DynMeridian, a DynCorp company, is a professional services firm that provides comprehensive analytical and technical support in arms control, national security affairs and related high technology to U.S. government and industry clients. With more than $31 million in revenue, DynMeridian’s highly successful approach enables clients to swiftly and effectively respond to changing national and domestic security environments. The ability to integrate new technologies is the key for the company to meet the needs of the clients in information management and decision support.

The client, a major government agency, was evaluating the needs. The project involved exploring recent technological advancements in data mining (DM) and extensible
markup language (XML) in order to determine the feasibility of leveraging these techniques in support of early identification of and/or improved prediction of personnel retention trends, as well as the efficient transfer of the information to the decision makers. PolyAnalyst was used to data mine the personnel database to generate insightful information and entity-level propensity-to-lose results. PolyAnalyst is currently being incorporated into the main database to handle the tasks of periodically remodeling the data and scoring records. PolyAnalyst will be the key element of the complete information management system.

PolyAnalyst Pro 4.4 runs on Windows NT platform at DynMeridian. A 866 MHZ Pentium machine with 256 MB RAM was used as the baseline desktop configuration.

PolyAnalyst is a comprehensive suite of Data Mining algorithms utilizing the latest achievements in automated knowledge discovery. A broad selection of exploration engines allows the user to predict values of continuous variables, explicitly model complex phenomena, determine the most influential independent variables, solve classification and clustering tasks, and find associations between events. PolyAnalyst offers a range of features for data access, dataset manipulation, machine learning, visualization, and reporting, as well as simple integration with external systems. PolyAnalyst has a unique capability of applying created models to data in any external source
through a standard OLE DB protocol.

PolyAnalyst offers a suite of powerful modeling tools that can be utilized quickly by relatively inexperienced users. Its decision tree algorithm produces numerical rules for boolean target attributes indicating the degree of certainty of the classification decision made in each tree node. SQL-mode In-Place Data Mining makes processing very large databases possible. Exporting created models to XML/ PMML format makes it easy to integrate the results of data mining with other information management and decision support components.

It would be helpful to be able to cut and paste from PolyAnalyst reports to other applications. Exporting the decision tree model to html format could be easier to manipulate. Also, enhanced data editing and cleaning functionality is desirable.

The efficient decision tree algorithm with an easy to use interface, large selection of other useful data mining algorithms, and reasonable pricing made PolyAnalyst the top choice for our needs. Additional strengths of the system were provided by its ability to analyze very large databases through in-place data mining and to export the created models to XML format to simplify their integration with other components of our decision support system. The availability of a fully featured evaluation copy of PolyAnalyst for downloading from the Megaputer website allowed DynMeridian to try the system hands-on prior to purchasing.

The PolyAnalyst Decision Tree proved to be easy to use and very efficient in generating rules and models. Numerical prediction rules generated by PolyAnalyst can be easily applied to other datasets in order to predict the personnel propensity to stay. This result is much more valuable for the management than simply predicting Yes/No for each case without knowing the degree of certainty of this prediction. In this project, scoring test data with the developed classification model produced results where the top 10% cases of the predicted most loyal personnel contained over sixty percent of all people who indeed served for a long time. This represents an over six-fold lift and allows to better target expensive personnel loyalty programs on the most qualified candidates. Furthermore, the entity-level propensity-to-lose results provide very valuable information for the managers. The entity-level data can be easily aggregated to predict separate personnel attrition rates for different groups.

Megaputer Intelligence support has been very timely. The company was quick in meeting the clients’ needs and incorporating client feedbacks into the new version of the software.

A comprehensive User Manual and Online Tutorial were intuitive and easy to follow. Supplementary documentation from Megaputer is very useful and easily available via their web site.

Reviewer

Heyun Su
Senior System Analyst
DynCorp
http://www.dyncorp.com