User Guide

Modeva is an integrated tool for model development and validation. This guide covers the full workflow: data processing, model wrapping and training, calibration, explainability, interpretable models, the diagnostic test suite and the low-code interface. Methods introduced in this distribution are mapped into their corresponding sections above.

Getting Started

• Introduction
  • Key Modules
  • Special Features

Data Processing

• Data Processing
  • Basic Data Operations
  • Exploratory Data Analysis
  • Feature Selection
  • Outlier Detection
  • Subsampling and Data Drift
  • Feature Engineering

Models

• Interpretable Models
  • Generalized Linear Models
  • Decision Tree
  • Gradient Boosted Decision Trees
  • Linear Tree and Gradient Boosted Linear Trees
  • Neural Tree
  • GAMI-Net
  • ReLU Neural Network
  • Mixture of Experts (MoE)
  • DirectRS
  • ICL-MoE
  • GBDT Leaf Kernel
  • FuseKernel
• Model Wrapping
  • Model Wrappers
  • Model Zoo and Leaderboard
  • Model Tuning
• Model Calibration
  • Model Probability Calibration
  • Interval Calibration for Regression
  • Interval Calibration for Classification

Explainability and Diagnostics

• Model Explainability
  • Global Explainability
  • Local Explainability
• Diagnostic Suite
  • Performance and Residual Analysis
  • Weakness Detection
  • AMIF Weakness Region Diagnostics
  • Error-Aware Clustering
  • Weak-Cluster Detection and Repair
  • Underfitting and Overfitting
  • Reliability
  • Robustness
  • Resilience
  • Fairness

Low Code

• Low code
  • Registry Hub
  • Data Summary
  • EDA 2D Charts
  • EDA 3D Scatter
  • EDA Multivariate
  • Data Processing
  • Model Training
  • Model Tuning
  • Model Test
  • Model Comparison
  • Model Explainability
  • Weakness Test