Predictive Analytics - it's not just for scientists

Predictive Analytics

It’s not just for scientists

Sponsors Gold

Silver and Bronze Sponsors

Lee Hawthorn, Acma, CGMA, Ba(Hons)

Many roles over the years Software engineer

Finance Manager

Business Analyst

Motivated to learn and share, to solve problems

Current : Data Analyst @ Payzone UK

uk.linkedin.com/in/leehawthorn/

@lee_hawthorn

Blog at Leehbi.com

Agenda

Analytics today

Data Mining with CRISP-DM

Demo’s

Best Practices

Next Steps

3 months to the year end…

MARKETING

HR

SALES

FINANCE

OPERATIONS

Disconnected Data

NEW HIRES

COMPETITOR

MARKETING BOOST

POLICY CHANGE

Connections

EVENTS

ACTIONS

PATTERNS

OUTCOME

RELATIONSHIPS

Stories in the data

WHY DO

CUSTOMERS

CHURN

WHERE IS THE

OPTIMAL

LOCATION FOR

ASSETS

HOW CAN WE

IMPROVE THE

MARKETING

CAMPAIGNS

HOW CAN WE

REDUCE FRAUD

Information Explosion

ANALYSIS GAP

Data Mining History

1990’s

Academia

2000-2010

Largest Companies

2010+

Increasing

demand

2013+

Cloud/Mature

Apps/Services

2015

Black box?

Data revolution

AZURE MACHINE

LEARNING

RAPID MINER

POWER

QUERY/PIVOT

COMMUNITY

DEVELOPMENT

DATA MOVING TO

THE CLOUD

ROBUST

ALGORITHMS1

110 Types of regression, which one to use?

How do we begin?

Cross Industry Standard Process for Data Mining

CRISP-DM

Demo 1

We want to predict the number of

transactions in potential stores that we are

considering to recruit.

Quantitative number leads us to regression.

Keep it simple – Linear Regression

Demo 2

We want to predict the category of customers that will purchase our new e-book reader

For marketing purposes we should classify Early Adoptors to Laggards.

This is a classification problem – we’ll use decision trees.

Best Practices

Executive buy-in

Get the subject matter experts involved early

If you need IT help give them plenty of notice

Document the whole process

Make it repeatable

Challenge and test the model yourself

Use third-party support if you need it

Enjoy yourself

Next Steps

Do you have a skill gap?

Database skills - http://geekgirls.com/category/office/databases/

Statistics skills – Learn R : https://www.coursera.org/specialization/jhudatascience/1?utm_medium=dashboard

Coding skills - https://www.codeschool.com

Join a user group –

http://bavc.sqlpass.org/

http://www.londonr.org/

Conclusion

Data is on the rise

New Techniques are needed

Patterns – Events – Actions

Interesting questions lead to new knowledge

Its not just for scientists USE CRISP-DM – Modern Software & Services

Don’t forget to enjoy yourself.