Predicting company decisicions using knowledge graphs

In a previous post Understanding Company Decisions Using Knowledge Graphs I argued that “using some Graph algorithm” it is possible to predict Google’s acquistion of data analytics company Looker. This post shows how I used “some Graph algorithm” to do that. Translating business reasoning into technical reasoning In the above-linked post I reasoned that: Google acquired Looker because Google’s Cloud competitors are expanding into a segment which is adjacent to Cloud important to Google a segment where Google is not active As you can see, the acquistion happened in a context. The context is that Amazon, Google, and Microsoft are active in the same industries except “data analytics and visualization” where Google is not active but Amazon and Microsoft are. In a graph, this context looks like this: The meaning of the nodes is as follows: yellow nodes: companies red nodes: products green nodes: industry segements With the data model in mind, let’s break the business requirements down and translate them into their technical counterparts. Finding strategic differences between companies using graph visualiziation and calculated ompany similarity Google’s Cloud competitors are expanding into a segment where Google is not active. This means that based on industry segments Google, Amazon, and

Analzing News articles using graph algorithms

When analyzing company investments it is useful to group investors to understand industry patterns. The image below shows such a grouping for some recent invesments in the FinTech-space. The company clusters are immediately visible. Whereas useful, these relationships only capture the most basic data: companyA invests in companyB. More realistically, however, such investments happen in a context; for instance, the above image does not capture the fact that the investors and companies have implict relationships with each other. The image below takes these implict relationships into account: Using Neo4J to analyze groups in this version shows different groups As an example consider Google and Amazon. In the basic version they were not connected… … however, Neo4J’s grouping algorithm put them into the same bucket When we drill down we see why this grouping makes sense: Google and Amazon are both active in the insurance industry. It is worth noting, that this relationship is not explicit but rather through respective partnerships and sub-companies: Google own Verily which in turn has launched Coefficient Insurance Company, a company in the insurance industry. Amazon, on the other side, has partnered with Acko, also a company in the insurance industry.