Credit Rating Model
Student Name: T.Kamal Kanth
In this project an unsupervised learning technique
was implemented to understand the process of bond ratings provided by rating
agencies like CRISIL, ICRA or CARE. Again the standard financial ratios were
used as the indicators or independent variables for ratings. However the
ratings by a rating agency were not used for their modeling, unlike other
studies conducted on similar topic under my supervision. Instead here the
objective was to discover the kind of rating a class of companies with like
financial ratios should get, from the available data. Towards this end, a
cluster analysis of financial ratios of manufacturing firms was conducted,
and significant difference between the mean vectors of the proposed clusters
were confirmed using MANOVA. Then the CRISIL and ICRA ratings of the firms
falling in the same cluster were observed and a remarkable consistency was
found to exist between these rating agency ratings and the formed clusters.
This in a way validated and established a new relationship between ratings
of rating agencies and financial ratios, apart from providing an alternative
new framework and model of predicting or providing a rating.