Clustering underlying stock trends via non-negative matrix factorization

Building a diversified portfolio is an appealing strategy in the analysis of stock market dynamics. It aims at reducing risk in market capital investments. Grouping stocks by similar latent trend can be cast into a clustering problem. The classical K-Means clustering algorithm does not fit the task of financial data analysis. Hence, we investigate Non-negative Matrix Factorization (NMF) techniques which, contrary to K-Means, turn out to be very effective when applied to stock data. In particular, recently developed NMF techniques, which incorporate convexity constraints, generate more disjoint latent trend groupings than the traditional sector-based groupings. In this paper, the NMF technique and its variants are applied to NASDAQ stock data (i.e., daily closing prices). Experimental results confirm that (convex ) NMF techniques are highly recommended to produce trend based assets and build a good diversified portfolio.