The effects of data mining techniques on software cost estimation

Current research at JPL incorporates data mining and machine learning techniques to see whether a better software cost model can be developed. 2CEE is a tool developed for developing new software cast estimation models using data mining techniques. The accuracy of these models has been validated! infernally through leave-one out cross validation. However, the newly generated models have not been validated to see how well they predict in the real world. Our study seeks to find out how well these machine learning based models perform against standard models for eighteen new flight and ground software projects. The accurate performance of the models against current real world projects is extremely important for practitioners to adapt new techniques.