| With the rapid development of computer technology,the role of machine learning in the medical field has gained widespread attention.The clinical experience of physicians has long played a decisive role in the treatment and prognosis of breast cancer.However,the accumulation of experience requires a large amount of clinical practice,and as the volume of data grows exponentially each year,manual processing of the data to find patterns is time-consuming and unsatisfactory.In order to improve the accuracy and efficiency of the clinician’s diagnosis,the data must be processed and mined by computer using more efficient algorithms.The data in this paper were obtained from real-world follow-up data of 3088 women with breast cancer between 1995 and 2000.The following studies were conducted.(1)In this paper,a double-weighted Bayesian algorithm is proposed to address the generally unbalanced nature of medical data.This paper proposes a double-weighted Bayesian algorithm to address the general imbalance in medical data.Firstly,it is validated on a public dataset;secondly,the double-weighted Bayesian algorithm and the plain Bayesian algorithm are used to predict the recurrence time of patients based on the follow-up data of breast cancer patients from the real world,and the results show an improvement in the prediction of recurrence time of breast patients compared to the plain Bayesian algorithm.(2)In this paper,the breast cancer patient data were processed to balance the data;four machine learning algorithms,namely K-nearest neighbour,support vector machine,decision tree and deep feed-forward network,were used to construct a multi-classification learner for experiments;then integrated learning was used and compared with Bagging algorithm and Adaboost algorithm;it is fully demonstrated that integrated learning of voting method can improve the effectiveness of single machine learning in cancer medical data.Among the three integrated learning algorithms,the Ada Boost algorithm performs best in classifying breast cancer patient follow-up data.This paper uses this as the basis for establishing a reliable and stable recurrence time prediction model for breast cancer patients.(3)In this paper,estrogen receptor-positive patients who underwent breast-conserving surgery were selected as the study group,and the KM method in survival analysis was used to investigate the benefit of endocrine therapy on patients’ recurrence;a Cox multi-factor risk model with inverse probability propensity score weighting was used to analyse breast cancer patients to obtain patients’ independent influencing factors of recurrence,and a column line graph model was established based on these independent influencing factors of recurrence,which has good accuracy and discrimination,resulting in visual predictions of patients’ disease-free survival at three,five and ten years.The double-weighted Bayesian algorithm proposed in this paper for medical data has a substantial improvement in the classification and prediction of unbalanced data compared to the plain Bayesian algorithm,and the machine learning model established is of clinical use in predicting the time to recurrence of breast cancer patients;according to the survival analysis it enables patients with tumour nucleus grade III to avoid recurrence due to under-treatment and patients with tumour nucleus grade The model provides clinicians with decision support for precision medicine and personalised treatment in clinical treatment and prognosis,and enables patients to benefit from recurrence prediction through early intervention,which has certain guidance and reference significance for clinical medicine. |
