Predicting good deadlines is hard. Wouldn't it be great if you could just plug in some details about a current project and a magic machine would tell you when it would be done? The Providence project is to use machine learning algorithms to try and build such a machine - or at least as close as we can manage.
It's a work in progress and it currently produces inconsistent estimates. The main resource the project needs to improve is data. If you keep detailed logs of features and effort in projects please consider uploading the data to an open data repository such as the PROMISE repository. It would be a great help to me and others working in the software effort field!
The core algorithm is fairly simple it just takes your best estimate and doubles it. Well it's a little more complex than that but not too much. It's based on work by Shan He and Leandro Minku who applied machine learning algorithms to software effort datasets to build models that could predict the effort of future projects. The algorithm uses a multi-objective algorithm to optimise for different error metrics simueltaneously which smooths over problems with the individual metrics. After running it for some number of generations the best non-dominated solution for each metric goes into an ensemble. When a new project comes in the project is fed to each of the members in the ensemble and the average effort over all of the predictions is produced.
The project comes bundled with a gradle helper so you can compile and run the tool with a single command.
gradlew execute −x test ./gradlew execute −x test