Different result for exponential and power compare to excel

[ektadesai22]

How to reproduce,

1. Have this data set
   [1, 2],
   [1, 3],
   [4, 4],
   [4, 5]
   [3, 6]
   [3, 7]
   [5, 8]
   [5, 9]
   [1, 10]

2. Excel results are
   

3. Whereas library give us below result
   

[kapil1201]

I faced the same issue, apparently there are two different method to find non linear regression coefficients. One is by converting the equation to linear and then minimizing it, and other one is minimizing the non linear function.
Σi(ln(yi)−(A+B xi))2
Σi(yi−a exp(b xi))2
Excel uses the former one, whereas this library uses the latter one. There is no mean involved in this method, but the linear way includes means.
I had to update the exponential code to meet this expectation.

[ektadesai22]

I faced the same issue, apparently there are two different method to find non linear regression coefficients. One is by converting the equation to linear and then minimizing it, and other one is minimizing the non linear function.
Σi(ln(yi)−(A+B xi))2
Σi(yi−a exp(b xi))2
Excel uses the former one, whereas this library uses the latter one. There is no mean involved in this method, but the linear way includes means.
I had to update the exponential code to meet this expectation.

@kapil1201 - Thank you for the reply.

Did you get similar issue for Power and Polynomial regression also?
Would you be kind to help us by sharing your algorithm changes or make a pull request?

[ektadesai22]

@kapil1201 - Could you please respond to this?

[kapil1201]

We have used MLjs library for this purpose.