Research On Automatic Diagnosis Of Disease Based On Machine Learning

With the rapid development of artificial intelligence and big data,a large amount of medical data is effectively utilized,and the diagnostic analysis of diseases has entered the era of intelligence.On the one hand,the traditional diagnosis of diseases mainly relies on the direct experience of doctors.The diagnosis results are greatly affected by factors such as human and environment.On the other hand,the distribution of medical resources in China is uneven,top medical institutions are overcrowded and there is a shortage of medical resources and low service levels in remote areas of the grassroots.In order to solve the problem,machine learning algorithm is applied to the automatic diagnosis of human diseases,which not only can help patients to find diseases early and get treatment in time,but also can assist doctors to make correct diagnosis of diseases effectively and reduce the probability of misdiagnosis and missed diagnosis.It can also break the information barrier between urban and rural areas,expand the advanced diagnosis and treatment capacity to the grassroots level,and improve the medical level at the grassroots level.The continuous development of image inspection technology in the medical field has promoted the process of big data analysis in medical images,and the deep learning method has been effectively applied to the recognition and processing of medical image data.This article took two types of data sets as an example.One is small-scale text data—the Indian liver patient data set ILPD,and the other is large-scale pathological image data—the public breast cancer patient data set BreakHis.Different application strategies was proposed for different dataset features,and some results had been achieved in theoretical research and practice.This paper used two research methods that are very effective and commonly used in current machine learning applications: comparative research and improved research.Based on the ILPD dataset,a comparative study was carried out using a variety of machine learning models.Firstly,the construction of three conventional machine learning methods,such as logistic regression,SVM and decision tree,was successfully applied to the ILPD dataset and achieve the classification and prediction of liver diseases;afterwards,a variety of evaluation methods was used to measure the experimental performance of the three algorithm models and the algorithm parameters are continuously optimized,and the optimal liver disease diagnosis model was finally obtained through comparative analysis.Based on the BreakHis dataset,the powerful image processing capability of deep learning was used to analyze the pathological images of breast cancer,and the feature extraction was further efficiently performed.And the data training was improved by improving the algorithm model.This paper proposed an improved DenseNet network breast cancer diagnosis model based on the current excellent DenseNet network to improve the network structure,combined with data enhancement and migration learning strategies,training samples after data enhancement and then feature extraction,and through optimization parameters and Fine-tuning learning had achieved automatic diagnosis of breast cancer for images of different multiples and different tumor types.The method we proposed has strong robustness and generalization ability when compared with previous research results,and its accuracy has also been greatly improved,which is an accurate and effective diagnosis model for breast cancer.