Analysis of Software Deformity Prone Datasets with Use of AttributeSelectedClassifier

Software Deformity Prone datasets models are interesting research direction in the era of software world. In this research study, the interest class of software deformity prone is defective model datasets. There are different techniques to predict the deformity prone datasets model. Our proposed solution technique is AttributeSelectedClassifier with selected evaluators and searching method for reducing the dimensionality of training and testing data provided by defected models NASA datasets by attribute selection before being passed on classifiers. We have used three evaluators and search methods. These evaluators are CFSSubsetEval, GainRatio and Principal Component Analysis (PCA). The search methods are BestFirst and Ranker. We have used 12 different classifiers for analyzing the performance of these three evaluators with search methods. The experimental results and analysis are measured with True Positive (TP-Rate), Positive Accuracy, Area under Curve (ROC) and Correctly Classified Instances. The results showed that that CFSSubsetEval and GainRatio performance is better in almost classifiers. Hoeffding tree, Naive Bayes, Multiclass, IBK and Randomizable filtered class increased performance in Positive Accuracy in all techniques. Stacking has worst performance in positive accuracy and True Positive tp-rate in all over technique.

@article{Malik2019,
  title     = {Analysis of Software Deformity Prone Datasets with Use of AttributeSelectedClassifier},
  journal   = {International Journal of Advanced Computer Science and Applications},
  volume    = {10},
  number    = {7},
  year      = {2019},
  publisher = {The Science and Information Organization},
  author    = {Maaz Rasheed Malik and Liu Yining and Salahuddin Shaikh},
  doi       = {10.14569/IJACSA.2019.0100703},
  url       = {https://doi.org/10.14569/IJACSA.2019.0100703}
}