Outlier detection in non-elliptical data by kernel MRCD

The minimum regularized covariance determinant method (MRCD) is a robust estimator for multivariate location and scatter, which detects outliers by fitting matrix to the data. Its regularization ensures that well-conditioned in any dimension. MRCD assumes non-outlying observations are roughly elliptically distributed, but many datasets not of form. Moreover, computation time increases substantially when number variables goes up, nowadays with common. proposed Kernel Minimum Regularized Covariance Determinant (KMRCD) addresses both issues. It restricted elliptical data because it implicitly computes estimates kernel induced feature space. A fast algorithm constructed starts from kernel-based initial exploits trick speed up subsequent computations. Based on KMRCD estimates, rule flag outliers. performs well simulations, illustrated real-life