In the course of this research, a bespoke function was developed to strike a balance between privacy preservation and data utility. This function is specifically designed to ensure the usability of location-based data while simultaneously safeguarding user privacy. The function employs a gradient descent optimization process to identify the optimal transformation for the dataset that is to be perturbed. This is a crucial step as the original data, if left unaltered, could be vulnerable to privacy breaches. To achieve the perturbation, Laplacian noise was introduced into the original data, with the custom function serving as the objective function for the transformation of the original location based dataset the effectiveness of this perturbation operation was then evaluated using a suite of 26 machine learning algorithms. This was premised on the assumption that potential attackers might utilize machine learning models to infringe upon user privacy. The evaluation was based on a variety of metrics, leading to the conclusion that the methodologies employed in this research were successful. This conclusion is substantiated by a series of rigorous experiments and a detailed case study of homogeneity attack conducted as part of this research.