ModelMAX®

Category Intelligent Software>Data Mining Systems/Tools

Abstract ModelMAX is a data mining and analysis software application. It's used for predictive modeling, descriptive modeling, cluster analysis, profiling, and segmentation.

ModelMAX automates the vast majority of work that is required, making it a favorite among business users.

ModelMAX is also very fast and technically proven, making it equally popular and respected among statisticians.

ModelMAX features/capabilities include:

What makes ModelMAX different from other data mining tools --

1) No need to be a guru - Knowing which modeling method to use for each situation and following all the associated modeling practices, rules, and conventions requires a great deal of specialized knowledge. With ModelMAX, the complexity is handled for you.

ModelMAX uses a combination of data mining technologies including statistics, a CHAID (CHi-squared Automatic Interaction Detector) and CART-like process (see G6G Abstract Number 20053B60), and neural networks (NNs).

All statistical validations are handled automatically, behind the scenes.

2) Automated variable reduction - Modelers typically spend countless hours systematically reducing the raw list of variables down to just a handful of the most significant ones.

ModelMAX uses an advanced proprietary pre-processing variable analysis technique called UDT™ (Universal Data Transformation™).

This technique transparently prepares and transforms input data, and ultimately selects the best variables that should be used in any given model.

3) Complete projects in record time - Completing projects using ModelMAX is extremely fast. A modeling project can be completed in just a few hours. ModelMAX users report a 10X gain in productivity over other modeling products.

4) Uncomplicated deployment – What good is a predictive model if you can’t put it into production easily. This is the Achilles heel of many modeling tools. ModelMAX makes model deployment easy.

You can score directly to a database, produce code to embed in production applications, or generate PMML (Predictive Model Markup Language). Model deployment is simple and straightforward with ModelMAX.

Types of models ModelMAX can create --

ModelMAX is capable of creating predictive models and descriptive models. Predictive models typically are used to determine an action or behavior such as response to a promotion; or an amount or value someone will spend.

Descriptive models are typically used to understand qualitative aspects such as how certain individuals group together, or what attributes are important with regard to certain actions and outcomes.

What differentiates ModelMAX from other predictive modeling approaches --

If you currently use statistics, you are aware of how important it is to employ the proper technique for each situation based on the type of data being used, the type of output being determined, the distribution of the data, etc.

ModelMAX embeds this knowledge into the product so users can be sure that all the proper statistical conventions and rules are adhered to.

ModelMAX® also relieves you of time consuming data preparation tasks, and even recodes and transforms categorical variables.

Additionally, ModelMAX uses a non-linear technique, so it identifies patterns, Not just trends, or central tendencies.

What differentiates ModelMAX from other 'descriptive modeling' approaches --

The type of clustering and profiling possible with ModelMAX® is highly mathematical. It would take 6 to 12 weeks of very involved statistical analysis to replicate what ModelMAX® can do in just a small fraction of the time.

The kind of data you need for modeling --

The best data for modeling contains information that describes the desired behavior. While it’s hard to know what knowledge is embedded in the data, generally, good information is stored in three (3) kinds of data.

The best source is typically transactional data such as what was purchased, how often, the price, etc.

Promotional data, such as how often someone was mailed, what offers were accepted, outbound and inbound calls, etc, can also be valuable.

Lastly, associative data, such as demographic and geographic elements, can add some value.

Product doesn't need a lot of data to successfully perform the analysis --

ModelMAX uses sample based methods so you don't have to work with a lot of data unless you want to.

The manufacturer suggests a minimum of 20,000 to 30,000 test records to create a Response (discrete) model, and 2,000 to 3,000 test records to create a Performance (continuous) model.

If you are accustomed to using more records you may do so. You also can use additional records to automatically validate a model. Once a model has been created, you can use it to score as many records as you like.

Similarly, the number of records necessary for clustering or profiling is Not large either.

Typically, these types of projects use far less records than what is available in total. In fact, it is quite common to work with only a few thousand records.

Reports produced by ModelMAX --

ModelMAX provides a variety of reports and information that can be viewed on screen, printed, or exported to a spreadsheet package.

Reports for predictive projects vary, from revealing which factors were important to the model, to detailing how a model performs on test or independent validation data.

ModelMAX recommends a cut-off at the point of diminishing returns.

Descriptive models generate a different series of reports that emphasize data visualization.

For example, perception maps depict the spatial relationship between like-groups or the subsets of data being analyzed.

Reports produced include a Gains Table, Lift Chart, Distribution, Variable Selection, Variable Analysis and Segment Comparison, Full File Scoring, etc.

Models You Can Build With ModelMAX --

1) Direct Marketing Response [Higher lift than RFM (Recency, Frequency, and Monetary) virtually 100% of the time];

2) Segmentation (Clustering);

3) Value - Profitability, Revenue, Lifetime Value;

4) Next Most Likely Product;

5) Collections Likelihood;

6) Prospecting;

7) Loyalty; and

8) Fraud, bankruptcy, loan default.

System Requirements

Manufacturer

Manufacturer Web Site ModelMAX

Price Contact manufacturer.

G6G Abstract Number 20351

G6G Manufacturer Number 100347