PROCESSING METHOD FOR TIME SERIES DATA AND DECIDING METHOD FOR SIMILARITY OF TIME SERIES DATA

摘要

PURPOSE:To process many similar data without losing the features of a time series and to automatically analyze the degree of similarity of data by picking up the data value of a maximum frequency and a maximum value and a minimum value of data of frequency 0 and plotting these data in accordance with time and generating the chart where they are mapped as a locus on a plane of the time base. CONSTITUTION:A time series data analyzer consists of an automatic analyzer 1 and a picture display device 2. Time series data is subjected to histogram processing at intervals of a prescribed time to pick up the data value of a maximum frequency and a maximum value and a minimum value of data of frequency 0, and these data are plotted in accordance with time to generate the chart where they are mapped as a locus on a plane of the time base. Thus, features of data over a long time {(the prescribed time)XN(an integer)} are decided The locus of the chart is stored as a recognition pattern in a neural network, and the degree of similarity to the pattern of another chart is decided by the character recognition method.