Mapping 132 stocks of OMXH in three minutes


Besides algorithmic trading, what other purposes could machine learning serve when making investment decisions?


While making a fundamental analysis of companies, we usually consider that the balance sheet structure varies between different industries. But how we can study this phenomenon? It is an interesting problem where we can use unsupervised learning to our advantage.


We gathered financial statements of almost all the companies listed in Helsinki stock exchange and segmented the companies using a Self-organising map (SOM). Thirteen financial characteristics were used and we expected that companies operating on a similar industry would group up together.


This is what we found out:

Click the image to download it in higher resolution.


The neural network assigns every company in a node, where companies with similar characteristics are being positioned close to each other. I.e., neighboring nodes will most likely contain companies operating on the same industry.


The image above can be interpreted as follows. The bigger the area of each sector of each pie, the larger the value is compared to others. For example, in the bottom left corner we can see nodes that contain highly profitable companies with a tiny amount of operating expenses and salaries. This kind of clustering is more illustrated in the image below.


PINK CLUSTER: Lots of operating expenses, high salaries, low profitability.


ORANGE CLUSTER: High profitability, low salaries, high amount of properties but a small amount of current assets. Almost no cash and inventories.


GREEN CLUSTER: Small to a medium relative amount of debt, lots of cash and inventories, medium operating expenses.


At least, this is what we found while looking at the map.


In the next graph, we'll take a look at how different companies are positioned in the SOM. We can immediately notice that the companies in the real estate business are located in the orange cluster, which inevitably makes sense when considering their business model. 


In the pink cluster, we have a diverse set of law, consulting and communication companies. 


Finally, the green cluster contains big, stable and asset-heavy companies, such as Kesko, SSAB, Ponsse. One interesting finding is the location of Rovio (top-left), which stands out from the rest of the companies located in the green cluster.

Click the image to download it in higher resolution.


These examples we presented above was just our interpretation. They are not absolute. Hopefully, this helps you to gain knowledge on how companies differ by their financial characteristics. Maybe you can even find your own clusters useful for your investment decisions.


Stay tuned for more!




The dataset contained 132 out of 141 companies listed in OMX Helsinki (last available data). The SOM was taught using R.


Share on Facebook
Share on Twitter
Please reload

Featured Posts

Building a trading strategy around VIX

June 13, 2018

Please reload

Recent Posts
Please reload

Please reload

© Bald & Grey Hair 2018​