the usefulness of feature selection in auditors opinion type prediction

abstract: despite the importance of predictive variable in prediction, in most of the research in the field of auditors’ opinion the purpose was rendering the suitable models. meanwhile, less attention was paid to the selection of optimal predictive variable and appropriate models of these selection. therefore, in most of these research the predictive variables were chosen randomly and according to the prior research. the process of selecting variables could be used as a preprocess for omitting irrelevant variables and selecting optimal variables before creating the model. in this regard, this study investigates the usefulness of correlation-based features selection (cfs) in auditors’ opinion prediction of listed companies in tehran stock exchange. the classifiers including artificial neural networks (ann) and logistic regression were used. in overall, the experimental results of investigating 1214 firms-years from 2008 to 2015, confirmed the usefulness of cfs method in predicting auditors' opinion. in other words, the application of the cfs method, increases the mean of accuracy in comparison with using all variables, and reduces the occurrence of type i and type ii errors. furthermore, the results indicated that ann outperforms the logistic regression.