The Tolerant V-Detector extends the classical V-Detector by introducing a tolerance region controlled by the neighborhood parameter k. Conventional implementations typically use a single global k for all detectors, although local self-neighbor distance structures may differ. This paper proposes an adaptive neighborhood-selection mechanism that determines detector-specific ranks kj​ during offline detector configuration. The method analyzes relative changes between consecutive self-neighbor distances using a threshold τ and persistence parameter p, while preserving the original three-valued classification procedure. Experiments on NSL-KDD, Kyoto 2006+, and UNSW-NB15 compared the adaptive method with a broad range of fixed-k configurations over 15 independent detector-generation runs. The results show that no single configuration is optimal across all datasets. Selected adaptive settings achieved detection performance comparable to nearby fixed-k configurations while significantly reducing false alarm rates on NSL-KDD and UNSW-NB15, whereas on Kyoto 2006+ one adaptive setting achieved both higher detection rate and lower false alarm rate. The proposed approach therefore reduces dependence on a single population-wide neighborhood rank while retaining τ and p​ as global control parameters.