Intelligent agents applied to software management.

摘要

Intelligent Agents (IAs) are an area of Artificial Intelligence (AI) that has recently attracted the attention of many researchers.; In this research, three IAs were applied to Software Management. An IA uses a set of rules to determine the best action to take according to the inputs received. To implement the rules for the three agents two things are needed: First, a set of software metrics, and second, for each metric, ranges of admissible values (minimum, average, and maximum).; The set of metrics used by the IAs were selected from the following sources: Software Engineering Institute (SEI); IEEE; Army; Department of Defense; and PAMPA.; To find ranges of admissible values for the metrics in the set, three databases containing data from past software projects were analyzed. The first was the National Software and Information Repository (NSDIR) sponsored by the Air Force; the second was the Data Analysis Center for Software (DACS) sponsored by the Department of Defense; and the third was the Software Engineering Laboratory (NASA - SEL) sponsored by the National Aeronautics and Space Administration/ Goddard Space Flight Center NASA/GSFC. In addition, a Software Engineering student project developed at Texas A&M University was measured and analyzed.; The first IA is used to estimate the phase of a project. The second IA is used to estimate the effort, development time, staff, paperwork, and number of defects in a software product. The third IA is used to diagnose problems in software products.; The IAs will give a warning to software managers if they detect that a metric is out of the admissible range of values. Hopefully, this warning will lead to actions that solve problems before they grow and endanger the software project.