I am looking to implement D-S Theory in my (computer science) research, I'll be using it to determine the probability that a triggered sensor event is a true positive.

How would you calculate an initial belief value without having the ability to perform data mining on a dataset (a cold start)?

One solution that has been postulated is to use the manufactures performance statistics that the sensors are working correctly and then adjust this over time to take account for false positives. Although this will result is a very high initial belief rate for some sensors (95% belief).

  • 1
    $\begingroup$ Why is this not a mathematics question? $\endgroup$ – Raphael Nov 18 '15 at 12:15
  • $\begingroup$ I assumed it was fine here as it was being applied in a computer science context. $\endgroup$ – Colin747 Nov 18 '15 at 12:31
  • $\begingroup$ Then please make this connection; it's not visible in your post as it is now. $\endgroup$ – Raphael Nov 18 '15 at 12:44
  • $\begingroup$ Sorry I thought it was implied when referring to data mining approaches. $\endgroup$ – Colin747 Nov 18 '15 at 14:23
  • $\begingroup$ There are DST questions on Cross-Validated too. $\endgroup$ – E. Douglas Jensen Aug 9 '18 at 15:39

If the 95% (initial) belief corresponds to the probability of a true positive according to the performance statistics, then that seems appropriate to me.

| cite | improve this answer | |

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.