Abstract:
Communication in Agile software development plays an important place for task coordination,
clarifying requirements and sprint activities supervision. The sentiment shown in these communication
messages can provide useful information about the progress of sprint, and how well
the team is working together. This study presents a machine learning based sentiment analysis
approach to classify the sentiment expressed in sprint communication and to define a numeric
value ranging from 0 to 1, where value closer to 1 indicate more positive communication
patterns, while values approaching 0 indicate increasingly negative compositions within each
sprint. These sentimental outputs can be generated at the end of the sprint or at any point during
the sprint. This enables real-time tracking of the sentiment during a sprint and identifying
communication problems that could inhibit the productivity of teams or impact the quality of
software. Various machine learning models were trained. Logistic Regression provided the best
compromise between generalization, interpretability, and stability and was therefore selected as
the final model for this study. A prediction pipeline was developed to evaluate the sentiment
on a sprint basis. This utilizes a trained Logistic Regression model and an integrated weighted
probability method. The pipeline produces predictions of sentiment category distributions and
forecasts of the score of the sentiments of a sprint. The patterns of communication that appeared
to suggest that there were coordination issues or disruptions of the work process were
often linked to low sentiment scores whereas the ones that appeared less problematic allowed
the process to proceed smoothly. These findings show that sentiment scoring is a feasible and
actionable metric for Project Managers and Scrum Masters to use in lightweight but effective
sprint health monitoring to support proactive sprint management, strengthens task coordination
and helps identify coordination issues that may influence team productivity and software quality
in Agile development.
