Research And Analysis Of Information Prediction Based On Machine Learning

Information is the representation of the interrelation and interaction between objective things.It reflects the actual state of the movement of objective things and the changes of the substance.The genes that can regulate protein expression are biological information,and people's daily communication and news are all information.If we can predict this information,it can make our life better.This dissertation starts with three practical problems,predicting breast cancer diseases,lung image recognition and hits of the news information prediction,and studying information prediction with machine learning.The main work of this dissertation is as follows:First of all,(1)the dissertation makes a deep research on the prediction of breast cancer on the early with the bioinformatics marker,and analyzes the predicting results of various machine learning methods on breast cancer.A classification model with high prediction accuracy and high recall rate is proposed.(2)In order to reduce the number of model features,we use three different feature selection methods to screen for microRNAs that have a greater impact on disease prediction.We propose a model of breast cancer with a combination of 5 microRNAs for prediction.The recall of the model only reduced 0.5%than the model used all features.In the second issue of the dissertation,we use the convolution neural network AlexNet to recognize the lung image,at the same time,in order to enhance the generalization of the model,we combine the adversarial generative generation network with AlexNet.The precision of the model can reach about 81.6%.Then,the paper deeply studies the click rules of network news and the prediction model of news popularity by capturing Sina Weibo and Sina news data.(1)The article first analyzes the relationship between the click-delay of a message and the amount of click-through,the relationship between the release time of a message and clicks,and the impact of big news on the click curve.(2)By the low-pass filter,it finds that the click sequence of a message is weighted by multiple gamma function.Therefore,the paper proposes a model of click sequence using a mixture of gamma distribution to fit the message,and the fitting result is good.(3)In predicting the influence of news,the paper puts forward two different methods to predict the popularity of news,one is based on the current more common method of feature engineering,the prediction accuracy can reach about 83%.The other method is based on the deep learning method.TextCNN which is suitable for the task of text classification is selected on the classification model.The accuracy of this method is also about 73%.