Outlier detection is a fundamental task in data analysis, whenever there is the need to asses if anomalies are present in an otherwise coherent dataset. In this work, we consider the Local Outlier Factor, a density based algorithm that gives an outlier score to each point in the dataset: the focus is on tackling the problem of tuning the locality of the algorithm search, which is determined by the number of neighbors k chosen for scoring. A novel approach is presented, based on integrating a clustering procedure, the Bayesian Bagged Clustering algorithm: the neighborhood parameter selection is guided by selecting a percentage of each cluster’s total number of points as k. Preliminary results of the proposal are shown on simulated data, demonstrating robustness with respect to the new parameter definition.