A Metamodel Recommendation System using Meta-Learning

Abstract

The importance and value of statistical predictions increase as data grows in availability and quantity. Metamodels, or surrogate models, provide the ability to rapidly approximate and predict information. However, selection of the appropriate metamodel for a given dataset is often a task, and the choice of the wrong metamodel could lead to considerably inaccurate results. This research proposes and tests the framework for a metamodel recommendation system. The implementation allows for virtually any dataset and preprocesses data, calculates meta-features, evaluates the performance of various metamodels, and learns how the data behaves via meta-learning, thus preparing and bettering itself for future recommendations. Testing on over 500 widely varied datasets, the framework provides positive results, often recommending a metamodel with similar performance as the actual best metamodel.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Mar 01, 2020
Accession Number
AD1101498

Entities

People

  • Megan K. Woods

Organizations

  • Air Force Institute of Technology

Tags

Communities of Interest

  • Autonomy
  • Energy and Power Technologies

DTIC Thesaurus Topics

  • Air Force
  • Algorithms
  • Anomaly Detection
  • Change Detection
  • Data Mining
  • Data Science
  • Data Sets
  • Information Science
  • Information Systems
  • Machine Learning
  • Network Science
  • Neural Networks
  • Operating Systems
  • Standards
  • Statistical Analysis
  • Supervised Machine Learning
  • United States

Fields of Study

  • Computer science

Readers

  • Computational Modeling and Simulation
  • Neural Network Machine Learning.
  • Systems Analysis and Design