Penalized Logistic Regressions Predict Up And Down Trends Of Stock

In recent decades,people’s ability to collect,transmit,and process data has evolved rapidly.At the same time,with many new problems and challenges,machine learning methods provide a viable approach.For example,a large number of gene sequence information,transcription data,protein data emerge.If this data is fully utilized to screen for benign and malignant tumors,it can accurately and effectively screen out whether the tumor is benign or malignant.Early diagnosis and targeted therapy for patients are essential to reduce mortality in humans.Among common cancers,both liver and lung cancer have extremely high incidence and mortality.In this paper,machine learning methods are mainly used to study the benign and malignant problems of liver tumors and lung tumors.For the prediction of benign and malignant liver tumors,this paper selects the data of 579 liver tumor samples on the Kaggle big data website.Eight classifiers were established with age,sex and 8 liver function indicators as predictors,and benign and malignant liver tumors as response variables: logistic regression,support vector machine,artificial neural network,limit gradient boosting algorithm,Ridge penalty logistic regression,Lasso penalty logistic regression,Elastic-Net penalty logistic regression and Adpative Lasso penalized logistic regression.The predictive model was learned from 492 patients as training samples,and the predictive performance of the model was tested by 87 patients,and the different prediction accuracy and prediction performance of the eight methods were compared.Ridge penalty logistic regression had the highest prediction accuracy,obtaining an accuracy of 0.812,a sensitivity of 0.790,and a specificity of 0.843.Therefore,the use of Ridge punishment logistic regression can effectively predict the benign and malignant conditions of liver tumors.For the prediction of benign and malignant lung tumors,this paper selects 603 lung tumors sample data from the UCI website of the University of Wisconsin,in which age and whether the patient is transferred to the general ward（yes,no）,intensive care（yes,no）,hemoglobin and platelets and other 22 indicators are used as predictors,and the benign and malignant lung tumors are used as the response variables.The ratio of benign and malignant tumors in the original data is 13:2,and due to the large difference in the ratio of categorical variables,the ratio of benign and malignant tumors in the data becomes 3:1 after the data processing using the SMOTE function of R software,so that the benign and malignant data of tumors are more balanced.Eight classifiers are established for the benign and malignant lung tumors as response variables: logistic regression,support vector machine,artificial neural network,Ridge penalty logistic regression,Lasso penalty logistic regression,Elastic-Net penalty logistic regression,Adpative Lasso punitive logistic regression,and group Lasso punitive logistic regression.706 patients were used as training samples to learn the prediction model,and 176 patients were used to test the predictive performance of the model to compare the different prediction accuracy and prediction performance of the eight methods.Among them,the prediction accuracy of the artificial neural network is the highest,obtaining an accuracy of 0.925,a sensitivity of 0.898,and a specificity of0.933.Therefore,the use of artificial neural networks can effectively predict the benign and malignant conditions of lung tumors.Logistic regression,support vector machine,artificial neural network,limit gradient enhancement algorithm and four kinds of punishment logistic regression models are used to predict benign and malignant problems of liver tumors,which concludes that Ridge punishment logistic regression predicts the best performance with an average accuracy of 0.812;logistic regression,support vector machine,artificial neural network and five kinds of punishment logistic regression models are used to predict benign and malignant lung tumors,among which the artificial neural network is the best performance to predict with an average accuracy of 0.925.A Ridge punishment logistic regression model can be used to predict benign and malignant problems of liver tumors,and an artificial neural network model can be used to predict benign and malignant problems of lung tumors.