The challenge of data clustering, especially for text data in low-resource languages such as Albanian, is exacerbated by the lack of linguistic resources and research. This paper evaluates the performance of three unsupervised machine learning clustering algorithms: K-Means, DBSCAN, and HDBSCAN, on datasets sourced from social media posts in Albanian, where clients expressed their opinions about services of two local telecommunication companies: first and second company. The methodology involves comprehensive preprocessing steps: cleaning, tokenization, and vectorization using TF-IDF. The results indicate that the K-Means algorithm achieved optimal clustering with an accuracy of 70.1% for the first company dataset and 76.4% for the second company dataset. DBSCAN, though effective in identifying dense clusters, showed lower accuracy due to outliers. HDBSCAN improved data density clustering but had reduced overall effectiveness. The study concludes that while K-Means is reliable for clustering Albanian text data, the unique attributes of low-resource languages require further refinement.