Predictive Analytics

How can you forecast behavior? Predictive analytics, a collection of scientific techniques, is the answer! Global high-tech organizations increasingly recognize the value that it offers to many of their solutions and products.

Predictive analytics allows you to automatically learn from past behavior present in large amounts of data to better predict the future. Contrary to expert knowledge, this data-driven knowledge is based on the ability to automatically recognize patterns in data that are not obvious to the expert eye. These so called predictive models present immediate, objective feedback which ensures consistent operational results and increases the effectiveness of existing processes, since they give you the ability to literally forecast behavior.

Our ADAPA® decision engine executes a multitude of pattern recognition models. It uses PMML (Predictive Model Markup Language) to upload and execute these models. PMML, a vendor-driven, open standard for predictive models, is an XML-based language which provides a way for applications to define statistical and data mining models and to share these between PMML compliant applications. ADAPA® supports an extensive collection of PMML elements, including the following predictive techniques:

• Back-Propagation Neural Networks
• Radial-Basis Function Networks
• Neural-Gas Networks
• Support Vector Machines
• Linear Regression
• Binary Logistic Regression
• Multinomial Logistic Regression
• Ordinal Multinomial Regression
• Stepwise Polynomial Regression
• General Linear Models
• Generalized Linear Models

Once a model is uploaded in ADAPA®, it can be executed in batch or real-time. Please note, ADAPA® is not a PMML producer, but a PMML consumer. Most importantly, by leveraging an open standard, models can be developed in your favorite statistical package or proprietary in-house software and then simply deployed in ADAPA®. In this way, ADAPA® works as a scoring engine, providing a reliable and fast way for you to manage, deploy, and execute a multitude of models and decision strategies.

Besides offering a web-based console to manage models, ADAPA® includes capabilities to test models for accuracy of scoring. Following best-practices upon deployment of a new model, a few data records (or several thousand) can be executed to ensure that the model has been properly deployed in a production environment.

ADAPA® is available as a decision engine in which predictive analytics and business rules work in tandem or solely as a scoring engine, in which predictive models can be deployed and executed by themselves. ADAPA® relies on many industry-proven technologies that make it a robust, scalable, and lightweight deployment choice. We invite you to visit the other ADAPA® pages which highlight its complimentary features, such as SOA-based integration as well as reporting capabilities.