An Improved Genetic Reduction Algorithm And Its Application In Detecting The Illegal Radios

Genetic reduction algorithm is one of the rough set theory reduction algorithms.Without changing the classification capability of the knowledge system,attribute reduction aims at deleting the attributes that are redundant,which will generate decision rules or classification rules with better generalization abilities.Attribute reduction plays a key role in avoiding dimensionality curse and raising the efficiency.Those radios broadcasting fake,degraded contents are called illegal radios.Illegal radios will influence the wireless order,and may disturb the normal utilization of the frequency band of aviation.Intelligent monitoring is an effective way to investigating and treating the illegal radios.With no time announcing and the contents are mainly converications are primary features of illegal radios.The core concepts of this thesis are given as follows:(1)The algorithm of time announcing recognition.After preprocessing and feature extraction,the HMMs of the time announcing and non-time announcing frequency modulation broadcast signals are established.The validity of the proposed method is testified by utilizing the practical monitoring data.Conjoined with the routine monitoring parameters,the information system for detecting illegal radios is established.(2)The genetic reduction algorithm with variable-length chromosomes based on equivalence relation information system.Under the framework of genetic algorithms with variable-length chromosomes(VCL-GA),whose fitness function is constructed according to the row-column importance of the simplified binary discernibility matrix.By introducing a novel adjustment function,the crossover operation and mutation operation can be performed adaptively and selectively for different individuals.Also,the validity and superiority are proved by comparing the experiment results of several UCI data sets with the existed method.(3)The genetic reduction algorithm with variable-length chromosomes based on dominance and equivalence relation information system.The fitness function is designed based on the positive region of dominant and equivalent relation information system,which accomplishes the attribute reduction of the decision table whose condition attributes are with dominance relation and the decision attributes are with equivalent relation.The validity of the novel method is tested by comparison with the traditional method.(4)Attribute reduction of the information system for detecting illegal radios by applying the proposed reduction algorithm,which generates more practical and concise judging conditions that are consistent with the monitoring experience.