Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Prophet: Automatic Forecasting Procedure

github.com

Post reply on HN
Print view View original

71–80 of 90 posts

Re: Prophet: Automatic Forecasting Procedure

#71

Post by GraceCat123 »

rdli wrote: post #30

As others have pointed out, Prophet is not a particularly good model for forecasting, and has been superseded by a multitude of other models. If you want to do time series forecasting, I'd recommend using Darts: https://github.com/unit8co/darts . Darts implements a wide range of models and is fairly easy to use. The problem with time series forecasting in general is that they make a lot of assumptions on the shape of…

Can we employ stochastic processes like the Poisson process to represent irregular data points? Are there any existing models for this?

Re: Prophet: Automatic Forecasting Procedure

#72

Post by GraceCat123 »

Terretta wrote: post #25
whymauri wrote: post #4

Also relevant: https://news.ycombinator.com/item?id=27695574

This is the HN comment thread on a well-written skeptical article with this zinger: “You can imagine my disappointment when, out-of-the-box, Prophet was beaten soundly by a ‘take the last value’ forecast.”

This example is super classic! XD

Re: Prophet: Automatic Forecasting Procedure

#73

Post by ImageXav »

mochomocha wrote: post #41
nighthawk454 wrote: post #27

Model development on Prophet stopped this year: https://medium.com/@cuongduong_35162/facebook-prophet-in-202... They recommend checking out these for cutting-edge time series forecasting: https://neuralprophet.com/ https://nixtla.github.io/statsforecast/

Fun fact: if you don't care about the auto-regressive aspect of NeuralProphet (it's turned off by default), you can implement the core of NeuralProphet/Prophet (piecewise linear trend + Fourier on weekly/daily seasonality) in about 60 LOCs with no other dependency than either torch or numpy+scipy.optimize, and without having to deal with Stan or the very poorly chosen heuristics of neuralprophet. Another thing that b…

This is interesting to me. Do you use a library to estimate the fourier series of a data series or have you implemented it from scratch? I've searched for this in the past but always got results RE. Fourier transforms, not series.

Re: Prophet: Automatic Forecasting Procedure

#74

Post by GraceCat123 »

tech_ken wrote: post #13
nick0garvey wrote: post #3

Can someone explain why the "no free lunch theorem" does not cause problems here? https://en.wikipedia.org/wiki/No_free_lunch_theorem

Two explanations First: Prophet is not actually "one model", it's closer to a non-parametric approach than just a single model type. This adds a lot of flexibility on the class of problems it can handle. With that said, Prophet is "flexible" not "universal". A time series of entirely random integers selected from range(0,10) will be handled quite poorly, but fortunately nobody cares about modeling this case. Second:…

These insights are really awesome! It reminds me of the common aphorism in Statistics: 'All models are wrong, but some are useful.'These insights are really like a wake-up call, thank you!

Re: Prophet: Automatic Forecasting Procedure

#75

Post by GraceCat123 »

chubs wrote: post #45

On this topic, does anyone know of a suitable time-series forecaster for multivariate analysis? Eg 8 independent/input variables, and one output variable? I've been using multiple linear regression (which works impressively!) but it doesn't take into account the time series, only the single prior day of inputs. Thanks :)

In Machine Learning conference papers, a common approach is to model relationships between variables using Graph Neural Networks (GNNs). Using GNNs is a powerful and flexible way to go. Maybe you can give it a try!

Re: Prophet: Automatic Forecasting Procedure

#76

Post by pachico »

It's the mechanism used by Grafana's forecasting feature. It's still not greatly explained and in many cases it makes it hard for the user to understand it's results as might be below zero for data that can only be positive numbers (requests per second, for instance).

Re: Prophet: Automatic Forecasting Procedure

#77

Post by rdli »

GraceCat123 wrote: post #71
rdli wrote: post #30

As others have pointed out, Prophet is not a particularly good model for forecasting, and has been superseded by a multitude of other models. If you want to do time series forecasting, I'd recommend using Darts: https://github.com/unit8co/darts . Darts implements a wide range of models and is fairly easy to use. The problem with time series forecasting in general is that they make a lot of assumptions on the shape of…

Can we employ stochastic processes like the Poisson process to represent irregular data points? Are there any existing models for this?

A common strategy is interpolation. The challenge is that forecasting itself is a form of interpolation. So you're forecasting based on forecasted data.

Re: Prophet: Automatic Forecasting Procedure

#78

Post by clircle »

GraceCat123 wrote: post #71
rdli wrote: post #30

As others have pointed out, Prophet is not a particularly good model for forecasting, and has been superseded by a multitude of other models. If you want to do time series forecasting, I'd recommend using Darts: https://github.com/unit8co/darts . Darts implements a wide range of models and is fairly easy to use. The problem with time series forecasting in general is that they make a lot of assumptions on the shape of…

Can we employ stochastic processes like the Poisson process to represent irregular data points? Are there any existing models for this?

I think a common “solution” is to use a Gaussian process model.
Post reply on HN