Hi Everyone This looks very promising and I think we should discuss this together. Suresh- can we meet and go through what has been done with Ramyaa, Adam and Nadeem through Skype? We should also discuss how we can turn this into a publishable paper. Except M W afternoons-evenings I am available at all other times Arvind
On Tue, Sep 10, 2013 at 10:29 AM, Nadeem Anjum <[email protected]>wrote: > Hi All, > > I have enabled the agent to commit crimes, which is decided by a number of > factors like reward associated with the crime, risk involved, number of > times the spot has been visited, whether police is present in vicinity, > whether the guardian of the spot is away . The simulation is being run for > two weeks. For the first week the agent commits no crime and just gets > acquainted with the spots. In the second week the agent commits crime based > on the above factors. > > You can have a look at it here: http://gw76.iu.xsede.org/criminfo/ > > A sample simulation has been saved by the name - "crimetest" . You can > click on load and enter - crimetest - to view it. You can also run a new > simulation on cities in taiwan (say taipei). > > We have four global variables: > > - risk_avoidance : a value between 0 and 1 (generated randomly) > - profit_seeking : a value between 0 and 1 (generated randomly) > - maxprofitpossible : set to 100 for this simulation > - maxriskpresent : set to 100 for this simulation > > Crime Spots are labelled as "*CS-n [reward, risk, p ,g ,f]*" on the map, > where > > where, > > - n is the id of the crime spot, > - reward is profit value associated with the spot - this is > initialized as a random number generated between 0 and > maxprofitpossible > - risk = inherent_risk + > risk_due_to_the_number_of_times_the_spot_is_visited > - inherent_risk is initialized as a random number generated between 0 > and maxriskpresent. > - risk_due_to_the_number_of_times_the_spot_is_visited = > num_times_visited * 0.02 * maxriskpresent > - p = probability that police is present in the vicinity of the spot > - g = probability that the guardian of the spot is not present at the > spot > - f = number of times crime has been committed at this spot > > The algorithm for deciding whether the agent commits a crime is as follows: > > //check patrolling police presence > tmp = random num between 0 and 1 > if(tmp > p){ //patrolling police not present > tmp = random num between 0 and 1 > r = risk > if(tmp < g){ // guardian is away > r = r/2 //risk is reduced to half . this is only for this > particular case. the risk value //stored for > this spot remains unchanged > } > tmp = random num between 0 and 1 > if( tmp > risk_avoidance*r/maxriskpresent.){ //agent decides > to take the risk > tmp = random num between 0 and 1 > if(tmp < profit_seeking*reward/maxprofitpossible){ > //reward is good enough > *commit_crime* > } > } > } > > Thanks, > Nadeem. > >
