Submissions will be evaluated using multiclass log-loss, based on the probabilities which you supply on each product class:
\[ logL = -\dfrac{1}{N}\sum_{i=1}^{N}\sum_{j=1}^{K} \textbf{1}_{y_{i} \in j} \log(p_{ij}) \]
where \(N\) is the number of observations in the test set,  \(K\) is the number of classes, \(\textbf{1}_{y_{i} \in j}\) is an in
dicator function, which is 1 if a product \(i\) is in class \(j\) and 0 otherwise. \(p_{ij}\) is the predicted probability of a product \(i\) to belong to class \(j\)