A fitting model based intuitionistic fuzzy rough feature selection

Abstract

Feature subset selection is an essential machine learning approach aimed at the process of dimensionality reduction of the input space. By removing irrelevant and/or redundant variables, not only it enhances model performance, but also facilitates its improved interpretability. The fuzzy set and the rough set are two different but complementary theories that apply the fuzzy rough dependency as a criterion for performing feature subset selection. However, this concept can only maintain a maximal dependency function. It cannot preferably illustrate the differences in object classification and does not fit a particular dataset well. This problem was handled by using a fitting model for feature selection with fuzzy rough sets. However, intuitionistic fuzzy set theory can deal with uncertainty in a much better way when compared to fuzzy set theory as it considers positive, negative and hesitancy degree of an object simultaneously to belong to a particular set. Therefore, in the current study, a novel intuitionistic fuzzy rough set model is proposed for handling above mentioned problems. This model fits the data well and prevents misclassification. Firstly, intuitionistic fuzzy decision of a sample is introduced using neighborhood concept. Then, intuitionistic fuzzy lower and upper approximations are constructed using intuitionistic fuzzy decision and parameterized intuitionistic fuzzy granule. Furthermore, a new dependency function is established. Moreover, a greedy forward algorithm is given using the proposed concept to calculate reduct set. Finally, this algorithm is applied to the benchmark datasets and a comparative study with the existing algorithm is presented. From the experimental results, it can be observed that the proposed model provides more accurate reduct set than existing model.