The original dataset has 26 columns. I am interested in the
following columns for this project:
Incident Datetime: The date and time when the
incident occurred.
Incident Category: A category mapped on to the
Incident Code used in statistics and reporting.
Incident Subcategory: A subcategory mapped on to
the Incident Code used in statistics and reporting.
Analysis Neighborhood: The neighborhood where the
incident occured.
Latitude: The latitude where the incident
occured.
Longitude: The longitude where the incident
occured.
Wrangling
Prototypes
The original dataset contained 337K rows. I used the data filtering
tool on the DataSF website to narrow the dataset down to only the
rows I am interested in. First, I filtered the dataset to only show
rows where the incident category contains the word "theft". I
furthered narrowed it down to rows where the incident subcategory
contains the word "vehicle". This left me with all the rows that
were an incident of motor vehicle theft or larceny from vehicle.
These are the types of incidents I am interested in.
Next, I filtered the dataset down using the DataSF filtering tool to
only show incidents that happened in the last 30 days. These are the
incidents which are most relavent for this project. I did this by
filtering to only include rows in which the incident datetime is
between 2020/03/29 and 2020/04/27. For the future, I will use the
DataSF API and dynamically filter rows to show the last 30 days
instead of hardcoding the dates.
Finally, I filtered the dataset using the DataSF filtering tool to
remove rows which have empty data. I noticed I was getting some
errors because certain rows did not have a value for latitude or
longitude. To fix this, I filtered out any rows in which latitude or
longitude are empty.
Final Visualizations
For the final visualizations, I wrangled the data using the same
filters as the prototypes. However, for the date, I dynamically
filter the last 28 days of data to get a snapshot of what is
currently happening. I also use the API for the data, as opposed to
the filtering tools on the DataSF website. Filtering is done by
editing the URL for the API request.
I am a Software Developer and Data Scientist based in San
Francisco. My research interests are in Machine Learning,
Computer Vision, Brain-Computer Interfaces, and Human-Computer
Interaction. I enjoy writing songs, playing basketball,
hiking, and baking.
