I Found this on Zindi Rule for the competition
If the error metric requires probabilities to be submitted, do not set thresholds (or round your probabilities) to improve your place on the leaderboard. In order to ensure that the client receives the best solution, Zindi will need the raw probabilities. This will allow the clients to set thresholds for their own needs.
Since the competition is about predicting 1's and 0's can we set a threshold before when predicting the probabilities before appending it to 1's and 0's
Better you predict 1's and 0's instead to predict probabilities.
The rule said if it's a probability prediction
Since it isn't required to predict probability then just predict 1s and 0s and don't bother about that rule as it doesn't apply to this competition. Just go ahead and predict 1s and 0s not probabilities
Get the probabilities and do np.argmax