Labeling Data is time-consuming and boring. Active Machine Learning may help reduce that labeling process down to 10~20%.
The basic workflow is as follows:
1. Label Data Partially
2. Train Model only with the labeled data
3. Predict using the non-labeled data
4. Check confidence score outputted from the model
– If score is LOW: Pass it to human for labeling
– If score is HIGH: Accept Machine Label
1. Return data and repeat process
