Geek Answers Handbook

How to find tops and bottoms in time series?

At first this question may sound really stupid, but it is not fundamental. Maybe it may seem insoluble in exactly any algorithm, but I pretend to be so.

So the question is. I have a chart like gold. I need to find where the tops and bottoms are axial in time. The problem is that I need to find where the major ups and downs start. The problem is that there are many small irrelevant ups and downs.

Here is an image for a better understanding - the red spots that I want to find (NOT EXACTLY, but somehow).

enter image description here

, , , , . . java .., .

+5
3

  • , / . .

  • / , , .

  • , " " " ". , . . 2.3.2 . ( , Google " ", "EMD" "Hilbert-Huang" Transform").

EMD :

enter image description here

, EMD , " " " ". ( , , -, - .)

+8

- , , N , N (.., yoy , ).

. , ( , - ).

0

, , , ? / , , x?

, /. , . : X , ( ), ( . 2-3 , 2- ).

EDIT: sorry, I missed the understanding, you are trying to find a range. Well, if you have data, then why not find the lowest point (s) first, and then go to the next day and determine the% change, then the next, etc. And rewrite the day you find a higher range. You will need to add a little more logic later (for example,% of this peak or something else), otherwise you will report, as a rule, take a stock of ipo and take it all high and say that it is one period.

0
source

More articles:


All Articles