Skip to content

sebastianangermund/kalman_filter

Repository files navigation

Use

For simple testing, set up a test in simulate_and_test.py. Here you fake the values you want to measure (y_true) and add noise to them (y_observed = y_true + noise). You then run y_observed through the Kalman filter. This is neat because you can plot true/noise/filtered to see the effect of the filter.

The standard deviation doesn't really matter, the filter will converge pretty quickly.
Here is a plot of a function (red) with added noise (yellow) that has been filtered (blue): image info

Filter bitcoin one minute data

Here are some different filter settings for different timespans of one minute bitcoin data. The filter step length is adjusted for the time range: you typically want more filtering for larger timeranges.

image info image info image info image info

About

Filter noisy data

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published