Practical Machine Learning with Prediction APIs

Post on 14-Apr-2017

469 views 2 download

transcript

Practical Machine Learning with Prediction APIs

 Joseph Ferraro, Salesforce MVP, CTO, Mavens Consulting  Vincent Reeder, Technical Architect, Mavens Consulting

Joseph Ferraro Salesforce Developer MVP

CTO, Mavens Consulting

@joeferraro

@mavensmate

Vincent Reeder Technical Architect, Mavens Consulting

@vincentreeder

About Mavens Consulting

 Preferred Healthcare & Life Sciences implementation partner for salesforce.com and Veeva

 50+ Salesforce1 & Heroku consultants located across North America and Europe

 12 Mavens in attendance at #Dreamforce15, speaking in 8 different technical sessions

 #mavensDF15

Agenda

 What is machine learning?

 Practical applications

 How does it work?

 Sales Demo with Code

 Service Demo with Code

 Summary

 Q&A

Machine Learning A type of artificial intelligence (AI) that provides computers with the ability to learn without being explicitly programmed

Machine Learning: What is it?

 Teaching a computer to “do stuff”

 Automatic

 Algorithmic

 Pattern recognition

 Artificial intelligence

http://www.cs.princeton.edu/courses/archive/spr08/cos511/scribe_notes/0204.pdf

Machine Learning: Practical applications

 Spam filtering

 Fraud detection

 Optical character recognition

 Medical diagnoses

 Anticipate customer needs

 Route requests efficiently

 Anticipate deals

Machine Learning: Common algorithms

 Decision trees

 Boosting

•  Spam filtering

 Support vector machines

•  Character recognition

http://www.cs.princeton.edu/~schapire/talks/picasso-minicourse.pdf

Machine Learning Output

  Regression: Estimating or predicting a response

•  “There is a 92% chance this is spam”

  Classification: Identify members of a group

•  “This text is English”

Machine Learning: How does it work?

Develop Model

Idea

Data Train

Model Test/

Analyze Go live!

Machine Learning: How does it work?

 Starts with an idea, but depends on data

 Datasets:

•  Training – optimal dataset, relationship discovery

•  Validation – compare algorithm performance

•  Testing – Apply selected algorithm to real-world data

 Learn!

Demo Use Cases

 Sales Cloud

•  Predict when leads will covert to opportunities

 Serive Cloud

•  Gauge customer mood

Application

 Node.js application running on Heroku

 Exposes endpoints for common machine learning tasks, such as training and prediction

 Subscribes to Salesforce push topics for Leads and Cases via the streaming API

Use Case: Sales

Use Case: Support

Summary

 Machine learning teaches computers to do stuff; cool stuff!

 Collect data, train your model, validate and test it, deploy!

 Not a perfect science

 Powerful when used correctly!

 @joeferraro

 @vincentreeder

Q&A

Thank you