We build a predictive model using the prior commit history as a training set to predict the risk of new commits going forward. This model leverages over 50+ metrics that we derive from each commit.
With this model we can do the following: * Correlate code commits/changes to features, stories, and work items to understand the quality and risk impact that those things have on the codebase * Utilize clustering to provide insights around developer productivity and quality impact to the codebase * Utilize clustering to help with sizing of future features based on historical actuals
These are just a few of the areas where we are using ML within Codemonkey.ai today. In the future our goal is to tie additional information from production environments into this model as well.