Predictive Modeling: Types, Benefits, and Algorithms

Predictive modeling is a method of predicting future outcomes by using data

modeling. It’s one of the premier ways a business can see its path forward and
make plans accordingly. While not foolproof, this method tends to have high
accuracy rates, which is why it is so commonly used.

What Is Predictive Modeling?

In short, predictive modeling is a statistical technique using machine learning and
data mining to predict and forecast likely future outcomes with the aid of historical
and existing data. It works by analyzing current and historical data and projecting
what it learns on a model generated to forecast likely outcomes. Predictive
modeling can be used to predict just about anything, from TV ratings and a
customer’s next purchase to credit risks and corporate earnings.

A predictive model is not fixed; it is validated or revised regularly to incorporate

changes in the underlying data. In other words, it’s not a one-and-done
prediction. Predictive models make assumptions based on what has happened in
the past and what is happening now. If incoming, new data shows changes in
what is happening now, the impact on the likely future outcome must be
recalculated, too. For example, a software company could model historical sales
data against marketing expenditures across multiple regions to create a model for
future revenue based on the impact of the marketing spend.

Most predictive models work fast and often complete their calculations in real
time. That’s why banks and retailers can, for example, calculate the risk of an
online mortgage or credit card application and accept or decline the request
almost instantly based on that prediction.

Some predictive models are more complex, such as those used in computational
biology and quantum computing; the resulting outputs take longer to compute
than a credit card application but are done much more quickly than was possible
in the past thanks to advances in technological capabilities, including computing
power.
Top 5 Types of Predictive Models
Fortunately, predictive models don’t have to be created from scratch for every
application. Predictive analytics tools use a variety of vetted models and
algorithms that can be applied to a wide spread of use cases.

Predictive modeling techniques have been perfected over time. As we add more
data, more muscular computing, AI and machine learning and see overall
advancements in analytics, we’re able to do more with these models.

The top five predictive analytics models are:

1. Classification model: Considered the simplest model, it categorizes data

for simple and direct query response. An example use case would be to
answer the question “Is this a fraudulent transaction?”
2. Clustering model: This model nests data together by common attributes.
It works by grouping things or people with shared characteristics or
behaviors and plans strategies for each group at a larger scale. An
example is in determining credit risk for a loan applicant based on what
other people in the same or a similar situation did in the past.
3. Forecast model: This is a very popular model, and it works on anything
with a numerical value based on learning from historical data. For example,
in answering how much lettuce a restaurant should order next week or how
many calls a customer support agent should be able to handle per day or
week, the system looks back to historical data.
4. Outliers model: This model works by analyzing abnormal or outlying data
points. For example, a bank might use an outlier model to identify fraud by
asking whether a transaction is outside of the customer’s normal buying
habits or whether an expense in a given category is normal or not. For
example, a $1,000 credit card charge for a washer and dryer in the
cardholder’s preferred big box store would not be alarming, but $1,000
spent on designer clothing in a location where the customer has never
charged other items might be indicative of a breached account.
5. Time series model: This model evaluates a sequence of data points
based on time. For example, the number of stroke patients admitted to the
hospital in the last four months is used to predict how many patients the
hospital might expect to admit next week, next month or the rest of the
year. A single metric measured and compared over time is thus more
meaningful than a simple average.
Predictive Modeling and Data Analytics
Predictive modeling is also known as predictive analytics. Generally, the term
“predictive modeling” is favored in academic settings, while “predictive analytics”
is the preferred term for commercial applications of predictive modeling.

Successful use of predictive analytics depends heavily on unfettered access to

sufficient volumes of accurate, clean and relevant data. While predictive models
can be extraordinarily complex, such as those using decision trees and k-means
clustering, the most complex part is always the neural network; that is, the model
by which computers are trained to predict outcomes. Machine learning uses a
neural network to find correlations in exceptionally large data sets and “to learn”
and identify patterns within the data.

Benefits of Predictive Modeling

In a nutshell, predictive analytics reduce time, effort and costs in forecasting
business outcomes. Variables such as environmental factors, competitive
intelligence, regulation changes and market conditions can be factored into the
mathematical calculation to render more complete views at relatively low costs.

Examples of specific types of forecasting that can benefit businesses include

demand forecasting, headcount planning, churn analysis, external factors,
competitive analysis, fleet and IT hardware maintenance and financial risks.

Challenges of Predictive Modeling

It’s essential to keep predictive analytics focused on producing useful business
insights because not everything this technology digs up is useful. Some mined
information is of value only in satisfying a curious mind and has few or no
business implications. Getting side-tracked is a distraction few businesses can
afford.

Also, being able to use more data in predictive modeling is an advantage only to
a point. Too much data can skew the calculation and lead to a meaningless or an
erroneous outcome. For example, more coats are sold as the outside
temperature drops. But only to a point. People do not buy more coats when it’s -
20 degrees Fahrenheit outside than they do when it’s -5 degrees below freezing.
At a certain point, cold is cold enough to spur the purchase of coats and more
frigid temps no longer appreciably change that pattern.

And with the massive volumes of data involved in predictive modeling,

maintaining security and privacy will also be a challenge. Further challenges rest
in machine learning’s limitations.

Limitations of Predictive Modeling

According to a McKinsey report, common limitations and their “best fixes” include:

1. Errors in data labeling: These can be overcome with reinforcement

learning or generative adversarial networks (GANs).
2. Shortage of massive data sets needed to train machine
learning: Apossible fix is “one-shot learning,” wherein a machine learns
from a small number of demonstrations rather than on a massive data set.
3. The machine’s inability to explain what and why it did what it
did: Machines do not “think” or “learn” like humans. Likewise, their
computations can be so exceptionally complex that humans have trouble
finding, let alone following, the logic. All this makes it difficult for a machine
to explain its work, or for humans to do so. Yet model transparency is
necessary for a number of reasons, with human safety chief among them.
Promising potential fixes: local-interpretable-model-agnostic explanations
(LIME) and attention techniques.
4. Generalizability of learning, or rather lack thereof: Unlike humans,
machines have difficulty carrying what they’ve learned forward. In other
words, they have trouble applying what they’ve learned to a new set of
circumstances. Whatever it has learned is applicable to one use case only.
This is largely why we need not worry about the rise of AI overlords
anytime soon. For predictive modeling using machine learning to be
reusable—that is, useful in more than one use case—a possible fix
is transfer learning.
5. Bias in data and algorithms: Non-representation can skew outcomes and
lead to mistreatment of large groups of humans. Further, baked-in biases
are difficult to find and purge later. In other words, biases tend to self-
perpetuate. This is a moving target, and no clear fix has yet been identified.