Bangladesh is one of the countries in the north-eastern part of South Asia, and it covers an area of about one hundred forty-eight thousand (1.48 × 10 5 ) square kilometers. It is located in the subtropical monsoon climate regime and is one of the most climate-vulnerable countries in the world. Due to the high impact of climate change, climate information is highly demandable to the government and policymakers. The different four types of temperature data from thirty-four locations in Bangladesh are considered in this study to find the homogeneous region and the distributional patterns of the temperatures are revealed as negatively skewed. The non-hierarchical k -means clustering technique is one of the unsupervised machine learning techniques is applied to reveal homogeneous regions in Bangladesh based on different types of temperature (dew point temperature, minimum temperature, maximum temperature, and temperature) separately, and jointly in the present research study. The elbow algorithm is used to find the optimal number of clusters, and the optimal number of clusters is obtained as nine with the different cluster solutions for different temperatures in the present study. The clustering solution produces different clustering results for dew point temperature, minimum temperature, maximum temperature, and temperature. The − regions (Dinajpur, Rangpur, Saidpur) from the northern part of Bangladesh always are the members of a cluster for four types of temperatures. Also, similar results are found for the regions Chuadanga, Ishurdi, and Rajshahi. Besides, the area Mymensingh forms cluster with different members for different temperatures, which indicates that the unique solution is not found. To find a significant and homogeneous cluster solution before applying the non-hierarchical clustering method, the linear principal component analysis is applied to produce a score based on dew point temperature, minimum temperature, maximum temperature, and temperature for each location to find a similar region. The northern regions (Bogra, Dinajpur, Rangpur, and Saidpur) and west-middle regions (Chuadanga, Ishurdi, Jessore, and Rajshahi) construct unique clusters with nearest regions for principal component analysis (PCA) scores. Similarly, the southern area (Cox’s Bazar and Teknaf) constructs a cluster for PCA scores or linear combination of different temperatures in this study. The map for the clustering solution based on PCA scores shows that the nearest and similar regions like hill regions, coastal regions, and plain lands have formed significant nine clusters. Finally, the clustering results based on the first principal component scores based on temperatures for each location are the final clustering solution in the present research work. In addition, the Ward linkage clustering algorithm is brought into play to the clustering solutions of PCA scores and found very significant results. The seven regions from the northern part of Bangladesh (Bogra, Dinajpur, Mymensingh, Rangpur, Saidpur, Sylhet, Tangail) are found in a cluster, and the nearest locations from the middle-west regions (Chandpur, Chuadanga, Ishardi, Jessore, Khulna, Maijdi Court, Mongla, Rajshahi, Satkhira) have formed another cluster with second largest members.