/HAR

Human activity recognition using smartphone data

Primary LanguageR

HAR

Human activity recognition using smartphone data.

Description

This repository includes code from a data mining project. The data set can be accessed from the UCI Machine Learning Repository. The goal of the project is to predict 6 classes of activity (walking, walking upstairs, walking downstairs, sitting, standing, and laying) based on data gathered from a cell phone accelerometer and gyroscope.

Results

SVM probably gives the best results. More coming.

To do list

  • Trees
    • basics
    • picture of 2-category case
  • SVM
    • basic
    • tune SVM
  • naiveBayes
  • Neural networks - Parker
  • K nearest neighbor
  • Silly walks
    • basic comparing 2 walkers with trees
    • cross validation of at least one model
    • apply svm
    • multiple walkers
  • Understand more about accelerometers and gyroscopes