CSE457: Process Book - Final

ST. LOUIS CAR CRIMES GEO-DISTRIBUTION


Jimin Lee — jimin.l@wustl.edu - 508714

Yiqi Chen — c.yiqi@wustl.edu - 520373


Background and Motivation


Our motivation for this project stems from a personal encounter shared by a close friend, which highlighted the severe issue of car thefts in St. Louis. On a recent October afternoon, my friend recounted his alarming experience of having his car robbed. Adding to the intrigue, he described another almost cinematic incident where some other thief, after stealing his friend’s car, was caught dancing atop it at a gas station. These incidents were not isolated—several acquaintances in St. Louis have faced similar misfortunes—prompting questions about the safety of parking locations. The recurring narrative in these discussions—“Where did you park?”—highlights a widespread concern about the security of various locales.

After a quick search, we found out that Missouri was one of the Top 10 states for most stolen vehicles nationwide, reaching an absurd number of 27,279 incidents in 2023. Driven by these concerns and the frequent stories of vehicular theft in St. Louis, we decided to investigate this issue further. Our project aims to visualize car crime locations across St. Louis over the past four years. In addition, we are including multiple relevant factors that might influence the safety of parking spaces. This includes the amount of vacant and/or abandoned buildings in neighborhoods, as well as the proximity to police stations and grocery stores. Through this analysis, we hope to uncover patterns or correlations that could inform safer parking practices and perhaps even guide local policy decisions regarding urban safety and planning. This initiative not only addresses a personal connection to an issue affecting the local community but also ties back to our academic and research interests in applying data visualization to solve real-world problems.

Project Objectives


The primary questions we want to address with our visualization: From this project, we seek to accomplish the following: By achieving these objectives, we aim to contribute to a reduction in car theft incidents, enhance the safety of the local community, and demonstrate the power of data visualization in addressing urban issues.

Data


Data Processing


For this project, our major data set is crime records of St. Louis city, which serves as the central focus of our analysis. For preprocessing, we stacked all csv files and filtered out incidents to include only those categorized as motor vehicle crimes. This makes sure that our visualization has a clear focus on the relevant incidents. Additionally, we handled missing data by removing any records that are lacking crucial geographic information such as latitude, longitude, or neighborhood. Then, we converted the date attribute to a consistent format allowing area map with brush function later.

In addition to the major data set, we supplemented 2 data sets — St. Louis land use, neighborhood, and 2020 census data. We used Python to convert the shape file to geoJson. For the neighborhood data, we converted the projection coordinate to EPSG:4326 - WGS 84 allowing leaflet to process it. Land use data is integrated into our analysis as separate layers on the map by selection. This integration will allow us to gain insight into potential influences or correlations these factors may have on car crime rates. 2020 Census data provide an overview of demographics information of each neighborhood. Neighborhood data facilitates effective visualization at different zoom levels. For instance, when viewing the map at a zoomed-out level, data will be aggregated by neighborhood to display a Choropleth map, which will provide a clear view of crime distribution across different neighborhoods in St. Louis.

Visualization Design


Sketches

Milestone 1

marker

Markers at each location of car crime records are visible when zoomed in. The detail of each car crime record are presented in a pop-up. Markers that are close by are clustered and represented as a circle with the number of clustered markers.


landUse

The map has multiple layers of different land use, including various residential districts and commerical districts. The legend and description of each layer will be included in milestone 2 submission


choropleth

The choropleth map will only be visible when zoomed out to show the overall crime counts with red indicating higher car crime counts.

Milestone 2

infoPanel

By click selection on the neighborhood polygon, the information panel provide a quick overview of the neighborhood. This includes a bar chart comparing the neighborhood car crime count to the St. Louis city average, a percentage bar chart showing the demographics information and how diverse the neighborhood is, and a pie chart showing the land use in terms of area. We can see obviously, the car crime frequency is higher when a neighborhood has a greater percentage of business district.


brush

The brush filter allows user to interaction with the map and filter the date relevant to the user's interest. The bar charts of time in the day and weekday provide additional insight into when the car crimes have occurred in the past. For example, the user may be interested in the car crime that took place on monday mornings in Downtown from the most recent 3 months. The functionality of our map will perfectly achieve such objective and facilitate informed parking.


legend

The legend provides additional information to the land use layer. It shows the corresponding category of each color in the selected layer.


userManual

The user manual gives user an instance and short guide of the St. Louis car crime visualization. It provides an overlook of our major website functionality, reducing friction in learning how to use the tools.

Final Design

From the Milestone 2 submission and in-class user studio, we collected valuable feedback and polished our webpage design accordingly. The final design is included the third and fourth pages of the above PDF.


final-overview

Here is an overview of our webpage after implementing the final design. Next, we'll explore the details of all updated and new design elements.


updated-user-manual

First of all, upon loading the webpage, the user's manual appears, providing a quick introduction to the functionality. The content has been updated based on our new design, and the popup width has been expanded for a better reading experience.


search-bar

Then, in the top-left corner, we implemented a search bar. Often, users are primarily concerned with their own neighborhood or the one they plan to visit. The search bar allows them to quickly locate the desired neighborhood, saving those who are less familiar with navigating St. Louis neighborhood the effort and time of dragging and clicking.


updated-filter

Below the search bar, we have updated the look of our filter. The filter can be hidden by clicking the small triangle circled in red. If users do not need the brush function, they can click on the triangle to hide the filter and better view the information panel below it. Clicking the triangle again will bring the filter back. In addition, the 'Selected Date' at the top displays the date range selected by the brushes.


toggle

Further down the sidebar, as circled in red, a toggle function has been added. It separated the map-specific information from neighborhood-specific information. Previously, users had to scroll extensively to determine the relevance of specific charts. Now, the information is more clearly organized into separate sections, reducing the need for excessive scrolling.


box-around

In the map-specific side of information panel, boxes with rounded corners and shades have been added around each chart, making them more distinctive and aesthetically appealing. Based on the feedback we received during the presentation, a bar chart showing types of car crimes -- stealing, tampering, and damaging properties -- is implemented, with stealing being the most prevalent category.


offence-cat

In the neighborhood-specific information panel, titles have been added to improve the clarity of the chart contents. They are responsive to the neighborhood selected on the crime map. The pie chart Offence Categories presents the car crime category by neighborhood. Hovering over it displays the actual counts of each category below the pie chart.


updated-layer-legend

During the user studio, we surprisingly found out many of our classmateswere unaware of the map layers or confused by the variety of colors coming along with the original data. Therefore, the layer legend is now shown by default, and can be hidden by clicking on the small triangle. In addition, the data has been further processed to merge similar categories, such as combining Single-Family Dwelling District and Two-Family Dwelling District to just Residential District. The color scheme is also unified, and category names are colored corresponding to what will appear on the map once the box is checked.

Features


Must-Have Features

Optional Features

Evaluation


The car crime data exhibit distinct spatial and temporal patterns. The interactive visualizations facilitate user engagement by allowing the toggling of various layers, such as residential, commercial, and business districts. This functionality has revealed a pronounced frequency of car crimes within business districts, thus emphasizing the relationship between district type and the incidence of crime. Moreover, the choropleth map effectively differentiates neighborhoods by crime intensity, pinpointing areas like Downtown and Downtown West as zones with elevated crime rates.

The visualizations also uncovered a significant escalation in car crimes during the evening hours, peaking around 10 PM. This pattern suggests a robust correlation with evening activities and potentially with commuter behavior. Both Question 1 and Question 2 of our study are answered through these findings. Furthermore, the neighborhood-specific information panels further delineate that the majority of car crimes pertain to vehicle thefts. To address Question 3, the implementation of filtering tools in time-based charts permits users to examine how car crime frequencies fluctuate by time of day and date, thereby augmenting the comprehension of temporal patterns.

Our visualizations effectively meet the project's objectives by fostering dynamic data interaction and exploration. The strategic use of color coding, detailed pop-ups, and filtering options significantly enhances user engagement and understanding. Following the user-study, Milestone 2, and presentation feedback, we have refined the usability and responsiveness of our visualizations to improve user experience. While most conceivable enhancements have been implemented, additional features could include an embedded search bar to facilitate specific street searches and diverse icons to represent different categories of car crimes. However, due to time constraints, these enhancements were not realized within the project's timeframe.

Project Schedule