Category Archives: Predictive Analytics

Back to Civ Tech

I have the great pleasure to report that I am back in civic technology.  I my new role as a data scientist with the State of Colorado, I will be merging siloed datasets from different agencies to create usable, anonymized data that researchers and policy makers can use to inform public policy recommendations!

Just a bit more about my new position….I am working for the University of Denver as a contractor to the Office of Information Technology for the State of Colorado.  I will be working across the various agencies and with IT and the Colorado Evaluation and Action Lab’s LINC (Linked Information Network of Colorado project).  This is a great opportunity to do important work for the state of Colorado.

Here are a few examples of the kinds of projects that I will be working on:

– Does rapid intervention with runaway children improve outcomes (graduation, criminal convictions/arrest, etc.)?

– What are the primary factors affecting non-custodial parents in the timely payment of child support payments?

What factors are associated with multi-system utilizers of the Colorado health care system?

Google IO Developer Keynote

The Denver Google Developer Group set up a meeting to watch the simulcast broadcast of the keynote by Sundar Pichal and related speeches.  The meeting was held over at GoSpotCheck, a startup doing retail inventory verification using AI and machine learning.

The keynote echoed the expected themes: vision, audio, and AI.  Google is developing the building block technologies to increase the amount of AI and machine learning tools that can be deployed.  This means that they not have AI training boards that can do petaflops of computations, they have new libraries for audio and vision, they have expanded TensorFlow to include more than just tensors/neural nets (i.e., other machine learning algorithms like random forest and support vector machine), and they are creating TensorFlow libraries that will run with javascript.  This means that AI models can be pushed down to Android cell phones and tablets and can also call upon their off-the-shelf tools that they have developed.

Bottom line: Between Caffe2,  TensorFlow, Python, and R, there are an abundance of open source tools to do machine learning.

On the meeting side, I met some great people doing interesting development, and I even won a Google laptop bag as one of the raffle prizes!

Using Predictive Analytics for Flu Prediction

The NYTimes did a piece today about the use of predictive analytics in analyzing the spread of the flu virus.  The source of the data?  Google searches.  The algorithm could accurately predict when the virus would peak 65% of the time.  That’s not too bad, and just think about it–the analysts are predicting this without drawing any blood, having any doctor intervention, etc.  This is based purely on Google searches!

Click here for the story.

Started my Masters in Predictive Analytics

Today I started the program at Northwestern University.  My first two courses are a strategy course in Predictive Analytics (business side) and a Statistical Methods course (numbers side).  I am really looking forward to the deep dive into the subject after a number of years of using some of these technologies.

The other students look highly qualified and accomplished, with many coming from senior positions at Fortune 500 companies, consultancies, predictive analytics vendors, and others.  Click here to see the program.