datumbox/NaiveBayesClassifier

Overflow in FeatureExtraction

Ginta opened this issue · 1 comments

Line 102 creates an overflow (chisquareScore sometimes is negative more precisely), I recomend to change int to double in 102. The issue was discovered when changing chisquare from 0 to negative changed the results.

Best,
Ginta

How did you manage to overflow it? I don't think you can store 2147483647 records in memory. Also can you clarify what you mean by "changing chisquare from 0 to negative"?