Stacked Generalization for Human Activity Recognition

TL;DR

This paper evaluates classical machine learning models, specifically Extra Trees and Stacked Classifier, for Human Activity Recognition, emphasizing best practices and heuristics to optimize their performance.

Contribution

It introduces and assesses two models, Extra Trees and Stacked Classifier, highlighting effective strategies for improving HAR model accuracy.

Findings

Stacked Classifier improves HAR performance with proper heuristics.

Extra Trees model shows competitive accuracy in HAR tasks.

Best practices significantly enhance classical ML model effectiveness.

Abstract

This short paper aims to discuss the effectiveness and performance of classical machine learning approaches for Human Activity Recognition (HAR). It proposes two important models - Extra Trees and Stacked Classifier with the emphasize on the best practices, heuristics and measures that are required to maximize the performance of those models.