A Reproducible Log-Driven AutoML Framework for Interpretable Pipeline Optimization in Healthcare Risk Prediction

2026-05-23 · Source: cs.LG updates on arXiv.org · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Data Science & Analytics, Health & Medical Research · Depth: Expert, extended

Summary

The yvsoucom-iterkit framework is a deterministic, log-driven automated machine learning (AutoML) system for interpretable pipeline optimization in healthcare risk prediction. It formulates pipeline optimization as a fully reproducible, configuration-level system, encoding each pipeline as a traceable log entity. Experiments on the Pima Indians Diabetes and Stroke datasets, involving over 18,000 pipeline configurations, revealed a structured and partially redundant search space. Performance is primarily driven by a small subset of interacting components: augmentation (0.454), model choice (0.198), and imbalance handling (0.101) on Pima, and imbalance handling (0.406) on Stroke. Ensemble models achieved strong and stable performance (Weighted-F1 0.89, Macro-F1 0.88 on Pima; Weighted-F1 0.94 on Stroke), demonstrating lower variability (0.023–0.026) than SVM.

Key takeaway

For AI Scientists and Machine Learning Engineers developing healthcare risk prediction models, you should adopt a pipeline-centric AutoML approach that jointly optimizes preprocessing and modeling. Prioritize ensemble models like XGBoost for their balance of performance and stability, and rigorously evaluate configurations across multiple random seeds to ensure robustness. Focus optimization efforts on high-impact components such as data augmentation and imbalance handling, as many preprocessing steps offer limited additional gains.

Key insights

Reproducible, log-driven AutoML reveals healthcare risk prediction pipeline component importance and search space redundancy.

Principles

• Pipeline optimization benefits from joint preprocessing and modeling.
• AutoML search spaces are often structured and redundant.
• Ensemble models balance performance and robustness.

Method

yvsoucom-iterkit deterministically enumerates pipeline configurations, logs all experimental metadata, and performs multi-level statistical analysis including component importance and similarity.

In practice

• Focus AutoML efforts on high-impact components like augmentation.
• Prioritize ensemble models for stable healthcare risk prediction.
• Evaluate pipelines across multiple random seeds for robustness.

Topics

• Automated Machine Learning
• Healthcare Risk Prediction
• Pipeline Optimization
• Reproducible Machine Learning
• Class Imbalance
• Model Interpretability
• Ensemble Models

Code references

• yvsoucom/itekit-examples
• yvsoucom/yvsoucom-iterkit

Best for: AI Scientist, Machine Learning Engineer, Research Scientist

Related on AIssential

• Explainable AI for Data-Driven Design of High-Dimensional Predictive Studies — Explainable AI can optimize high-dimensional predictive study design by recommending feature engineering for interpretable models.
• MLS Part 14: XGBoost Explained — The Powerful Gradient Boosting Algorithm Behind Winning ML Models — XGBoost enhances gradient boosting with second-order optimization and regularization for superior speed and accuracy.
• Auto-FP: An Experimental Study of Automated Feature Preprocessing for Tabular Data — Automating feature preprocessing for tabular data can be framed as HPO or NAS, with evolution-based algorithms performing best.
• Mask to Concept: Auto-Promptable SAM3 via Efficient Test-Time Concept Embedding Search for Few-Shot Annotation — M2C adapts SAM3 for few-shot medical annotation by searching for transferable visual concepts within its frozen architecture.
• Starting Off on the Wrong Foot: Pitfalls in Data Preparation — Statistically informed data preparation significantly improves model robustness and efficiency for imbalanced, real-world datasets.
• Comparing decentralized machine learning and AI clinical models to local and centralized alternatives: a systematic review — Decentralized learning offers clinically acceptable performance in healthcare, especially for privacy-constrained contexts, outperforming local learning.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by cs.LG updates on arXiv.org.