# Fitting a curve to $N$ dimensional data

I have a data set with $N$ independent variables and one dependent variable(function of all the $N$ independent variables). The dependent variable is either $0$ or $1$ (like a step function). I am looking for a functional form to estimate this data. Is doing a multidimensional linear regression correct for such data or should I look for higher polynomial fit? Doing multidimensional linear regression gives a RMSE of $0.35$ .

P.s :I had read a document where a linear fit has been suggested for the $N=1$ case ,so I went ahead with it first.

• You will probably want to estimate a multiple regression, i.e., $y_i=f(x_1,x_2,...)$. The usual functional forms are either linear (ordinary least squares regression), or nonlinear (Logit, or Probit regression). – Nameless Sep 15 '18 at 18:55