Algorithm 1: The self-learning algorithm using relabeled strategy

Input:$L$ is the set of labeled data.     $U$ is the set of unlabeled data.     $M$ is the base model.     $ConLev$ is the Confidence level. Output: Trained classifier.

Step 1: Pretrain the model $M$ with labeled data $L$ and obtain the pre-trained model $M^{\prime }$.

Repeat:

Step 2: Predict the unlabeled data $U$ using $M^{\prime }$.

Step 3: Select the instances with the predicted probability more than $ConLev$ per iteration ($U^{\prime }$) using relabeled strategy.

Step 4: Expand $L$ with $U^{\prime }$ i.e., $L+U^{\prime }\to L$ and remove from U.

Step 5: Train the model $M^{\prime }$ with L.

Until: some stopping criterion is met or U is empty.