基于LDA和XGBoost算法的乳腺癌预测模型构建研究

Breast cancer is the leading cause of cancer death in women, and the number of male breast cancer patients can not be ignored. Therefore, using information technology to predict the disease is an important way to improve the rate of disease diagnosis. This experiment carries out dimension reduction to the multi index characteristics of the breast cancer dataset provided by the kaggle database, analyzes the medical test indexes of the 498 groups of 30 dimensional breast cancer patients, uses the linear discriminant analysis (LDA) to merge the characteristic attributes, and projects the data to the low dimensional space, and proposes the extreme gradient lifting algorithm (eXtreme Gradient Boosting). Xgboost), which uses grid search for cross validation to obtain the optimal parameters, constructs xgboost prediction model, and uses AdaBoost, random forest and naive Bayes algorithm as performance comparison classifiers; The experimental results show that the classification accuracy of the prediction model trained after dimensionality reduction is 2.7% higher than that before dimensionality reduction, and the classification effect of the prediction model constructed by xgboost is the best, reaching 98.7%.