maxmc

7 ms·

1.
by maxmc 3y ago
Hi, Max from Nixtla here. We are surprised that this has gained so much attention and are excited about both the positive and critical responses. Some important clarifications: The primary goal of this first version of the paper is to prese
2.
by maxmc 3y ago
We love ARIMAs. That is why we put so much effort into creating fast and scalable Arimas and AutoArima in Python [1]. Regarding your valid concern. There are several reasons for the high computational costs. First, ARIMA and other "sta
3.
Open Source FinOps GPT: Using GPT-4, TimeGPT and Vantage(github.com)
8 points by maxmc 3y ago|0 comments
4.
Statistical methods outperform BigQuery's Auto ML(github.com)
4 points by maxmc 4y ago|0 comments
5.
Statistical methods outperform Amazon’s ML Forecast(github.com)
3 points by maxmc 4y ago|1 comments
6.
by maxmc 4y ago
Using 30,490 series of daily sales at Walmart (M5) we show that, Amazon Forecast is 60% less accurate and 669 times more expensive than running an open-source alternative in a simple cloud server.
7.
by maxmc 4y ago
A lot of M3 datasets we use are high-frequency, with large seasonal inputs. Considering Gaussian Processes (GP) complexity is O(N^3), a careful study of their performance would be challenging. Also... I'm not aware of any efficient GP
8.
by maxmc 4y ago
Bonferroni's correction on hold-out data is an excellent suggestion. To adapt it into time series forecasting, one could perform temporal cross-validation with rolling windows and follow the performance's variance through time. Un
9.
by maxmc 4y ago
Thanks for the comment! In Machine Learning literature, the variance of accuracy measurements originates from different network parameters initialization. Since the deep learning ensembles already use aggregate computation in the hundreds o
10.
by maxmc 4y ago
Comparison of several Deep Learning models and ensembles to classical statistical models for the 3,003 series of the M3 competition.
11.
Statistical vs. Deep Learning forecasting methods(github.com)
222 points by maxmc 4y ago|81 comments
12.
Simple Seasonal Naive: faster and more accurate than Prophet and NeuralProphet(github.com)
1 points by maxmc 4y ago|1 comments
13.
by maxmc 4y ago
Imagine you want to forecast the next day of electricity load. We ran the test with PJM data. Results: SeasonalNaive is 20% more accurate than NeuralProphet and 366 times faster. StatsForecast's MSTL is 24% more accurate than NeuralPr
14.
Show HN: Probabilistic hierarchical forecasting with statistical methods(github.com)
2 points by maxmc 4y ago|0 comments
15.
Probabilistic hierarchical reconciliation for time series(github.com)
3 points by maxmc 4y ago|0 comments
16.
Trends and Opportunities in Time Series(gradientflow.com)
1 points by maxmc 4y ago|0 comments
17.
Complex Exponential Smoothing (CES) is now available in Pyhton(github.com)
2 points by maxmc 4y ago|0 comments
18.
Python implementations of time series forecasting and anomaly detection(robjhyndman.com)
2 points by maxmc 4y ago|0 comments
19.
by maxmc 4y ago
In the tutorial, we show how to Forecast 1M Time Series in 15 Minutes with Spark, Fugue, and Nixtla’s Statsforecast.
20.
Show HN: Scalable Time Series Modeling with open-source projects(nixtla.github.io)
5 points by maxmc 4y ago|1 comments
21.
by maxmc 4y ago
Figure 7 of the mentioned paper evaluates FB-Prophet in a extremely convenient environment of long horizon h in {30,60,90,120,150,180}. It is known that ETS and ARIMA models concatenate errors and degrade in performance with longer forecast
22.
by maxmc 4y ago
This comparison was not intended as an academic paper. However, we are confident that predicting and evaluating performance over more than 55k series from the M international forecasting competitions (standard benchmarks in the field) is a
23.
by maxmc 4y ago
Be surprised. Here is the original Neuralprophet paper: https://arxiv.org/pdf/2111.15397.pdf It only compares itself to Prophet.
24.
by maxmc 4y ago
Optimized using numba. 400% faster than statsmodels and 1.5x faster than R, with improved accuracy and robustness.
25.
Exponential Smoothing (ETS) for Python(github.com)
1 points by maxmc 4y ago|1 comments
26.
Show HN: Fast and robust Auto Arima in Python(github.com)
2 points by maxmc 5y ago|0 comments