Experiences with Start-Stop Operation of Reaction Wheel Ball Bearings

Lessons Learned and Results from Applying Data-Driven Cost Estimation to Industrial Data Sets

“…The second impression is that this criterion is probably not very helpful when selecting the most appropriate method, because the reader has to cope with contradicting outcomes of empirical investigations. For instance, [1] and [28] present contradicting results, although they evaluate the same data set.…”

“…Among the data-intensive methods, some require past project data in order to build customized models (define-your-own-model approaches), others provide an already defined model, where factors and their relationships are fixed (fixed-model approaches). The popular COCOMO models [1], [3] or SLIM [20] are examples of fixed models. (COCOMO also allows calibrating the initial model with one's own project data.)…”

“…However, similarity measures are intolerant of noise and of irrelevant cost factors. This might be dealt with by applying additional factor selection techniques [1].…”

“…Recently, data-intensive estimation methods (that make intensive use of data to compute estimates) have been gaining more and more interest from both research and industry communities [1]. One reason is the increasing availability of data in industry that have been collected in a systematic way (e.g., motivated by companies' efforts to reach higher maturity levels).…”

“…The lessons learned and recommendations are derived from a case study that was conducted with Toshiba Information Systems (Japan) Corporation (TJ) in the context of a cooperation project between the Software Engineering Center of the Japanese Information-technology Promotion Agency (IPA-SEC) and the Fraunhofer Institute for Experimental Software Engineering (IESE). The estimation technique Optimized Set Reduction (OSR ®,1 ) [1] was selected for the study, because it copes with numerous practical problems of industrial data sets. For example, it does not make any assumptions about the distribution of the underlying data, copes with missing data, and can operate on project characteristics on a nominal and continuous scale (like application type and effort, respectively).…”