Research On Operational System Target Grouping Method Based On Complex Networks Community Detection

Target grouping is an indispensable part of situation analysis. In face of information war, it has brought new challenges that using information technology to support the system confrontation. And it has become an inevitable trend by conducting target grouping from the perspective of operational system.Target grouping in operational system is a new emerging problem in situation analysis, which seeks to make a reasonable group division of the operational targets in enemy’s operational system by analyzing their individual attributes and relationship with an aim to reduce the pressure of command officers, helping them have comprehensive and accurate understanding of the composition of the enemy’s operational system on the whole. This dissertation carries out research for the problem by using the method of complex networks community detection, which specifically includes:1. Depth analysis and mathematical modeling of target grouping in operational system was carried out. The problem of target grouping in operational system was elaborated, and its procedures was put forward, further more, it’s summarized that there were eight cases when acquiring operational targets’information and two different ways of target grouping. By discussing the relationship between complex networks community detection and target grouping in operational system, the research approach of using the method of complex networks community detection was established, and a network description model of the enemy’s operational system was set up based on the complex networks model, accordingly, a mathematical model of target grouping in operational system was created. In accordance with the topology features of target groups in operational system, the complex networks modularity was selected as evaluation index of the target grouping result in operational system, which helped to establish a complete problem-solving solution.2. The algorithm HSAPOTDG was proposed for target disjoint grouping when there were full individual attributes of operational targets. In view of different types of individual attributes date of operational targets, divided them into three categories, numeric, nature descriptive and degree descriptive, unified the types using designed method, normalized them adopting Min-max normalization method. By referring to structure equivalence, operational targets’individual attributes and relationships were fused so as to use both to calculate their similarity. The algorithm took all the operational targets as a whole target disjoint group based on hierarchy-split thought at first, and each time deleted the edge between two targets with minimum similarity in enemy’s operational system network model. Then it computed complex networks modularity corresponding to the newly formed target disjoint grouping and the similarity of targets in each target disjoint group. Repeated above steps until all the edges in enemy’s operational system network model were deleted, and then selected the target grouping with maximum complex networks modularity as the final result. Both the experiment analysis and case analysis had showed that HSAPOTDG had relatively strong autonomy and objectivity and its result was reasonable and effective for military use.3. The method for target disjoint grouping without operational targets’individual attributes was brought up. In such case, target disjoint grouping was characterized as NP combinational optimization problem, and the genetic algorithm GAPOTDG, using complex networks modularity function as its fitness function and target function, was designed to solve it. To avoid defects of string encoding and graph-based encoding, it adopted matrix encoding to code each individual which realized individual single-point crossover and needed no decoding. The algorithm PIOTNTS was designed to generate initial individuals by using targets’topological structure similarity in enemy’ operational network model and combining traditional clustering method. Its result had a certain degree of precision and diversity, which reduced the search space and accelerated algorithm convergence. The method to compute the quality of individual’s genes was also provided, and the individual’s optimal gene was exchanged to realize individual single-point crossover. For the invalid solution might be generated during individual crossover, a correction scheme was developed. Mutation individual’s worst gene was randomly split or fused into other genes to realize non-uniform mutation. In addition,μ+λ strategy was used to select progeny population. Both the experiment analysis and case analysis had proved GAPOTDG’s superiority in concise steps, convergence speed and solution’s precision as well as its availability in military application.4. The method for target overlapping grouping without operational targets’ individual attributes was put forward. Based on the topological structure of target groups in enemy’s operational system network model, operational targets were divided into three categories, internal, external and boundary operational target. It was pointed out that a target group with clear group structure should have clear boundaries, and the boundary sharpness of target group was defined as well as overlapping ratio of target group by explaining the meaning of a clear boundary. To quest for an answer and lower the difficulty of algorithm designing to the minimum, the algorithm POTOGBIDG was put up based on the boundary information of target disjoint groups. It used the result of operational target disjoint grouping, and calculated the influences of each boundary operational target in each target disjoint group on the boundary sharpness of other target disjoint groups which connected with the boundary operational target, so to conclude if they could belong to multiple target disjoint groups simultaneously which realized operational target overlapping grouping. In the formula of calculating target groups’ boundary sharpness, there was a controlling parameter r, and its value could be adjusted flexibly to control the overlapping ratio of target groups, making it possible to expose the potential overlapping operational targets existing among target disjoint groups in different hierarchies. POTOGBIDG was proved to be ingenious in design and easy in practice by experiment analysis and case analysis. It was able to realize operational target overlapping grouping in a reasonable and flexible way and was also effective in military application.The problem of target grouping in operational system was systematically studied in the dissertation, which would offer theoretical and methodological support and reference for future research and practice of target grouping in operational system.