In Topcoder’s many marathon challenges, I have faced my thought and ability limitation. But I know well that there is no fixed solution and ability limitation. In this contest, I managed to develop a solution, but I can’t test it.
In this challenge, contestants use tree decision and neural network methods for this solution. But yet I haven’t reviewed them.
In my case, after relationship analysis based on meaning of factors and target features, a probability graphical model is deveoped including Hidden Markov Model and a simple probability relationship learning model, I use standard neural network. I think that this problem’s important point is to predict next events according to previous events. By the way, in contrary to the genenral Markov chain case, events after 1 ~ 3 years based on previous events should be predicted. This problem is very relevant to many order’s hidden Markov chain. It is the core of this contest.
Furthermore, it can be solved with RNN (Recurrent Neural Network).
In such a point, this problem is very interesting.
I didn’t complete my solution. By the way, I want to continue to develop and test this challenge.
So may I get total data including system case’s sociopolitcal data and atrocity data.
Atrocity events fundamently result from paricipants that are human beings but not nations, regions, so and on.