Portfolio theory with Apache

Josh Kalina from Refinitiv

May 14th, 2019, 4:10 PM (duration: 20 minute)

Location: Revolution Brewing- Brewpub

Markowitz portfolio theory was a revolution in the way that Finance viewed risk and reward in Capital Markets. However, the theory’s pure application is usually relegated to theory in favor of more computationally lighter ways of reinterpreting Markowitz's original ideas. Within the formula, there is a covariance matrix that needs to be generated relative to the size of the number of assets. This is usually where Python, R or SPSS fail. To give context here, there are close to 110,000 publicly traded companies worldwide. That means creating a covariance of returns as large as 110,000 by 110,000. This creates issues too large for the likes of Python or R. In practice, individual assets are generally substituted out for sector funds or broad indices like the Russell 2000 and Russell 3000. However, lets give Markowitz his due. The goal of this presentation is to show how using Spark and the cloud, we can perform this computation within memory to get a better view of how a global portfolio should be weighted.