Prediction Of Diabetes Based On Convolutional Neural Network

Diabetes is a common chronic non-communicable disease.Metabolic disorders in patients cause long-term high blood sugar levels in patients.Because the disease is not easy to cure,the long-term high blood sugar level in the patient brings serious harm to the kidney,cardiovascular and nervous system of the patient,which causes many complications and brings great harm to the physical and mental health of the patient.The big harm seriously reduces the patient's life happiness.However,due to the aging of the population and people's bad habits,the incidence of diabetes is still high,and the prevalence rate is on the rise.The prevention and treatment of diabetes has become a global health problem that needs to be solved urgently.With the development of medical informatization,various medical institutions have accumulated a large amount of patient medical electronic data during the patient's diagnosis and treatment.Make full use of a large amount of medical data,and explore valuable treatment rules,which can assist doctors in diagnosis and treatment,thus alleviating the shortage of medical resources in some areas;it also helps to adjust treatment plans according to the patient's condition and improve treatment efficiency.As a multidisciplinary interdisciplinary subject,machine learning has been greatly developed in recent years.As a branch of machine learning,deep learning can be used to explore deeper features of data because of its powerful feature extraction ability.Therefore,it is widely used in many practical applications such as face recognition and speech recognition.How to apply the powerful information mining function of deep learning to medical data mining,and discovering potentially valuable information between data will be of great practical significance and social value.Based on the data of hospital admission records of patients with diabetes after data analysis and feature processing,this paper uses the improved convolutional neural network algorithm to establish a model for predicting diabetes changes.The model can assist the doctor to use the model to judge the development of the disease according to the case record of the patient's diagnosis and treatment,and to judge the treatment effect of the patient,thereby discriminating the effectiveness of the treatment plan.The main work done in this paper is as follows:1.Statistical analysis and feature processing of case data are proposed.Since the dataset used in this study is real data obtained from the hospital,the dataset has a large amount of data,a high data dimension,and a missing value.In order to ensure that the next analysis can be built on good structured data,this experiment first analyzes the data set.This paper cleans the missing values firstly;Distributes the data distribution of the dataset to determine the distribution characteristics of the data.For the phenomenon of unbalanced distribution of dataset categories,data balancing is performed by using a few types of weight oversampling techniques;Abnormal data detection is performed by using isolation forest algorithm,thereby obtaining statistically smooth data;Using Xgboost.The algorithm performs feature weight ordering,removes features that do not contribute to the classification result,and performs feature reduction.2.This study proposes a CNN-EI algorithm based on convolutional neural networks and integrated learning.This study analyzes the characteristics of this data model,selects the convolutional neural network as the predictive model,and combines the excellent data classification ability of integrated learning to improve the algorithm of convolutional neural network,hoping to obtain better prediction effect to assist doctors in diagnosis and treatment.3.In order to verify the necessity of the data processing method proposed in this paper and the validity of the model,a comparative experiment was added in the research process.The experimental results show that compared with the conventional convolutional neural network and the traditional classification algorithm,this study does have good prediction effect and universality,and the data processing part is also beneficial to the improvement of the model performance.4.Based on the above research results,this study constructs a predictive model for the development of diabetes disease,which can be used to judge the development trend of patients' diseases and judge the effect of diagnosis and treatment,so as to be used for auxiliary diagnosis and treatment.