Donna

Your partner in planning

Personal Story

As my project at work grew, I found tickets in the backlog EPIC increasing some even leading to 1.5 years old. Such large time delay also causes context loss and forces the PM/BA to open individual ticket and note priority.

Here comes BacklogAnalyser that analyses the tickets and classifies them into similar categories for effective decision making.

For Presentation slides visit here

Tech Stack

• Python
• scikit-learn
• sentence-transformers

Main Files

1. app.py: Main application script orchestrating the workflow.
2. preprocess.py: Loads and cleans ticket data.
3. embed.py: Generates text embeddings using SentenceTransformer.
4. cluster.py: Clusters ticket embeddings using KMeans.
5. mock_jira_response.json: Sample JIRA ticket data.

Data Flow

1. Load Data: load_data() in preprocess.py reads tickets from mock_jira_response.json.

2. Preprocess Tickets: preprocess_items() combines title and description, cleans the text, and adds a clean_text field to each ticket.

3. Generate Embeddings: embed_texts() in embed.py converts each ticket’s clean text into a vector embedding using a pre-trained model.

4. Cluster Tickets: cluster_embeddings() in cluster.py applies KMeans clustering to group similar tickets.

5. Output Results: The script prints each ticket’s clean text with its assigned cluster/category.

Example Usage

python app.py

How It Works

• The tool reads JIRA tickets from a JSON file(JIRA API integration as part of future enhancement)
• It cleans and preprocesses the text data.
• Each ticket is embedded into a vector space.
• Tickets are clustered into categories using KMeans.
• The output shows which tickets belong to which category.