search

sketch-city / project-ideas

Running list of all project ideas - pick one and run with it!
http://sketch-city.github.io/project-ideas/
89 stars 7 forks source link

Houston Crime Prediction Engine #100

Open dylanjbarth opened 7 years ago

dylanjbarth commented 7 years ago

The idea would be to use Houston crime reports to create a predictive model about where and when crime is most likely to happen in the future.

Potential use cases:

It would also be interesting to augment the crime reporting data with weather, event, and demographic data.

randy7771026 commented 7 years ago

UCR data on the HPD site and therefore on the Odata and dataworld are always 45 to two months behind. I have been pulling geojson from http://cohgis-mycity.opendata.arcgis.com/datasets/8d515a90e80840b3bc7a3ada352b0d15_0?uiTab=table&filterByExtent=false pretty much every 30 days. When you move that to github and open it you get https://github.com/randy7771026/jan212017/blob/gh-pages/HPD_RecentCrime02212017.geojson more comment coming

randy7771026 commented 7 years ago

We did a data jam https://houstondatavis.github.io/data-jams/#!/december-2016

randy7771026 commented 7 years ago

I had been making monthly maps of crimes close to businesses for Adrian Garcia to make better use of officers in the Greater Northside Mangement District https://randy7771026.github.io/jan212017/

All of that to say your #100 is a fun idea and a lot of folks have played with it but not from a project sense. Also that some earlier limits to data were removed and the city then built this beta http://mycity.houstontx.gov/recentcrime/recentcrime_v2.html and that I plan to update the UCR and am also working on making the data that right now is just on my machine and on github available in some form for the hackathon.

randy7771026 commented 7 years ago

Also that the 311 data we have may have some relevant data patterns and the notes in that data may give us more.

dylanjbarth commented 7 years ago

Thanks @randy7771026 for all the resources and feedback! To add something else to the list, I'm also trying to get in touch with the developers behind https://github.com/VincentHPD/. It looks like that project was started with exactly this in mind, but not much movement on it the past couple of years.

poprox commented 7 years ago

FYI - police don't like the term "predictive" so much, my understanding is they prefer the term "risk-based policing" for what it's worth.

You may want to take a look at Risk Terrain Modeling (RTM) developed by Dr. Joel Caplan, who is the Deputy Director of the Rutgers Center on Public Security. Their website is http://www.rutgerscps.org

Risk Terrain Modeling (RTM) identifies spatial crime vulnerabilities by identifying and communicating environmental attractors of crime incidents. It is evidence based, data-driven analysis that focuses on the spatial dynamics of crime and excludes demographic information. RTM goes beyond “Hot Spot” analysis that shows where crime has occurred by visually and statistically showing areas of elevated risk of crime occurring. Many police organizations around the world are using these tools.

It is my understanding HPD uses similar tools, but I'm not sure of the exact toolset they use.

dylanjbarth commented 7 years ago

@poprox thanks for the heads up around the terminology, somebody should have told the founders of http://www.predpol.com/ that before they named their company 😆

Thanks for the link, I'll have a read.

I'd be really interested to find out what tech HPD currently uses.

poprox commented 7 years ago

@dylanjbarth also check out the Esri open source toolset http://solutions.arcgis.com/local-government/help/repeat-and-near-repeat-analysis/ ---- it performs similar analysis and is free, you will however need a Esri desktop license and can get a free 60 day eval here - https://www.arcgis.com/features/free-trial.html

randy7771026 commented 7 years ago

There have recently been a lot of changes at HPD http://www.chron.com/news/houston-texas/article/Wave-of-promotions-shakes-up-Houston-Police-11037589.php If You have a local PIP meeting I would definitely encourage your attendance.

Within http://randy7771026.github.io/connect/ and http://randy7771026.github.io/connect/ you will find ways to call divisions and offices. I have found that if once you have defined an assistant chief in charge of a project and you call the Chief's office and ask for them I was directed to the wrong links for information but that allowed me to deconstruct and find the right links then ask our friendly city people for some changes which were part of getting the full geojson datasets.

http://www.houstontx.gov/police/phone.htm

I have to wonder about Crime Analysis & Command Center 713 308-1900 Criminal Intelligence 713 308-8700 from http://www.houstontx.gov/police/phone.htm

On Fri, Apr 28, 2017 at 10:17 AM, Nick Popovich notifications@github.com wrote:

@dylanjbarth https://github.com/dylanjbarth also check out the Esri open source toolset http://solutions.arcgis.com/ local-government/help/repeat-and-near-repeat-analysis/ ---- it performs similar analysis and is free, you will however need a Esri desktop license and can get a free 60 day eval here - https://www.arcgis.com/ features/free-trial.html

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/sketch-city/project-ideas/issues/100#issuecomment-298026290, or mute the thread https://github.com/notifications/unsubscribe-auth/ADMLJw0c0PdKwSj6HUDyiZ2j2TXn2jaIks5r0gLvgaJpZM4NKZjc .

randy7771026 commented 7 years ago

OOPS here is the pretty one http://randy7771026.github.io/connect-restyle/

On Fri, Apr 28, 2017 at 10:54 AM, Randy Baxley randy7771026@gmail.com wrote:

There have recently been a lot of changes at HPD http://www.chron.com/news/houston-texas/article/Wave-of- promotions-shakes-up-Houston-Police-11037589.php If You have a local PIP meeting I would definitely encourage your attendance.

Within http://randy7771026.github.io/connect/ and http://randy7771026.github.io/connect/ you will find ways to call divisions and offices. I have found that if once you have defined an assistant chief in charge of a project and you call the Chief's office and ask for them I was directed to the wrong links for information but that allowed me to deconstruct and find the right links then ask our friendly city people for some changes which were part of getting the full geojson datasets.

http://www.houstontx.gov/police/phone.htm

I have to wonder about Crime Analysis & Command Center 713 308-1900 <(713)%20308-1900> Criminal Intelligence 713 308-8700 <(713)%20308-8700> from http://www.houstontx.gov/police/phone.htm

On Fri, Apr 28, 2017 at 10:17 AM, Nick Popovich notifications@github.com wrote:

@dylanjbarth https://github.com/dylanjbarth also check out the Esri open source toolset http://solutions.arcgis.com/lo cal-government/help/repeat-and-near-repeat-analysis/ ---- it performs similar analysis and is free, you will however need a Esri desktop license and can get a free 60 day eval here - https://www.arcgis.com/feature s/free-trial.html

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/sketch-city/project-ideas/issues/100#issuecomment-298026290, or mute the thread https://github.com/notifications/unsubscribe-auth/ADMLJw0c0PdKwSj6HUDyiZ2j2TXn2jaIks5r0gLvgaJpZM4NKZjc .

LawExplorer commented 7 years ago

In The Silo Effect, author Gillian Tett included a chapter about a similar project that was successful for a time in Chicago ([New York: Simon & Schuster, 2015], 139-63). The protagonist of the project, Brett Goldstein, subsequently led a project called WindyGrid.

dylanjbarth commented 7 years ago

Just to capture some of the discussion from the meetup last night, Ameet had a cool idea about predicting specific types of crime that is less likely to get attention, eg environmental crime.

randy7771026 commented 7 years ago

https://github.com/CivicScape/CivicScape

randy7771026 commented 7 years ago

I had a few distractions this morning but Hopefully

randy777 10:36 AM ok https://data.world/houston/houston-uniform-crime-reports should be as up to date as it gets data.world Houston Uniform Crime Reports - dataset by houston Crime incidents that have occurred in the City of Houston, 2015-2016 randy777 11:16 AM And though the format and likely the extent changes after February you should find enough more recent data to build anything that comes to your minds in https://github.com/randy7771026/jcrime/tree/gh-pages GitHub randy7771026/jcrime jcrime - first thirty day grab which is better formated on the #resources channel in the slack

judopp commented 6 years ago

I would be curious whether the storm and all the disruption associated with it shows up as a discontinuity in the crime data. Will the sectors of beats with the high crime numbers change? And, if so, what does it say about the new "hot spots"?

randy7771026 commented 6 years ago

I had the same thought as I tried to search for an incident reported on Nextdoor two nights ago in http://mycity.houstontx.gov/recentcrime/recentcrime_v2.html and as the year to date stats were presented of the last month versus a year ago last month showed overall lower numbers. Presentation was at our Northside Police Interaction Presentation.

On Thu, Sep 21, 2017 at 8:53 AM, judopp notifications@github.com wrote:

I would be curious whether the storm and all the disruption associated with it shows up as a discontinuity in the crime data. Will the sectors of beats with the high crime numbers change? And, if so, what does it say about the new "hot spots"?

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/sketch-city/project-ideas/issues/100#issuecomment-331163234, or mute the thread https://github.com/notifications/unsubscribe-auth/ADMLJ8Hao3CS8EkZiZr2hCOJwU0hHeouks5skmpPgaJpZM4NKZjc .

alankjackson commented 6 years ago

I've been studying crime in the Heights, for what it's worth. Current analysis lives at http://www.ajackson.org/crime I spent an inordinate amount of time cleaning up the HPD datasets. They were pretty ugly. I plan to add the constable data as well.