Home

scomp-link Wiki

The Astromech Arm for Your Python Projects — an end-to-end ML toolkit that automates the complete machine learning workflow.

Table of Contents

• Installation
• Quick Start
• CLI Reference
• Python API
• Feature Engineering
• Model Selection
• Explainability
• Drift Detection
• Fairness Metrics
• Time Series Forecasting
• Pipeline Persistence
• Data Quality
• Configuration & Logging
• Architecture
• FAQ

---

What is scomp-link?

scomp-link is a Python package (3.10+) that handles the full ML lifecycle:

Data Profiling → Feature Engineering → Model Selection → Training → Validation → Explainability → Deployment → Monitoring

It provides:

• 13 CLI commands for zero-code workflows
• Python API for programmatic use
• .scomp format for portable pipeline artifacts
• Interactive HTML reports at every stage

Why scomp-link?

Problem	scomp-link Solution
Spending hours on boilerplate preprocessing	FeatureEngineer handles it in one line
Not knowing which model to pick	Decision-tree-based auto-selection
Black-box predictions	SHAP + LIME built in
No idea if production data has shifted	DriftDetector with PSI + KS test
Model bias against protected groups	FairnessMetrics with 4/5 rule
Scattered model files and configs	.scomp bundles everything
Need results without writing Python	13 CLI commands

Supported Tasks

Task	Methods
Regression	Linear, Ridge, Lasso, ElasticNet, SVR, GBM, SGD, Random Forest
Classification	Logistic, SVC, KNN, Naive Bayes, Decision Tree, GBM, Random Forest
Clustering	KMeans, Hierarchical, Mean-Shift
Time Series	ARIMA, SARIMA, Exponential Smoothing
Anomaly Detection	Isolation Forest, LOF, TabNet, Transformer Autoencoder
Text	BERT Contrastive, Sentence-Transformers, TF-IDF
Images	CNN (ResNet/Inception), Feature Clustering