Figure 16: Examples of attributes in databases
Suppose you run an HMO, or a steel tempering process, or a 7 degree-of-freedom dynamic robot arm. In each case, you have dozens of variables that may interact with each other as shown in fig. 16. You would like an intelligent assistant to spot patterns and regularities among pairs and triplets of the variables in your database. In particular, you would like to find more than just linear correlations. You would like to find significant non-linear relationships as well. Memory based learning is ideal for this application. Because it has no training phase (other than loading the data set), it can quickly check for relationships between many different subsets of variables in a database. Because it is locally weighted, it will even identify non-linear relationships in the data. These properties make it ideal for many data mining applications where the main goal is to extract previously unidentified, but interesting, relationships in data.