Acquisition Program Problem Detection Using Text Mining Methods

This research provides program analysts and Department of Defense leadership with an approach to identify problems in real-time for acquisition contracts. Specifically, we test the abilities of statistical algorithms using text mining techniques to detect unusual changes in acquisition programs’ cost estimates at the completion of the programs. Currently, the government purchases monthly written reports, an informational tool on status of an acquisition program, but has not been integrated into problem prediction analysis. We center our research on the following two questions: First, can we quantify the qualitative written reports? Second, can we use these quantifications of the texts to predict cost growths in acquisition programs? Through using text mining techniques, we validate the worth of the written reports by creating algorithms that identify 80% percent of problems in acquisition programs, while increasing the probability of a problem existing given our algorithm detects by 56% from the current methods. These positive results for this analysis provide program offices with a method to detect potential problems in acquisition contracts; furthermore, this research helps the government more efficiently manage their resources as well as reduce cost and schedule overruns