Research And Application Of Data Mining Artificial Neural Network In Prognosis Of Colorectal Cancer

[Object] 1.Comparison of BP artificial neural network model and traditional method COX proportional hazards model,logistic regression model for colorectal cancer patients after 5 years of survival time prediction performance.2.The BP artificial neural network model was used to predict the survival time of patients with colorectal cancer,and the method of predicting the survival time of colorectal cancer was explored.[Method]We retrospectively analyzed the medical records of 327 patients with colorectal cancer treated from January 1,2010 to December 31,2011 at the First Affiliated Hospital of Fujian Medical University.1.Cox multivariate and binary logistic regression were used to screen the factors related to the survival of patients with colorectal cancer.By mapping and comparing the area under the Roc curve,BP artificial neural network and logistic regression model were used to determine the survival of patients with colorectal cancer The prediction effect;2.The predictive performance of BP artificial neural network and Cox proportional hazards model in colorectal cancer patients was compared by using coherence index C.3.using the paired T test to compare the predicted survival time and the actual survival time of the BP artificial neural network model to test the prediction effect of the model.[Results]1.Cox regression analysis showed that CEA,surgical method,postoperative pathology,postoperative chemotherapy,perioperative blood transfusion were the prognostic factors of postoperative colorectal cancer patients(P <0.05);by paired T test analysis and comparison BP neural network and Cox proportional hazards model of 50 pairs of coherence index C suggest that BP neural network prediction performance is better than the Cox risk ratio model(P = 0.018).2.Logistic regression analysis showed that CEA,surgical procedure,postoperative pathology and tumor invasion depth were the prognostic factors of colorectal cancer patients(P <0.05).ROC curve analysis showed that the area under the curve of BP neural network model 0.930)was greater than the Logistic regression model(0.908),and Z test showed statistically significant difference(P = 0.041).3.BP neural network model in predicting the survival time of patients with colorectal cancer and the actual survival time of patients with paired T test analysis showed that BP artificial neural network has a good predictive performance,the difference was not statistically significant(P = 0.997).[Conclusion]1,BP neural network model is superior to traditional Cox proportional hazards model and Logistic regression model in the treatment of non-linear medical data.2,BP neural network model in the prognosis of survival time to provide a more accurate prediction,worthy of clinical data and other non-linear data processing to promote.