I have a data set with historical information of some events (let's say event A and event B),these events describe the discovery of land mines, the coordinates of the event and the date of the event; is there a way I can use this historical information to predict points (coordinates) where event A or B could happen i.e. where might be still land mines that haven't been found?
3
-
$\begingroup$ How much data do you have about the events? Can you predict if they are going to happen at all? $\endgroup$– solarflareCommented May 22, 2018 at 1:53
-
$\begingroup$ I think this topic falls under Game theory...and so i have added a relevant tag $\endgroup$– user9947Commented May 22, 2018 at 4:31
-
$\begingroup$ I'm facing a similar problem. I have lat,long,time as the three main dimensions of my event points. $\endgroup$– Avinier NakamotoCommented Jun 5 at 19:10
Add a comment
|
1 Answer
$\begingroup$
$\endgroup$
3
Leaving aside the time aspect, you could do a cluster analysis on the event coordinates. If you use an algorithm that gives you a medoid (ie centre) of the clusters, you can then look at other points, and work out how close they are to the centres of the event clusters. It might be possible from this to predict which event could happen at those coordinates (which is the closest cluster medoid), and how likely it is (distance from the medoid).
This, however, depends very much on the shape of the data. If there is no discernible structure contained in it, then this will not work. But it is definitely worth trying.
-
$\begingroup$ But how are you going to represent historical info mathematically? Like wars will never occur between same countries..if there are different countries then size, population, resources, past history also comes into play...I dont think this is a simple ML problem as your answer indicates $\endgroup$– user9947Commented May 22, 2018 at 4:34
-
$\begingroup$ You could go to three dimension, taking time as the third. $\endgroup$ Commented May 22, 2018 at 8:01
-
$\begingroup$ @DuttaA My answer tries to deal with the information given in the question. Of course it might be a much more difficult issue, but at least my suggestion is something the OP can try out. I am fully aware (and have mentioned that) that it might not work if the data doesn't fulfil certain requirements. $\endgroup$ Commented May 22, 2018 at 8:33