Research And Application On Data Mining Based On Genetic Algorithms

Data mining is a new subject formed with the development of theinformation technology and is a new research point in database area.Classification is one of the important themes in Data mining.Decision Tree is a method of Data mining that is used widely to miningclassification model. It has studied widely and made a great progress.However, Decision Tree always leads to be over-fitting, to have largescales and induce longer classification rules In that it adopts greedyalgorithm. According to these shortcomings, an approach to data miningwhich integrates genetic algorithm and association rule algorithm isproposed in this paper.Associate rule is an important theme in the data mining to discoverpreviously unknown, interesting relationships among attributes from largedatabases. In view of the Apriori algorithm's limitation: firstly, it needs agreat of candidate itemset when the frequency itemset is very large;secondly, it needs a lot of I/O operation when the database is very large.An improved Apriori algorithm based on partition technology is proposedin this paper, and the same time we established an Association rulediscovering model combining the improved Apriori algorithm and thefeature of genetic algorithm. There are four main points in this thesis asfollows:1. The conception in data mining, basic process, main technology and thedevelopment of the Situation are described in this theme. And the sametime, some useful classification algorithms are discussed such as decisiontree, genetic algorithm, and neural networks.2. Biological Principle, mathematic principle, search principle and thefeature are completely described in the paper.3. The approach to association which integrates genetic algorithm isdescribed in the paper, besides, through one application example in thismodel, proved that this model is feasible and effective in the associationrule discovering.4. The approach to classification which integrates genetic algorithm andassociation rule algorithm is described in the paper, besides, comparedwith the traditional algorithms by experiment.Finally, the characteristics of genetic mining are summarized in thispaper. Then it expects what we can do in the further reserarch work.