Using Stock Clustering to Support Portfolio Diversification
Summary
The document proposes clustering stocks using financial statement ratios such as price-to-book, price-to-earnings, and return on equity, potentially alongside market measures such as beta and annual price change. The suggested workflow is to form groups from these features, select stocks across multiple clusters, and then assign portfolio weights using a conventional method such as mean-variance optimization. The aim is to avoid relying solely on a stock’s recent return and volatility when planning a portfolio for the next period.
This is a hypothesis posed for discussion, not a tested strategy. The document presents no empirical results or detailed clustering procedure. It raises feature selection as an open issue but does not resolve practical concerns such as scaling, correlated inputs, report timing, look-ahead bias, cluster stability, or whether cluster membership improves diversification out of sample.
Key ideas
- Cluster stocks using selected accounting and market features before choosing portfolio holdings.
- Combine representatives from different clusters, then apply a separate portfolio-weighting method.
- The proposed approach seeks to complement reliance on recent returns and volatility.
- Feature design and out-of-sample validation are unresolved in the document.
Tags
Full text
# Stock clustering as a part of portfolio diversification # Stock clustering as a part of portfolio diversification I have a research hypothesis and now I'am trying to look at it from different angles.Now I am a bit puzzled.Maybe someone is also interested in machine learning application(especially clustering) in financial area and can advise something. There are many portfolio theories, for example, Markowitz portfolio theory, that assign weights for every share in portfolio.Imagine, that at the end of a certain year I am estimating results and planning my asset portfolio for the next year. Shares, that perfomed good, can fall during the next year and vice versa. So, it seems reasonable not to rely entirely on the yield and volatility of a security's price over the year. So, my hypothesis is that we can analyse annual financial reports of stock issuers. For example, choose group of financial multipliers (P/B,P/E,ROE) also maybe compute some annual metrics such as beta coefficient, average price change over the year (in %) and group shares into different segments. After that, we can pick elements from all clusters and only after that assign weights using traditional theories.Ideally, we will get a group of stocks with different performance and seems that it could have a positive impact on portfolio sustainability. What do you think? Does it sound adequately? Are there any pitfalls in my theory? What to look for when selecting features for clustering? Thanks everyone in advance!
Shown in full with attribution under the source's licence. Licence: CC BY-SA 4.0 (Stack Exchange)
This summary was written by Stratmill's research agent from the original; it is not a copy of the source.