building a beer recommender with yhat (papis.io - november 2014)
DESCRIPTION
Building the predictive aspect of applications is the fun, sexy part. New tools like scikit-learn, pandas, and R have made building models less painful, but deploying/embedding models into production applications is challenging. We'll show how Yhat makes deploying predictive models written in Python or R fast and easy by building a beer recommendation system and an accompanying webapp.TRANSCRIPT
Building a Beer Recommender with Yhat
PAPIs.ioNovember 2014
1) Intro2) The Problem: Building predictive apps3) Solutions4) Case Study: Beer Recommender5) Demo6) Q/A
Here I am on the Internet.
Founder/CTO @ Yhat
Hi, I’m Greg!
Founders Company
InvestorsGreg Lamp, CTO
Austin Ogilvie, CEO ● Launched in 2013● HQ in New York
Why is Data Science so talked about?
regression
Get Raw Data
Strategic Insights
Real World Scoring
Data Driven ProductsBusiness Impact
Clean Data
Stages of the Analytics Project Life Cycle
Expert data teams
Management
Customers & Front Line Employees
What makes building analytical apps hard?
Hi, I’m Trey.
Meet Trey, the Data Scientist
We need to reduce churn. Okay. I'll look into it.
I figured out that....some complex stuff about vector space that'll improve...
....and that's how we'll reduce churn.
Sounds good. Let's do that...
The "a ha" moment isn't the end.
Any of you know what Gradient Boosting is?
So when can we go live with the new model?
Now what?
Options
1) Translate Code
2 Rebel Policemen 2
2) PMML
?
3) Batch Jobs
use your tools
use your tools move quickly
use your tools move quickly
any workflow
use your tools move quickly
any workflow no translating
Case Study
+ = ?
A Beer Recommender in Python
The Data
http://snap.stanford.edu/data/web-BeerAdvocate.html
Beers
Users
Ratings
Distance
vs
vs
measuring similarity
eeny
? ?
eeny meeny
?
?Cosine
eeny meeny miny
?Cosine
moe
pick one.you can always
change
Thank you,
Scoring
Aggregate
Sort
Filter
Return
Deployment
What does this mean?
Import Yhat
Create a YhatModel
Define execute
Grab incoming data
Call your function
Format and return results
deploy your [email protected]:yhat/beer-rec.git
Python/R Source on github
We’re [email protected]
Thanks!@yhathq
Appendix
アーノルド坊やは人気者Little Arnold Boy is Popular
Diff’rent Strokes
Unitless y-axis
Time