This repo provides code to develop estimated Wi-Fi analytics using both data collected by 'wardriving', and a predictive model developed using national statistics.
According to Wikipedia:
Wardriving is the act of searching for Wi-Fi wireless networks by a person usually in a moving
vehicle, using a laptop or smartphone. Warbiking, warcycling, warwalking and similar use the
same approach but with other modes of transportation.
The etymology of the name originates from wardialing, a method popularized by a character played by Matthew Broderick in the film WarGames and named after that film. Wardialing consists of dialing every phone number in sequence to search for modems.
Wardriving has become a widely used method for researchers focusing on mobile computing and networking security, but there has been very little examination of the statistical issues with this approach.
This code provides us with the ability to compare Wi-Fi wardriving data with a predictive model to help us quantify underlying statistical properties. With this knowledge, we can help to improve future wardriving studies as researchers can introduce methodological designs which better account for selection bias, sample representativeness and spatial statistical issues such as the modifiable areal unit problem.
The preferred approach is to use Anaconda to handle packages and virtual environments. Hence,
you can first create a conda environment called wawpy:
conda create --name wawpy python=3.7
Activate it (run this each time you switch projects):
conda activate wawpy
Install any required packages (mainly visualization-related):
conda install geopandas matplotlib seaborn pykml
To preprocess the statistical area data run the following from the scripts folder:
python scripts/preprocess.py
This processes all local statistical areas for England, Wales and Scotland for both premises and households. After calculating the population in each area, the 'oa_lookup.csv' file is written which contains information for each statistical unit, such as the population density or urban-rural geotype.
There is a set order in which to run the code from the scripts folder, as follows:
python scripts/oa_list.py
python scripts/prems.py
python scripts/sc.py
The oa_list.py processes all collected WiGLE (.kml) data files and exports the
all_collected_points.shp file to the data/intermediate folder. It finally writes
out the oa_list.csv to the same folder.
Next, the prems.py script processes the ITRC premises-level data into the
data/intermediate folder for each statistical area.
All self-collected data are then processed via the sc.py script which adds a set buffer to
each data point and intersects this shape with other APs and buildings. Data are written out
to the results folder.
To make an estimate of fixed broadband adoption and Wi-Fi availability, for all areas
using national statistics (ns), run the ns.py script:
python scripts/ns.py
Finally, the vis.py script takes self-collected data for all areas, along with the
national estimates from the predictive model, to produce a set of plotted visualizations:
python vis/vis.py