Research On Cancer Detection Based On Machine Learning

Cancer is one of the most dangerous diseases that threaten human health which is difficult to prevent,prone to recurrence,easy to transfer,and high in mortality.Studies indicate that diagnosing cancers comparatively early can increase survival probabilities of patients.Hence,the early detection screen of cancers is very significant for patients’survival time and quality.In recent years,Metabolomics and Proteomics are the key research directions in the field of bioinformatics.Metabolomics studies the dynamic development of the body by analyzing differences in metabolic substances in the body,which can provide new ideas for the early cancer diagnosis.Proteomics can reflect the change of human health condition by using the change of the body’s protein levels,which has also made significant advances in cancer detection.Based on the above,this paper proposes a method for cancer diagnosis using Metabolomics and Proteomics combined with Machine Learning algorithms,and the following researches include:(1)A cancer diagnosis method using the model Extreme Learning Machine(ELM)is proposed and researched.The data of thyroid cancer and paraganglioma patients provided by a hospital in Shanghai is used to compare the classification performances of Extreme Learning Machine,Back Propagation Neural Network(BPNN),Support Vector Machine(SVM),k-Nearest Neighbor((kNN)and Random Forests(RF)and verify the feasibility and advantages of these methods.The experimental results show that the extreme learning machine is not only feasible but also has certain advantages in the classification of cancer.(2)A Multi-dimensional mass spectrometry data feature selection method based on Particle Swarm Optimization(PSO)combined with extreme learning machine classifier is proposed.The particle swarm optimization algorithm is used to find the optimal solution of feature selection.Statistical analysis was performed on the features obtained by particle swarm optimization.The experimental results show that the mass spectrometry lines of healthy samples are quite different from the mass spectrometry lines of cancer samples.(3)The particle swarm optimization algorithm is used to optimize the features and the biomarker points are selected in these features,and the selected biomarker points are further classified and evaluated in the next step.The experimental results show that the classification accuracy of the biomarker points extracted by the particle swarm algorithm in the cancer sample set is as high as 98.44%.In summary,the proposed cancer classification diagnosis method based on metabolomics and proteomics combined with machine learning has a good recognition effect on cancer cases,and provides new research ideas for cancer diagnosis,which has important clinical significance.