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

61–70 of 90 posts

Re: Prophet: Automatic Forecasting Procedure

#61

Post by laichzeit0 »

ImaCake wrote: post #47
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As others have pointed out, Prophet is not a particularly good model for forecasting, and has been superseded by a multitude of other models.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…

> they expect that your data comes at a very regular interval Does prophet rely on this assumption? For health timeseries data the tool of choice is survival analysis - typically using Cox proportional hazards regression or similar regression tools that are able to handle irregular or censored data. I've seen some moves towards using fancy bayesian or fancier machine learning stuff for clinical trials but a big issue…

There’s also the auton survival library. I’ve used it for very big survival models with time varying coefficients:

https://autonlab.org/auton-survival/

Re: Prophet: Automatic Forecasting Procedure

#62

Post by em500 »

mochomocha wrote: post #59
Donald wrote: post #43

Earlier quoted context omitted.

Do you have an example implementation of reimplementing the core of these?

It's literally what I did at work last week, which is why I found this submission timely. I'd have to check with my employer if it can be made public. I don't see any reason why not, there's not much to it.

What did you use to implement the regularization of the trend breakpoints? Prophet by default uses a regular grid and thins them out with STAN. I couldn't find a quick regularization replacement in numpy/scipy/statsmodels with equivalent performance. (I don't want to drag in another huge dependency with Torch or TF).

Re: Prophet: Automatic Forecasting Procedure

#63

Post by hackernewds »

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As others have pointed out, Prophet is not a particularly good model for forecasting, and has been superseded by a multitude of other models.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…

What does Dart do that a multibillion dollar entity with an excellent open sourcing track record misses doing? Perhaps it addresses a niché case well. Genuinely curious

Re: Prophet: Automatic Forecasting Procedure

#64

Post by hackernewds »

alexmolas wrote: post #12
wendyshu wrote: post #10

Earlier quoted context omitted.

Isn't recommended by whom?

Every time I, or someone at work with more experience than me, have tried Prophet it has ended up in changing the approach and trying a different technique. In my experience with time series hand-crafted recipes tend to work much more better than out-of-the-box solutions.

What do you instead? I think Prophet does get close to an answer that isn't "depends" like everyone else is suggesting as an alternative

Re: Prophet: Automatic Forecasting Procedure

#65

Post by hackernewds »

xs83 wrote: post #55

Im a Data Engineer in a large consulting company and I have been incredibly impressed with AutoGluon for forecasting. You can build and train a model in around 10 lines of code and it frequently gets into the top 3 or 4% of competitions on Kaggle without much data pre-processing

Any good tutorials on how to implement it?

Re: Prophet: Automatic Forecasting Procedure

#66

Post by hackernewds »

ttt333 wrote: post #7
pantsforbirds wrote: post #2

Has anyone else struggled with Prophet? I've experimented with it on a few real world datasets and I've had very inconsistent results.

Yes. I've tried using it for pretty straightforward time series forecasts, and I struggled to make it into something useful in a business context. I'll disclaim that I'm just a finance dude and not a data scientist or programmer. But the documentation leads me to believe that I am in the target audience. I felt like I could grasp the basic mechanics after reading the paper, but I wish the documentation could help som…

I'm sad to see no one has responded with a solution to your problem. You are absolutely the target audience, and in my experience, Prophet is "as good as it gets" to generalized forecasting.

Re: Prophet: Automatic Forecasting Procedure

#67

Post by hackernewds »

beckingz wrote: post #42
simonhughes22 wrote: post #38

Wondering how many people are now downloading this and other libs like Dart and trying to do stock market prediction or crypto price forecasting. Most of the devs i know, myself included, have dabbled in coding up trading algorithms at some point in time.

It's the classic data nerd trap. "I'm pretty good at statistics and can predict things using software... I bet I could make money in the stock market" And then they realize just how hard it is.

If it were easy, quants wouldn't be getting paid $1M in TC

Re: Prophet: Automatic Forecasting Procedure

#68

Post by hackernewds »

SpaceManNabs wrote: post #46
beckingz wrote: post #42

Earlier quoted context omitted.

It's the classic data nerd trap. "I'm pretty good at statistics and can predict things using software... I bet I could make money in the stock market" And then they realize just how hard it is.

the hard part isn't the stats. it is all the information that people buy and setting up those ingest pipelines! If i had a satellite telling me when a certain big company has a lot of cars in the lot parked after hours, I could make a zillion bucks too!

buying up and setting up those ingest pipelines seems easy? it's trivial to do #cars detection automatically.

in fact, this sort of alternate data is pretty commonplace in firms I've worked at.

Re: Prophet: Automatic Forecasting Procedure

#70

Post by specproc »

hackernewds wrote: post #63
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…

What does Dart do that a multibillion dollar entity with an excellent open sourcing track record misses doing? Perhaps it addresses a niché case well. Genuinely curious

As others have mentioned in the thread, Prophet has been abandoned, and in my experience anyway, wasn't all that great.
Post reply on HN