Welcome to LogML’s documentation!

LogML is a modeling and data analysis automation utility. It was created as a tool for bioinformaticians to quickly run data analyses and highlight important results about their data.

Although LogML provides several kind of analyses, its primary tool is Feature Importance Analysis which results in selecting a handful of features which are most important. This is a multi-step process, which starts with data preparation, models selection and then using combination of statisitical methods and ML models to find features which are highly essential to the relation between features and targets.

Other kinds of analyses: Survival Analysis and EDA. See full list of supported analyses: LogML Pipeline.

LogML is build on top of Python ML ecosystem (pandas + sklearn frameworks).

Result of LogML run is an HTML report, which is to some extent flexible, and with use of javascript (plotly) plots, provides filtering and navigation through large data.

User Manual

• Quick Start
  • Cond Environment
  • First Run
  • First Survival Analysis
• About LogML
  • Terms and abbreviations
  • Pipeline Workflow
• Configuration Overview
  • Input metadata configuration
  • Modeling
  • Preprocessing
  • Configuration utilities
  • Dataset Queries
  • Random States
• Running LogML
  • Command Line Parameters
• Output Artifacts
  • Output Package
  • Report
• How-to Topics

Internals

• LogML Pipeline
• Data Preprocessing
• Feature Importance Analysis
• Exploratory Data Analysis
• Survival Analysis
• Registry

Development

• Dev How-to
• Running - Advanced Topics
• Dev Environment
• Code Reference
• Release Notes