Back in February we highlighted a way to use Estimize: visualizing the financial network created by its shared analyst coverage. In short, we created a correlation matrix that described the analyst overlap between two stocks. If this user correlation was 0, then the two stocks shared no analysts in common, but if the value was 1.0 then they shared every analyst in common. As you might expect, most stocks had values somewhere in between.
Previously we showed how to use Minimum Spanning Trees (MSTs) to visualize this network. This provided a unique look at the strongest links between stocks, but imposing that the graph be a tree (i.e. with no loops) can sometimes filter out too much information.
Correlation filtering provides another way to visualize the same correlation matrix. This method retains more information from the original matrix, but still filters out weak links that obscure the true correlation structure. The following graph was created by filtering the full user correlation matrix and retaining links that were above a threshold value:
Node sizes and colors are determined by judging the overall importance of the stock to the network; in general, more connected nodes have bigger sizes. Edge sizes are modulated by the strength of the user correlation between stocks: edges are thicker if two stocks have stronger user correlation values and thinner for weaker relationships.
The graph immediately shows the importance of GE, many analysts who cover that stock are active in other stocks. We can also see the split between the highly connected left side of the graph and the loosely coupled right which is centered around MSFT.
Want to learn how to mine social data sources like Estimize, StockTwits, Twitter, and Google Trends? Make sure to download our new book Intro to Social Data for Traders by our very own Thomas Pendergrass