Predictive analytics

Predictive analytics is a method of analyzing data in an effort to predict future activity, behavior, and trends. Data scientists examine both historical and new statistical information to develop a predictive model that forecasts the likelihood of an event occurring. Early attempts used computer-generated models in estimating scientific and military outcomes. One of the first predictive applications in the business world involved an analytical approach to determining credit scores. Modern predictive analytics relies on a variety of methods to interpret a vast amount of data. They have been used to influence decisions in business, sports, medicine, marketing, and numerous other fields.

Background

Some of the earliest attempts to use statistical data to predict future behavior occurred during World War II (1939–1945). American mathematician Norbert Wiener attempted to predict the movement of German aircraft by calculating a plane's trajectory and pilot's likely evasive maneuvers. His goal was to forecast where a plane was likely to fly so as to be targeted by anti-aircraft fire. Wiener's effort was ultimately unsuccessful.

Later attempts to develop a predictive model to target aircraft used primitive forms of computer technology to examine flight and trajectory data. Scientists working on building the atomic bomb also used computer models to predict the behavior of nuclear chain reactions—the process necessary to produce the energy needed for a bomb.

In the late 1950s, the Fair Isaac Corporation (FICO), a California-based analytics firm, began using a statistical model to predict a lending company's risk in credit decisions. By the 1980s, the company had developed FICO scores, an analytics model used as a primary decision-making tool in determining who receives loans, credit cards, and mortgages. The system works by assigning everyone a standard score of 670. Points are then added or subtracted based on a number of statistical factors. For example, if a person has a full-time job, 28 points are added to the score; if a person has been on the job less than a year, 25 points are subtracted; an unemployed person loses 42 points. Final credit scores generally range between 300 and 850; the higher the score, the more likely it is believed a person will pay what they owe.

Overview

In the twenty-first century, businesses and government agencies collect an enormous amount of data on customers and individuals. Much of this data is generated through electronic interaction, such as credit card transactions, social media activity, and online purchases. To analyze such large data pools, predictive analytics uses a variety of methods; most of which require the use of computer technology.

One of the most common is predictive modeling, a technique that collects and analyzes data as a way to forecast outcomes. Smaller amounts of data can be examined with the use of a simple equation, while larger amounts of data require advanced analytical software. A common example of predictive modeling is data analyzed by the insurance industry to determine the price of a driver's policy. Companies take into account variables such as age, location, gender, type of car, and past driving history and assign a risk value to the driver. Drivers considered at a higher risk of accident are typically charged more. Predictive modeling is also used in meteorology to forecast the future movement of weather systems, or in computer programming to gauge the likelihood an email is spam.

Predictive modeling can be further broken down into classification modeling and regression modeling. In classification modeling, a person is assigned a value and placed in a classification based on that score. A person with a 350 credit score, for example, would be placed in a "high-risk" classification. Regression modeling examines data to predict a specific numerical value, such as the expected revenue a product will generate or the length of time a machine part will last before it needs to be replaced.

An offshoot of predictive modeling called data mining relies on computer programs to examine large stores of collected data. The program looks for elements such as statistical frequency and the relationship between data in an attempt to identify a pattern of behavior. A more advanced method of predictive analytics uses machine learning, a form of artificial intelligence (AI) that searches for data patterns. This method is useful for extremely complex data sets or in the absence of a known predictive mathematical formula.

A type of predictive analytics known as a decision tree breaks down data into smaller sets and examines variables to trace a logical path to a decision. A decision tree may begin with one set of data. Common variables in the results split the data into subsets resembling the branches on a tree. After analyzing the results, researchers may be able to predict which variable prompted a specific decision.

Businesses use predictive analytics to better detect fraud, increase marketing efficiency and revenue, improve company operations, and reduce risk. In the medical field, it can be used to determine the best course of treatment for a patient. In the twenty-first century, the process has been increasingly used in the sports world to improve athletic and team performance.

For example, Billy Beane, the former general manager of the Oakland Athletics, changed the way Major League Baseball assigned value to players by introducing predictive analytics to the scouting process. Rather than seek out players based on athletic prowess or ability to hit home runs, Beane identified players who reached base most often. The predictive models he used showed that players who got on base—no matter whether it was by a walk or a hit—scored more runs, which led to more wins. Beane's use of analytics was soon adopted by other professional leagues and tailored to fit the individual sport.

In 2021, Forbes reported that AI's integration into business logistics, particularly when used in predictive analytics, led to a widespread reduction of operating costs across a number of disciplines. At the same time, the publication reported that the AI industry, which experienced vast growth in the late 2010s and early 2020s, was set to be worth an estimated $309 billion by 2026. By 2023, Forbes reported that the use of AI had revolutionized a number of industries and that generative AI programs, which utilize neural networks to mimic human thinking, were being used to process vast amounts of information at a rate impossible for humans.

Bibliography

Beasley, Kevin. "Unlocking the Power of Predictive Analytics with AI." Forbes, 11 Aug. 2021, www.forbes.com/sites/forbestechcouncil/2021/08/11/unlocking-the-power-of-predictive-analytics-with-ai. Accessed 12 Dec. 2024.

Beatman, Andy. "Revolutionizing Business with AI: Real-World Examples of Transformative Impact." Forbes, 13 June 2023, www.forbes.com/sites/microsoft‗/2023/06/13/revolutionizing-business-with-ai-real-world-examples-of-transformative-impact/. Accessed 29 Aug. 2023.

Davenport, Thomas H. "A Predictive Analytics Primer." Harvard Business Review, 2 Sept. 2014, hbr.org/2014/09/a-predictive-analytics-primer. Accessed 22 Dec. 2017.

Finlay, Steven. Predictive Analytics, Data Mining and Big Data: Myths, Misconceptions and Methods. Palgrave Macmillan, 2014.

"Predictive Analytics." SAS, www.sas.com/en‗us/insights/analytics/predictive-analytics.html#. Accessed 22 Dec. 2017.

Rouse, Margaret. "Predictive Analytics." TechTarget, searchbusinessanalytics.techtarget.com/definition/predictive-analytics. Accessed 22 Dec. 2017.

Siegel, Eric. Predictive Analytics: The Power to Predict Who Will Click, Buy, Lie, or Die. John Wiley & Sons, 2016.

Steinberg, Leigh. "Changing the Game: The Rise of Sports Analytics." Forbes, 18 Aug. 2015, www.forbes.com/sites/leighsteinberg/2015/08/18/changing-the-game-the-rise-of-sports-analytics/#31143c744c1f. Accessed 22 Dec. 2017.

Van Rijmenam, Mark. "The History of Predictive Analytics." Datafloq, 24 June 2017, datafloq.com/read/history-predictive-analytics-infographic/438. Accessed 22 Dec. 2017.

"What Are the Main Types of Predictive Analytics?" FICO, www.fico.com/en/predictive-analytics/understanding-predictive-analytics/what-are-the-main-types-of-predictive-analytics. Accessed 22 Dec. 2017.