Confusion matrix mention sentiment.
The results of the classifier (parameter values: epoch = 25, learning rate = 0.7, n-grams = 3) after applying the simple certainty rule (neutral if certainty < 0.8): confusion matrix with counts (left), normalized by the true labels (middle) and normalized by the predicted labels (right). The values in the diagonals of the middle matrix are the precision rates, and the values on the diagonals of the right matrix are the recall rates. Recall rates here are reduced due to the certainty rule, but the most important errors (classify positive if the true value is negative and classify negative i... Mehr ...
The results of the classifier (parameter values: epoch = 25, learning rate = 0.7, n-grams = 3) after applying the simple certainty rule (neutral if certainty < 0.8): confusion matrix with counts (left), normalized by the true labels (middle) and normalized by the predicted labels (right). The values in the diagonals of the middle matrix are the precision rates, and the values on the diagonals of the right matrix are the recall rates. Recall rates here are reduced due to the certainty rule, but the most important errors (classify positive if the true value is negative and classify negative if the true value is positive) are reduced.