Decision Tree And It's Application In The Predicting Of Colorectal Cancer After Operation

Objective:1. To explore the application of decision tree of data mining in medical data analysis of colorectal cancer.2. To predict the postoperative survival time of colorectal cancer patients through the decision tree model.Methods:1. Collecting clinical data of colorectal cancer patients that had been confirmed by imaging and postoperative pathology from January 2010 to December 2011 of the first affiliated hospital of Fujian medical university gastrointestinal surgery II Ward,and set up a database software platform.2. Improving the postoperative follow-up, and selecting cases with complete clinical and postoperative follow-up data, then building excel spreadsheet by the factors associated with the patients prognostic, such as gender, age, CEA,perioperative lose suspended red blood cell volume, operation method, operation time, intraoperative blood loss, tumor growth, tumor differentiation degree,primary tumor(T), regional lymph nodes(N), distant metastases(M), the time of distant metastasis, postoperative chemotherapy and other treatment. Assignment and calculating the information gain value of each variable, screening predicting variables that closely related with prognosis.3. Building data flow of model, then, building decision tree model by c5.0algorithm based on the information gain. By using the computer random distribution, processing training and validation for 10 times with 90% for training data set and 10% for validation set.4. Outputting prediction result, comparing and verifying the actual survival time and predict survival time by scatter diagram.Results:1. Establishing a database software platform of colorectal cancer from January2010 to December 2011 of the first affiliated hospital of Fujian medical university gastrointestinal surgery II ward?2. Established the decision tree model of postoperative survival time of colorectal cancer, the output shown: the accuracy of raining sample is 69.8%, the validation sample is 75.8%.3. By comparing scatterplot, the predict survival time and the actual time in a line nearly.4. Verifying the result by scatter diagram of actual survival time and predict survival time, and shown: there is a high conformity degree of predicting survival time and the actual survival time.Conclusion:1. The decision tree model can be used to predict postoperative survival time of patients with colorectal cancer.2. The decision tree model can be predicted for individual of postoperative survival time of patients with colorectal cancer.