Improved Rule Based Classifier Based on Decision Trees (IRBC-DT) for Gastric Cancer Data Classification

摘要

Objectives: To design and develop an improved rule based classifier based on decision trees (IRBC-DT) for Gastric Cancer data classification with increased accuracy, hit rate and substantial reduction of elapsed time. Methods/Analysis: At the initial stage, IRBC-DT mingles a pair of techniques, namely the boosting and arbitrary sub-space, in order to build rules based on classification. As a result, the subsequent level divides the dataset into two parts in which the first set for training the data and the second one for pruning. Then a decision tree is built for analyzing the misclassified instances. Findings: Each feature is tested and assigned with precise weight for which the k-nearest neighbor classifier is applied, based on the weighted features. As a final point, the algorithm will get updated with the instances which contain the misclassified class labels. Once subsequent to analysis and updating the instances of misclassified class labels, the conflicting rules are checked and the same are removed. Attribute bagging is a set of classifier that operate on a sub-space of the original element space, and produces the class corresponds to the result of those unique classifiers. Random subspace scheme has a striking option of data classification that ensemble with considerably more number of features, such as cancer data. Also, boosting is modeled particularly for classification, which alters the weak classifiers into strong ones by means of an iterative process. Boosting mechanism makes use of selecting the apt classification in order to coalesce the complete classifier results. Applications/Improvements: IRBC-DT is implemented in MATLAB and can be applied in healthcare sector. From the results it is perceived that the method gains better performance than that of the existing algorithms for gastric cancer data classification.