This repository contains the implementation of the course project for the MOOC Getting and Cleaning Data on Coursera. The purpose is to create a tidy dataset from another (less tidy) dataset.
The dataset is used for Human Activity Recognition Using Smartphones. It contains data on experiments where people moves where recorded with the sensor signal (accelerometer, gyroscope) of a Samsung Galaxy S.
The dataset is split across several files:
| File | Content |
|---|---|
| activity_labels.txt | Contains the activities labels (WALKING, WALKING_UPSTAIRS, WALKING_DOWNSTAIRS, SITTING, STANDING, LAYING) |
| features_info.txt | Feature description, how they are calculated... |
| features.txt | List of all features. It will be used to extract column names |
Then there are two dataset (train and test), and for each:
| File | Content |
|---|---|
| subject_{train,test}.txt | List of subject by observation |
| X_{train,test}.txt | The observation with as many features as described in the features.txt file |
| y_{train,test}.txt | The observed activities |
subject_{train,test}.txt, X_{train,test}.txt and y_{train,test}.txt have the same length.
Variables are described in the CodeBook. The main objective was to extract variables that match only the mean or std, so we can focus on means and standard deviation. Then variables were slightly renamed to remove parenthesis and replace '-' by '.'.
At last, the tidy data set contains means for each of those extracted variables for each activity then by each subject.
Clone this repository, then open the script with R:
source('run_analysis.R')
You might need to install the following packages first: data.table, plyr.
The data is downloaded from https://d396qusza40orc.cloudfront.net/getdata%2Fprojectfiles%2FUCI%20HAR%20Dataset.zip, and unzip'ed in the data/ directory.
The run_analysis writes the result in a file Meandata.txt in the current working directory. Format of this file is described in the CodeBook file.