Decision Tree Pruning Method Based On RST

Data Mining is a process to mine available, valid and comprehensible pattern fromlarge-scale data in way based on computer including other new technologies.There are many popular methods for Data Mining, such as statistic analysis method,neural network, decision tree method, genetic algorithm and so on. As a widely used methodfor classification, decision tree method induces the rules that are showed by a tree that canfind some valuable mechanic and potential information.Now, much decision tree building method can get a decision tree with good precision,but most of them need large computation and, have limitation with generalization ability.Mathematician Z.Pawlak from Poland proposed Rough Set Theory that is another new toolafter the probability theory, fuzzy set and evidential theory to deal with the uncertainknowledge, and its validity has been conformed from the successful uses in various scienceand engineering domains in recent years. Rough Set Theory is introduced in decision treesand gets better result through the analysis on theory and experimental validation.We can discover that most of the post-pruning method only pays attention on the wholeand, proposed a new pruning strategy-decision tree pruning method based on leaf .The mailcontent of this dissertation is as below: 1 Decision Tree Building: Introduce decision tree simply. Depict the famous decisiontree building method ID3 and its improved method C4.5 that is now widely used. 2 Decision Tree Pruning: Decision tree pruning is an effective method to improve thedecision tree extensiveness and to avoid over-fitting. There are pre-pruning and post-pruningstrategies. In this section, we will analysis six post-pruning method. 3 Decision Tree Building and Pruning method based on Rough Set Theory: Rough SetTheory can deal with uncertain knowledge and so is introduced into decision tree building andpruning method to eliminate the noise, by less computational complexity can we get adecision tree with high precision.4 Decision Tree Pruning method based on Leaf Node: Mostly, Post-pruning methodinstitute pruning strategy by comparing the effects before and after pruning the non-leaf nodein decision tree, it neglects the contribution of separate leaf node so that affect the effect ofpruning. To avoid this phenomenon, a pruning method based on leaf node is proposed to makepruning strategy upon the significance of every leaf node. The new idea is proved feasible byexperiment.