Quantitative clustering techniques solely rely on grouping a bunch of stocks based on returns and their correlations, seems to fail while trying to classify a set of stocks in an industry. I ran a test on set of 20 stocks from utilities and couldn’t find meaningful clusters just based on this data analysis alone.
Another technique which helped me though was the fundamental classification method. I created a vector of factors for this sector and rated each stock’s exposure to this factor. Then grouping stocks which had very similar factor weights seemed to show a lot of promise.
I should take examples from Indian stocks and elucidate this more clearly in the next post on this topic.