Predicting Risk Matrix in Software Development Projects using BERT and K-Means

This study aimed to create a risk matrix model for a software development project with 60.000 datasets from the Stack Overflow Question on Kaggle.com. BERT has been used for sentence embedding techniques to represent entire sentences and their semantic information as vectors, K-Means has been used to group sentences and calculate the frequency of occurrence so that the probability is expressed in the vertical direction, and TextBlob estimates the severity of the impact with sentiment analysis represented in the horizontal direction. Validation with BERT on the cluster formed by K-Means by creating a classification text with 86% accuracy results with 30 epochs. The proposed risk matrix formation model is a novelty that can be integrated into the risk management framework. The model determines risk priorities, especially in software development projects.