I created an algorithm to fill missing values in a list of numbers. I have a list for example [2,4,5,3,6,8,4,NaN,5,NaN,4,6,4,2,4,6,NaN.....]. Think of it much much bigger.
I am looking for tools where they can do the same thing. Fill those missing values. But not in random. In my case, I use a neural net to find the most likely number to put there. I have 90% correct with my algorithm and I want to compare it with other tools in the same data.
Can you suggest any tool to try it?
You're not going to find a piece of software that only interpolates, so you're going to have to get something like Matlab, R, etc, and perhaps also additional toolkits or packages (for Matlab and R, respectively).
If you get R, base interpolation methods include approx (linear) and spline (cubic, Hermite). R package akima has aspline (a unique algorithm that results in "hand drawn" results), package signal has interp (filter-based) and interp1 (linear, nearest-neighbor, pchip, cubic, cubic spline with smoothing), package splines has interpSpline (B-spline).
You may need to slightly reformulate your data to use some of them, giving an x (position of the value in the series) and a y (value) for non-NaN's, then asking them to interpolate at the x values that had NaN's.
That's just what I found by searching for 'interpolate'. R has many other packages to do neural nets, wavelets, time series, and a whole host of other methods that would interpolate your data. Some of these other methods may require some domain knowledge (what kind of data series this is).
So it might be wise to first figure out methods that can do interpolate-ish things (NN, state space, time-series methods, wavelets, splines, etc, etc) and then find a package that does many of them and proceed from there.
