I am working with a dataset contains the daily return time series of 50 different stocks, I want to divide these stocks into several different groups.

My idea was to make a new dataset contains some indicators of these 50 stocks, and use a cluster algorithm to classify them, for example Hierarchical Clustering Method and plot dendrogram to decide how many groups make most sense. The indicators I am using are: Mean, Standard Deviation, Skewness and Sharpe ratio.

First of all, my economic background is weak and I was wondering if there are any other well used indicators could be added to this dataset to make the classification make more sense.

Secondly, too many indicators can make it difficult to interpret the stocks being classified.

Can someone share some ideas?


Question 1: Geograhic location, market sector, it all depends on what you're trying to achieve with the clustering.

Question 2: Yes. Adding too many indicators will cause your data to overfit and at some point you'll end up with every cluster having 1 member. I would use a regularisation penalty term to counter this. Either L1 (Lasso Regression) or L2 (Ridge regression) would work for this. Or you can use Elastic Net which would use a portion of both.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.