Tksrivastava/stable-customer-segmentation
Stable Customer Segmentation is an end-to-end ML pipeline that learns latent customer representations using autoencoders before applying clustering, enabling more stable and interpretable customer segmentation compared to traditional feature-space clustering approaches.
What's novel
Stable Customer Segmentation is an end-to-end ML pipeline that learns latent customer representations using autoencoders before applying clustering, enabling more stable and interpretable customer segmentation compared to traditional feature-space clustering approaches.
Code Analysis
6 files read · 2 roundsA production-oriented customer segmentation system that learns stable latent representations via a deterministic autoencoder and clusters FMCG retail data using HDBSCAN in both raw and latent spaces.
Strengths
Strong separation of concerns with reusable core modules; substantive feature engineering (CV, entropy, YoY growth); deterministic design choices (seeds, L2 regularization) for production stability; clear pipeline orchestration.
Weaknesses
No test suite present; minimal input validation and error handling; hardcoded artifact paths reduce portability; Kaggle dependency adds friction for local testing.
Score Breakdown
Signal breakdown
Innovation
Craft
Traction
Scope
Evidence
Commits
21
Contributors
1
Files
16
Active weeks
5
Repository
Language
Python
Stars
1
Forks
0
License
NOASSERTION