Land Value - Predictive Analytics
Land Value - Predictive Analytics
Land Value - Predictive Analytics
Dr Shidan C Murphy
Global Director, Banking and Financial Services
Location: Singapore
Biography
Dr. Shidan Murphy is an award-winning scientist, educator, pilot and data scientist with significant business experience in
the Asia-Pacific. He has published several papers published in international journals and written software to teach
statistical computing in R.
Beginning his career as a Research Scientist with the Canada Government, Dr. Murphy developed numerical methods to
predict how climate change, shipping patterns and shoreline modifications affect fish communities in the Great Lakes of
North America and in the Canadian Arctic.
Today, Dr. Murphy works towards solving the analytical issues facing financial institutions, retailors, utilities, insurance and
consumer-packaged good companies across Asia, India and Australia. Dr. Shidan Murphy oversees the pre-sales and
technical delivery of Data Analytics for Altair Engineering in the Asia-Pacific.
Altair-at-a-Glance
$572M 74
FY22 Revenue In 27 Countries
3,000+ 150+
Engineers, Scientists, Altair and Partner
and Creative Thinkers Software Products
13,000+
Customers Globally
© Altair Engineering, Inc. Proprietary and Confidential. All rights reserved.
Government & Defence Heavy Rail Industrial Goods Life & Earth Sciences Education Material Suppliers
© Altair Engineering, Inc. Proprietary and Confidential. All rights reserved.
Advanced Analytics
Data Preparation Machine Learning Stream Processing & Unified Enterprise Platform
Visualization
Automate data collection, Leverage advanced Unified low code data prep,
data cleansing, and data machine learning and AI. Visualize real-time operational machine learning & visualization
transformation. Use trusted Understand what is performance. Spot anomalies, platform for experimenting and
and accurate data to make impacting your business. trends, and clusters in operationalizing AI/ML apps.
the right decisions seconds.
Analytics Modernization
SAS Language Data Governance & IDE SAS Language &
Compiler Deployment Open Source
Run SAS Language & Streamline Dev-ops, Maintain, edit, and run new
Open-Source codes. govern data & access, and existing SAS language
create & manage data programs.
Altair® SLC® pipelines.
Cloud + Server
Altair® Analytics
Altair® SmartWorks Hub® Workbench®
6
Cloud + Server Desktop
© Altair Engineering, Inc. Proprietary and Confidential. All rights reserved.
Altair Units
Revolutionary business model to enable customers to
get more from Altair software. D AT A
AN AL Y S T
ENGINEER DESIGNER
7
© Altair Engineering, Inc. Proprietary and Confidential. All rights reserved.
RapidMiner
8
© Altair Engineering, Inc. Proprietary and Confidential. All rights reserved.
Enterprise Data
PDFs, Excel,
Core Reports
Monitoring and
Reporting
IIoT
Enterprise
Applications
10
© Altair Engineering, Inc. Proprietary and Confidential. All rights reserved.
Monitoring Data
Understanding
Academy builds expertise
Teaches data science foundations and principles through self-paced,
persona-based learning courses, pathways, and certifications.
DO THESE
RESULTS MAKE
Evaluation Modeling
SENSE?
12
© Altair Engineering, Inc. Proprietary and Confidential. All rights reserved.
13
© Altair Engineering, Inc. Proprietary and Confidential. All rights reserved.
Contents
• Brief Introduction to Data Science
• Comments on Data
• Data Preparation and Model Dataset
• Model and Validation
• Future Work
© Altair Engineering, Inc. Proprietary and Confidential. All rights reserved.
• Fact
• User-driven process that uses computers to help wade through enormous amounts of data in
order to discover useful patterns
© Altair Engineering, Inc. Proprietary and Confidential. All rights reserved.
16
© Altair Engineering, Inc. Proprietary and Confidential. All rights reserved.
• The predator is
staring at you.
• Our environment is
evolutionarily new
• Decision-making
process hasn’t
caught up
© Altair Engineering, Inc. Proprietary and Confidential. All rights reserved.
19
© Altair Engineering, Inc. Proprietary and Confidential. All rights reserved.
20
© Altair Engineering, Inc. Proprietary and Confidential. All rights reserved.
WW2 Fighter Jets: Too much Armor is bad and so is too little
Armor:
• Increases Protection
• Reduces Speed
• Reduces Endurance
• Reduces Manoeuvrability
• Reduces Useful Load
• Increases Chance of
Getting Hit
21
© Altair Engineering, Inc. Proprietary and Confidential. All rights reserved.
22
© Altair Engineering, Inc. Proprietary and Confidential. All rights reserved.
23
© Altair Engineering, Inc. Proprietary and Confidential. All rights reserved.
24
© Altair Engineering, Inc. Proprietary and Confidential. All rights reserved.
25
© Altair Engineering, Inc. Proprietary and Confidential. All rights reserved.
Statistical Significance
“The race is not always to the swift nor the battle to the strong, but that's the way to bet”
~ Murphy’s Law
26
© Altair Engineering, Inc. Proprietary and Confidential. All rights reserved.
• “A large portion of replications produced • Ioannidis JPA (2005) Why Most Published
weaker evidence for the original findings” Research Findings Are False.
https://doi.org/10.1371/journal.pmed.002012
• “There is increasing concern that most
4
current published research findings are
false.” • Estimating the reproducibility of
psychological science
• “Failed papers circulate through the literature
https://www.science.org/doi/10.1126/science.
as quickly as replicating papers.”
aac4716
“Stand on the shoulders of giants”
• Camerer et al. Evaluating the replicability of
• More than 70% of researchers have tried social science experiments in Nature and
and failed to reproduce another scientist's Science between 2010-2015
experiments. https://doi.org/10.1038/s41562-018-0399-z
30
© Altair Engineering, Inc. Proprietary and Confidential. All rights reserved.
Model
Dataset
Validation
Data Machine Learning Predictions
On-going (algorithms, tools)
monitoring
© Altair Engineering, Inc. Proprietary and Confidential. All rights reserved.
CRISP-DM
Determine business Collect,
objectives and data describe,
• A Cross-Industry Standard mining goals explore and
verify
Process for Data Mining
quality of
• A tool - and application - data
neutral model
Champion/
• Encourages best Challenger Select,
clean,
practices
construct,
• Offers structure integrate
Integrate and format
• Encourages better, faster into data
results from data mining business
processes Select,
generate,
Evaluate how results build and
achieve business evaluate
objectives models
© Altair Engineering, Inc. Proprietary and Confidential. All rights reserved.
33
© Altair Engineering, Inc. Proprietary and Confidential. All rights reserved.
DATA
© Altair Engineering, Inc. Proprietary and Confidential. All rights reserved.
Location of Offices
Plotted the Offices in Altair Panopticon
36
© Altair Engineering, Inc. Proprietary and Confidential. All rights reserved.
37
© Altair Engineering, Inc. Proprietary and Confidential. All rights reserved.
38
© Altair Engineering, Inc. Proprietary and Confidential. All rights reserved.
39
© Altair Engineering, Inc. Proprietary and Confidential. All rights reserved.
Suggestions
• Don’t add rows before data
• Column Names on only one line. This can confuse data types on import
• Don’t add a column number label (yellow line)
• Probably better to split address into parts
40
© Altair Engineering, Inc. Proprietary and Confidential. All rights reserved.
• Number
41
© Altair Engineering, Inc. Proprietary and Confidential. All rights reserved.
42
© Altair Engineering, Inc. Proprietary and Confidential. All rights reserved.
Outliers
Influence is greater on small sample sizes
43
© Altair Engineering, Inc. Proprietary and Confidential. All rights reserved.
Outliers
Influence is greater on small sample sizes
44
© Altair Engineering, Inc. Proprietary and Confidential. All rights reserved.
Outliers
Influence of one affects the many!
45
© Altair Engineering, Inc. Proprietary and Confidential. All rights reserved.
46
© Altair Engineering, Inc. Proprietary and Confidential. All rights reserved.
47
© Altair Engineering, Inc. Proprietary and Confidential. All rights reserved.
48
© Altair Engineering, Inc. Proprietary and Confidential. All rights reserved.
49
© Altair Engineering, Inc. Proprietary and Confidential. All rights reserved.
50
© Altair Engineering, Inc. Proprietary and Confidential. All rights reserved.
Supervised Learning
• Key Features
• Basis of most ML techniques
• Datasets with known outcomes (labelled)
• Patterns in data are based on predicting the known outcome
53
© Altair Engineering, Inc. Proprietary and Confidential. All rights reserved.
Data Preparation
Careful - what are we trying to predict?
Type of
Rental Rental
ON_KPKNL Area Rental Rental Price/m2
space (m2) Price/year
Object
54
© Altair Engineering, Inc. Proprietary and Confidential. All rights reserved.
55
© Altair Engineering, Inc. Proprietary and Confidential. All rights reserved.
57
© Altair Engineering, Inc. Proprietary and Confidential. All rights reserved.
3.5x 1.5x
1.5x
1.1x 2x
58
© Altair Engineering, Inc. Proprietary and Confidential. All rights reserved.
59
© Altair Engineering, Inc. Proprietary and Confidential. All rights reserved.
60
© Altair Engineering, Inc. Proprietary and Confidential. All rights reserved.
61
© Altair Engineering, Inc. Proprietary and Confidential. All rights reserved.
62
© Altair Engineering, Inc. Proprietary and Confidential. All rights reserved.
63
© Altair Engineering, Inc. Proprietary and Confidential. All rights reserved.
64
© Altair Engineering, Inc. Proprietary and Confidential. All rights reserved.
100
y= 0.58x + 21.21
My Weight Gain
80
60
R-squared = 0.25
Y Axis (Effect)
y=0.58x + 21.21
R−squared=0.25, p=0.02
p = 0.02
40
20
0 20 40 60 80
X Axis (Cause)
100
y= 0.58x + 21.21
My Weight Gain
80
60
R-squared = 0.25
Y Axis (Effect)
y=0.58x + 21.21
R−squared=0.25, p=0.02
p = 0.02
40
20
0 20 40 60 80
X Axis (Cause)
80
My Weight Gain
60
Y Axis (Effect)
40
20
0 20 40 60 80 100
X Axis (Cause)
70
© Altair Engineering, Inc. Proprietary and Confidential. All rights reserved.
Annual Rent
Variable 1 …. Variable X
(Predicted)
New data without 0.137 390 2
known outcomes 125,000
0.177 23 .5 89,000
0.145 10 9 882,591
© Altair Engineering, Inc. Proprietary and Confidential. All rights reserved.
Machine
Learning
Linear
Logistic Support vector
Discriminant
Regression machines
analysis
Neural
GLM Decision trees
networks
Boosting Bagging
Random
73 forests
© Altair Engineering, Inc. Proprietary and Confidential. All rights reserved.
74
© Altair Engineering, Inc. Proprietary and Confidential. All rights reserved.
75
© Altair Engineering, Inc. Proprietary and Confidential. All rights reserved.
Future Work
77
THANK YOU
altair.com
#ONLYFORWARD
©©Altair
AltairEngineering,
Engineering Inc. Proprietary and Confidential. All rights reserved.
1985 13,000+
Founded & Headquartered Customers
in Troy, MI U.S. Globally
$532M 86
FY21 Offices in
Revenue 25 Countries
3,000+ 150+
Engineers, Scientists, Altair and Partner
and Creative Thinkers Software Products
79
© Altair Engineering, Inc. Proprietary and Confidential. All rights reserved.
80
© Altair Engineering, Inc. Proprietary and Confidential. All rights reserved.
Robust Preparation
Turn difficult data into smart data. Dedupe, redact,
generate calculated fields, split/merge, filter, join,
pivot/unpivot. Auditable.
Easy to Use
Built for non-technical users. Click-driven prep with
80+ prebuilt functions. No scripting. No coding.
Automate
Wizard-based approach to building automated tasks
and scheduled, repeatable processes.
Scale
Support any number of users in a governed, secure
environment. Pooling and failover. Cloud-ready.
Advanced Scorecards
Automate loan applications and similar processes.
Utilize weight of evidence, logistic regression, and
reject inference methods.
Operationalize in Seconds
Upload models to SmartWorks Analytics with one click.
Export Python, R, SQL, SAS Language, or PMML code
to third party systems.
One Platform
Data Engineering for Everyone
Connect, acquire, explore, prepare,
catalog and pipeline your data
Model Building
Automated Data Science Create models through automated,
visual, and code-based approaches
Fully automated – upload data, get
predictive insights True Team
Model Ops Transparency
Simplify operations and deploy
models wherever they bring value
REST API
Seamless Migration Create API
+ Protocols
Manage models inproduction
Test
Altair’s code analysis tools can analyze thousands of programs in
REST API
minutes. Professional services to support the complete migration
process, including assessment, proof of concept, and rollout.
Deployment
Program Analytics
Code
Enterprise Scalability
Deploy on-prem or in the cloud (AWS, GCP, Azure,
Oracle). Embed dashboards in your own
applications. Control external systems directly
from Panopticon dashboards.