Hacker News new | comments | show | ask | jobs | submit login

A good implementation will have a bunch of ready-to-use functions for - polynomials of various degrees, at least.

Polynomial curve fitting can be "dangerous". As in, extrapolating results can be unpredictable if not done right.

I recall an example where 7-8 points, trending up, are fitted with a polycurve that sharply dropped right after the last data point thereby giving erroneous prediction.

Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | DMCA | Apply to YC | Contact