Predictive Analytics For Dummies, 2nd Edition

Chapter 10

Building a Predictive Model

IN THIS CHAPTER

Defining your business objective

Preparing your data

Developing, testing, and evaluating the model

Deploying and maintaining the model

Some claims are fraudulent. Some customers will churn. Some transactions are fraudulent. Some investments will be a loss. Some employees will leave. But the burning question in everyone's mind is: Which ones?

Building a predictive analytics model can help your business answer such questions. The model will look at the data you have about your customers, for example, and tell you the probability of customer churning. But such questions merely touch upon the surface of what predictive analytics can do; the potential applications of this fascinating discipline are endless.

As mentioned earlier in the book, a model is a mathematical representation of a real-world phenomenon we're interested in making sense of. For example, you can use the data you have to build a model that mimics the stock market in which your firm is actively engaged in all sort of trades — and then your job is to sort out the winning trades from the losing ones. In such a case, your model helps you select a strategy to make money by trading on the stock market.

Building a predictive analytics model becomes vital when the consequences of not acting — or of making the wrong decision — would be costly. Fraudulent transactions, for example, can drain resources such as time, money, and personnel; they can break the financial health ...