AL QBC diversity density balance committee-weighting

This active learning is about Query-by-committee algorithm (QBC) , while the basic QBC algorithm has been modified to performs better in resolving data imbalances and reducing labeling costs.In this project,qbc.ipynb file is used to test algorithm performance,and qbc_ddbcw.py file is used for practical application,which almost written according to sklearn standard.At the same time,qbc_ddbcw is a QBC algorithm that combines diversity,density,balance and committee-weighting.The other four pictures are about the performance of various QBC algorithms on CTG datasets(UCI). Comparing with the random sampling,the recall of QBC was increased by 1.67%, F1 by 1.03% and accuracy by 0.43%,while the maximum difference was 6.69%, 3.89% and 1.61% respectively.On the other hand,QBC achieves the best performance of random sampling only needs 30% to 40% of its sample size.Among these algorithms, DCW-qbc(diversity&committee-weighting) performs best in recall,balance-qbc performs best in f1 and accuracy.