Being able to optimize testing resources in software development isvaluable when time and resources are scarce. The aim of this projectis to help testing departments improve the distribution of testing resources by identifying modules susceptible of having defects. Danfoss has logged all code changes made to their software modules thelast decade. The logs contain information about each specific change,whether it is a defect fix, feature implementation or maintenance.Using the change history of modules and module imports, indirectmodule connections can be discovered to find additional modules atrisk for having defects. The susceptible modules can be representedby software metrics in order to train a machine learning model. Thepurpose of the machine learning model is to prioritize the susceptible modules from modules requiring most to least testing resources.Once obtained, the priority list of modules should help the testing department distribute their resources, giving the most attention to themodules with a high number of predicted defects and focusing lesson modules with no predicted defects. The regression models werechosen based on popularity and performance in related work in thefield of software defect prediction. The models were evaluated basedon prediction error, coefficient of determination, recall and specificity.The results of this project indicate that a linear regression model canperform well when attempting to predict the number of defects basedon software metrics. Multi-layer perceptron regression models alsoperformed well but more work is required to match the performanceof linear models.