Three major aspects need to be considered in comparing different data mining tools applicable to a given business:

• Accuracy - The data mining tool should provide an accurate prediction and/or description of the database.


• Interpretation - The results and processes of the data mining tools should be easily understandable.


• Integration - The data mining tool should be easily integrated into the existing IT structure and the current business processes of the company.

The voting strategy technique is an improvement of the basic nearest neighbor technique, which invloves taking a vote from the K-nearest neighbors for assigning a prediction to the new record. The vote can be the majority or the average values of the K-nearest neighbors.

In order to predict a missing value or an expected event for a new record in a database, the nearest neighbor technique looks for most similar records in the historical database and assigns their value to the new record. The key issue is how to define similarity.

A model can be a mathematical equation, a set of rules, a neural network or a process to derive certain features from a database. It describes some essential structure of a historical database and can be successfully applied to new data for the prediction of missing values or expected events.