基于随机森林模型的肺部肿瘤浸润性预测与分析研究

探讨肺部肿瘤浸润性的独立危险因素,建立预测分类模型,并针对现下肺部肿瘤预测模型多使用传统Logistic回归方法,容易导致拟合过度并且预测准确度有限的问题,提出了基于随机森林(RF)算法,建立肺部肿瘤浸润性的风险预测模型,与Logistic回归、决策树(CART)、支持向量机(SVM)、XGBoost从模型区分度、模型校准度和临床适用度3方面进行比较,结果表明,随机森林分类器对于肺部肿瘤浸润性数据集有着更高的预测准确性、一致性和临床适用性。对变量特征重要性进行分析发现,恶性概率对判断肺部肿瘤浸润性有着显著影响,对于促进早发现、早诊断、早治疗肺癌有着重要应用价值。

基于随机森林的个人信用评估模型研究及实证分析

信用评估是商业银行控制和防范信贷风险的关键途径,针对当前个人信用评估模型多使用单一分类器,容易导致过拟合且预测精度有限的问题,提出了基于随机森林组合分类算法的个人信用评估模型,并在实证分析中与KNN、RBF-NET、SVM等单分类器模型以及组合模型GBDT比较,发现基于随机森林组合分类器模型,在个人信用评估的应用中,具有更高的预测精度和稳定性。通过对特征变量评价发现,贷款者个人信息中现有账户状态(透支或有余额等情况)、信贷期限、信贷历史记录、贷款金额对信用风险预测准确率有显著的影响。

Expression of HIF-1α in breast cancer and precancerous lesions and the relationship to clinicopathological features

Objective： The aim of this study was to observe the expressions and clinical significance of HIF-1a in breast cancer and precancerous lesions, and analyze the relationship between the expressions and clinicopathological features in breast cancer. Methods： We analyzed the HIF-1a expression in 128 cases of invasive ductal carcinomas, 146 precancerous lesions patients including 89 cases of ductal carcinoma in situ and 57 cases of atypical ductal hyperplasia. 53 cases of usual ductal hyperplasia breast tissues were selected as a control group. The specimens were evaluated for HIF-1a, estrogen receptor （ER） ＆ progesterone receptor （PR）, epidermal growth factor receptor type 2 （HER2/neu） and Ki-67. Immunoreactivity was semi-quantitatively evaluated in at least 1000 cells examined under the microscope at 40 x magnification and recorded as the percentage of positive tumor cells over the total number of cells examined in the same area. The percentage scores were subsequently categorized. The express of HIF-1a and their relationship with multiple biological parameters including ER ＆ PR, HER2/neu and Ki-67, the biomarkers levels of CA153, CA125 TSGF, and CEA in blood serum and nipple discharge, histological grade, region lymph node metastasis, distant metastasis and recurrence on files were also assessed. Results： Compared with usual ductal hyperplasia, the positive expression rate of HIF-1a in atypical ductal hyperplasia, ductal carcinoma in situ and invasive ductal carcinomas group was significantly increased （P 〈 0.01）. The positive rates of HIF-1a in invasive ductal carcinomas were 68.75%, which were significantly higher than that in ductal carcinoma in situ （43.8%）, atypical ductal hyperplasia （31.6%）, usual ductal hyperplasia （9.4%; X2 = 13.44, 22.27, 52.79, respectively, P 〈 0.01）. Statistical analysis showed that difference of abnormal expression rate of HIF-1a between ductal carcinoma in situ and usual ductal hyperplasia （X2 = 18.37, P = 0.00）, atypical ductal hyperplasia and usual ductal hyperplasia （x2 = 8.14, P = 0.00） was significant （P = 0.00）. However, no significant difference in the positive expression rate of HIF-1a was found between atypical ductal hyperplasia and ductal carcinoma in situ tissue （X2 = 2.19, P = 0.14）. There was a significantly difference in the mean HIF-1a frequency between ER ＆ PR positive invasive ductal carcinomas group and negative group, epidermal growth factor receptor type 2 （HER2/neu） positive and negative groups, Ki-67 proliferation index 〈 14% and 〉 14% groups, histological grade （I ＋ II） and grade III invasive ductal carcinomas groups, with lymph node metastasis, distant metastasis and recurrence groups （P 〈 0.05） and without groups （P 〈 0.05）. However, there was not difference in the mean HIF-1a between age （〈 50 years vs 〉 50 years）, tumor diameter （〈 2 cm vs 〉 2 cm; P 〉 0.05）. The nipple discharge and serum levels of CA153, TSGF, CA125 and CEA in invasive ductal carcinomas HIF-1a positive patients were significantly higher than those in the negative patients （P 〈 0.05）. Conclusion： In breast cancer, HIF-1a expressibn was abnormally increased. The aberration of HIF-1a may play a key role during oncogenesis （atypical ductal hyperplasia or ductal carcinoma in situ） and promote breast cellular transformation into malignancy, a finding useful for further understanding of tumorigenesis. The abnormal expression of HIF-1a may be as an early event in the development of breast tumor. The over-expression of HIF-1a might be important biological markers for invasion, metastasis and recurrence of breast cancer.