Research On Disease Prediction Method Based On Deep Learning Optimization Algorithm

With the development of artificial intelligence,the ability of computers to process and analyze data has increased exponentially.Medical data has particularity and complexity.To dig out the potentially useful information behind the medical data,there are higher-level requirements for the speed and accuracy of the algorithm.In order to provide more scientific and effective medical services,this subject conducts research from three aspects: feature engineering,gradient descent optimization algorithm and prediction model,and establishes a cardiovascular disease risk prediction model to assist doctors in decision-making.First,in the feature engineering stage,a strong feature vector is constructed and random forest is merged for feature selection.At this stage,based on the background of disease prediction,three new strong feature vectors are constructed and added to the original data set,and the correlation coefficient graph is used to analyze the relationship between the feature vectors.At the same time,the more classic random forest feature selection algorithm in the ensemble learning model is introduced to perform feature selection work and obtain the importance ranking of features so that the selected features are more suitable for the model.Secondly,aiming at the problem that the logistic regression algorithm is easy to fall into the local optimum,a cardiovascular disease prediction model based on LR-N is proposed.The model uses Nesterov accelerated adaptive moment estimation algorithm and adaptive learning rate optimization algorithm(Nadam algorithm)to optimize and improve the loss function of the logistic regression algorithm.Analyze the accuracy,recall,F1-score,MCC value and other evaluation indicators of the LR-N model to verify the effectiveness and usability of the LR-N model.Finally,in order to further improve the prediction effect of the model,a cardiovascular disease prediction model based on R-Lookahead-LSTM is proposed.This model is based on the optimization of the stochastic gradient descent algorithm of the Fast weight part of the Lookahead algorithm to the Rectified Adam algorithm;the Tanh activation function is further improved to the Softsign activation function to promote model convergence,and the R-Lookahead algorithm is used to further optimize the long and short-term memory network model.As a result,the long and short-term memory network model is better improved,so that the model stabilizes as soon as possible;and it is applied to the cardiovascular disease prediction model.