The iconic city of Chicago (Illinois), is widely known for its crime.
It was surnamed 'Chirak' because of its crime rate comparable to the war in Irak.
Also known for Al Capone and it's 'Chicago Outfit' gang during the prohibition,
Fred Hampton's assassination by the city's police or the sudden rise of crime in 2015, Chicago's crime is an interesting subject to study.
Through this study of 20 years of crime, we want to bring insights on the crime evolution, geographical distribution,
and effects on the city with an unbiased and apolitical approach to give people a fair idea of the safety of the city.
Our analysis is divided into three parts. We will first look at the crime inside the city of Chicago.
Then, based on our first analysis, we try to observe relationships between crime and aspects of life quality such as health and safety of restaurants.
Finally, we will provide two predictive models for the crime in Chicago that could be used by the police, citizens, and others.
Those models aim to evaluate potential arrest probability or most likely crime given a location and a time.
The Police Department of Chicago reported more than 7 millions crime since 2001. Those crimes fall
into multiples categories (eg homicide, prositution...).
The bigger the text, the more frequent this type of crime is in Chicago.
We can see what are the most frequent kind of crime.
Here is an explanation of some of them that might not be familiar for everyone:
More information can be found in the sources. [1] [2].
In order to reduce violence, Chicago Police Department adopted new crime-fighting techniques in 2004 thanks to cooperation with the LAPD (Los Angeles) and NYPD (New York City). Let's see how crime has evolved since the Police started to collect crime data in 2001.
Indeed, crime started decreasing after 2004. However did the new techniques employed by the Police lead to more arrests?
Even though Police adopted new techniques in 2004, we see that the proportion of arrests did not increase. An hypothesis is that the new techniques are mostly for preventing crime. They reduced the number of crime rather than the number of cases closed with a criminal arrested.
Sometimes looking at the whole data at once is not representative of sub-categories. In order to see if HOMICIDE crimes have followed the same trend we will now show the homicide evolution.
It is very interesting because we can see a huge drop in 2004 that might be correlated with the new crime-fighting techniques used by the Chicago Police. However the trend is now different than when we consider all kinds of crime together. Indeed we can clearly see the rise of homicides in 2016. This sudden rise of homicides was widely cover in the newspapers in the U.S.A. This huge homicide rise in Chicago was apparently responsible for half of the cases that led to a rise of homicide in the USA for 2016.
Let's now try to have a spatial analysis of crime in Chicago. It is interesting to relate specific areas to other factors such as the number of crimes, type of crimes to better understand where violence is happening. We will split the city into community areas. There are 77 community areas in Chicago. Our first approach is to consider crime in the North vs in the South of Chicago.
We cannot tell if the South is more violent than the North or vice versa. We need to split the city in smaller areas to obtain better insights. The following map displays the number of crime reported per community area and hoovering on an area will tell you what are the top 3 most frequent crimes for that area.
We can deliver interesting insights from this map:
Let's have a look at what kind of crime those OTHER OFFENSE, in the airport area, correspond.
You can observe that the most important one is OTHER WEAPONS VIOLATION. It seems that, as it is strictly
forbidden to have a gun in an airport in Illinois, this might lead to a lot of such cases. Indeed,
there are generally no restrictions for other areas. [3]
We should compare the crime map with the homicide one as we realized earlier than homicide crime were following
another trend than crime in general. You can see below a heatmap of the homicides in Chicago since 2001.
This map confirms our previous analysis. As we can see the areas with a high density of homicides are
usually in community areas with narcotics as one of the top crime. It suggests that violence in
Chicago is related to drug dealing and more generally to criminal gang activity.
Moreover, areas without a lot of homicide cases are usually in the border of the city. Those
neighborhoods are probably residential.
We can also think that some central areas without a high number of crimes are well protected. Indeed,
there is not a lot of homicide cases near the Downtown or the University. Those important areas are
usually well protected. As an example, it is interesting to know that the University of Chicago has
its own Police Department.
We will now see if there is a relationship between violence and demographic evolution.
We thus took into consideration the population of each community area to obtain a crime rate per 100,000 inhabitants.
The following map shows the crime rate per community area in 2002, 2010 and 2017. Adding the layer sequentially (using the button on the right)
can help you understand how violence evolved over the past 20 years. The first layer shown is the crime rate for 2002.
What can be seen is that the Downtown was the most violent zone
in 2002 in term of crime rate. We have to remember that people usually do not live in the Downtown
as this area is dedicated to businesses so the crime rate might be high for this reason.
However, we can see that the crime rate for the downtown area
decreased a lot between 2002 and 2010. It might correspond to a huge effort made by the Police to
fight violent in the busiest area of the town.
We can, by adding layers successively that violence is spreading in the West of Chicago with
community areas Austin, Humboldt Park, West Garfield Park, East Garfield Park and North
Lawndale and in the South of Chicago with community areas Fuller Park, Englewood and West
Englewood for example.
Now let's see how the population of each area evolved between 2000 and 2010, and between 2010 and 2017. We might see
trends such as people leaving violent areas, or joining new areas. The button on the right can be used to change between 2000-2010
and 2010-2017.
As we can see the most violent area Austin had a huge population drop between 2000 and 2010. In
general, areas that were associated with violence in our analysis correspond to areas with a
population drop.
Moreover, we can clearly see that the downtown area gained population between 2000 and 2010. As we
realized, the crime rate dropped in that area between 2002 and 2010 even though there were more and
more people living there. It really means that the Chicago Police tried to reduce violence in the
downtown area.
However, from 2010 to 2017, we can see that lots of areas are concerned with the population's decrease.
Especially the violent ones and the ones in the border of Chicago. Finally, it seems that areas
around downtown are still attracting people.
After looking at the most dangerous areas in Chicago in terms of crime rate, we will try to see if there is a relationship between the high crime rate of an area and other factors that correspond to the quality of life. We will look at the number of flu clinics and the quality of restaurants.
Below is a map with the density of clinics that gave flu shots in Chicago.
By looking at the geographical distribution of the clinics in Chicago, we can see that there is a bigger amount of clinics that gave flu vaccines in the Northern West part of the city. There is a clear difference between North and South by means of the number of clinics. We can then suppose that the Northern part is wealthier than the Southern part. Moreover, the wealthiest areas are probably the one near the lake as it is usually where we find the most number of clinics that gave flu shots. The most violent area 'Austin' does not have a lot of clinics even though it is close to downtown. We can think that this area is one of the poorest of Chicago on top of being one of the most violent ones.
We are now going to look at food inspections passing rate in Chicago.
We can see that the violent areas are usually linked with a lower inspection passing rate. Indeed we can speculate that those areas are the poorest of the city. The areas where restaurants are passing the inspection the most are near the downtown, or at the border of the city. Those are probably the richest residential areas of Chicago.
Overall when looking at life quality factors, we can clearly see that violent areas are the one suffering from a lower quality of life. We can speculate that those areas are the poorest of Chicago. The richest part of the city are probably the ones that benefit the most from Police action and are therefore less prone to violence.
Now that we have extracted meaningful data from our crime dataset, and got a better understanding of
the situation,
we would like to build tools that could help police and citizens by detecting patterns in the crime
of Chicago.
To do so, we are modeling our dataset for different purposes using a variety of machine learning
methods. The idea is to build two kinds of models:
In order to present those two models, we give the ability for the user to select a few
parameters and observe the results of our models. The other parameters have been set such that for categorical features,
it is the most frequent one, and for continuous features, we took the average. Finally, the date chosen is the latest one with sufficient data (2018).
Also, the most violent area determined in our analysis is Austin area (code 25).
As we can observe, people arrested for Narcotics had to be caught red-handed because no-one may report drug-related crimes. Thus, our model is likely to predict that a narcotics-related crime will result in an arrest. Whereas for most of the thefts, the culprit is not arrested. This isn't a surprise as most theft reports are made by the victim after the event. We finally observe that the month of the year doesn't affect the arrest probability.
We can observe that for the second model, the summer months have different main crime types than the rest of the year. Therefore, during this period, the Police should adapt to this change.