Platform Infrastructure

The platforms we use for all of our data mining and predictive analytics projects are actually made up of a number of component parts designed to perform certain functions and in totality are the platforms we refer to on this website. The platforms include:

• IBM SPSS Modeler software with its Association, Classification and Segmentation modules
• LityxIQ intelligent, hosted analytics suite
• Systat SigmaPlot graphical and data analysis software
• Microsoft Office Suite — Excel, Access, etc. and SQL DBMS
• Various other text mining, statistical, graphical visualization software packages
• Access to other data mining and predictive modeling tools including SAS
• Box.net communications hub used for the portal interface to the platforms

The platforms and portal structure is illustrated in the following diagram:

Platform Capabilities

The CRS LityxIQ platform enables CRS or our customer (in the hosted service solution) to build sophisticated data analyses, models and optimization solutions while providing easy-to-view data exploration tables, graphs and dashboards. The design of the platform enables quick delivery of the models and can easily scale to support the need for hundreds of models. Our built-in model libraries keep everything in order and tracks model objects, versioning, and performance history.

The platform enables models to be refreshed and rebuilt automatically over time on a schedule or as new data becomes available. This functionality is built into the hosted platform at delivery to our customer. In addition, models can be rescored against new data, which can also be automated and scheduled. Scored results are available in multiple formats for use in the customer’s execution engines.

The CRS Business Analytics platform has many of the same features but is not provided in a hosted environment. Instead, the platform is exercised for the customer’s benefit by a CRS consultant. This platform is used more for initial data exploration, model investigation, and runs most of our pre-built solutions.

The components in our platforms enable the user to analyze and model many different business scenarios across multiple functions of a business including sales, marketing, customer service, operations, finance, and the supply chain. All of these business functions are interrelated and thus contribute to a comprehensive solution.

Unstructured (Text) Data Inputs

In addition, we can incorporate information from unstructured sources such as call center logs, free-form surveys, web content, blogs, social media dialogues, and emails to learn what the customer is actually saying and their sentiment. When this content is understood and modeled along with the structured data, the insights enable our customers to select the most profitable marketing programs or take the most appropriate action.

Platform Customer Benefits

In summary, the platforms provide the following benefits:

• Build models that predict and forecast business results within a singe department or across multiple departments and functions.
• Build models incorporating multiple information types (structured and unstructured data).
• Build models incorporating the widest and most sophisticated statistical and mathematical algorithms and approaches including C & R Tree, QUEST, CHAID, Decision List, Linear regression, K-Means, GRI, Binary Classifier, Numeric Predictor, Neural Net, C5.0, Logistic regression, Bayesian Network, Cox regression, SVM, Time Series, Kohonen, TwoStep, Apriori, CARMA, etc.
• Apply multiple metrics of model quality and fitness such as quality scores, lift charts, side-by-side model comparisons, and automated best-model selection.
• Validate data and models using various approaches such as outlier analysis, decision trees, clustering, association rules, holdout, and cross-validation.
• Score and implement models for embedding within existing business processes.

The platforms with our Customer Portal provide a complete structure for delivering the desired results in a secured (log-in, password protected) environment.