Software Development Efficiency Metrics Prediction Using Fed-Layered Self Attention Grid
Purab Ojha · 2024
Prediction of software development and operational efficiency metric Lead Time for Change (LTFC) using novel Fed-LSAG (Layered Self Attention Grid) in federated learning setup utilizing Transformer based deep learning model on clients. It helps the developer/organization to predict and improve the cost, time to market and future state of software development and management. Our proposed Fed-LSAG is a novel approach, where we have used a layered self-attention grid mapped to a self-attention layer of local transformer model of each client. The proposed Federated learning model is hosted on Google Cloud platform, and it’s trained to predict Lead time for change of a program. It also helps to identify the key phases from cycle time of lead time to change which carries the highest impact on the faster code delivery to production. This is the first attempt in the industry to predict the software development efficiency metrices.