Global Positioning System (GPS) signals are often unreliable in indoor environments due to attenuation and multipath effects caused by surrounding structures. As an alternative, Wi-Fi fingerprinting-based indoor positioning systems (FPIPS) have gained traction, leveraging Received Signal Strength Indicator (RSSI) values and the widespread availability of Wi-Fi infrastructure without requiring additional hardware. However, improving localization accuracy while maintaining low computational cost remains a persistent challenge. A region-based K-Nearest Neighbour (RB-KNN) algorithm has shown promise, yet the process of defining regions and selecting their optimal number is still non-trivial. This study investigates the use of four unsupervised machine learning algorithms–K-Means, DBSCAN, Gaussian Mixture Models, and Agglomerative Clustering–to automatically form fingerprint regions for RB-KNN. By systematically evaluating these approaches, the work aims to identify the most effective clustering strategy for enhancing localization accuracy and efficiency in campus-scale indoor environments. Experimental results demonstrate that Agglomerative Clustering outperforms other methods, achieving a mean positioning error of 4.44 m, a 38% improvement over the baseline RB-KNN approach. The findings highlight the potential of machine learning-based region segmentation to significantly enhance the performance of Wi-Fi fingerprinting-based indoor positioning systems.