Question Description
I'm working on a computer science writing question and need an explanation to help me learn.
1 Explain the relationships among data mining, text mining, and sentiment analysis.
2. What should an organization consider before making a decision to purchase text mining software?
3. Discuss the differences and commonalities between text mining and sentiment analysis.
Go to fairisaac.com. Download at least three white papers on applications. Which of these applications may have used the data/ text/Web mining techniques discussed in this chapter?
Explanation & Answer
View attached explanation and answer. Let me know if you have any questions.
1
Data Analysis
Student’s Name
Institutional Affiliation
2
1. The relationship between data mining, text mining, and sentimental analysis is that
data mining involves extraction of data from a specific source, text mining involves
analysis of unstructured text to retrieve relevant information and transforming them to
business intelligence, and sentimental analysis evaluates the positivity or negativity of
the extracted data (Isah et al., 2014). Text mining analyzes big volumes of textual data
to extract relevant information and in most cases encompasses text structuration.
Sentiment analysis will then evaluate how positive or negative or neutral the extracted
information is (Santos et al., 2018). Text mining is used for sentiment analysis,
concept extraction for analytics, behavioral pattern recognition, and all these are the
procedures that follow after data mining.
2. Several factors must be considered when deciding to purchase text analysis software.
For instance, the software must have the capacity to solve the business need for text
analysis. Also, the existing talent must be able to effectively understand and use the
software; or the cost of hiring new talent should not be higher (Santos et al., 2018).
Besides, the cost of the software must be worth the purchase, both the original and
subsequent costs (Alamanda et al., 2019). The organization must have the required
amount of data for text analysis to make the most out of the software (Isah et al.,
2014). Industry-based factors such as the ability to attain competitiveness using the
software must be considered.
3. Text mining and sentimental analysis both are ways of extracting meaning from
consumer data. They also form part of a successful customer experience improvement
program. However, they work differently. The sentimental analysis focuses on
deriving meanings of words and sentences their opinion judgments while text analysis
concerns with analyzing big volumes of textual data to extract relevant information
(Santos et al., 2018). They provide different insights. Text analysis indicates what is
3
trending and that is being talked about frequently whereas sentimental analysis shows
if whatever is being written or talked about the topic is positive or negative (Isah et
al., 2014), and to the degree of negativity or positivity.
Attached are the three applications that may have used the data/ text/Web mining
techniques discussed in this chapter
4
References
Alamanda, D., Ramdhani, A., Kania, I., Susilawati, W., & Hadi, E. (2019). Sentiment
Analysis Using Text Mining of Indonesia Tourism Reviews via Social
Media. International Journal Of Humanities, Arts And Social Sciences, 5(2), 72-82.
https://doi.org/10.20469/ijhss.5.10004-2
Isah, H., Trundle, P., & Neagu, D. (2014). Social media analysis for product safety using text
mining and sentiment analysis. 2014 14Th UK Workshop on Computational
Intelligence (UKCI). https://doi.org/10.1109/ukci.2014.6930158
Santos, C., Rita, P., & Guerreiro, J. (2018). Improving international attractiveness of higher
education institutions based on text mining and sentiment analysis. International
Journal of Educational Management, 32(3), 431-447. https://doi.org/10.1108/ijem01-2017-0027
Decision Management
Q& A
AI & ML
Predictive modeling
in a disrupted
environment
Strategies and tools to
support customer needs
Chris Frothinger
Senior Global
Solutions Architect
Chris helps FICO customers take a strategic
approach to building analytic solutions, offering
broad expertise in predictive scorecard
development, decision trees, artificial intelligence,
and machine learning modeling and analysis.
“Because this current period of
upheaval will have finite start
and end, it is more effective to
try to understand how existing
models will behave under very
different conditions, and respond
with rapid strategy changes.”
Chris Frothinger
Senior Global Solutions Architect, FICO
© 2020 Fair Isaac Corporation. All rights reserved.
Predictive modeling in a disrupted environment:
strategies and tools to support customer needs
Decision Management
AI & ML
Concurrent with creating a once-in-a-century health crisis, COVID-19 has driven
dramatic shifts in consumer spending and other financial activity, sounding an alarm
bell for deeper visibility to credit risk. Beyond the pandemic, social justice, equality, and
corporate responsibility have entered a new era, with many organizations renewing their
commitment to ensure fairness and transparency in their interactions with consumers.
In this Q&A, Chris Frothinger, Senior Global Solutions Architect for FICO, explains
how scorecards, machine learning (ML), and explainable AI (xAI) capabilities
in FICO® Analytics Workbench™ can help credit risk professionals to achieve
two key objectives: gain deeper insights into credit risk, and ensure that each
decision made is ethical, explainable, and supports the responsible use of AI.
Q:
With consumer
behaviors changing
so much, shouldn’t
new credit risk
models be built?
A:
A sports analogy is useful here. With the rapid onset of COVID, the main shift in credit
risk modeling has been from offense to defense—evolving the game plan by adjusting
score cutoffs and stress-testing the models that are already in place. In many cases
there’s not even enough time to deploy new decisioning tools; to address change in real
time, analysts are creating new strategies to apply to existing tools and models.
Here’s why: the nature of developing new models is such that, in the time it will take
to capture associated data and filter it back into the models, we may have already
emerged from the depth of the COVID economic crisis. Economic signals indicate that
the near future will be an outlier period in credit risk modeling compared to, say, the 2008
financial crisis, the effects of which many consumers felt for years. Because this current
period of upheaval will have finite start and end, it is more effective to try to understand
how existing models will behave under very different conditions, and respond with rapid
strategy changes.
© 2020 Fair Isaac Corporation. All rights reserved.
2
Predictive modeling in a disrupted environment:
strategies and tools to support customer needs
Q:
How can COVIDrelated financial
conditions be
tested with
current models?
Decision Management
AI & ML
A:
Here’s an example: In March 2020, 30% more renters used credit cards to pay rent than
in the previous month, likely due to widespread unemployment. Customers making
rent payments with cards will show a jump in credit utilization, with some moving from
a purchasing mode—paying off their entire balance each month—to using the card for
revolving credit. These are second-order effects that can be picked up in the model and
reflected in customer treatment strategies.
Scorecards are a powerful tool for rank-ordering just about anything. If the credit risk
model is unbiased, testing it with variables that reflect higher credit utilization, and/or a
larger proportion of revolving credit to overall debt, will shift an entire score distribution
band downward, indicating increased risk.
In the current COVID financial environment, when affected customers request credit
line increases, originations teams can be prepared with answers and options. They
can be armed with strategies from the credit risk organization that include custom
scoring models, but also strategies that incorporate the FICO® Resilience Index, which
is designed to rank-order consumers with respect to their sensitivity to an economic
downturn.
Even within a narrow FICO® Score band—for example, near the common FICO Score
cutoff of 680—a range of resilience can be seen. When this information is used to further
segment customer groups, credit line increases can be prudently extended to more
people who request them.
Q:
How do you get
from the scorecard
to the appropriate
treatment strategy?
A:
With the objective of helping customers weather short-term financial distress, robust
decision support tools play an important role in estimating the impact of all kinds of
actions on decision trees. This is where “stress testing” a model comes into play, with
many iterations of variable testing. You can change all kinds of data inputs—such as
credit utilization, number of late payments within the last 6 or 12 months, and many
others—pour the data through, and see which leaf nodes of the tree grow and shrink.
In this way, for example, customers with more sensitivity to financial stress might
be landing on collections actions. Instead, a large segment of these customers
can be shifted to a new decision, such as a three-month payment holiday, instead
of a flat “no” to a credit limit increase. This produces a more positive customer
experience, and can improve outcomes for both the customer and the bank.
© 2020 Fair Isaac Corporation. All rights reserved.
3
Predictive modeling in a disrupted environment:
strategies and tools to support customer needs
Q:
How can machine
learning help
banks support
customers whose
financial lives have
been disrupted?
Decision Management
AI & ML
A:
There’s keen interest in seeing how credit risk models will behave in reflecting current
consumer behaviors to help guide socially supportive lending. FICO’s xAI toolkit enables
you to recognize how model outcomes are derived, so you can more deeply understand
the features that drive the predictions.
Since it’s not always feasible to build new models, get comprehensive new data, or react
to changes in the existing data, xAI tools allow credit risk analysts to see how different
scenarios impact decisions. This is accomplished by using the machine learning
capabilities of xAI to tweak the values at the margin of the model, and then look at how
the resulting adverse actions, and other dynamics within the model, affect customers.
These same tools and techniques can be used to examine how model shifts may cause
outputs to suffer from a palatability standpoint, such as indicating a large portion of
otherwise on-time paying customers be put into collections status.
In other words, with rapid and severe change in the economic environment, you won’t
be able to change the model. But with xAI tools and machine learning, you will be
better able to see how the model will react, and adjust treatment strategies to support
customers in this period of financial distress.
Q:
Do you have any
parting advice?
A:
Working with xAI and machine learning may be new territory for many credit risk
analysts. The xAI techniques I’ve discussed can be applied to multiple model types to
help ensure the best explainability possible for any given approach. Scorecards are
the ultimate in explainable human intelligence, so when xAI and ML learnings filter into
them, analysts will find themselves in a familiar place at which to segment customers
appropriately, and apply treatment strategies that preserve financial lifelines.
In tumultuous times, scorecards, decision trees, and xAI tools can give analysts
the confidence to make prudent lending decisions that, most importantly, support
customers whose financial lives have been disrupted.
Find out more about how
FICO® Analytics Workbench™ can help your organization:
www.fico.com/en/products/fico-analytics-workbench
More Precise
Decisions
FICO and Analytics Workbench are trademarks or registered trademarks
of Fair Isaac Corporation in the United States and in other countries. Other
product and company names herein may be trademarks of their respective
owners. © 2020 Fair Isaac Corporation.
All rights reserved.
4862EX 07/20 PDF
FOR MORE INFORMATION
NORTH AMERICA
LATIN AMERICA & CARIBBEAN
www.fico.com
www.fico.com/blogs
+1 888 342 6336
info@fico.com
+55 11 5189 8267
LAC_info@fico.com
EUROPE, MIDDLE EAST, & AFRICA
ASIA PACIFIC
+44 (0) 207 940 8718
emeainfo@fico.com
+65 6422 7700
infoasia@fico.com
ANALYTICS
Better risk management for Marlette Funding
Professional
“isScorecard
intuitive, allowing
users to spend more
time analyzing the data
than writing code.
”
Arnold Pravinata,
Chief Decision Science Officer
at Marlette Funding
An analytic upg...