Research On Classification Of Time Series Data Set Based On Improved Multi-Scale Entropy Algorithm

Functional Magnetic Resonance Imaging(f MRI)has been widely used in human and animal brain or spinal cord experimental research due to the characteristics of non-invasive,high spatial and temporal resolution.The principle of f MRI is to use magnetic resonance imaging to detect changes in blood oxygen level dependent(BOLD)signals in functional areas of the brain.At present,the way to perform statistical analysis of BOLD signals has become a research hotspot,and its research results will be of great significance to the analysis of brain functional activity.This paper studies the characteristics of BOLD signal non-stationary and non-linear properties.Based on the membership function in fuzzy mathematics,an improved Multi-Scale Entropy(MSE)algorithm is proposed,and the algorithm is analyzed theoretically and numerically.Firstly,the first-order coarse graining is extended to the second-order coarse graining.The first-order coarse graining quantifies the fluctuation of the local mean value,and the second-order coarse graining quantifies the dynamic characteristics of local volatility on multiple scales.This paper uses two coarse-grained processes to classify and identify chronic urticaria patients and healthy groups based on the multi-scale characteristics of BOLD signals.Secondly,for the part of the algorithm thats find the distance between two vectors,the hard threshold function turn to the membership function,which solves the problem that the multi-scale entropy algorithm itself will have a lot of uncertain values in the second-order coarse-graining process.In order to avoid uncertain values in the calculation process,the purpose is to expand the threshold within a reasonable range,and to avoid noise interference.The improved threshold function avoids the emergence of uncertain entropy to a certain extent,improves the stability of the algorithm,and promotes the accuracy of classification.Finally,to solve the problem of selecting the threshold value by the improved multi-scale entropy algorithm,this paper uses the Receiver Operating Characteristic(ROC)as the determination criterion,and selects the appropriate threshold after several experiments based on the length of the data,which provides a reliable parameter selection basis for the application of future algorithms.This article uses the measured data of the project to achieve an effective classification of BOLD signals in patients with chronic urticaria.Compared with existing public studies,more regionss of interest have been found.In addition,this paper also uses four sets of open source data to test the improved algorithm,and experiments show that the improved algorithm is universal.