Just add Imagination

© 2016 KNIME.com AG. All Right Reserved.

KNIME User Meetup Mountain ViewJan 30, 2017

© 2016 KNIME.com AG. All Right Reserved.

Just add Imagination …

Rosaria Silipo

KNIME

© 2016 KNIME.com AG. All Rights Reserved.

The Evolution of Data Analytics

3

Classic Analytics of CRM Data

More Data: Customer Intelligence

Change Objective: Restocking Strategy in Retail

No Historical Examples: Anomaly Detection in IoTPredictive Maintenance

Change Perspective: User Experience

© 2016 KNIME.com AG. All Rights Reserved. 4

Classic CRM Analytics: Churn Prediction



5

CRM SystemData about your customer• Demographics• Behavior• Revenues

Model

• Churn Prediction• Upselling Likelihood• Product Propensity /NBO• Customer Segmentation• Campaign Management• …



6


Churn Score for each Customer

7

YouTube: “Building a basic Model for Churn Prediction with KNIME” https://www.youtube.com/watch?v=RHsO10q7e2Y

https://www.youtube.com/watch?v=RHsO10q7e2Y


Classic CRM Analytics: Customer Segmentation


Classic CRM Analytics: Customer Segmentation

9


KNIME Web Portal: Deploying Configurable Workflows

10


Classic CRM Analytics: CRM Data, Train a Model, Use the Model

11

Daily CallersNon-Daily Callers

No Voice Messages


Include other Data Sources: Social Media

12

Why only CRM Data?Web Pages, Google APIs, Sensors, Social Media, …


Expand your DataCustomer Intelligence on Social Media


The Problem

A major European Telco

14

• Can you tell us what people say about our new product?

• Can you tell us who is supporting the product and who trashing it?

• Of those, can you tell us who is an influencer?

Its Forum Site


The Techniques

15

Text Analytics Network Mining


The Data

16

• The Data Set unfortunately cannot be shared

• Slashdot Forum Data are!

• Slashdot was a public forum built in 1997 and hosting a number of discussions: from software to philosophy, from science fiction to politics.

• Politics was the biggest discussion group

• So, politics is what we analyzed to find out:– What users were thinking about a political issue

– Who was pro and who was con

– Who was an influencer


Text Analytics and Network Mining: Workflow

Read Data

Network Mining

Sentiment Analysis

Joiner

Scatter Plots


Text Analytics: Results

Most negative user pNutz

Most positive and most

talkative user dada21


Network Mining: Results

Dada21

Carl Bialik from the WSJ

Doc Ruby


Text Analytics and Network Mining: Results

pNutz

Carl Bialik

dada21

Doc Ruby

99BottlesOfBeerInMyF

WebHosting Guy

Tube Steak

Catbeller

from the WSJ


New Analysis Domains: Text, Networks, Images, …

21

Why only Prediction? Text Analysis, Sentiment,

Network Graphs, …


Social Media for User Experience:Twitter & Online Dating


You have got Options!

23


Most Popular Online Dating Sites

24


As a Topic Monitoring Tool

25


New Prediction Objectives: Prices, Numbers, Failures, Restocking, …

26

Why only Churn? We could predict Stock Prices, Iot Mechanical

Failures, Traffic Numbers, Bike Restocking Needs, …


From Churn to Bike Numbers:Restocking Strategies


IoT

28

Household

Energy

Wearables

Health

City

Industry


The Challenges

• Handling very large amounts of data created over time

• Forcing sensor-equipped objects (house or city) to learn, and therefore to become smarter


This Use Case

Capital Bikeshare in Washington DC


The Business Challenge:


Even MORE of a Business Challenge

• Any Station without bikes for 1 hour:

$XXXX Per Violation

• Any Station with no free slots for 1 hour:

$XXXX Per Violation

© 2016 KNIME.com AG. All Rights Reserved.http://bikeportland.org/2013/03/10/behind-the-scenes-of-capital-bikeshare-84006

Capital Bikeshare Response

Over 3 years

307 Stations2963 Bikes19.4% Casual Bikers5.9m Bike Moves


Top 250 Routes


The Goal

• Restocking alert signal

• 1 hour warning! Lag(flag-1)


Input Attribute Impact


Lean Restocking Alert System

The input feature subset with the smallest error (81% accuracy):

• Hour of the day

• Working day (Y/N)

• Current Bike Ratio

• Terminal (station code)

Past of bike ratio and weather infos do not seem to be relevant!

Is this because most bikers are registered members?


The hardest Case: No Class Examples

38

Can I still learn to predict if I have no Historical Examples for that

class?


The ultimate Challenge: No ExamplesAnomaly Detection in Predictive Maintenance


3. Dynamic Unsupervised Anomaly Detection.

Here some measures change over time till their values are not normal anymore. For example, while a motor is slowly deteriorating, one of the measurements might change till it gets out of control and the motor breaks. We want to stop the motor before it completely breaks producing even more damages.

This problem is similar to number 2 but slightly more challenging because it is not pattern based and it changes slowly over time.

Anomaly Detection: The Approach


Until Now! A Motor and its Sensors

28 time series from 28 sensors on 8 different parts of a mechanical engine.

A1 (input shaft vertical)

A2 (second shaft horizontal upper bearing)

A3 (third shaft horizontal lower bearing)

A4 (internal gear 275 degrees)

A5 (internal gear 190,5 degree)

A6 (input shaft bearing 150)

A7 (input shaft bearing 151)

M1 (torque KnM)


The Data

• Time Series are FFT-derived Spectral Amplitudes

• There is only one motor breakdown episodes on July 21, 2008

• The breakdown is visible only from some sensors and only in some frequency bands

• The engine was substituted with a new one after the breakdown


Data Visualization: Time Plots by Frequency Bands

A1-SV3 [500, 600] Hz

New motor piece

Old motor piece

Breaking pointJuly 21, 2008

Only some Spectral Time Series shows the break down

A1-SV3 [0, 100] Hz


Learn “normal”: Training Set

A1-SV3 [0, 100] Hz

A1-SV3 [500, 600] HzBreaking pointJuly 21, 2008

Only some Spectral Time Series shows the break down

31 August 2007

Training Set

Predictive Maintenance


Time Series Production: Second Level Alarms R Stacked Plot

Sep 2007

Mar 06 2008

May 05 2008

MA

(alm

(t))

Jul 2008

A7, SA1 [200,300]Hz


Time Series Production: Use KNIME to Send Alarms !!!

Using RESTful Services:


… just add Imagination!


Any New Ideas?

• Find Analytics Areas for your Business: Customers, IoT, User Experience, …

• Adjust the use case to your Data

• Invent your Analytics Goals!

• Learn about the new techniques

• Spread the Knowledge through Communication

48


Thank You!

49

See you March 15-17 in Berlin!


Thank You

Free copy of the e-book “KNIME Beginner’s Luck”From KNIME Press https://www.knime.org/knimepress

Promotional Code:

KNIMEBayArea_Meetup2017

https://www.knime.org/knimepress


The KNIME® trademark and logo and OPEN FOR INNOVATION® trademark are used by KNIME.com AG under license from KNIME GmbH, and are registered in the United States. KNIME® is also registered in Germany.

Just add Imagination

Data & Analytics