To get a first approximation on this, assume everyone in the US has access to a gun, then take a look at the FBI's yearly extended homicide tables, which breaks down US homicides by demographic factors, (and means of death)

The first, obvious conclusion from FBI data is that a group making up approximately 4% of the US population commits about half the US homicides, mostly against others in that group, and the other 96% of the US has a homicide rate lower than England - and also much lower violent crime rate than England.

I think that's a very generous assumption considering that strict gun control laws mostly correlates with population density. Gun laws vary a lot from state to state and in most states that have some sort of licensing requirement anyone below the top half of the middle class would have to invest a non-trivial (a little more/less than a driver's license) amount of time and money into owning any sort of firearm. I know in the Boston area the unwritten policy is "municipalities inside I495 typically don't issue concealed carry permits unless you're a cop or politically connected" Basically state laws can skew things a lot.

you can't just take the well behaved subset of the US population and then compare it to an other country on the whole. Thats ridiculous.

You would need to do the same segmentation of the English population, and compare like segments.

Not publicly, so far as I know.

Any time someone purchases a gun from a store, or from most dealers at gun-shows, a background check gets run. This, + paperwork and records residing with individual sellers are your only real sources of data, neither of which are public.

The concept of a national gun registry has been debated hotly for quite a while, but that would definitely not be public if it existed.

oh ok.. then maybe credit card transactions if they dont buy it with cash instead

These aren't clear either. Is that $1000 At Guns R' Us a rifle, or bulk ammo? Is that $25,000 at Guns, Fishing and Other Stuff a rifle with fancy inlay or a new fishing boat.

there are so many gun owners in the US. If you look at that sheer number and compare it to the number of horrible incidents involving guns, I think it would be hard to draw conclusions on gun purchase data.

I think building a model on other factors about the people that commit these acts would perhaps yield better results.

I think you may be going into precog territory while looking for a needle in a haystack. Some of the common patterns seem fairly well documented: disaffected male, mental issues, history of abuse, affiliation with radical islam. But most experts seem to believe finding a key indicator before hand is really impossible since you will also label so many people who will actually resist the urge to violence. The Orlando killer popped on the FBIs radar twice and they couldn’t make the call with all the expertise and experience the organization has. Machine learning a data model to predict violence will have so many false positives as to be either useless or used as a totalitarian hammer. [1, 2, 3, 4]

[1] http://www.huffingtonpost.com/2012/12/19/mass-shooting-psych...

[2] http://www.livescience.com/21787-predicting-mass-shootings.h...

[3] http://andrewgelman.com/2016/02/25/probability-paradox-kills...

[4] https://bayesianbiologist.com/2013/06/06/how-likely-is-the-n...

It's not going to be very interesting, but the entire lists of NYC residents who got handgun and long gun licenses was published a while ago, less than 60K each for a city of 8 million (I counted the totals myself, easy with a fixed number per all but the first and last pages).

Also check the Federal government statistics as organized by the US gun industry lobby, the National Shooting Sports Federation: http://nssf.com/ Look up the ammo production and importation numbers, well over 12 billion per year, and I'm not sure which include the seconds from the DoD Lake City plant that ATK sells to us (the Pitman-Robinson point of sale tax data will).

Maybe also check the raw numbers of concealed carry licenses, the GAO publishes these, but 10 states now don't require a license. I'll add that at the local level of my home county, Jasper in Missouri, as of a year and a half ago, 5% of the age eligible population had a $$$ Missouri concealed carry license ("$$$" because our's is one of the most expensive, and any state's will do if you want to save money). All these numbers are rapidly increasing as the population ages, as well as the other obvious reasons.

Following the [X] I was wondering if we can build a model [Y] and some other data sources like [Z], etc ?

(X = Reichstag Fire, Y = of potential communists, Z = communist party affiliation) (X = 9/11 Attacks, Y = of terrorists, Z = mosque attendance)

I guess you could, but has history taught us nothing? If I am a communist, am I a threat to you? If I am a muslim, am I a threat to you? If I have a gun, am I a threat to you?

I don't own a gun or have a gun license, nor do I want to be the victim of gun violence (or violence of any kind, having experienced it), but I don't know what you would gain from this information if you are looking to profile a potential disaster.

thats why i used both causation and correlation. Correlation probably doesnt help here but until the model is built you can't really tell right?

not sure why i got a down vote for this

Firearms Dealers vs. Burgers, Pizza, and Coffee http://flowingdata.com/2016/06/14/firearms-dealers-vs-burger...

ATF Listing of Federal Firearms Licensees (FFLs) - 2016 https://www.atf.gov/firearms/listing-federal-firearms-licens...

This might offer a start.

I'm not sure about gun purchase data, but the Bureau of Alcohol, Tobacco, Firearms and Explosives has some interesting information on their website (https://www.atf.gov/resource-center/data-statistics). I'm not sure if this is useful for you, but I hope this helps you out.

I read a well done statistical analysis on state-by-state gun ownership correlation.

You might find some of the data sources listed valuable.

http://slatestarcodex.com/2016/01/06/guns-and-states/

If you watch last week's John Oliver you will see that due to heavy lobbying there is no proper electronic record keeping. Gun stores keep record but on paper and when they close the data is put on microfilm stored somewhere.

I heard on NPR that the National Institute of Justice, which is an arm of the Justice Department, has funded some studies. Might want to try there.

to add, there is also training data available in form of people who have committed shootings or have any crime record. its then just about feature selection that could prove some causation or correlation

what do you mean by "training data"?

in machine learning we could use some supervised technique if we have training data instead of unsupervised. sorry for not being clear

what I thought, just making sure.