Relevant subspace, which is one of efficient means in high-dimensional data analysis, can be used to effectively uncover local outliers hidden in sparse data regions from high-dimensional dataset. However, existing relevant subspace construction approaches are inappropriate for high-dimensional outlier detection due to the assumption of uniform attribute distribution, attribute correlation, and incomparable outlier degree among the relevant subspaces. In this study, a novel relevant subspace-based outlier detection approach is proposed by using Chebyshev Coulomb resultant force, which is an up-to-date similarity measure for high-dimensional data. Firstly, relevant subspace is redefined by adopting Chebyshev Coulomb resultant force, which quantifies the deviation of data objects on each attribute from neighboring or local dataset, so that assumptions of uniform attribute distribution and attribute correlation is avoided. Secondly, we construct Coulomb outlier factor using the magnitude of Chebyshev Coulomb resultant force to address sparse distribution diversity of local dataset in relevant subspace, thus alleviating the comparability of outlier degrees among various relevant subspaces. Thirdly, a contextual outlier detection algorithm, which uses relevant subspace attributes and local dataset as valuable information to explain outliers, is proposed by utilizing relevant subspace and Coulomb outlier factor. In the end, experimental results on UCI datasets validate that our approach can effectively revamp detection performance, – and outliers originated by the algorithm are interpretable.