved.

In the actual process of locating facility, the decision-maker always faces the following constraints: they must determine where to locate the initial facility (or facilities) when the number of edges to be served is uncertain, and, the constructed facilities can not be removed when the new facility is built. The existed models and algorithms are always pointed to a static facility location process, but here we need a dynamic facility location model with above constraints. Considered the online vertex covering problem in this paper, we present a structural lower bound which does not need any complexity assumption, and prove that the analysis is tight by giving a competitive algorithm as well as competitive ratio proof for a restricted version of the online vertex covering problem, and we also obtain that this algorithm is optimal if we are permitted to use non-polynomial time.

As for coalition structure being strict to matroids, two distribution principles of rationality and equality were designed for strictly cooperative games on matroids. Moreover, τ value for strictly cooperative games on matroids was constructed according to τ value for the typical cooperative games, which is characterized by two distribution principles of rationality and equality with the help of strictly mathematic reasoning. Finally, the τ value compared with Shapley value for cooperative games on matroids in personal rationality.

This thesis analyzed the significance of prognostic and health management (PHM) in network maintenance support. The key technologies of network PHM: management mechanism, health evaluation, prognostic, fault localization, fault recognition and health management were emphatically discussed. Problems in key technologies of recent network PHM were summarized while the future developing trends were introduced. Based on these analyzes, the thesis proposes a hierarchical distributed structure on mobile agent and an integrated management framework for the network PHM system.

In order to solve the uncertain disturbance during supply chain operation based on networked manufacturing (NM), we use robust H_∞ control method to inhibit bullwhip effect in supply chain system and make sure the stability of system. A dynamic time-varying supply chain deviation system model is set with state variables and control variables. By applying linear matrix inequality (LMI), we get H_∞ robust control strategy. Through feedback controller u_k design, it effectively reduces fluctuation of product, order and inventory caused by customer uncertain demands. The real case analysis shows that our presented robust control strategy can inhibit bullwhip effect caused by different type of customer demands with inhibition ξ and make supply chain system robust stabilization.

The production management of traditional Chinese medicine manufacturing enterprise relies heavily on the experience of manager, and delayed one-sided information transfer leads to the disjointedness between management and production as well as the delay of production command. Considering the above shortness and specific characteristics of traditional Chinese medicine production, this paper designed the knowledge-based MES for traditional Chinese medicine enterprises to optimize production processes. In this paper, function model of the system was given and the function modules of the system were described in detail. This paper also focused on the research of the system's key technologies including knowledge extraction, knowledge integration, knowledge reuse, process optimization. The MES uses an open, scalable, modular, object-oriented architecture to achieve functions such as optimization and schedule of productive process, on-line quality inspection of medicine, batch tracking of products and predictive maintenance of equipments. It has been applied in a pharmaceutical companies of Guizhou with good results.

Customer clustering is a fundamental issue in the logistics distribution study. Considering that retailers are bounded on certain geographic location and the clustering process is highly influenced by the distribution scale (that is the goods scale the customer demand) and loading scale of vehicles, we proposed a customer clustering method constrained by road network and distribution scale, for the clustering process with the location of distribution center unknown. The concept of “shortest arterial road distance (SARD)” takes distribution road information into account, which is ignored by the traditional Euclidean distance. Furthermore, a constrained clustering method for customers based on improved CLARANS was given, which consider the geographic information and fulfill the constraints of distribution scale. Finally, numeric experiments on both synthetic and real datasets justify the effectiveness of the method.

To improve the operation efficiency of container terminals, the yard crane scheduling problem in one mixed stacking block is studied in this paper. In one mixed stacking block, the container operation time windows constraint, sequence-depended setup time for yard crane and storage space constraint for containers are defined and an integer programming model is proposed for the yard crane scheduling problem. A improved tabu search heuristic algorithm is developed and the numerical experiments show the algorithm effective and efficient, particularly when large-scale problems.

Considering that the global optimal solutions often locate on or near the boundary of the feasible region for many constrained optimization problems, a novel genetic algorithm was proposed in this paper. The basic idea of the proposed algorithm was to put feasible solutions and infeasible solutions into two different containers respectively. Subsequently, a new designed crossover operator(named inwardoutward crossover operator) was used to a feasible solution and a infeasible solution, then a line search along a potential decent direction was used to improve the offspring so as to find a good solution on or near to the boundary of feasible region. By this search procedure, the possibility for obtaining the globally optimal solution is obviously enhanced, and similarly, the convergent speed is also strengthened. The “particle swarm mutation” inherited the advantages of Particle Swarm Optimization (PSO) algorithm and searched for the potential solution along the direction of the best current particle and the direction of the best individual of the whole swarm in the past. Selection operator retained a constant rate of infeasible solutions. Numerical results indicate that the proposed algorithm can be efficient to get global optimal solutions or near to them in smaller population and less iteration times.

The paper took different regions and regional industries as initial water right allocation objects in the basin; the compound system for basin initial water right allocation was given; the connotation and objective of the compound system was analyzed; and the bi-level optimization model of the compound system optimization was established to allocate water right of different regions and regional industries, and eliminate the water conflict and contradiction among regions, embody the fairness and efficiency of different regions' water right, prevent different industries from the vicious water right competition, optimize comprehensive benefit of society-economy-ecology in the basin. Meanwhile, according to the interactive decision-making theory, the model algorithm was put forward. The case analysis verified the effectiveness of the model.

The psychological mechanism and intervening strategies of rumor propagation were analyzed firstly. Using mean field approximation, we gave a new model in which rumor attribute was considered. By introducing strictly first order stochastic dominance and strictly second order stochastic dominance, we analyzed the relationships among rumor attribute, intervening intensity, network structure and diffusion effect, then derived and proofed four propositions that were very useful for us to explore rumor propagation. Finally, some propositions were verified by computer simulation.

Aiming at the complexity of modern battlefield, diversity, uncertainty and illegibility of battle damage assessment (BDA) data sources, the BDA fuzzy Bayesian networks is constructed by integrate the fuzzy set theory into Bayesian networks. The continuous node of this network is converted to corresponding membership degree, and then the membership degree is converted to probability knowledge which can be used to Bayesian networks inference by possibility transformation theory. The effect on BDA of weather, electronic jamming, optical jamming and weapon-target matching degree are considered in this model. The combat decision-maker can use the model to simulate on the target damage under the varied kinds of attack conditions, and make quick and correct assessment on target damage grade. It can get true target damage information by syncretizing the recon-naissance intelligence collected after the attack, and it also can make use of assessment accuracy to modify the assessment model. Finally, it will provide decision-making evidence for second strike and properly making fire-power configuration scheme.

In this paper a novel method based on SPEOS/CATIA to evaluate glare of flight deck was proposed. The secondary development work was carried out from SPEOS/CATIA framework, the glare sources could be detected, and the key parameters of quantifiable evaluation were calculated as well, such as source luminance, background luminance, position index and solid angle of source. And then, the desired glare index could be computed automatically and quickly. Finally, unified glare rating (UGR) and daylight glare probability (DGP) equation were utilized as example to predict the glare pollution in flight deck. The results indicate that the proposed method is convenient for safety design of flight deck.

In order to improve the agility of virtual cellular manufacturing in the functional layout manufacturing system for multi-product and small-batch manufacturing enterprise, it presents a novel method for virtual cell formation based on process modularization. The integrated method which combines both parts-based clustering analysis and process-based clustering analysis decomposes parts process based on process modularization. And a process module warehouse for the manufacturing system is formed, then we can construct virtual manufacturing cells. The superiority of the novel clustering analysis is that it can reduce the control difficulty of manufacturing system, improve the construction speed of virtual cell of manufacturing system. Besides, the method also can reduce lead time and improve the agility for virtual cellular manufacturing system.

Firstly, it makes an in-depth study on the econometric relationship between the quarter increment of the foreign exchange reserve in China and the main macro-economic variables. It finds that the macro-economic situation have great effect on the increase of foreign exchange reserves. Secondly, for the first time, it puts forward the combined forecasting method of R² weight, which uses R² statistics as the combined forecasting weight to combine the different models' results. With this method, we got the stable and accurate result that the foreign exchange reserves will increase continuously. Finally, it gives some suggestions.

By unit root test with two unknown structural breaks, analyzing the two monthly time series of trade balances and foreign exchange reserves since 1994, the result is that both sequences are trend-stationary processes with two unknown structural breaks. And based on the time points of breaks, the features and development histories of trade balances and foreign exchange reserves were characterized. The analysis showed that, global economic crisis caused by the problem of American sub-prime mortgage made the mean of trade balances and foreign exchange reserves drop down, but it didn't change the rising trends of the two series. It is predictable that the trade surpluses and foreign exchange reserves will continue to expand at least in the short term.

We give a single vendor and single buyer VMI model in which raw materials are no deteriorating, items are deteriorating and buyer faces partial backlogging for unmet demand. It's not necessarily the best optimal way that vendor employs lot-for-lot strategy to yield items for the impact of setup cost. Therefore, we allow that vendor employs integer ratio strategy to procure raw materials and yield items, which means that each batch of raw materials procurement meet integer time of production, and each batch of items yield meet integer time of replenishment to buyer. We get these decision variables of time that meet production of each batch of raw materials procurement, time that meet replenishment to buyer of each batch of items yield, replenishment interval and buyer's service level by solving steps based on genetic algorithm. The result of numerical example indicates that the VMI system can get the satisfied solution of minimum per unit time total cost when vendor employs integer ratio strategy to procure raw materials and yield items. The respective change of setup cost, deterioration rate and losing sale ratio has an affect on vendor decision to produce and the minimum per unit time total cost of VMI system.

For the global modeling and parameter optimization of manufacturing process featured with multi-extreme quality characteristics, a global sequential design (GSD) approach is proposed to reduce the experimental sample size and improve the predictive performance as well. Firstly, the proposed approach sets up an elementary model of the process using support vector regression (SVR) under a certain kind of initial experiment manner. Secondly, after identifying support vectors of the initial experiment sample, new experiments are added simultaneously at the 45 degree or axial direction of support vectors in each dimension of process parameters. Thereafter a sequential model of the process is set up. Finally, the sequential steps mentioned above are iterated until the lower prediction mean square error (MSE) or the upper sample size acceptable is reached. The simulation and empirical study show that, GSD approach can rationally allocate the experimental runs around the multi-extreme quality characteristics within the whole feasible region of process parameters, which overcome the shortcoming of only finding single extreme using traditional line steepest ascent sequential design and therefore improve the experimental efficiency. Furthermore, compared with uniform grid design, Latin hypercube sampling and uniform design manner, under the same sample size, the prediction MSE of GSD based SVR model decreases 30% at least; while under the same prediction MSE of SVR model, the sample size of GSD decreased 12.0% at least.

By means of the theory of CVaR (condition value at risk) in financial engineering, the author proposed the concept and calculation formula of CVaR of distribution network, which was then utilized to evaluate the risk level of distribution network. A multi-optimization model of distribution network design taking DNCVaR, benefit and customer satisfaction as objectives was developed, which can not only reflect the risk preference of supply chain core enterprise, but also resolve combinational decision problem of distribution network risk level-construction of distribution center-purchase quantity of distribution center-serving object of distribution center-purchase quantity of distribution outlet. Based on polynomial optimization technique, this model can deal with the combination of multi-objectives. These work provide helpful exploration for both the CVaR and the decision-making problem of distribution network design optimization. The result of the optimization decision model of an example indicates that the model constructed in this paper can effectively resolve the combinational decision problem mentioned above.

A multi-item joint replenishment problem is discussed for a two-echelon supply chain, which consists of a MTS (make to stock) manufacturer and a retailer selling multiple items. The three-echelon inventory system includes manufacture's raw material and finished product and retailer's product. Three replenishment models are formulated to study the impact of multi-echelon cooperation and multi-product joint replenishment on the total cost of the channel and the parties. The properties of the optimal solutions for the proposed models are then analyzed. Numerical simulations show that model JC (both the manufacturer and the retailer make the joint replenishment policy jointly) performs better in reducing the average total cost of supply chain, as well as the manufacturer, as compared to model JN (the manufacturer and the retailer make joint replenishment policy individually). But it is just reverse for the retailer. In addition, permissible delay in payment is adopted to achieve the coordination of supply chain.

Projects tender institution has high void rate very often. Taking it as an example, this paper provides a measure combination optimization method in design for reducing void rate of institution (MCOM in DRVRI), which includes the first optimizing problem and the second optimizing problem. Contrasted with those speculation methods, the method provided here has the advantage of calculability. Contrasted with those mathematic model methods, the method is structurelized and formlized, which makes it more practical in use. So the method can be widely used in various institution design.

Users' personal information is the basis of online personalized services. However, because of the risks in cyber world, online users are more concerned about their privacy security than ever, which results in either their unwillingness to provide personal information to websites or falsifying some of the required personal information during the registration. Based on the review of relevant literatures, a theoretical model is proposed, which show the influencing factors affecting online users' intention to provide personal information and privacy. Three kinds of factors, related to individuals, websites and situations, are adopted in this model. Furthermore, the theoretical model and relative hypotheses are verified by the empirical research by means of Structural Equation Model. The results show that individual's online privacy concerns will have negative impact on people's willing to provide personal information, while trust tendency can have positive impact on people's willing to provide personal information by influencing user's trust in the web site. As far as the site feature, good reputation, web security control and privacy policy, to a great extent, through establishing user's trust to have further positive impact on people's willing to provide personal information. In addition, the will is also affected by user's perception of site's serviceability and information sensitivity, while the impact by perception of site's usability and involvement degree haven't been supported. The results will be beneficial to websites with regard to the online users' personal information collection and a better registration procedure.

Much attention has been drawn to the concepts of intuitionistic fuzzy number and its applications in fuzzy multi-attribute decision making problems, some of the scholars also extend the intuitionistic fuzzy number to generalized intuitionistic fuzzy number. However, very little research has been completed on the extension of traditional interval-valued intuitionistic fuzzy numbers. Therefore, in this paper, the definitions of generalized interval-valued intuitionistic fuzzy number are proposed, based on the concepts and properties of intuitionistic fuzzy numbers and interval-valued fuzzy numbers. In order to help solve the fuzzy multi-attribute decision making problems, the arithmetic rules on the set of generalized interval-valued intuitionistic fuzzy numbers are further introduced, and the closeness of the operators in the domain of definition is proven. Then three ranking principles are proposed so as to implement the aforementioned arithmetical methods into the numerical example of workstation assessment, which demonstrates the validity of the proposed approach.

In order to reduce amount of computation, speed up convergence and restrain premature phenomena of quantum evolutionary algorithm, a hybrid chaotic quantum evolutionary algorithm is presented. The algorithm uses the chaotic initialization method to generate initial population that have better diversity, the simple quantum rotation gate to update non-optimal individuals of population to reduce amount of computation, and the hybrid chaotic search strategy to speed up its convergence and enhance its global search ability. A large number of tests show that the proposed algorithm has higher convergence speed and better optimizing ability than quantum evolutionary algorithm, real-coded quantum evolutionary algorithm and hybrid quantum genetic algorithm. Tests also show that when chaos is introduced to quantum evolutionary algorithm, the hybrid chaotic search strategy is superior to the carrier chaotic strategy, and has better comprehensive performance than the chaotic mutation strategy in most of cases. The proposed algorithm is the only one all of whose tests are convergent, and is easily implemented and applied in practice. It is satisfied in solving traffic signal timing optimization problem of urban road.

Most of existing text feature selection methods are serial and are inefficient timely to be applied to Chinese massive text data sets. So, it is a hotspot of text mining how to improve efficiency of text feature selection by means of parallel thinking. Combining genetic algorithm with parallel collaborative evolutionary, a parallel collaborative evolutionary genetic algorithm (PCEGA) based on rough sets was designed and used to select text features. The presented method took advantage of genetic algorithm to select features and employed parallel collaborative evolutionary to enhance time efficiency, so that the more representative feature subsets was acquired quickly. Experimental results show that the method is effective.