Plateau detection

[Former user]

Hi!

Thanks for your great work. I have some date which does not have a peak but several plateaus (consecutive peaks with the same value). Is there a possibility to detect them with your functions?

[Lucas Hermann Negri]

Hi,

Currently no, as peakutils.indexes will not detect plateaus (with exact same values). I think this is a needed feature, and may be added.

[Lucas Hermann Negri]

Fixed by pull request #2.

[Nathan Gilford]

Was the above ever resolved? In my data plateaus are still not detected as peaks:

[Lucas Hermann Negri]

It was. Can you share the data and the exact code snippet use to generate this plot?

[Nathan Gilford]

The data is real-time 1-minute ticks of EURUSD forex charts, using closing prices to draw the line graph. The data is collected through the Oanda API, but should be the same for any real-time source.

import peakutils
from peakutils.peak import indexes
import matplotlib.pyplot as plt

col1 = df.index.values
col2 = df['rates']
pkin = peakutils.indexes(col2, thres=0.05, min_dist=1)
plt.plot(col1, col2, lw=0.4)
plt.plot(col1[pkin], col2[pkin], 'ro')
plt.show()

[Lucas Hermann Negri]

Can you get me a sample that is problematic for you?

[Nathan Gilford]

Here is an example, I've managed to export it all as a csv, but I'm unsure if that is necessary?

ind.  rates
...     ...
24  1.16305
25  1.1631
26  1.1631
27  1.1631
28  1.163

As you can see index 25, 26, 27 in this example all have the same closing rate. I have an above average rate of repeating numbers because I take the close rates from bullish candlesticks and the open rate from bearish ones for a very specific purpose; normally I don't think forex data would have this many repeats, but they're still likely to happen.

What this looks like graphed is:

[Lucas Hermann Negri]

From the data points you suplied:

import peakutils as pu
import peakutils.plot as p
import numpy as np

data = np.array([1.16305, 1.1631, 1.1631, 1.1631, 1.163])
x = np.arange(len(data))
ind = pu.indexes(data, thres=0.05, min_dist=1)
p.plot(x, data, ind)
p.plt.show()

[Bastien BRETEAU]

Works fine for me with version 1.1.0 even with a plateau with 3 consecutives indexes with the exact same value.

Thank you for that very usefull library by the way!

[Nathan Gilford]

Oh I see. It was not an issue with peakutils, but my fault as a beginner to python. While I thought imports always brought in the newest version of a module that was not the case, but chasing down the version and updating fixed the issue right away for me. Thank you so much for your help!

[Lucas Hermann Negri]

I'm glad that it is working now. Special thanks to Christophe Vestri for implementing the plateau detection.