SQL Server Data Mining—Creating a Model for Product Recommendations

Microsoft .NET Framework, ASP.NET, Visual C# (CSharp, C Sharp, C-Sharp) Developer Training, Visual Studio

Creating a Model for Product Recommendations

Our approach for product recommendations is based on the idea that we can use a mining model to look at every customer and the DVDs that they have bought, and then look for patterns of DVDs that often occur together. The Association Rules mining model is often used for this kind of analysis (sometimes called market basket analysis) and produces a set of rules that say, for example, if the customer is buying a DVD of The Godfather, what are the other movies that other buyers of The Godfather have purchased?

Each of these rules has a probability associated with them. For example, many customers may also have bought similar films, such as The Godfather Part II or Goodfellas, so the rules that relate The Godfather to these DVDs would have a high probability. If only a single customer bought It's a Wonderful Life and The Godfather, this rule would have a low probability. In data mining terminology, the number of times that a set of items occurs together is called the support, so the example of It's a Wonderful Life and The Godfather appearing together would have a support of 1.

We can use these rules to make a prediction: For a specific DVD, give me a list of the most probable DVDs that a customer might also enjoy.