Use of PageRank for Tech Companies
What are the uses of Page Rank apart from being used by Google for their search algorithm? In this article from Travis Addair, a Stanford University graduate, and now an engineer at Uber, talks about how Page Rank can be used for ranking Tech Companies. The links are now the employees moving from one company to the other. According to Travis, “Applying PageRank, we can rank tech companies to see which have the highest concentration of top talent.” After applying the ranking algorithm, it can help determine more information about each company, and we can decide which company we think suits us more.
The data collected results to “a graph of 34,910 company nodes with 40,999 job transfer edges” (Addair 2) With Google having the most profiles with 1462 and other big companies like Microsoft, Amazon, Apple and Intel all in the top 10. After applying the page rank algorithm to the data set, the top 10 companies results were hardly surprising. Google is number 1 with 0.0222 PageRank score, while other big companies filled out the top 9, with number 10 being Carnegie Mellon University. This is due to the fact that interns at large tech companies went back to do research at the university. The result showed almost every company with more profiles have higher page rank, but it is not always the case. Modifications were also made to the PageRank algorithm, such as adding an edge for every year spent at the company before an employee transfers out, but the results stayed more or less the same.
However, when a recency bias was added, which is when more edges are added to the employees’ current company, the result changed a lot. Uber moved up to 2nd while Airbnb moved up to 7th from 13th. After some more adjustments, the final results still show Google at the top, which compared to LinkedIn’s proprietary algorithm, “LinkedIn Attractors” and Paysa’s “Paysa CompanyRank”, makes a lot of sense. The top 10 between Addair’s result and LinkedIn’s are very similar with Paysa’s being a little different. The conclusion was that the algorithm works well, except for those companies with less than 100 profile samples. This goes to show how PageRanks is actually very useful in a lot of cases, and can be used to determine a lot more things than just website rankings.
http://snap.stanford.edu/class/cs224w-2016/projects/cs224w-23-final.pdf