Sun Jul 22 08:52:46 EDT 2012
Smoothing financial data
Using the heuristic that large amounts span larger time frames, it
seems best to parameterize the data based on span ~ amount in such a
way that the integral is equal to the size of the transaction.
I.e. "shape-invariant" smoothing.
How to implement? This needs an evaluation function that sums over
all data points, evaluating the interpolating function, maybe a
To determine a proper scale it's probably best to sync to income,
i.e. a quadratic spline spanning 2 months corresponds to the average
Maybe negative and positive amounts should also be covered
differently: income lags but expense trails.
What do I need
- CSV -> Haskell
- Haskell -> Gnuplot