Density-Based Geometric One-Class Classifier Combined with Genetic Algorithm

One of the most prospective issues in recent machine learning research is one-class classification (OCC), which considers datasets composed only one class and outlier. It more reasonable than traditional multiclass dealing with problematic or special cases. Generally, accuracy interpretability for users are considered to have a trade-off OCC methods. A classifier based on hyperrectangle (H-RTGL) can alleviate such uses H-RTGL formulated by conjunction geometric rules (called an interval). This interval form basis since it be easily understood user. However, existing H-RTGL-based classifiers following limitations: (i) they cannot reflect density target class, (ii) using primitive generation method, (iii) there exists no systematic procedure determining hyperparameter classifier, influences its performance. Therefore, we suggest descriptor 1 − H R D d elaborate including parametric nonparametric approaches. Specifically, design genetic algorithm that comprises chromosome structure operators id="M2"> through optimization hyperparameter. Our study validated numerical experiment several actual different sizes features, result compared algorithms along other classifiers.