FACTORS FOR HIGH PERFORMANCE BUSINESS
COMPUTING (HPBC)
• Explosion of data becoming available, both internal
and external, to organizations
• Availability of cost-effective and accessible systems
(in terms of computational speed, data storage,
memory) to be able to do something useful with it.
• Methodologies to analyze and make sense of this
vast amount of data are being developed and
improved every day
DATA:
Customers by
identifying
exactly who
they are
Real estate
holdings
Their net worth
Where they live
Demographics
and more
Lifestyle
Spending
patterns
Income
TECHNIQUES USED :
Decision
trees
Linear
modeling
Regression,
Rules-
based
algorithms
Simplest
technique
Machine
learning
Neural
networks
More
Complex
Text
analysis
Social
network
analysis
Newer
Technique
MODELING :
• NET LIFT (OR UPLIFT) MODELING : where two or
more scenarios are analyzed simultaneously to
trace all possible outcomes and choose the right
treatment (or lack of treatment) for a particular
situation
• ENSEMBLE MODELING : in which a suite of
models are run and the final response comes from a
weighting of the individual models’ results, and
where the model-weighting can also be refined
based on the situation.
HYBRID FRAUD FRAMEWORK PROCESS
FLOW :
CASE STUDY 1:
Money laundering is a serious problem in the financial
industry. For example, here is segmentation to clusters
that indicates a suspicious cluster based on the collected
data (Fig.1). This information helped a financial institution
to see the abnormal patterns and check them to reveal
money laundering.
Besides money laundering there are many other financial
crimes such as credit card fraud, insider fraud, mortgage
fraud, insurance fraud and other. Taking this into account,
a financial institution must have a high-performance risk
management system. This can be achieved with
sophisticated algorithms of Predictive Analytics and
Machine Learning.
Contd….
Fig.2 is a clustering example of the
dependence between different
clients' categories and their bank
activity. This analysis helps to
define clients with anomalous
behavior which can potentially
indicate fraudulent activity.
CASE STUDY 2 :
1. Developed statistical models
on historical data to predict
charge-off/ early pay-off and
other loss producing behavior
propensity.
2. Developed an easy to use MS
Excel dashboard tool with the
model algorithms built in.
3. The tool takes an application’s
profile details as input and
gives scores to it on various
risk propensities.
4. Based on the scores, the tool
also categories the applications
into Low Risk/ Medium Risk
and High Risk segments
Contd………..
1. The business has an informed
way to accept/ reject
applications based on what
behavior to expect.
2. By using the recommendations
to improve risk profile, they
have a way to improve the
portfolio quality without
increasing rejection rates.
3. Charge-off rates have gone
down by 9% (with no change in
rejection rates) within six
months of implementation.
HELPS IN:
Saving time and
money by allowing
to focus on top
prospects
Identifying new
prospects with
connections to
customers
Maximizing business
development
success with
standard and
custom training
Applying data-driven
strategy to all
marketing efforts
Segmenting and
ranking order
customer base
Finding new
prospects that mirror
most profitable
customers
BENEFITS TO FIRMS :
MARKETING
•To boost cross-sell
and upsell revenue
because sales and
marketing staff can
use predictive
models to
anticipate the
needs of
customers
•Product
Introduction
•Trigger offer
recommendations
based on life
events, such as
closing on a home
or the birth of a
child
FORECASTING
•Churn/Attrition
Management
•Forecasting
•Derivatives
Forecasting
•Credit Scoring
•Collection Strategy
Optimization
RISKMANAGEMENT
•Fraud Detection &
Prevention
•Securities Pricing
Product Purchase
Propensity
•Risk Modeling &
Analysis

Predictive analytics in financial service

  • 2.
    FACTORS FOR HIGHPERFORMANCE BUSINESS COMPUTING (HPBC) • Explosion of data becoming available, both internal and external, to organizations • Availability of cost-effective and accessible systems (in terms of computational speed, data storage, memory) to be able to do something useful with it. • Methodologies to analyze and make sense of this vast amount of data are being developed and improved every day
  • 3.
    DATA: Customers by identifying exactly who theyare Real estate holdings Their net worth Where they live Demographics and more Lifestyle Spending patterns Income
  • 4.
  • 5.
    MODELING : • NETLIFT (OR UPLIFT) MODELING : where two or more scenarios are analyzed simultaneously to trace all possible outcomes and choose the right treatment (or lack of treatment) for a particular situation • ENSEMBLE MODELING : in which a suite of models are run and the final response comes from a weighting of the individual models’ results, and where the model-weighting can also be refined based on the situation.
  • 7.
    HYBRID FRAUD FRAMEWORKPROCESS FLOW :
  • 8.
    CASE STUDY 1: Moneylaundering is a serious problem in the financial industry. For example, here is segmentation to clusters that indicates a suspicious cluster based on the collected data (Fig.1). This information helped a financial institution to see the abnormal patterns and check them to reveal money laundering. Besides money laundering there are many other financial crimes such as credit card fraud, insider fraud, mortgage fraud, insurance fraud and other. Taking this into account, a financial institution must have a high-performance risk management system. This can be achieved with sophisticated algorithms of Predictive Analytics and Machine Learning.
  • 9.
    Contd…. Fig.2 is aclustering example of the dependence between different clients' categories and their bank activity. This analysis helps to define clients with anomalous behavior which can potentially indicate fraudulent activity.
  • 10.
    CASE STUDY 2: 1. Developed statistical models on historical data to predict charge-off/ early pay-off and other loss producing behavior propensity. 2. Developed an easy to use MS Excel dashboard tool with the model algorithms built in. 3. The tool takes an application’s profile details as input and gives scores to it on various risk propensities. 4. Based on the scores, the tool also categories the applications into Low Risk/ Medium Risk and High Risk segments
  • 11.
    Contd……….. 1. The businesshas an informed way to accept/ reject applications based on what behavior to expect. 2. By using the recommendations to improve risk profile, they have a way to improve the portfolio quality without increasing rejection rates. 3. Charge-off rates have gone down by 9% (with no change in rejection rates) within six months of implementation.
  • 12.
    HELPS IN: Saving timeand money by allowing to focus on top prospects Identifying new prospects with connections to customers Maximizing business development success with standard and custom training Applying data-driven strategy to all marketing efforts Segmenting and ranking order customer base Finding new prospects that mirror most profitable customers
  • 13.
    BENEFITS TO FIRMS: MARKETING •To boost cross-sell and upsell revenue because sales and marketing staff can use predictive models to anticipate the needs of customers •Product Introduction •Trigger offer recommendations based on life events, such as closing on a home or the birth of a child FORECASTING •Churn/Attrition Management •Forecasting •Derivatives Forecasting •Credit Scoring •Collection Strategy Optimization RISKMANAGEMENT •Fraud Detection & Prevention •Securities Pricing Product Purchase Propensity •Risk Modeling & Analysis