Research And Implementation Of Lung Prediction Methocd Based On Machine Learning

Lung cancer is the most common malignant tumor of the respiratory system.In China,the morbidity and mortality of lung cancer are in the first place.Lung cancer lacks the typical symptoms at an early stage.once found,patients are mostly in the middle and late stage and are easily accompanied by distant metastasis,making treatment very difficult.The other reason for the high mortality rate of lung cancer is its poor prognosis,a considerable number of patients will have recurrence after treatment.Early diagnosis and prognostic analysis of lung cancer play an important role in improving the survival rate.The development of high-throughput sequencing technology and the emergence of analytical methods have brought a new direction for the prediction and treatment of lung cancer.In the process of tumor formation,multiple genes are required to participate and interact with each other.the whole gene expression data of lung cancer can be obtained by high-throughput sequencing,and the molecular mechanism can be used to study the occurrence and development of lung cancer,it can provide theoretical support for early detection and prognostic diagnosis of lung cancer.In this study,we obtained data sets from two public databases,TCGA(Cancer Genome Map)and GEO,and established a set of "diagnosis+prediction" system for lung cancer.The system can screen out the susceptibility genes for lung cancer as features to construct a diagnostic model of lung cancer,and judge the subtypes of lung cancer.At the same time,the system predicts the malignant degree of patients with lung cancer and the possibility of rehabilitation,which can achieve higher accuracy..Aiming at the early diagnosis of lung cancer,this study proposes a set of lung cancer susceptibility gene screening and subtype classification model based on machine learning.Firstly,the transcriptome data and clinical information of lung cancer were obtained from TCGA database,and the differential expression was analyzed by R language combined with limma,EdgeR and DEseq.Then,the lung cancer susceptibility genes are accurately screened by GO function and KEGG pathway analysis.The hierarchical cluster analysis was performed and the heatmap was drawn.Lung cancer prediction models were constructed by four machine learning methods:logical regression,decision tree,support vector machine and soft voting,and the model was verified on the data sets obtained by different methods.In addition,according to the results of the discriminant model of lung cancer susceptibility genes,the models with strong sensitivity to lung cancer were selected to construct the classification model of lung cancer subtypes.Through the analysis of the function of genes in important pathways,combined with hierarchical clustering heatmap,we can see that the susceptibility genes identified in this study can clearly distinguish between the two types of samples and play a key role in the occurrence and development of lung cancer,it is very important to understand the pathogenesis and early diagnosis of lung cancer.Aiming at the prediction of the cure possibility of lung cancer,this study proposes a set of prediction model of malignant degree and rehabilitation of lung cancer based on depth neural network.Lung cancer samples are obtained from GEO database,and the feature genes are fed into the depth neural network for training selected by statistical method and model method,and the parameter gradient is updated by reverse gradient descent algorithm.The training time of this model is longer than machine learning algorithms,but it can achieve higher prediction accuracy.By predicting the malignant degree and rehabilitation possibility of lung cancer,it can assist doctors to arrange the next treatment plan,which has guiding significance for clinical treatment.The work of this study is based on the above five modules:screening of lung cancer susceptibility genes,construction of lung cancer diagnosis model,lung cancer subtype classification,lung cancer malignant degree classification and lung cancer rehabilitation prediction model,a set of lung cancer comprehensive prediction system is formed,which can not only analyze and process the data,select the susceptibility genes of lung cancer,but also complete the diagnosis of lung cancer and the prediction of rehabilitation after treatment.It can help patients to detect the disease early,assist doctors in clinical diagnosis and guide the formulation of the next treatment plan.