This repository contains the processing code for producing tidy datasets from the raw dataset at http://archive.ics.uci.edu/ml/datasets/Human+Activity+Recognition+Using+Smartphones
In order to produce the tidy datasets, I do the following:
- Read the measurement features from the file
features.txt, recognizing the ordered index and name of each measurement - Subset the recognized measurements to those whose names contain either the text
meanor the textstdwhich signify that they are either an arithmetic mean or standard deviation of the corresponding series. - Read the descriptive textual activity labels from the file
activity_labels.txt - For the
testdataset, read in the subject column data for all the test observations from the filesubject_*.txt - From the
testdataset, read in the measurement feautres from the fileX_*.txt. The measurement values should correspond to the features read in step 2 before the subsetting. The measurement values are then projected onto only the features left after the subsetting in step 2. - The activities associated with all the records are read in from the file
y_*.txtfor thetestdataset, and the numerical activity qualitative values are replaced with the textual labels acquired from step 3. - The columns from steps 4, 5, and 6 are then combined into one table where the rows are the observations and the columns are the variables.
- The steps 4:7 are repeated for the
traindataset. - The outcome datasets from steps 7 and 8 are combined vertically (the rows from the latter are appended to rows from the former) and this is our final tidy dataset:
tidy_data.csv - The separate dataset
tidy_data-averaged.csvis produced by further modifying the dataset from step 9 by computing the arithmetic mean of each measurement variable across all observations, while grouping by the subject and activity.