FAQ

What is Sift Agent technology?

Sift AgentTM  technology is the proprietary method used by our data mining software to extract English "if-then" rules from data. The rules can then be used to make predictions and also to generalize about the most important factors.

Extracting rules is called "rule induction" and is considered by many experts as the preferred way to do predicitive analytics.

How does “Sift Agents” technology work?

It employs a techniques of Artificial Intelligence called "machine learning" to search for the patterns found in the input data.

So why should I care about Sift Agents?

Because it is an extremely efficient way to find the patterns in the data. In virtually all data mining problems the number of possible useful patterns is exponentially large. It is impossible to search for all of them exhaustively. Sift Agents zeroes in on them by using proprietary algorithms to sift the best ones out quickly.

How does your software save me money?

It builds models in less hours thus allowing you to try more options and does not require expensive and less accurate statistical packages. More importantly, you get better results which allows you to operate your business more efficiently and to get more useful information from data.

Why is Nuggets the most advanced Data Mining tool on the market today?

Nuggets uses Artificial Intelligence to find useful patterns in data.

This means that the user does not have to deal with complicated modeling software that use methods such as statistical regression, neural Networks or tree builders.

Building models that can find valuable and actionable information can now be created by the people in an organization that understand the business or scientific problem.

Statisticians and other modelers will now find that many of the restrictive requirements that formerly needed to be met by their data or method can now be ignored with no loss of accuracy.

Missing values do not have to be imputed (i.e. filled in) by using another model or the mean value. These techniques reduce accuracy.

Correlations among predictor variables do not have to be artificially reduced by complex mathematical transformations, which introduce accuracy reducing noise into the results.

Models that have large numbers of predictor attributes can now be handled by Nuggets. Other methods are limited to tens of predictor attributes. Nuggets can deal with thousands.

More importantly, these models are least as accurate in their predictions as other methods.

Easier to use and more accurate is just the beginning.

Since Nuggets models are a simply a set of If-Then rules, their predictions can be better understood by you. Each prediction is supported with an English language rule which explains why the prediction was made, how accurate it is and how often it occurred in the data that gave rise to the model.

Do you have a trial version?

Yes. See the downloads section.

Does Nuggets provide a “Confusion Matrix”?

Yes for mutiple value models. It is found in the model folder as a CSV file which can be edited or anayzed with a spreadsheet program such as Excel.

How Do I validate a Data Mining Model?

If you use the "Autobuild" feature Nuggets validates the model and places the report in the model folder. If you use the manual model building function, you must validate the model from the menu.

What is a Confusion Matrix Report?

For models with two values of the dependent attribute, these counts are false positives and negatives.

The Confusion Matrix Report is useful for validating classification models.

Nuggets creates the report when it validates a model. It shows the dependent attribute values predicted for each actual value of the dependent attribute. Thus you can see the correct and incorrect counts of the predictions. Hence the name "Confusion matrix"

The diagonal elements are the counts of the correct predictions and the off-diagonal elements the incorrect predictions.

The report is placed by Nuggets in the model folder as a CSV file.  It can be opened with a spreadsheet program such as Excel.

What do you mean when you call Nuggets a classification modeling system?

A classification modeling system such as Nuggets creates a model to predict which class one or a set of instances, belongs. An instance is one occurrence of the set of values of the predictor attributes and the prediction is the class to which the predicted attribute for that instance belongs.

It requires that the predicted attribute be "nominal" (also called "categorical"). The predictor attributes, also called independent attributes can be nominal or numeric.