CsAdaBoost integrated learning algorithm based on cost-sensitive improvement

The invention discloses an improved CsAdaBoost integrated learning algorithm based on cost sensitivity, and belongs to the field of monitoring. Analyzing characteristic values obtained by testing signals in the experiment by using an ensemble learning algorithm, and establishing a defect diagnosis model; cost-sensitive learning is introduced, the difference of misclassification costs of different types of samples is fully considered, and the learning target of a classification algorithm is converted from reduction of overall errors to reduction of classification costs; by changing a loss function optimization target, effective identification of minority class samples is realized under the condition of ensuring the overall accuracy. The CsAdaBoost algorithm is applied to the defect diagnosis process of the transformer, and accurate recognition of the generation position of a characteristic sound signal of transformer equipment is achieved. The method can achieve the quick and accurate recognition of a feature