Analysis of Quality of Living Data of Households of Indian Districts Using Machine Learning Approach of Fuzzy C-Means Clustering

Machine learning is used to analyse the 2011 census data of physical amenities and educational levels of Indian households in order to cluster and categorize Indian districts based on living standards and educational attainment. Household-level data on amenities such as electric lighting, television, mobile, car/scooter ownership; kitchen, toilet, bathing facilities and home ownership of families; as well as educational levels are used for 640 Indian districts, each consisting of 26 parameters (i.e. attributes). This makes the data set fairly large and complex for a clustering problem, and in order to preserve data granularity, fuzzy C-means (FCM) clustering algorithm has been chosen for analysis. The features of the algorithm are briefly presented. The analysis considers 4–10 clusters for the data set. The quality of clustering with larger clusters is discussed with appropriate indices. The results of computation yield the correlation between the variables which allow us to look at the relationships between them. The results also yield the classification of districts in a scale ‘well-off’ to ‘disadvantaged’ for various levels of clustering and show how the number of districts for each variable changes with the number of clusters. It is argued that these results can be used to orient investment plans for various sectors such as education and housing and establish ease-of-living ranking indices for districts which, in turn, can establish a rational basis for coordinated development of Indian regional economies.