| Input: a dataset, V, including all the positive and negative samples, and the negative samples are more than the positive samples. |
| Output: ensemble classifier based on integrating multiple classification instances. |
| (1) For to |
| (2) If is a positive sample |
| (3) , is the weight of |
| (4) else |
| (5) |
| (6) End If |
| (7) End For |
| (8) t is used to record the current iteration number, and its initial value is set as 1 |
| (9) While () |
| (10) = Null, the negative training set is emptied |
| (11) If equals 1 |
| (12) All negative samples are gathered into k clusters based on the -means method. Assume set P is composed of all the |
| positive samples and the parameter |
| (13) For each cluster, the sample locating closest to the center is selected and added into . Furthermore, the number of |
| negative samples is equal to that of positive samples |
| (14) else |
| (15) According to the weights of negative samples, k negative samples are selected in proportion to their weights. |
| These samples are added into and |
| (16) End If |
| (17) The training dataset is composed of and . A new classification instance based on SVM is constructed by using |
| the training dataset and integrating their weight distribution |
| (18) is used to classify all the samples in , evaluate its classification performance , |
| and compute its classification error rate |
| (19) The adjustment weight is calculated, and the weight of each positive and negative sample |
| is updated by using the rule |
| (20) |
| (21) End While |
| (22) An integrated classifier is constructed by integrating classification instances based on the voting mechanism. |
| The final classification result is obtained as follows. |
| |
