STATS — Software component trend analysis over time series

Abstract

Through initiatives such as open sourcing, software development organizations have embraced component reuse. Inner sourcing, in which organizations reuse internally created components, is gaining interest. Large companies such as Philips, PayPal and Ericsson have embraced inner source initiatives. Software components reuse can significantly reduce software development and testing time. A challenge when reusing components is to gauge their quality and projected reliability over time. Minor component changes can create unforeseen complexities. This work proposes STATS — Software Component Trend Analysis Over Time Series, a self-directed artificial neural network which uses historic performances to predict the performance of inner source components over time. Using time series-based learning instances, STATS can aid in the prediction of component trouble reports based on historic knowledge. This is accomplished through the history of components trouble reports over varying time series. Using STATS, system architects and developers predict the reliability of candidate open source and inner source software components in the medium and long term.