Non-Linear CURVE FITTING using PYTHON
Vložit
- čas přidán 18. 08. 2019
- A tutorial on how to perform a non-linear curve fitting of data-points to any arbitrary function with multiple fitting parameters.
I use the script package and the script.optimize.curve_fit function to perform an exponential fitting of data-points.
Documentation:
docs.scipy.org/doc/scipy/refe...
Code shown in the video:
www.bragitoff.com/2019/08/non...
Hope you find it useful!
Follow me on:
FB Page: / bragitoff
FB Profile: / ducktape07
FB Page2: / physwhizforum
IG: / ___physwhiz___
IG2: / ducktape07
BLOG: bragitoff.com
FORUM: physwhiz.bragitoff.com
ANDROID APPS:
CrysX: www.bragitoff.com/crysx/
Others: play.google.com/store/apps/de...
I saw at least five tutorials by now, asked three questions on stack-overflow cause that stuff never worked with my function. And finally I found a fellow scientist who gives this marvellous tutorial on curve fitting. Thank you a thousand times, it works!
Did you gibe him a sub and like? 😇
Same as I, this video answered all my questions.
Bless your soul, this was excellent. I was able to find non-linear relationships between temperature and electrical load of a city, for electricity load forecasting. THANK YOU!
I'm glad you found it useful 😊
How do you decide the best initial guess I also want to find a relationship between droplet diameter and height, but the r2 value is around 0.48 so how can I improve thank you
Dude your a god that was so well explained! Love you!
Thank you very much, friend, I've been suffering for two days what you can't imagine to achieve something like this.
Thank you very much! I hope I can help any viewer with this other tips.
# In this example, if you want to know the errors of the fitting parameters "a" and "b" you can add this line to your code:
a, b = np.sqrt(np.diag(pcov))
# If you want to know the r^2 score of the fit, use this lines in your code:
from sklearn.metrics import r2_score
r2 = r2_score(yData, func(xData, *popt))
I wish you the best of lucks!
Thanks for this nice tutorial. I have a question. How can I calculate the r2 (r-squared) of the curve fit? Is there any way to convert the pcov to r2? THANKS!
Is it possible to choose the size of the y-axis? I mean, my experimental y values are very little and i want the graph only to show them in comparision with the curve fitting ...
Awesome!! This helped me a lot.
Thank you for the video! I have one question, how do I determine the uncertainty of the coefficients generated by the curvefit?
what an absolute legend
absolute hero so much better than online forums/documentation explenations
Glad you think so!
How do you find the R squared value?
Thanks man! Very simple code.
Thank you...simple and effective
How can we compute the area under curve ( for exemple for an asymetric gaussian )
I have a question, how can you predict using the curve_fitted function for a specific value but no array, which isn't part of the dataframe? Let's say the x value is 6 and a prediction for this number is needed.
Thanks for the help :))
THANK YOU SO MUCH!!
how to get error ...??
Well done, thank you
How can we get the equation for curve?
U saved my life
thx sir
that was helpful :)
I have this problem, where the data I'm trying to fit is not of polynomial/exponential form.
lovely content! thank you. but your plot .. talk about overfitting the data :D :D :D
i got error saying " ImportError: cannot import name 'read_arguments' from partially initialized module 'popt' " could you please help?
sir kindly make video on improve the goodness(reduce SSE) of fitting curve, thank you
interesting topic. thanks.
Can u do a curve fit with more than one curve, example u have 2 function/model with same parameter and u want to fit those curves with 2datas u have of that function respectively, maybe u can weight the residual of the curves?
thanks
Do you know if it is possible to use scipy functions instead of defining the fitting function yourself? That is, for example, if I want to make a fit my data with lognormal or power law equation, do I have to write it explicitly and change the parameters for each?
This is the most automatic it can get if you want to do meaningful science. We only need to define a fitting function here which we do so by considering the type of problem we are considering.
thanks sir
can I still use curve_fit if I have errors in x but not in y?
Good question. curve_fit may not be the best choice in such a case. You may wanna try odr from scipy.
I tried to do same code, however python did not sketch the curve for me !
hi Mr., How do I know which initial values to consider for the initial conjecture of the parameters?
I will be grateful for the answer
Just an educated guess is usually enough. If you're modelling something you might have an idea about what to expect. In some cases even bad initial guesses may converge. If not then just try another one randomly if you have no odea about what to expect.
Each problem has a method to get initial values. In Process Simulation you must know where do you com from and what do you espect. There no exists a general method o mathematics for the initial value. Cheers
Hi, would you be able to help me I'm trying to fit a curve_fit through my data but it's not working?
Is the code running? Since the curve_fit method requires an initial guess, so that may be the reason if the optimization doesn't converge. You can try different values for initial guess.
The code does run and give a straight line , I have guessed the slope and intercept do I also need a start_slope and start_intercept?
You want to do a linear fit?
yeah a linear one , shall i send you my code?
Although the video is for non-linear fitting, you can still do linear fitting by modifying the func as:
def func(x,m,c):
return m*x+c
Can we do this on real dataset?
It depends on the amount of noise that your data might have. Also, if you have an idea about the what mathematical relationship should it follow.
Thank you so much for explaining in details. Can you explain why you wrote a=%5.3f, b=%5.3f in line 28 of your code?
It means it will print a floating point number with 5 characters, with 3 digits after the decimal point. It is just for formatting the output.
bhai tu gora nhe hy pr koshish poori kr rha hy k wsa boly mgr nhee
You know, You know kam bola kr
what ??
But what’s your problem
I would prefer you to speak in your original accent, please