Research And Application Of Colorectal Cancer Survival Prediction Model Based On Deep Belief Network

Colorectal cancer,including colon cancer and rectal cancer,is one of the most common malignant tumors in China.At present,the incidence and mortality of colorectal cancer ranks fifth among all malignant tumors globally.According to the latest statistical report of the National Cancer Institute,the 5-year survival rate of colorectal cancer patients is 64.4%,while the 5-year survival rate of colorectal cancer patients in China is less than 48%.Therefore,studying the survival time of colorectal cancer patients is of great significance for doctors to make clinical decisions,evaluate patient prognosis,and improve treatment options.However,the prediction of cancer patient survival usually depends on doctors using clinical data or scoring systems to make judgments based on subjective experience.The prediction accuracy is not high.With the rapid development of modern information technology,data mining and big data analysis based on machine learning are standardized Application,many disease prognosis models have been constructed to help improve such predictions.This article aims at the characteristics of colorectal cancer patients with complex index variables and large data volume,and proposes a prediction model of colorectal cancer patient survival based on deep belief network(DBN)to help doctors evaluate patient prognosis,improve treatment options,and then improve colorectal cancer Patient viability.The main contents of the study are as follows:(1)Based on the construction of DBN colorectal cancer survival prediction model.In this paper,Establishing a deep confidence network through the construction and stacking of Restricted Boltzmann Machine and the BP neural network is used on the top layer to perform reverse parameter tuning,and finally a DBN prediction model is generated.In view of the difficulty of network depth selection in the DBN model,the reconstruction error based on RBM training is used to autonomously determine the number of hidden layers,which improves the model fitting ability.(2)Data preprocessing and model tuning.According to the NCCN clinical guidelines,AJCC 7th edition cancer staging guidelines,and related research literature,the SEER database screens out data sets that are closely related to colorectal cancer.Preprocessing the extracted data includes: clustering,dimension reduction,numerical,Data cleaning and normalization.The thesis combines the initially constructed model and uses the processed data to optimize the network parameters,and then determines the optimal DBN model structure suitable for the prediction of survival of colorectal cancer patients.Finally,compared with the processing results of the classic BP neural network model,the experimental results show that the prediction accuracy based on the deep belief network model is higher.(3)System design and implementation.Combined with the DBN prediction model that has been constructed with the optimal structure in the paper,an online prediction system for colorectal cancer survival has been developed and implemented,which provides a practical method and way for doctors to evaluate the prognosis and clinical research of colorectal cancer patients.