Application Of Semi-Supervised Self-Training Algorithm In Breast Cancer Analysis And Prediction

The number of women suffering from breast cancer has increased significantly in recent years.Breast cancer has become the most common cancer among women,and its mortality rate is second only to lung cancer.So far,early detection and diagnosis are still the basis of breast cancer treatment.With the continuous development of artificial intelligence,using the machine learning methods to inductively analyze breast cancer medical data and explore the rules to establish an intelligent diagnosis system for breast cancer has become a research hotspot in the field of health care.At present,there are still many problems in establishing an intelligent diagnosis system for breast cancer.First,researchers can often collect a large number of suspected breast cancer samples,while labeled samples that can diagnose the benign and malignant tumors often require experts to spend a lot of time to consult.Intelligent diagnostic systems often have poor classification accuracy and poor generalization due to the lack of labeled samples.Secondly,a large number of unlabeled samples that in the hands of researchers are not fully utilized.How to choose valuable unlabeled samples to join the training is blind to the researchers.Finally,the original breast cancer data often has the problem of attribute association redundancy,which is not conducive to direct modeling and prediction.At the same time,the researchers also rarely carry out specific induction and analysis of the original medical data of breast cancer.In order to solve the above problems in the establishment of the intelligent diagnosis system for breast cancer,fully explore the value of breast cancer medical data,and establish a diagnostic system with high classification accuracy and high generalization,the work in this paper is as follows:(1)It applies the semi-supervised self-training method to the intelligent diagnosis of breast cancer,and proposes a semi-supervised self-training method based on density peak optimization fuzzy clustering.The method first performs density peak clustering on the unmarked sample set.After selecting out the cluster centers manually,the new cluster centers are used as the initial clustering centers of the fuzzy clustering to perform fuzzy clustering.The experimental results show that the classification accuracy of this method is improved compared with the self-training method combined with other clustering algorithms.(2)It analysis the breast cancer data from the University of Wisconsin Hospital in the United States.Through the method of data visualization and data interaction,the key features that determine the benign and malignant are selected out from the complex features.It provides an important detection direction for exploring the etiology and pathology of breast cancer and early diagnosis.(3)Using the improved self-training method to classify and predict the pre-processed breast cancer data,and establish a breast cancer intelligent diagnosis system based on semi-supervised learning.The system can improve the classification accuracy and the generalization of the classification through the iterative self-training in the case of less initial tagged data.It provides a new idea for the intelligent diagnosis of breast cancer and promotes the development of intelligent medical care.