I got this error.

Please someone help me if you know this.

Hi. Anybody had any issues with LightGBM regressor/ETS returning the same values for each forecasting step ?

Dear community, please share some love: https://twitter.com/MergenthalerMax/status/1596217596248047616?s=20&t=YBg78-kLy3AFiWT-ei85fA

Hey! As part of our #reinventforecasting week, we released the MSTL model. Show some love community ❤️ https://www.reddit.com/r/datascience/comments/z7etr8/on_the_law_of_similarity_or_how_10000yearold/

Dear community, here is our newest release: https://twitter.com/MergenthalerMax/status/1598397618572259328?s=20&t=G__jeLNo12N2j7uCbwQxBA

We created some memes that we did end up not using: feel free to take them lol

Ha, I for one love these memes @Max (Nixtla)

Yay https://medium.com/@valeman/-86655805a676

@Max (Nixtla) what is this DOT thing, is there an article link please somewhere?

Dynamic Optimzed Theta

The theta family of models has been shown to perform well in various datasets such as M3.

StatsForecast

includes four models of the family:

Theta

,

OptimizedTheta

,

DynamicTheta

,

DynamicOptimizedTheta

. The implementation is based on the work of Jose A. Fioruccia, Tiago R. Pellegrini, Francisco Louzada, Fotios Petropoulos, and Anne B.Koehlerf, and optimized using numba.

It was just to long to write jajajaja

I see, great. I think Theta author is someone else though. These two Assimakopoulos, V., & Nikolopoulos, K. (2000). The theta model: a decomposition approach to forecasting. International journal of forecasting,

Thought this post would be relevant for the Nixtla community here, some chatter about it at H2O.ai : https://huggingface.co/blog/time-series-transformers

@Max (Nixtla) is the 4 models combo the same that Svetunkov & Co used in M4? From what I recollect they branded is with name “SCUM” 😀 Or Nixtla combo is a bit different? This is a link to their paper https://researchportal.bath.ac.uk/en/publications/a-simple-combination-of-univariate-models

Hey there, just joined, not sure if we’re supposed to present ourselves in this slack but I’ll do it anyway 🙂 I work on an anomaly detection system called ThirdEye. We use a bunch of forecasting-based method for anomaly detection, even though I mostly work on matrix profile these days. Just discovered your libs with Statistical vs Deep Learning forecasting methods piece of work, and will definitely dig it. Thanks for the awesome work!

Just saw preview from another open source forecasting library that used conformal prediction to boost probabilistic prediction intervals capabilities. The results are amazing, stay tuned will be announced soon. @Max (Nixtla) @fede

For those who have not seen it, Amazon team no longer says Deep Learning has a priority lane when it comes to forecasting.

Quick question: any suggestions for Intermittent demand forecasts ? Like products that are ordered only maybe bi weekly or whatever ? One week yes, one week no ?. We tried lightgbm, binary forecasting with lightgbm, prophet, IMAPA and ADIDA and our results are not great. only 60% accuracy in line with our moving average baseline.

New paper from the team that published the first ever paper applying Conformal Prediction to forecasting Sequential Predictive Conformal Inference for Time Series https://arxiv.org/pdf/2212.03463.pdf

Hi @Max (Nixtla) @fede is Nixtla using Spark? From what I recollect Fugue was mentioned I presume Nixtla looked at Spark at some point and decided to use / not to use it. Was wondering if Spark is suitable for building forecasting system?

Yes, spark is great. We use fugue as a backend that connects with spark, ray and dask. Is a great tool. @Han Wang and @Kevin Kho have been great allies in our mission. :)

🎉🎉 Announcing Conformal Prediction in NeuralProphet 🎉🎉 The team at NeuralProphet has been busy building exciting innovations in NeuralProphet and recently released day v0.5.0 that now included Conformal Prediction for time series forecasting 🔥🔥🔥🔥🔥🚀🚀🚀🚀🚀 My Medium article ‘Probabilistic Forecasting with Conformal Prediction and NeuralProphet’ showcases using NeuralProphet to predict energy consumption. https://valeman.medium.com/probabilistic-forecasting-with-conformal-prediction-and-neuralprophet-af9c87901d94

Course on Conformal Prediction! https://mindfulmodeler.substack.com/p/e-mail-course-on-conformal-prediction

Hi. How can I see the actual splits of cross validation ? I would like to have an expanding window cross validation and I want to make sure all the datapoints are ok. But I see that Stats_forecast.cross_validation only returns the forecasts (over the forecast horizon) of the models (?)

Hi! I want to know where the fit function is done in cross-validation? maybe if I want to change the optimization function for training. Here is my implementation part: " models = [ADIDA(), IMAPA(), CrostonClassic(), CrostonSBA(), CrostonOptimized(), TSB(alpha_d=0.5, alpha_p=0.5)] fcst = StatsForecast(df=df_to_process, models=models, freq='W-MON', n_jobs=-1) crossvalidation_df = fcst.cross_validation( df = df_to_process, h = 8, step_size = 1, n_windows = 7, sort_df=True ) "

Hello. I just stumbled upon Nixtla and think it is the coolest thing ever! I do have a quick question: Do any of the Nixtla models allow you to customize the loss function? In particular, I am thinking about time series predictions where there is an unbalanced costs, ie it costs much more to overshoot the true times series value than to undershoot it. For example, in the famous data-driven newsvendor problem it costs c1 to overstock newspapers but costs c2 to understock newspapers and generally c1 >> c2. How do the Nixtla models handle such problems? Has their performance been tested on time series predictions with 'imbalanced' loss functions?

Over the summer, I worked on implementing a paper called How Big Should Your Data Really Be? Data-Driven Newsvendor: Learning One Sample at a Time. This paper derives a statical model which makes time series predictions on data with unbalanced costs and proves near-optimal theoretical guarantees on the performance of this model. I'm interested in possibly pushing my model to the statsforecast library. Would this be something @Max (Nixtla) or the team at Nixtla would be interested in?

I also love how you use nbdev! I use it for my open source projects too.