When you want to analyze your large database to derive some connections among different attributes of your customers, deduce unknown or future values from the current state, you need data mining algorithms. These algorithms use a subset of your data to build their "knowledge base"- a model, which contains rules, importance factors, connections among different attributes (field values). In some algorithms, even the model structure itself can be very informative to you, in other cases you want to use the model to predict an unknown customer attribute by feeding known information about him/her. Naturally, models do not give 100% percent sure prediction. You can evaluate the precision of your model by testing it on other segments of your database.

This process may sound complicated, but in DataScope a sophisticated Model Wizard helps you getting through the steps of model building, from selecting the model type, specifying input and target fields to selecting the training and test subsets and evaluating the model result. For fine-tuning the selected model, you can display the corresponding Model Properties dialog.

The model is an integrated component of your project. You can use all DataScope columns for building (not the imported ones only, but those created by the Expression Builder, type conversions and so on). The result columns are such columns again, so you can reuse the result for building other models, expressions or display them on DataScope charts.

Whenever the model structure itself is also informative (like a decision tree or a rule set), DataScope offers special charts for surveying them.

Changes in model input data affect the model. You can set whether you want the algorithm to "re-learn" the model if its input data changes; the columns calculated by the model are always refreshed when their input columns change. You can also re-train and re-test your model manually at any time.

The models can be exported in the standard PMML format for external evaluation tools or to be used in DataScope Model Executor.