Predictive Analytics for Software Project Management
Learn more about our work at CodeAnalytics.
In this project, my colleague Rossella Tritto and I developed a predictive analytics system aimed at enhancing software project management by providing accurate forecasts of project costs, durations, and complexity using machine learning models. This solution combines advanced data analysis with a user-friendly, cloud-deployed web application, making project estimation accessible to non-technical users.
Project Highlights
- Data-Driven Insights: We utilized historical datasets and key features like Adjusted Function Points (AFP) to train predictive models, achieving better accuracy in project estimations.
- Machine Learning Models: Implemented regression models, random forests, and neural networks, with the random forest model achieving a reliable R² score of 0.89.
- Microservices Architecture: Deployed a cloud-native, Dockerized application on Google Cloud Platform, ensuring scalability, security, and efficiency.
- User-Centric Design: Developed an interactive web interface using Streamlit for easy data input, predictions, and visualizations of project metrics, allowing users to understand historical trends.
Key Outcomes & Future Direction
The system demonstrated high accuracy in predictive metrics, offering reliable support for project managers in planning and decision-making. Looking ahead, we plan to tailor this application to organizational datasets, further improving its relevance and accuracy in real-world settings. This project highlights the potential of combining machine learning with effective software management tools.
You can read more details about the project in this paper:
Explore the code on GitHub: CodeAnalytics Repository.