Predictive Analytics
Definition: Predictive analysis, also known as predictive analytics is a process that uses current and historical data to generate predictions about the probabilities of future events.
Description: The main component(s) in predictive analytics is the
predictor. This can be one variable, or many variables, that are used to forecast future behavior. A commonly cited example is that insurance companies use the predictors of age, gender, and driving record to predict the likelihood of future claims.
The predictors are placed into a model that analyzes the data, and then provides a forecast of the likelihood of an event with a certain degree of reliability.
- Diagram of Predictive Analytics:
(Image Courtesy of "Predictive Analytics with Data Mining: How It Works" by Eric Siegel)
Sometimes, all of the data from a company, school, or other entity is input into the model. One of the goals of predictive analytics is to determine which of the predictors influence a future probability of an event to the greatest extent. This type of predictive analytics is called a decision model, as it is used to guide business decisions to use those variables which provide the greatest impact as predictors.
History: In the early stages, predictive analytics dealt with only one or two variables, and the variables were weighted according to which had the greatest influence on the desired outcome. Now, computers determine the most effective model, by examining all of the predictors from various data sources. In this way, the computer looks for correlations, and determines the most effective model to determine the future event.The process uses data mining, in which the multiple records of a company are investigated to locate patterns that will be accurate indicators of future events.
Uses: Predictive analytics is used in multiple environments, which include buying patterns of customers, probability of a client defaulting on a loan, and fraud detection. An example of its use in fraud detection involves credit card companies looking for patterns that are common to transactions with stolen credit cards. Multiple purchases over a very short interval of time, even for inexpensive items, are generally an indicator of fraudulent activity. Therefore, in order to protect themselves and their customers, credit card companies are involved in “data mining”, exploring the data to locate trends that could be indicators of fraud.
Author Guy Bellamy once wrote “Hindsight is an exact science”. The use of predictive analytics may make hindsight more valuable than anticipated. A proper exploration of historical and current trends in business, education, or government is invaluable to such agencies. Predictive analytics serves as a method to sift through the data, and determine the right combination of key predictors to build a model to optimize future success.
Associated links:
--
RobertMahoney - 16 Feb 2007
PredictiveAnalysis is kind of analysis based on historical facts. So the first thing we need to notice is that the facts should be true so that the result won't go in the wrong direction.
--
YansongLiang - 15 Feb 2007
to top
