Class prediction reversal during model training - xgboost4j-spark


#1

During my model training, I noticed that predictions and probabilities looks a bit odd. Conventionally, I have always seen that higher probabilities are positive class and lower probabilities are negative class. But in my trained model,

probability < 0.5 is categorised as “1” (Positive Class)
probability > 0.5 is categorised as “0” (Negative Class)

Is this odd ? And why would the prediction directions be reversed ?