Research On Quantitative Association Rules Algorithm Based On Niche Genetic Algorithm

With the continuous escalation of genetic testing methods and the continuous decrease in the cost of genetic testing,more and more diseases can be treated using gene technology.In gene therapy technology,the most important thing is to search for pathogenic gene loci that need treatment.The main purpose of this paper is to explore the pathogenic loci of polygenic diseases and find out the hidden pathogenic gene information through genetic data excavation.The specific research contents and research results are as follows:1.Research is made on the NICGAR algorithm for excavating association rules based on niche genetic algorithm,discovering three problems to be solved in the algorithm.The first problem is that the time complexity of the EP update process is too high;for this reason,we have introduced heap sorting and "distance" keeping method for improvement.The second problem is that the real distribution of the dataset cannot be reflected in the data initialization process;for this problem,we propose the concept of "data density" and approximately stimulate the distribution of the data to set the amplitude from the "data density"perspective.The third problem is that the mutation operator uses the fixed amplitude as the amplitude of variation;for this problem,we believe that if the data density is high,the amount of change in its range will decrease accordingly.If the data density is low,then the amount of change in its range will increase accordingly.2.In order to illustrate the feasibility and superiority of the solutions for the above three problems,four comparison experiments were conducted in this paper,including the comparison between the improved algorithm and the original NICGAR algorithm,the improved algorithm and the two types of NGAs algorithm,the improved algorithm and the four single-objective evolutionary algorithms,the improved algorithm and the two multi-objective genetic algorithms.In the comparison experiment with the original algorithm,we find that the improved algorithm is far superior to the original algorithm in some aspects,and even if the original algorithm is dominant,the advantage is relatively small;in the comparison with the two NGA algorithms(Clearing and ASCGA),the advantages of the improved algorithm are confirmed by the Wilcoxon signed rank sum test and the data values;in the comparison with the four single-objective evolutionary methods(EARMGA,GAR,GENAR,and Alatasetal),the Friedman non-parametric test and paired T-test also confirm that the improved algorithm outperforms all single-objective algorithms in terms of interest degree extraction;and in the comparison with two multi-objective genetic algorithms,the Friedman non-parametric test and paired T-test are also used,confirming that the improved algorithm has higher advantage in the extraction of high interest and short rules.3.The improved algorithm of this paper is applied to the mining of gene datasets.The experimental results show that the extracted rules are mainly based on short rules,which are more easily understood.At the same time,they also contain a small number of long rules.Such rules have certain specialities.Embodies the additive effect of genes.Therefore,the improved algorithm has certain feasibility in gene data mining.