Success Stories

Analyze

Solve

Deploy

Succeed

Technologies

Python, SVM, CNN, RNN, Ensemble Modeling, Keras, TensorFlow, XGBoost, NumPy, Scikit-learn

Software Bug Classifier
 

Picture1-1 copy.jpg

Background

The client is a community-driven software foundation running multiple products and solutions. The client’s software engineering organization includes numerous in-house staff as well as thousands of contributors from all over the world. Their multifunctional defect-tracking system allows developers to watch for notable bugs, improvements and other modification requests in product efficiency. The existing flow was time-consuming and contained many bug misclassifications which affected the work process.

Challenge

Create an intelligent model that will help optimize defects processing and ensure correct classification. The system is to provide the capability to automatically match the defects to specific products and components and assign them to the right team members.

Solution

A Self-learning system based on the "Support Vector Machine" method to convert defective descriptions and meta-data into numeric sequences that further progress through neural networks to provide classification with the feedback mechanism to improve system knowledge over time. We were able to reach 90% accuracy.