{"id":9504,"date":"2020-12-27T19:12:38","date_gmt":"2020-12-27T19:12:38","guid":{"rendered":"https:\/\/towardsdatascience.com\/understanding-deadly-police-encounters-with-data-science-3cf1192d9778\/"},"modified":"2025-04-24T16:01:46","modified_gmt":"2025-04-24T21:01:46","slug":"understanding-deadly-police-encounters-with-data-science-3cf1192d9778","status":"publish","type":"post","link":"https:\/\/towardsdatascience.com\/understanding-deadly-police-encounters-with-data-science-3cf1192d9778\/","title":{"rendered":"Understanding Deadly Police Encounters with Data Science"},"content":{"rendered":"

Data for Change<\/a><\/h3>\n

Gaining insight into police violence and systemic racism through classification<\/em><\/h3>\n\n

Between January 1, 2013 and December 15, 2020, 8,709 people were killed by police in the United States.<\/a> Fifty-one percent of them have been people of color – people who are Black, Latinx, Native American, Asian, Pacific Islander – this despite the fact that these racial and ethnic groups comprise less than forty percent of the overall US population. As Americans continue to confront systemic racism and its fatal manifestations in policing, I’ve sought to gain insight into where deadly police encounters are likely to happen where the victim was a person of color using the tools of data science. More specifically, I built a classification model that can identify where deadly encounters are likely to happen based on socio-economic characteristics of the communities in which they occur and created an interactive map of the United States to visualize their geographic spread using Tableau.<\/p>\n

Approach<\/strong><\/h3>\n

Since US law enforcement agencies do not provide public data on deadly encounters, I turned to the eminent Mapping Police Violence<\/a> project which has comprehensive data on nearly all deadly police encounters that have occurred in the US since 2013 – including information about the precise location of the killing as well as information about the victim’s race.<\/p>\n

I also used American Community Survey Data<\/a> from the US Census Bureau which provides detailed population information on a yearly basis on topics not covered by the decennial census. In my case, I am particularly interested in those relating to socio-economic indicators like education, unemployment, rates of health insurance coverage, use of food stamps, computer ownership and access to the internet – nearly all of which the ACS offers data on by racial\/ ethnic categories.<\/p>\n

Since the Mapping Police Violence data has zip code information for its entire dataset of all deadly encounters, my observations are individual zip codes<\/strong>. Therefore, for all zip codes in the entire United States, my model seeks to explain which zip codes have deadly encounters in which people of color are killed by police and which zip codes do not, by comparing their socio-economic characteristics.<\/p>\n

After gathering my data from the US Census API and Mapping Police Violence (MPV) dataset, I engineered a number of key features for my analysis. The most important of which was the presence or absence of one or more police killings of a person of color. The value of this feature comes from the MPV data which provides the "true" count which ranges from 0 to 10, however, I binarized it to a zero or one because I am approaching this topic as a classification problem. After I merged the MPV data with the census data, all zip codes from the census data not included in the MPV data were labelled zero. At that point, I discovered a class imbalance which has important implications for model evaluation.<\/p>\n

Importance of Recall<\/strong><\/h3>\n

Only 9% of all US zip codes account for all police killings of people of color between 2013\u20132020. Therefore, I decided to select the best model based on whichever model could offer the best recall score for the positive\/minority class, that is zip codes where people of color have been killed by police.<\/p>\n

\"Image
Image by author.<\/figcaption><\/figure>\n

I put special emphasis on ensuring that recall for the minority class is as high as possible even if precision is low because compared to the cost of a human life, which economists estimate is worth $10 million, and that person’s family would regard as priceless, the cost of incorrectly identifying a zip code as having a deadly encounter when it does not (false positive) is several orders of magnitude less important than the cost of incorrectly predicting that a zip code does not have a deadly encounter when it in fact has (false negatives).<\/p>\n

Note, however, that this choice to prioritize recall is a subjective one. I can imagine this same analysis being done by, say, a consortium of law enforcement agencies seeking to gain insight into this very same question being concerned about the reputational costs of incorrectly labeling a police department that has not had a deadly encounter as having had one, and therefore choosing to prioritize precision instead of recall. In this one choice of model evaluation metrics, we are acutely reminded of the myth of technical neutrality and the ethical implications of the way we build all machine learning algorithms.<\/p>\n

Iterative Modelling and Dealing with Class Imbalance<\/strong><\/h3>\n

Although my initial EDA showed that the classes are very imbalanced, I nevertheless ran a few baseline models without regard for class weight to motivate how much improvement I would need to make to build a useful model in the next stage. The models I ran included KNN, Logistic Regression, Random Forest Classifier, Bernoulli Na\u00efve Bayes and SVC. As expected, recall was pretty abysmal across the board (except for Na\u00efve Bayes).<\/p>\n

Next, I ran the same models using four different methods of oversampling and undersampling: Random Oversampler, Random Undersampler, ADASYN and SMOTE. After iterating through all twenty of these combinations, I decided to eliminate Random Forest and KNN as potential models. RandomForest only performs well on undersampled data, which is not ideal because it means I don’t get to train the model on as much of the data as I otherwise could, and even then, it is prone to overfitting. Meanwhile, KNN simply does not perform as well as the remaining three. As for over- and undersampling techniques, I get the highest recall scores using RandomOverSampler and ADASYN, so I continue to narrow down my model selection using these two oversampling techniques.<\/p>\n

After narrowing it down to three potential models and two potential sampling techniques, I next looked to see how well Logistic Regression, Bernoulli Naive Bayes and SVC would perform when I brought in additional classification scoring metrics starting with f_beta which allows me to calculate model accuracy while accounting for the relative importance of recall compared to precision. In my case I wanted to use a very high beta of at least 10 because recall is at least 10 times more important than precision in this context where we are predicting where a human life may be lost. I also deployed StandardScaler to prepare for trying different regularization techniques.<\/p>\n

I found that there is not much of a difference between beta=10 versus beta=20, however, ADASYN did provide better classification metrics than RandomOverSampler for all three of my model contenders. One great advantage of ADASYN is that the synthetic data it generates is adaptively shifted towards the decision boundary which enables the model to focus on classifying difficult examples. This makes it an ideal sampling technique in this context where I care a lot about distinguishing between zip codes that do and do not have deadly encounters.<\/p>\n

Ultimately Logistic Regression offers the best recall performance, but for one last check, I compared each of the three model’s ROC-AUC Curves.<\/p>\n

\"Image
Image by author.<\/figcaption><\/figure>\n

Once again, the differences in performance are somewhat marginal but Logistic Regression does take the lead and has the added benefit of high interpretability which is extremely desirable for the objectives of this project. Unlike, say, SVC, Random Forest or other more ‘black box’ models, Logistic Regression not only allows us to identify the most important features, but also whether they have a positive or negative impact on the target variable.<\/p>\n

Before fitting the final model, I tuned the hyperparameters for Logistic Regression. Among the solvers, sag and saga both fail to converge after 5,000 iterations – newton-cg, lbfgs and liblinear, have relatively comparable results, however, newton-cg and lbfgs only support L2 regularization while liblinear allows for both L1(Lasso) and L2 (Ridge). However, since there is likely multi-collinearity between many of my features due to the inherent nature of my socio-economic data, I wanted to make sure that I could use the best possible regularization technique, so I decided to use liblinear as my solver. The differences between the performance of L1 (Lasso) and L2(Ridge) are all but negligible, however L1 gives a marginally better recall score so I chose L1. Now to consider regularization strength, C. Once again, the differences between the C parameter options are marginal, but C=0.001 gives the best minority class recall so I went with that.<\/p>\n

Performance of Final Logistic Regression Model<\/strong><\/h3>\n

My final Logistic Regression model using ADASYN sampling along with the hyperparameters detailed above ultimately delivered good results on the holdout test data. It correctly classified 87.5% of deadly encounters<\/strong>. That is to say, of the 509 cases of deadly encounters in my holdout test data, it accurately classified 436 of them as such, and only missed 73.<\/strong> This is a pretty impressive improvement from the baseline logistic regression model which had a measly recall of 27%, as well as a nice bump from the initial boost that ADASYN alone offered which got recall up to 84%.<\/p>\n

\"Image
Image by author.<\/figcaption><\/figure>\n

As for classification metrics beyond recall (though recall is my most important), I’ll note that precision for the minority class remains quite low at 28.4%. That is to say, for the 1,534 zip codes my model predicted would have deadly encounters, only 436 actually did. However, the costs associated with these false positives are several orders of magnitude less important than the costs associated with the false negatives associated with failing to identify a zip code that has had a deadly police encounter as such since this means failing to identify the loss of a human life. When weighting the relative importance of recall as 10 times more than precision, the model has an f-beta score of 84%.<\/p>\n

What Socio-Economic Factors Affect the Likelihood of Deadly Encounters?<\/strong><\/p>\n

Of the 45 variables used in my model, the 10 most important features are as follows. Features with orange bars make deadly encounters more likely while those with gray bars make them less likely.<\/p>\n

\"Image
Image by author.<\/figcaption><\/figure>\n

There are a few counter-intuitive takeaways here – one is that police killings of people of color are more likely to occur in zip codes where people of color experience poverty and utilize social services like SNAP and food stamps as well as<\/strong> in zip codes with higher numbers of the overall population holding a Bachelor’s degree or higher. This may suggest the possibility of external, unobserved causes beyond zip code level socio-economic characteristics such as racial dynamics within<\/strong> zip codes. This could include, perhaps, conflicts associated with rapid gentrification where affluent white newcomers, at odds with the existing lower-income community, call the police to address these conflicts which increases the likelihood of deadly encounters. Though on this subject, further analysis is needed. It is also somewhat surprising that the greater the percentage of households in a zip code that do not have a computer, the less likely it is that it will have a deadly encounter. This may be a reflection of rural poverty, where many households in the community may be poor with limited access to technology, yet are also more likely to white and therefore have fewer police killings of people of color. But again, further analysis on this conjecture is needed.<\/p>\n

Now as for what this looks like geographically, view the original and interactive map of deadly encounters here<\/a>.<\/p>\n

\"Image
Image by author.<\/figcaption><\/figure>\n

On this map, zip codes where people of color have been killed by police are highlighted in red and all others in blue, if you hover over an individual zip code, you will see the number of actual people of color killed by police, my model’s prediction of whether or not such a killing would take place, as well as stats on the 5 most important features in my model.<\/p>\n

Concluding Thoughts<\/strong><\/h3>\n

This analysis has barely scratched the surface of the critical work necessary to expand our collective understanding of systemic racism and police violence in the US – as well as the role that data science may play therein. Many questions remain about the exact mechanisms through which racialized socio-economic inequalities translate into disproportionate rates of police violence.<\/p>\n

Where possible, future work would do well to examine characteristics of law enforcement agencies responsible for deadly encounters rather than characteristics of communities within which they exist, as well as measures of racial inequality within zip codes – for instance comparing differences between median income of white and black households. On the issue of law enforcement agency characteristics, Police Scorecard<\/a> and Campaign Zero have already compiled a great collection of data for the state of California. Another avenue for future work would be to look at more granular socio-economic data, such as by census tract which are much smaller than zip code areas.<\/p>\n

One thing that is clear, however, is the paramountcy of data itself. Although the Deaths in Custody Reporting Act of 2013<\/a> mandates that all states that receive federal criminal justice assistance grants must report all deaths that occur while the deceased was in the custody of law enforcement agents, including the context of being detained or arrested, seven years later we have yet to see whether or not police departments will comply with this Act. It is also unclear how long it may take for the data to be fully compiled and released to the public. In the meantime, while US law enforcement agencies continue to kill over 1,000 people a year, the work of the Mapping Police Violence project and other collaborative research groups remains essential in the quest to better understand and rectify the unequal toll police violence takes on communities of color.<\/p>\n

Please do check out the original and interactive deadly encounters map <\/a>to see how your community fits within this national picture.<\/p>\n

Anyone interested in the nitty gritty of this project should also check out my GitHub repo<\/a>. If you’d like to chat or learn more about me or my work, feel free to reach out on LinkedIn<\/a>.<\/p>","protected":false},"excerpt":{"rendered":"

Gaining insight into police violence and systemic racism through classification<\/p>\n","protected":false},"author":18,"featured_media":9505,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"is_member_only":false,"sub_heading":"Gaining insight into police violence and systemic racism through classification","footnotes":""},"categories":[44],"tags":[749,628,453,2103,874],"sponsor":[],"coauthors":[26721],"class_list":["post-9504","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-data-science","tag-classification","tag-data-for-change","tag-editors-pick","tag-police","tag-racism"],"yoast_head":"\nUnderstanding Deadly Police Encounters with Data Science | Towards Data Science<\/title>\n<meta name=\"robots\" content=\"index, follow, max-snippet:-1, max-image-preview:large, max-video-preview:-1\" \/>\n<link rel=\"canonical\" href=\"https:\/\/towardsdatascience.com\/understanding-deadly-police-encounters-with-data-science-3cf1192d9778\/\" \/>\n<meta property=\"og:locale\" content=\"en_US\" \/>\n<meta property=\"og:type\" content=\"article\" \/>\n<meta property=\"og:title\" content=\"Understanding Deadly Police Encounters with Data Science | Towards Data Science\" \/>\n<meta property=\"og:description\" content=\"Gaining insight into police violence and systemic racism through classification\" \/>\n<meta property=\"og:url\" content=\"https:\/\/towardsdatascience.com\/understanding-deadly-police-encounters-with-data-science-3cf1192d9778\/\" \/>\n<meta property=\"og:site_name\" content=\"Towards Data Science\" \/>\n<meta property=\"article:published_time\" content=\"2020-12-27T19:12:38+00:00\" \/>\n<meta property=\"article:modified_time\" content=\"2025-04-24T21:01:46+00:00\" \/>\n<meta property=\"og:image\" content=\"https:\/\/towardsdatascience.com\/wp-content\/uploads\/2020\/12\/1mg8EtCL1vTxbcb3ewlh1Iw.png\" \/>\n\t<meta property=\"og:image:width\" content=\"1600\" \/>\n\t<meta property=\"og:image:height\" content=\"900\" \/>\n\t<meta property=\"og:image:type\" content=\"image\/png\" \/>\n<meta name=\"author\" content=\"Dr. Amanda J. Cheney\" \/>\n<meta name=\"twitter:card\" content=\"summary_large_image\" \/>\n<meta name=\"twitter:creator\" content=\"@TDataScience\" \/>\n<meta name=\"twitter:site\" content=\"@TDataScience\" \/>\n<meta name=\"twitter:label1\" content=\"Written by\" \/>\n\t<meta name=\"twitter:data1\" content=\"Dr. Amanda J. Cheney\" \/>\n\t<meta name=\"twitter:label2\" content=\"Est. reading time\" \/>\n\t<meta name=\"twitter:data2\" content=\"11 minutes\" \/>\n<script type=\"application\/ld+json\" class=\"yoast-schema-graph\">{\"@context\":\"https:\/\/schema.org\",\"@graph\":[{\"@type\":\"Article\",\"@id\":\"https:\/\/towardsdatascience.com\/understanding-deadly-police-encounters-with-data-science-3cf1192d9778\/#article\",\"isPartOf\":{\"@id\":\"https:\/\/towardsdatascience.com\/understanding-deadly-police-encounters-with-data-science-3cf1192d9778\/\"},\"author\":{\"name\":\"TDS Editors\",\"@id\":\"https:\/\/towardsdatascience.com\/#\/schema\/person\/f9925d336b6fe962b03ad8281d90b8ee\"},\"headline\":\"Understanding Deadly Police Encounters with Data Science\",\"datePublished\":\"2020-12-27T19:12:38+00:00\",\"dateModified\":\"2025-04-24T21:01:46+00:00\",\"mainEntityOfPage\":{\"@id\":\"https:\/\/towardsdatascience.com\/understanding-deadly-police-encounters-with-data-science-3cf1192d9778\/\"},\"wordCount\":2302,\"publisher\":{\"@id\":\"https:\/\/towardsdatascience.com\/#organization\"},\"image\":{\"@id\":\"https:\/\/towardsdatascience.com\/understanding-deadly-police-encounters-with-data-science-3cf1192d9778\/#primaryimage\"},\"thumbnailUrl\":\"https:\/\/towardsdatascience.com\/wp-content\/uploads\/2020\/12\/1mg8EtCL1vTxbcb3ewlh1Iw.png\",\"keywords\":[\"Classification\",\"Data For Change\",\"Editors Pick\",\"Police\",\"Racism\"],\"articleSection\":[\"Data Science\"],\"inLanguage\":\"en-US\"},{\"@type\":\"WebPage\",\"@id\":\"https:\/\/towardsdatascience.com\/understanding-deadly-police-encounters-with-data-science-3cf1192d9778\/\",\"url\":\"https:\/\/towardsdatascience.com\/understanding-deadly-police-encounters-with-data-science-3cf1192d9778\/\",\"name\":\"Understanding Deadly Police Encounters with Data Science | Towards Data Science\",\"isPartOf\":{\"@id\":\"https:\/\/towardsdatascience.com\/#website\"},\"primaryImageOfPage\":{\"@id\":\"https:\/\/towardsdatascience.com\/understanding-deadly-police-encounters-with-data-science-3cf1192d9778\/#primaryimage\"},\"image\":{\"@id\":\"https:\/\/towardsdatascience.com\/understanding-deadly-police-encounters-with-data-science-3cf1192d9778\/#primaryimage\"},\"thumbnailUrl\":\"https:\/\/towardsdatascience.com\/wp-content\/uploads\/2020\/12\/1mg8EtCL1vTxbcb3ewlh1Iw.png\",\"datePublished\":\"2020-12-27T19:12:38+00:00\",\"dateModified\":\"2025-04-24T21:01:46+00:00\",\"breadcrumb\":{\"@id\":\"https:\/\/towardsdatascience.com\/understanding-deadly-police-encounters-with-data-science-3cf1192d9778\/#breadcrumb\"},\"inLanguage\":\"en-US\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"https:\/\/towardsdatascience.com\/understanding-deadly-police-encounters-with-data-science-3cf1192d9778\/\"]}]},{\"@type\":\"ImageObject\",\"inLanguage\":\"en-US\",\"@id\":\"https:\/\/towardsdatascience.com\/understanding-deadly-police-encounters-with-data-science-3cf1192d9778\/#primaryimage\",\"url\":\"https:\/\/towardsdatascience.com\/wp-content\/uploads\/2020\/12\/1mg8EtCL1vTxbcb3ewlh1Iw.png\",\"contentUrl\":\"https:\/\/towardsdatascience.com\/wp-content\/uploads\/2020\/12\/1mg8EtCL1vTxbcb3ewlh1Iw.png\",\"width\":1600,\"height\":900,\"caption\":\"Image by author.\"},{\"@type\":\"BreadcrumbList\",\"@id\":\"https:\/\/towardsdatascience.com\/understanding-deadly-police-encounters-with-data-science-3cf1192d9778\/#breadcrumb\",\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"name\":\"Home\",\"item\":\"https:\/\/towardsdatascience.com\/\"},{\"@type\":\"ListItem\",\"position\":2,\"name\":\"Understanding Deadly Police Encounters with Data Science\"}]},{\"@type\":\"WebSite\",\"@id\":\"https:\/\/towardsdatascience.com\/#website\",\"url\":\"https:\/\/towardsdatascience.com\/\",\"name\":\"Towards Data Science\",\"description\":\"Publish AI, ML & data-science insights to a global community of data professionals.\",\"publisher\":{\"@id\":\"https:\/\/towardsdatascience.com\/#organization\"},\"alternateName\":\"TDS\",\"potentialAction\":[{\"@type\":\"SearchAction\",\"target\":{\"@type\":\"EntryPoint\",\"urlTemplate\":\"https:\/\/towardsdatascience.com\/?s={search_term_string}\"},\"query-input\":{\"@type\":\"PropertyValueSpecification\",\"valueRequired\":true,\"valueName\":\"search_term_string\"}}],\"inLanguage\":\"en-US\"},{\"@type\":\"Organization\",\"@id\":\"https:\/\/towardsdatascience.com\/#organization\",\"name\":\"Towards Data Science\",\"alternateName\":\"TDS\",\"url\":\"https:\/\/towardsdatascience.com\/\",\"logo\":{\"@type\":\"ImageObject\",\"inLanguage\":\"en-US\",\"@id\":\"https:\/\/towardsdatascience.com\/#\/schema\/logo\/image\/\",\"url\":\"https:\/\/towardsdatascience.com\/wp-content\/uploads\/2025\/02\/tds-logo.jpg\",\"contentUrl\":\"https:\/\/towardsdatascience.com\/wp-content\/uploads\/2025\/02\/tds-logo.jpg\",\"width\":696,\"height\":696,\"caption\":\"Towards Data Science\"},\"image\":{\"@id\":\"https:\/\/towardsdatascience.com\/#\/schema\/logo\/image\/\"},\"sameAs\":[\"https:\/\/x.com\/TDataScience\",\"https:\/\/www.youtube.com\/c\/TowardsDataScience\",\"https:\/\/www.linkedin.com\/company\/towards-data-science\/\"]},{\"@type\":\"Person\",\"@id\":\"https:\/\/towardsdatascience.com\/#\/schema\/person\/f9925d336b6fe962b03ad8281d90b8ee\",\"name\":\"TDS Editors\",\"image\":{\"@type\":\"ImageObject\",\"inLanguage\":\"en-US\",\"@id\":\"https:\/\/towardsdatascience.com\/#\/schema\/person\/image\/23494c9101089ad44ae88ce9d2f56aac\",\"url\":\"https:\/\/secure.gravatar.com\/avatar\/?s=96&d=mm&r=g\",\"contentUrl\":\"https:\/\/secure.gravatar.com\/avatar\/?s=96&d=mm&r=g\",\"caption\":\"TDS Editors\"},\"description\":\"Building a vibrant data science and machine learning community. Share your insights and projects with our global audience: bit.ly\/write-for-tds\",\"url\":\"https:\/\/towardsdatascience.com\/author\/towardsdatascience\/\"}]}<\/script>\n<!-- \/ Yoast SEO plugin. -->","yoast_head_json":{"title":"Understanding Deadly Police Encounters with Data Science | Towards Data Science","robots":{"index":"index","follow":"follow","max-snippet":"max-snippet:-1","max-image-preview":"max-image-preview:large","max-video-preview":"max-video-preview:-1"},"canonical":"https:\/\/towardsdatascience.com\/understanding-deadly-police-encounters-with-data-science-3cf1192d9778\/","og_locale":"en_US","og_type":"article","og_title":"Understanding Deadly Police Encounters with Data Science | Towards Data Science","og_description":"Gaining insight into police violence and systemic racism through classification","og_url":"https:\/\/towardsdatascience.com\/understanding-deadly-police-encounters-with-data-science-3cf1192d9778\/","og_site_name":"Towards Data Science","article_published_time":"2020-12-27T19:12:38+00:00","article_modified_time":"2025-04-24T21:01:46+00:00","og_image":[{"width":1600,"height":900,"url":"https:\/\/towardsdatascience.com\/wp-content\/uploads\/2020\/12\/1mg8EtCL1vTxbcb3ewlh1Iw.png","type":"image\/png"}],"author":"Dr. Amanda J. Cheney","twitter_card":"summary_large_image","twitter_creator":"@TDataScience","twitter_site":"@TDataScience","twitter_misc":{"Written by":"Dr. Amanda J. Cheney","Est. reading time":"11 minutes"},"schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"Article","@id":"https:\/\/towardsdatascience.com\/understanding-deadly-police-encounters-with-data-science-3cf1192d9778\/#article","isPartOf":{"@id":"https:\/\/towardsdatascience.com\/understanding-deadly-police-encounters-with-data-science-3cf1192d9778\/"},"author":{"name":"TDS Editors","@id":"https:\/\/towardsdatascience.com\/#\/schema\/person\/f9925d336b6fe962b03ad8281d90b8ee"},"headline":"Understanding Deadly Police Encounters with Data Science","datePublished":"2020-12-27T19:12:38+00:00","dateModified":"2025-04-24T21:01:46+00:00","mainEntityOfPage":{"@id":"https:\/\/towardsdatascience.com\/understanding-deadly-police-encounters-with-data-science-3cf1192d9778\/"},"wordCount":2302,"publisher":{"@id":"https:\/\/towardsdatascience.com\/#organization"},"image":{"@id":"https:\/\/towardsdatascience.com\/understanding-deadly-police-encounters-with-data-science-3cf1192d9778\/#primaryimage"},"thumbnailUrl":"https:\/\/towardsdatascience.com\/wp-content\/uploads\/2020\/12\/1mg8EtCL1vTxbcb3ewlh1Iw.png","keywords":["Classification","Data For Change","Editors Pick","Police","Racism"],"articleSection":["Data Science"],"inLanguage":"en-US"},{"@type":"WebPage","@id":"https:\/\/towardsdatascience.com\/understanding-deadly-police-encounters-with-data-science-3cf1192d9778\/","url":"https:\/\/towardsdatascience.com\/understanding-deadly-police-encounters-with-data-science-3cf1192d9778\/","name":"Understanding Deadly Police Encounters with Data Science | Towards Data Science","isPartOf":{"@id":"https:\/\/towardsdatascience.com\/#website"},"primaryImageOfPage":{"@id":"https:\/\/towardsdatascience.com\/understanding-deadly-police-encounters-with-data-science-3cf1192d9778\/#primaryimage"},"image":{"@id":"https:\/\/towardsdatascience.com\/understanding-deadly-police-encounters-with-data-science-3cf1192d9778\/#primaryimage"},"thumbnailUrl":"https:\/\/towardsdatascience.com\/wp-content\/uploads\/2020\/12\/1mg8EtCL1vTxbcb3ewlh1Iw.png","datePublished":"2020-12-27T19:12:38+00:00","dateModified":"2025-04-24T21:01:46+00:00","breadcrumb":{"@id":"https:\/\/towardsdatascience.com\/understanding-deadly-police-encounters-with-data-science-3cf1192d9778\/#breadcrumb"},"inLanguage":"en-US","potentialAction":[{"@type":"ReadAction","target":["https:\/\/towardsdatascience.com\/understanding-deadly-police-encounters-with-data-science-3cf1192d9778\/"]}]},{"@type":"ImageObject","inLanguage":"en-US","@id":"https:\/\/towardsdatascience.com\/understanding-deadly-police-encounters-with-data-science-3cf1192d9778\/#primaryimage","url":"https:\/\/towardsdatascience.com\/wp-content\/uploads\/2020\/12\/1mg8EtCL1vTxbcb3ewlh1Iw.png","contentUrl":"https:\/\/towardsdatascience.com\/wp-content\/uploads\/2020\/12\/1mg8EtCL1vTxbcb3ewlh1Iw.png","width":1600,"height":900,"caption":"Image by author."},{"@type":"BreadcrumbList","@id":"https:\/\/towardsdatascience.com\/understanding-deadly-police-encounters-with-data-science-3cf1192d9778\/#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https:\/\/towardsdatascience.com\/"},{"@type":"ListItem","position":2,"name":"Understanding Deadly Police Encounters with Data Science"}]},{"@type":"WebSite","@id":"https:\/\/towardsdatascience.com\/#website","url":"https:\/\/towardsdatascience.com\/","name":"Towards Data Science","description":"Publish AI, ML & data-science insights to a global community of data professionals.","publisher":{"@id":"https:\/\/towardsdatascience.com\/#organization"},"alternateName":"TDS","potentialAction":[{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https:\/\/towardsdatascience.com\/?s={search_term_string}"},"query-input":{"@type":"PropertyValueSpecification","valueRequired":true,"valueName":"search_term_string"}}],"inLanguage":"en-US"},{"@type":"Organization","@id":"https:\/\/towardsdatascience.com\/#organization","name":"Towards Data Science","alternateName":"TDS","url":"https:\/\/towardsdatascience.com\/","logo":{"@type":"ImageObject","inLanguage":"en-US","@id":"https:\/\/towardsdatascience.com\/#\/schema\/logo\/image\/","url":"https:\/\/towardsdatascience.com\/wp-content\/uploads\/2025\/02\/tds-logo.jpg","contentUrl":"https:\/\/towardsdatascience.com\/wp-content\/uploads\/2025\/02\/tds-logo.jpg","width":696,"height":696,"caption":"Towards Data Science"},"image":{"@id":"https:\/\/towardsdatascience.com\/#\/schema\/logo\/image\/"},"sameAs":["https:\/\/x.com\/TDataScience","https:\/\/www.youtube.com\/c\/TowardsDataScience","https:\/\/www.linkedin.com\/company\/towards-data-science\/"]},{"@type":"Person","@id":"https:\/\/towardsdatascience.com\/#\/schema\/person\/f9925d336b6fe962b03ad8281d90b8ee","name":"TDS Editors","image":{"@type":"ImageObject","inLanguage":"en-US","@id":"https:\/\/towardsdatascience.com\/#\/schema\/person\/image\/23494c9101089ad44ae88ce9d2f56aac","url":"https:\/\/secure.gravatar.com\/avatar\/?s=96&d=mm&r=g","contentUrl":"https:\/\/secure.gravatar.com\/avatar\/?s=96&d=mm&r=g","caption":"TDS Editors"},"description":"Building a vibrant data science and machine learning community. Share your insights and projects with our global audience: bit.ly\/write-for-tds","url":"https:\/\/towardsdatascience.com\/author\/towardsdatascience\/"}]}},"distributor_meta":false,"distributor_terms":false,"distributor_media":false,"distributor_original_site_name":"Towards Data Science","distributor_original_site_url":"https:\/\/towardsdatascience.com","push-errors":false,"_links":{"self":[{"href":"https:\/\/towardsdatascience.com\/wp-json\/wp\/v2\/posts\/9504","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/towardsdatascience.com\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/towardsdatascience.com\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/towardsdatascience.com\/wp-json\/wp\/v2\/users\/18"}],"replies":[{"embeddable":true,"href":"https:\/\/towardsdatascience.com\/wp-json\/wp\/v2\/comments?post=9504"}],"version-history":[{"count":0,"href":"https:\/\/towardsdatascience.com\/wp-json\/wp\/v2\/posts\/9504\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/towardsdatascience.com\/wp-json\/wp\/v2\/media\/9505"}],"wp:attachment":[{"href":"https:\/\/towardsdatascience.com\/wp-json\/wp\/v2\/media?parent=9504"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/towardsdatascience.com\/wp-json\/wp\/v2\/categories?post=9504"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/towardsdatascience.com\/wp-json\/wp\/v2\/tags?post=9504"},{"taxonomy":"sponsor","embeddable":true,"href":"https:\/\/towardsdatascience.com\/wp-json\/wp\/v2\/sponsor?post=9504"},{"taxonomy":"author","embeddable":true,"href":"https:\/\/towardsdatascience.com\/wp-json\/wp\/v2\/coauthors?post=9504"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}