Symptoms Prediction Study Based On An Optimized Neural Network Algorithm

The unceasing innovation of computer technology brings great convenience to human’s daily lives.It not only change people’s thinking habits,but also facilitating the transformation and development of various industries.Nowadays artificial intelligence and big data have plugged the wings of medical informatics development.With the concept of big data being deep-seated and big data technology becoming more mature and diverse,a great deal of scientific scholars have combined machine learning with big data on disease prediction in the medical field.The thesis study put human breast cancer,a common disease on a big data platform,and use neural network algorithms to carry out prediction research on breast cancer conditions,to give a way to solve medical big data reasonably.First,the thesis excavates and visualizes the SEER database.Breast cancer trends over the past two decades were analyzed in terms of clinical onset / lethality and type of condition by dividing it into datasets with different ethnicities and ages according to the characteristics of the data.At the same time,regression analysis is used to analyze the entire breast cancer fulcrum and local characteristics,to obtain predictions of breast cancer patients for different surgical methods and make accurate assessment of risk factors.Second,the thesis introduces a genetic algorithm and an analog annealing algorithm for the shortcomings of artificial neural networks.In order to solve the "precocity " of traditional GA and the problem of low diversity caused by late iteration,an improved GA is proposed to optimize BP.When the loop is repeated locally,it makes a large mutation when it reaches a certain number of times,constantly determines whether the colony is stable,and if it is stable,it first mutates and then intersects.An improved saga-optimized bp model has been proposed along with a simulation annealing algorithm that allows the entire annealing to ensure that it deviates sufficiently from the local optimal solution after the end of the genetic calculation cycle.In addition to this,the thesis analyzes improved algorithms for parallelism,enabling their rational deployment onto distributed platforms.Finally,a breast life prediction is performed by working out the parameters of an improved neural network model for BP,building a Hadoop platform,and activating the spark grouping in operation,which complements the training and prediction.As proved by multiple experimental results,the optimized BP neural network is not only pinpoint accuracy but also enormously improved in operation efficiency under distributed platform,and prove that the optimized BP neural network model based on spark is feasible,which provides a new method for the research and prediction of breast cancer.