On top of basic R environment, your R must have:

• plyr package
• reshape2 package

The data we use can be downloaded from:

• https://d396qusza40orc.cloudfront.net/getdata%2Fprojectfiles%2FUCI%20HAR%20Dataset.zip

So basically, you just git clone this script git, and download the data from the location above. Here's the typical example command flow to run the script:

$ git clone <URL-to-this-git>.git scriptdir
$ wget https://d396qusza40orc.cloudfront.net/getdata%2Fprojectfiles%2FUCI%20HAR%20Dataset.zip
$ unzip *.zip
$ cp scriptdir/run_analysis.R UCI\ HAR\ Dataset
$ cd UCI\ HAR\ Dataset
$ R

> source("run_analysis.R")

The analysis is divided into 6 steps.

• Step 0: Read files from local directory and store them in RAM memory as data.frame

  • The definition of each file is found in UCI\ HAR\ Dataset/README.txt.

• Step 1: Combine the two datasets, "test" and "train", into one dataset called 'merged'

  • In this step, the column names are also set by reading features data.frame in the second column.

• Step 2: As we examin only the mean and std values, only the corresponding columns are fultered.

• Step 3: Map the activity number (1-6) to the name defined in activitylabels data.frame.

• Step 4: Make the column name of each variable more meaningful and better suited for R.

  • Remove the 't' character at the beginning as it does not hold special meaning in our analaysis.
  • Replaced 'f' with FFT to mean Fast Fourier Transformation
  • Do not use abbreviated words such as 'Acc' and 'Mag'
  • Do not use - and () as they can lead to some R unexpected behavior.

• Step 5: Melt the data once, and reshape the data in a tidy format.

  • As the data.frame now is too wide and hard to get the tidy result, the data.frame is made in a tall form by melt function from reshape2 library. This resulted melted table is easier to deal with; we can use dcast to reshape the data. The resulted casted table is yet unordered like 1, 10, 11, 20, 2..., it must be reordered by factor and arrange function from plyr package.