Similarity Measurement Of Multivariate Typhoon Time Series Based On DTW

The typhoon has brought enough rain to the vast areas,and has a close relationship with human life and production.However,typhoon always brings all kinds of damage and it has the characteristics of strong suddenness and destructive force.It is one of the most serious natural disasters in the world and one of the largest Marine disasters in China.The typhoon will bring strong winds,heavy rain,storm surge and other natural disasters.These disasters have high frequency and wide spread,resulting in a large number of casualties and great loss of property.Therefore,it is very important to forecast the typhoon effectively.Similarity measurement is a key step in ocean prediction and disaster prediction.It is of great significance to study the similarity measure of typhoon.Multi-element,dynamic,spatiotemporal,unequal,seasonal and predictable characteristics of typhoon data have caused difficulties for Marine applications,such as Marine forecasting and Marine assisted decision making,bringing the following problems:(1)In view of the multi-element and dynamic characteristics of typhoon data,how to describe it effectively to facilitate the quality assessment and repair of typhoon data;(2)In view of the spatiotemporal characteristics of typhoon data,how to evaluate and repair the data quality of typhoon data so as to improve the accuracy of typhoon similarity measure;(3)In the view of the unequal,seasonal and predictable characteristics of typhoon data,how to select a suitable method for the similarity measure in order to improve the accuracy of Marine forecasting and disaster prevention and mitigation.For this reason,this paper firstly proposes a method of typhoon data description based on multivariate time series.Based on it and in combination with the characteristics of typhoon data,the quality assessment and repair methods of typhoon time series are proposed.In the end,the similarity measurement method of typhoon time series based on weighted Dynamic Time Warping distance is proposed,which improves the accuracy of similar typhoon finding and provides auxiliary decision-making for relevant departments.This paper analyzes the characteristics of multi-element,dynamic,space-time,unequal length,seasonality and predictability of typhoon data,aiming to improve the accuracy of typhoon similarity measurement.This paper studies the method of similarity measurement of typhoon time series,and its main contents include the following three aspects:(1)this paper summarizes the current research status of data quality assessment and restoration method,time series similarity measure methods and typhoon similarity problem,and through the analysis found that the data quality assessment and repair method is seldom involved in Marine areas,rarely using the DTW(dynamic time warping)for typhoon similarity measure,similarity of typhoon most studies focused on the similar typhoon path,which is not conducive to judge the disasters causing by typhoon intensity.(2)in view of the quality problems prevalent in typhoon data,the integrity,consistency evaluation and repair method of typhoon time series are proposed.Firstly,several factors that influence the similarity measure of typhoon are sorted out,and the elements of the research are selected by the principal component analysis method,and the method of typhoon data description based on multi-time series is proposed.On this basis,the quality assessment method for typhoon time series is determined by analyzing the characteristics of typhoon data.Then in view of the problems of the integrity and consistency of typhoon time series,this paper presented the typhoon repair methods.For the integrity problem,the adjacent value repair,wind-pressure relationship repair and latitude and longitude calculation are used.For conformance problems,factor dependency repair is adopted.The experimental results show that the accuracy of the evaluation method and the method of repair is high,and the accuracy of the similarity measurement is an indispensable part of the similarity measurement.(3)in view of the unequal length of typhoon time series,a method of similarity measurement based on principal component analysis and dynamic time wraping distance is designed.According to the different importance of each element in the similarity measurement process,the hierarchical structure method and the principal component analysis method are used to calculate the weight of typhoon elements.At the same time,considering the different length of time series of different typhoons,the similarity measurement of typhoon time series is carried out by dynamic time wraping distance.The experimental results show that considering multiple factors,the weights of the typhoon elements are considered comprehensively,which improve the accuracy of the similarity measure.