I'm trying to fit data in Mathematica. I import data in Mathematica after using findfit function then I get some results
model1 = c + d Sin[2 \[Pi] n /T y];
f2 = FindFit[data, model1, {c, d, n, T}, y]
f1 = FindFit[data, model, {a, b, \[Omega], \[Phi]}, x]
Out [1]:= {c -> 0.0318682, d -> -0.00101215, n -> -98.6938, T -> -98.8437}
Show[Plot[Evaluate[model1 /. {f2}], {y, 0, 20}, PlotStyle -> {Red}], ListPlot[data, PlotRange -> All]]
Above result is not good so please suggest more methods to find curve fitting from data.
Did you try using the NMinimize method? The FindFit function minimizes the distance between the curve and the data points, this minimization can be done by NMinimize which is very accurate. To do this you just add the command Method-->NMinimize as the last entry of the FindFit[...] call, i.e.
FindFit[data, model1, {c, d, n, T}, y, Method -> NMinimize]
Does this help you at all?
You might also try increasing the MaxIterations. For example,
FindFit[data, model1, {c, d, n, T}, y, Method -> NMinimize, MaxIterations-->1000]
and see if that helps.
Also you won't be able to fit both "n" and "T", only the ratio "n/T", so you should remove one variable.
Dear Jonathan Parry ,
I already try this NMinimize method in Mathematica. If you have any idea about this question then tell me Please.
Dear Sanjay
I tried to do the fit myself, with data generated from a randomly chosen value of the parameters c, d, and n/T, and found that the FindFit function worked very well. I even added some random uncertainty or "noise" to the generated data points to make it more realistic, and again the fit was still very good.
Is there something wrong with your data points perhaps?
Edit:
If the ratio n/T is large then it can cause the fitting function to oscillate rapidly and make the fit rather difficult. You could try to add a constraint to n/T such as,
FindFit[data, {model1,nT>x}, {c, d, nT}, y, Method -> NMinimize]
then change the value of x starting at 0, 1, ..... and see if you can find a good fit for a certain x.
Also have a look at other options like "NormFunction" to see if they can help with the fit.
Dear Sanjay
I am a young maths beginner, I saw your answer, but you can't download the PDF of "How to make an effective search and downloads the about Mathematical switchable viewer from the Internet", if you have time can you send me email: [email protected]，thanks.
Mathematica has a lot of capability to fit functional forms to data. But the key is what does your data look like? Can you upload a PDF/Excel of your data so we can what the difficulty is.
Dear Sanjay, using FindFit in Mathematica has the strong request that one should already know the fitting model function, which is a thing not easy to manage. I suggest you using methods based on Moving Average (in particular, Savitzky–Golay filter is optimal, see attached paper) or Convolution Kernel; with these tools I have had good results on data approximation. Gianluca