Colonic Pressure Signal Analysis Method Based on Ensemble Support Vector Machine

Slow transit constipation is a common intestinal motility disorder. The colonic peristalsis can be well analyzed by the colonic pressure signal. However, medical physiological signal data are generally imbalanced, resulting in the low performance of traditional classification methods. To overcome this challenge, this paper proposes a Support Vector Machine classification method based on ensemble method. We extracted statistical features, nonlinear features and wavelet features from the colonic pressure signals, and integrated them as feature vectors to input the classifier. The experimental results show that, compared with the traditional support vector machine, the proposed method significantly improves the recognition effect of the minority class samples. Through grid search for model parameters optimization, ensemble support vector machine not only improves the recognition effect of minority class samples, but also maintains the recognition performance of majority class samples, which effectively alleviates the problem of biased toward the majority class in the classification process and improves the recognition accuracy of slow transit constipation patients.