About the Brookfield RPS AVM

Real Property Solutions
1
About the Brookfield RPS AVM
A national, high quality and cost effective Automated Valuation Model (AVM)
Introduction
The Brookfield RPS AVM delivers cost-effective commercial grade automated home valuations for
residential properties across Canada. Whether for use at origination or for portfolio reviews, the Brookfield
RPS AVM provides clients with the information they need to make better and more informed decisions.
About The Brookfield RPS AVM
Product Overview
The Brookfield RPS AVM is:
A national solution
National coverage, including areas where automated valuations typically
do poorly as a result of limited information.
Commercial grade
Accuracy rate of +/-10% across the country (72% of the time), which is in
line with commercial grade AVM standards across North America.
Sophisticated
Built using highly advanced machine learning systems and tree-based
models, which can take into account extensive and diverse data sets,
learn from the data in their environment as well as incorporate new and
different situations.
Helpful added
information
Supporting information provided such as multiple confidence scores and
market price trend charts.
Flexible
The Brookfield RPS AVM can be provided by batch or real time through a
browser or via an XML integration.
Part of a complete
solution
A full suite of valuation products are available via a customizable
cascade, including other market-leading AVM providers, desk-top
appraisals (desk-top, drive-by and full) to help balance cost, risk and
detail.
Inputs: Powerful Property and Neighbourhood Data and Information
•
Updated regularly with hundreds of thousands of records
added on an annual basis
•
Contains detailed property price and attribute information
on millions of unique residential properties across Canada
•
Incorporates important neighbourhood information and
proprietary analytics such as proximity indices,
neighbourhood homogeneity and segmentation models,
amongst others
Feb_14_2014 YVJ
Brookfield RPS maintains property and neighbourhood
information that influence house prices. This database is:
Real Property Solutions
2
Models: Machine Learning and Gradient Boosting Tree Approach
The Brookfield RPS AVM is built using
machine learning, which is a branch of
artificial intelligence and is rooted in early
research on cybernetics and robotics.
Today, it is mostly concentrated on the
mathematical algorithms and techniques
to derive inferences from big data that
are more susceptible to noise, nonnormal distributions, heteroskedasticity
and other non-linear problems with the
speed and efficiency to handle data with high levels of volume, velocity and variety. Machine learning is
not a single technique or technology, but rather a field of computational science that incorporates
numerous technologies to create systems that can learn from the data in their environment and take into
account different and new situations, which is critical to effectively pricing houses in a constantly changing
economic environment.
Because what drives property values differs from place to place, for different property types and over time,
traditional linear regressive functions tend to be too general for accurate estimates across the diverse
housing stock in Canada. To solve for this, we use a Gradient Boosting Tree (GBT) based approach. GBT
models differ fundamentally from conventional techniques that aim to fit a single parsimonious model.
Boosted regression trees combine the strengths of two algorithms: regression trees and boosting, an
adaptive method for combining many simple models to give improved predictive performance.
Three models are available:
1. Machine Learning Regression (MLR) Model – a direct address match is found and the property’s
specific attributes are used to compute an AVM value using Brookfield RPS’ most-advanced machine
learning regression model.
2. Index Model – a direct address match is found and the property’s original value is indexed forward
using a proprietary algorithm based on the Brookfield RPS House Price Index.
3. Local Market Average (LMA) Model – no direct address match can be found, so local market attribute
averages are used as inputs into the MLR model. Address matching would be done via postal code
with returned values representing an average property value for the micro-neighbourhood.
Outputs: Confidence Scores and Accuracy
In addition, Brookfield RPS has developed both a classifier and confidence function. The classifier
determines if the record can be accurately predicted (hit) or not (miss), while the confidence function
provides an indication of how confident the model is in the returned result.
Currently, the models being utilized at Brookfield RPS can predict property values across Canada within
+/- 10% of their actual value, 72% of the time, with an 80% hit rate, which is in line with commercial grade
AVM standards across North America.
Brookfield RPS AVM product summary:
 National coverage and robust hit rate
 Commercial grade, statistically accurate
 Cost effective
 Real-time or batch processing available
 Multiple confidence scores
 Flexible report views
 Market trending information utilizing the
Brookfield RPS House Price Index
 Can be combined with other industry AVMs, and
AVM cascade is customizable based on client
requirements
More Information
For more information, please visit http://www.brookfieldrps.com.
© 2014. All rights reserved. Brookfield RPS is a division of Brookfield Asset Management. All other trademarks are property of their
respective holders.
Feb_14_2014 YVJ
Summary of Features

Download Report