Join John Akred, the Chief Technology Officer of Silicon Valley Data Science, as he discusses using diverse and unstructured data types to predict events.
Event Details
How many times have you been running late and were excited by the sound of an approaching "El" train – only to realize it was the third train going in the opposite direction?
We too wrestle with the challenges of public transportation. In fact, the Silicon Valley Data Science headquarters are located next to the Bay Area's commuter line – a constant and disruptive reminder of our daily frustrations with the train.
We put our frustration into action by taking advantage of our location and integrating other publically available sources to create an event detection system that monitors the health of the train network in real-time. In order to gain understanding of a physical system we could only observe in limited ways, we created a variety of data streams including video and audio that detect the sound and image of trains as they pass by.
In this informative talk we will give a brief overview of the techniques we used for audio and video detection. We will then describe how we went about combining our physical observations both with each other and with external data sources such as Twitter and public APIs. By taking into account the different properties of each data source, we will show how to arrive at a prediction more accurate, and more robust, than any one of its parts.
We will conclude by discussing how event detection techniques are useful in other business settings such as industrial engineering and public health to make more informed decisions about the environment.
Program
6:30 PM-7:00 PM: Networking w/hors d'oeuvres and cash bar
7:00 PM-8:00 PM: Program Presentation
Questions
Roger Moore, '92
Senior Principal / Sagence
312-543-1319