Ubiquitous Scoring of 1000+ Warranty Categories Using Predictive Rules Derived from Text
Barry de Ville, Sas Global Forum · 2007
Volvo North America extracted warranty code predictive classification information from the text fields of over 1,000,000 records that represented warranty repair data over a 5-year period. This work demonstrated that thousands of warranty codes could be reliably assigned by computer agents with an accuracy rate that ranged from 60% to 90% (compared to human accuracy rates in the 50% range). Because the computer agents can be triggered from such applications as Microsoft Word and Excel (using SAS ® Add-In for Microsoft Office), reliable, text-based, decision-making information can be available on-demand and can be used by operators and business analysts who have no special information technology or engineering training.