Introduction

The following graphs present an analysis of Baltimore City crime data spanning from 2016 to 2024. The data reveals significant trends and patterns in crime occurrences, types, and demographics.

# R libraries used:
library(httr)
library(data.table)
library(ggplot2)
library(lubridate)
library(dplyr)
library(plyr)
library(scales)
library(ggthemes)
library(RColorBrewer)
library(ggrepel)
library(leaflet)
library(leaflet.extras)
library(plotly)

#Downloading and reading data

setwd("~/Master Degree/Data Visualization Decs Makin (DS-736-W01, GB-736-W01)/R")

fileURL <- "https://www.dropbox.com/scl/fo/va3maecms7cvdnq3ho4b9/ALyMQyLDHbMksLbhuYmX24Y/Baltimore%20Crime/Part1_Crime_Beta_5960161298247612570.csv?rlkey=wfotyhfgb6kutv8jgh028vwxu&dl=1"

response <- GET(url = fileURL, write_disk("Baltimore_City_Crime.csv", overwrite = TRUE))

filename <- "Baltimore_City_Crime.csv"

df <- fread(filename, fill = TRUE)

Visualization Tabs:

The visualizations presented here focus on crime events in Baltimore City from 2016 to 2024.

#Filtering and modifing main dataframe

df$CrimeDateTime <- year(mdy_hms(df$CrimeDateTime))
yearFilter <- seq(from=2016, to=2024, by=1)

newdf <- df %>%
  filter(CrimeDateTime %in% yearFilter) %>%
  data.frame()

Histogram

This histogram visualizes Baltimore City crime trends from 2016 to 2024. Here’s a summary:

  • Overall Trend: While there are some fluctuations, the total number of crimes generally decreased over this period.
  • Peak and Low: The highest crime count occurred in 2017 with 52,291 incidents. The lowest count was in 2020, with 36,305 incidents.
  • Recent Years: After the significant drop in 2020, crime counts have been gradually rising but still remain below the levels seen in 2016, 2017, and 2019.

Possible Explanations for the 2020 Dip: It’s worth noting that 2020 was the year of the initial COVID-19 pandemic and lockdowns, which may have influenced the significant decrease in crime. 

P1 <- ggplot(newdf, aes(x= CrimeDateTime)) + 
  geom_histogram(bins= length(unique(newdf$CrimeDateTime)),
                 color = "darkblue", 
                 fill= "lightblue") +
  labs(title = "Histogram Baltimore City Crimes Trends (2016-2024)",
       x="Year",
       y="Count of Crimes") +
  scale_y_continuous(label=comma, breaks = seq(0, max(count(newdf$CrimeDateTime)), by = 10000)) +
  stat_bin(binwidth = 1, geom='text', color='black', 
           aes(label=comma(after_stat(count))), 
           vjust=-0.4,
           size= 3) +
  scale_x_continuous(labels = yearFilter,
                     breaks = yearFilter) +
  theme_light() +
  annotate(
    "text", x = 2020, y = 40000,
    label = "Lowest Crime",
    color = "red",
    size = 2,
    fontface = "bold") +
  annotate(
    "text", x = 2017, y = 56000,
    label = "Highest Crime",
    color = "red",
    size = 2,
    fontface = "bold")

P1

Types

This stacked bar chart displays Baltimore City crime counts by crime type and year from 2016 to 2024.

Most Frequent Crimes:

  • Larceny consistently stands out as the most common crime throughout the entire period.
  • Common Assault and Aggravated Assault are the next most frequent, though their numbers are notably lower than Larceny.

Trends Over Time:

  • Larceny shows a general downward trend, although there are fluctuations year-to-year.
  • Common Assault also seems to have a slight downward trend overall.
  • Aggravated Assault appears to be more volatile, with some years showing increases and others decreases. 
stackdf <- newdf %>%
  select(Description, CrimeDateTime) %>%
  mutate(Description = case_when(
    Description %in% c("LARCENY FROM AUTO") ~ "LARCENY AUTO",
    Description %in% c("ROBBERY - CARJACKING") ~ "ROBBERY - CARJACK",
    Description %in% c("ROBBERY - COMMERCIAL") ~ "ROBBERY - COMM",
    TRUE ~ Description  )) %>%
  group_by(Description,CrimeDateTime) %>%
  dplyr::summarise(n = length(CrimeDateTime), .groups = 'keep') %>%
  data.frame()

stackdf$CrimeDateTime <- as.factor(stackdf$CrimeDateTime)

agg_tot <- stackdf %>%
  select(Description,n) %>%
  group_by(Description) %>%
  dplyr::summarise(tot = sum(n), .groups = 'keep') %>%
  data.frame()

max_y <- round_any(max(agg_tot$tot), 2000, ceiling) + 10000

P2 <- ggplot(stackdf, aes(x= reorder(Description, n, sum),
                          y = n,
                          fill = CrimeDateTime)) + 
  geom_bar(stat = "identity",
           position = position_stack(reverse = TRUE)) +
  coord_flip() + 
  labs(title = "Baltimore City Crimes by Crime Type and Years (2016-2024)",
       x = "",
       y = "Crime Count",
       fill = "Year") +
  theme_light()  +
  theme(plot.title = element_text(hjust = 0.5),
        axis.text.y = element_text(size = 8)) +
  scale_fill_brewer(palette = "Paired",
                    guide = guide_legend(reverse = TRUE)) +
  geom_text(data = agg_tot,
            aes(x = Description,
                y = tot,
                label = sprintf("%.2fk", tot/1000),,
                fill = NULL),
            hjust = -0.1, size = 3) +
  scale_y_continuous(breaks = seq(0, max_y, by = 10000 ),
                    limits = c(0, max_y),
                    labels = label_number(suffix = "k", scale = 1e-3)) 

P2

Weapon Usage

This set of pie charts provides a visual breakdown of the types of weapons used in Baltimore City crimes from 2016 to 2024.

Key Observations:

  • Firearms are consistently the most common weapon used across all years, with their proportion hovering around 65-70%.
  • Hands are the second most frequent “weapon” involved, indicating a significant proportion of crimes involve physical altercations.
  • Knives represent a smaller but notable portion of weapon usage.
  • Other weapon categories like blunt objects, asphyxiation, and personal weapons are present but less common.
  • There’s a category for “Drugs/Narcotics,” suggesting that some crimes are classified based on the involvement of substances.
  • Explosives and poison are rarely used in crimes. 
weaponfilter <- c(NA,"Y","","UNKNOWN","OTHER")

weapon_counts <- newdf %>%
  select(CrimeDateTime,Weapon) %>%
  filter(!Weapon %in% weaponfilter) %>%
  mutate(Weapon = case_when(
    Weapon %in% c("KNIFE_CUTTING_INSTRUMENT", "KNIFE_CUTTIN") ~ "KNIFE",
    Weapon %in% c("PERSONAL_WEA") ~ "PERSONAL_WEAPONS",
    Weapon %in% c("HANDGUN", "RIFLE", "SHOTGUN", "AUTOMATIC_HA",
                  "AUTOMATIC_FI", "AUTOMATIC_RI", "OTHER_FIREARM", "OTHER_FIREAR",
                  "AUTOMATIC_FIREARM",
                  "AUTOMATIC_SH") ~ "FIREARM",
    Weapon %in% c("FIRE_INCENDI", "FIRE_INCENDIARY_DEVICE") ~ "FIRE",
    Weapon %in% c("MOTOR_VEHICLE_VESSEL", "MOTOR_VEHICL") ~ "MOTOR_VEHICLE",
    Weapon %in% c("DRUGS_NARCOTICS_SLEEPING_PILLS", 
                  "DRUGS_NARCOT") ~ "DRUGS_NARCOTICS",
    TRUE ~ Weapon 
  )) %>%
  group_by(CrimeDateTime,Weapon) %>%
  dplyr::summarise(Total_Incidents = n(), .groups = 'keep') %>%
  arrange(CrimeDateTime, desc(Total_Incidents)) %>%
  group_by(CrimeDateTime) %>%
  mutate(Percent = round(100 * Total_Incidents / sum(Total_Incidents),1)) %>%
  slice_head(n = 10)

weapon_counts$CrimeDateTime <- as.factor(weapon_counts$CrimeDateTime)
  
P4 <-  ggplot(data = weapon_counts, aes(x = "", y = Total_Incidents, fill = Weapon)) +
    geom_bar(stat = "identity", position = "fill") +
    coord_polar(theta = "y", start = 0) +
    labs(fill = "Weapons",
         x = NULL,
         y = NULL,
         title = "Tracking Weapon Usage in Baltimore City Crimes Over Time",
         caption = "Weapons that were not clasified or Unknown have been removed") +
    theme_light() +
    scale_fill_brewer(palette = "Paired") +
    theme(plot.title = element_text(hjust = .5),
          axis.text = element_blank(),
          axis.ticks = element_blank(),
          panel.grid = element_blank()) + 
    facet_wrap(~CrimeDateTime, ncol = 3, nrow = 3) +
    geom_text(aes(x=1.7,
                  label = ifelse(Percent > 5,paste0(Percent, "%"),"")),
              size = 2,
              position = position_fill(vjust = .5))

P4

Race & Gender

This pie chart presents a breakdown of crimes in Baltimore City by the race and gender of the individuals involved.

Key Observations:

  • Race:
    • Black individuals are involved in the largest proportion of crimes (52.3%).
    • White individuals account for 19% of the crimes.
    • There is a significant portion where race is unknown (27.7%).
    • Asian and Native American individuals are involved in a very small percentage of crimes.
  • Gender:
    • Males are involved in a slightly higher proportion of crimes (39.8%) compared to females (41.3%).
    • There is a notable portion where gender is unknown (18.9%). 
Gendf <- newdf %>%
  select(Gender) %>%
  mutate(
    Gender = case_when(
      Gender %in% c("",NA,"U") ~ "Unknown", TRUE ~ Gender )) %>%
  mutate(
    Gender = case_when(
      Gender %in% c("M") ~ "Male", TRUE ~ Gender )) %>%
  mutate(
    Gender = case_when(
      Gender %in% c("F") ~ "Female", TRUE ~ Gender )) %>%
  group_by(Gender) %>%
  dplyr::summarise(Total_Incidents = n(), .groups = 'keep') %>%
  data.frame()

Racedf <- newdf %>%
  select(Race) %>%
  mutate(
    Race = case_when(
      Race %in% c("","UNKNOWN",NA) ~ "Unknown", TRUE ~ Race )) %>%
  mutate(
    Race = case_when(
      Race %in% c("BLACK_OR_AFRICAN_AMERICAN") ~ "Black", TRUE ~ Race )) %>%
  mutate(
    Race = case_when(
      Race %in% c("NATIVE_HAWAIIAN_OR_OTHER_PACIFIC_ISLANDER") ~ "Pacific Islander", TRUE ~ Race )) %>%
  mutate(
    Race = case_when(
      Race %in% c("AMERICAN_INDIAN_OR_ALASKA_NATIVE") ~ "Native American", TRUE ~ Race )) %>%
  mutate(
    Race = case_when(
      Race %in% c("WHITE") ~ "White", TRUE ~ Race )) %>%
  mutate(
    Race = case_when(
      Race %in% c("ASIAN") ~ "Asian", TRUE ~ Race )) %>%
  group_by(Race) %>%
  dplyr::summarise(Total_Incidents = n(), .groups = 'keep') %>%
  data.frame()

P5 <- plot_ly(Racedf) %>%
  add_trace(labels = ~Race, values = ~Total_Incidents,
            type = "pie", hole = 0.7,
            textinfo = "label+percent",
            textposition = "outside",
            insidetextorientation = "radial",
            marker = list(colors = c("#1f77b4", "#ff7f0e", "#2ca02c", "#d62728", "#9467bd", "#8c564b")),
            hoverinfo = "label+percent") %>%  
  add_trace(data = Gendf,
            labels = ~Gender, values = ~Total_Incidents,
            type = "pie",
            textinfo = "label+percent",
            textposition = "inside",
            insidetextorientation = "radial",
            domain = list(x = c(0.16, 0.84), y = c(0.16, 0.84)),
            marker = list(colors = c("#e377c2", "#7f7f7f", "#bcbd22")),
            hoverinfo = "label+percent") %>%  
  layout(title = "Crimes by Race and Gender", showlegend = FALSE)

P5

Heatmap

This heatmap visualizes the concentration of violent crimes in Baltimore City. The map uses color intensity to represent the density of crimes, with warmer colors (red, orange) indicating higher concentrations and cooler colors (blue, green) indicating lower concentrations.

Key Observations:

  • Central Baltimore: The downtown and central areas of Baltimore show the highest density of violent crimes, indicated by the bright red and orange zones.
  • Southern Baltimore: The southern part of the city also exhibits a significant concentration of violent crimes, though slightly less intense than the central areas.
  • Northern and Eastern Baltimore: The northern and eastern sections of the city generally have lower violent crime densities, as shown by the cooler colors.
  • Specific Hotspots: Some localized areas within the broader zones show particularly high concentrations, suggesting specific neighborhoods or districts with elevated violent crime rates. 
violnewdf <- newdf
violnewdf$Description <- ifelse(grepl("^ROBBERY", vnewdf$Description), "ROBBERY", vnewdf$Description) 

crimeFilter <- c("AGG. ASSAULT","ROBBERY")
LocationFilter <- c(NA,0)

violnewdf <- violnewdf %>%
  filter(Description %in% crimeFilter) %>%
  filter(!Latitude %in% LocationFilter, Latitude >= 0 ) %>%
  select(Description, CrimeDateTime, Latitude, Longitude, Neighborhood) %>%
  group_by(Description,Neighborhood,CrimeDateTime,Latitude, Longitude) %>%
  dplyr::summarise(n = length(CrimeDateTime), .groups = 'keep') %>%
  data.frame()

m1 <- leaflet() %>%
      addTiles() %>%
      addHeatmap(
        lng = violnewdf$Longitude,
        lat = violnewdf$Latitude,
        data = violnewdf,
        intensity = violnewdf$n,
        blur = 20, 
        max = 10,  
        radius = 15
        ) %>%
      addMarkers(lng = violnewdf$Longitude,
           lat = violnewdf$Latitude, 
           popup = paste(
             "Description:", violnewdf$Description, "<br>",
             "Neighborhood:", violnewdf$Neighborhood, "<br>",
             "Date:", violnewdf$CrimeDateTime, "<br>",
             "Count:", violnewdf$n),
           clusterOptions = markerClusterOptions()
      ) %>%
      setView(lng = -76.6122, lat = 39.2904, zoom = 12) %>%
      addControl(
        html = "<h4 style='font-size: 12px; padding: 3px 8px; margin: 3px;'>Violent Crime Heatmap in Baltimore City</h4>",
        position = "topleft"
      ) %>%
      addControl(
        html = "<p style='font-size: 10px; padding: 3px 8px; margin: 3px;'>Violent Crimes Defined as: Homicide, Rape, Agg. Assault and Robbery</p>",  
        position = "bottomleft"
      )

m1

Wrap up

The visualizations presented have provided a multifaceted view of crime in Baltimore City between 2016 and 2024. We’ve seen that overall crime trends have fluctuated, with a notable decrease in 2020 followed by a gradual increase. Larceny consistently remains the most common offense, while violent crimes like aggravated assault and robbery show concerning trends. 

The geographic distribution of crime highlights the concentration of violent incidents in central and southern Baltimore, indicating areas requiring focused attention. The analysis of weapon usage reveals the persistent problem of firearm violence, while the demographic breakdown underscores the disproportionate impact on Black individuals.

These insights offer valuable information for policymakers, law enforcement agencies, and community organizations working to address crime in Baltimore. By understanding the patterns, trends, and demographics of crime, we can develop more targeted and effective strategies for prevention and intervention.