"use client";

// Load the Inkeep script
const inkeepScript = document.createElement("script");
inkeepScript.type = "module";
inkeepScript.src =
  "https://unpkg.com/@inkeep/widgets-embed@0.2.263/dist/embed.js";

document.body.appendChild(inkeepScript);

// find the Mintlify search container
const buttons = Array.from(document.getElementsByTagName("button"));
const searchButtonContainerEl = buttons.find((button) =>
  button.textContent.includes("Search or ask...")
);

const clonedSearchButtonContainerEl = searchButtonContainerEl.cloneNode(true);

// replace with dummy div, required to remove event listeners on it
searchButtonContainerEl.parentNode.replaceChild(
  clonedSearchButtonContainerEl,
  searchButtonContainerEl
);

// Once the script has loaded, load the Inkeep chat components
inkeepScript.addEventListener("load", function () {
  // Settings for the components
  const sharedConfig = {
    baseSettings: {
      apiKey: "c99c3754a78cf9f7d3d389ef9fa7c3d49cdeb18b95050d10",
      integrationId: "cltexfnd9000iy0ufy9oqo6h6",
      organizationId: "org_VKqv6JAjdI85T7aJ",
      primaryBrandColor: "#0c0c0c",
    },
    aiChatSettings: {
      chatSubjectName: "Nixtla",
      botAvatarSrcUrl:
        "https://storage.googleapis.com/organization-image-assets/nixtla-botAvatarSrcUrl-1709853122420.png",
      botAvatarDarkSrcUrl:
        "https://storage.googleapis.com/organization-image-assets/nixtla-botAvatarDarkSrcUrl-1709853121631.png",
      getHelpCallToActions: [
        {
          name: "Ask on Slack",
          url: "https://join.slack.com/t/nixtlaworkspace/shared_invite/zt-135dssye9-fWTzMpv2WBthq8NK0Yvu6A",
          icon: {
            builtIn: "FaSlack",
          },
        },
        {
          name: "View Repositories",
          url: "https://github.com/Nixtla",
          icon: {
            builtIn: "FaGithub",
          },
        },
      ],
      quickQuestions: [
        "How do I train my own model using mlforecast?",
        "How was TimeGPT trained and what's it best for?",
        "How do I make multivariate scoring for hierarchical forecasting?",
      ],
    },
  };

  // for syncing with dark mode
  const colorModeSettings = {
    observedElement: document.documentElement,
    isDarkModeCallback: (el) => {
      return el.classList.contains("dark");
    },
    colorModeAttribute: "class",
  };

  // add the chat button
  const chatButton = Inkeep().embed({
    componentType: "ChatButton",
    colorModeSync: colorModeSettings,
    properties: sharedConfig,
  });

  // insantiate Inkeep modal
  const searchButtonWithCustomTrigger = Inkeep({
    ...sharedConfig.baseSettings,
  }).embed({
    componentType: "CustomTrigger",
    colorModeSync: colorModeSettings,
    properties: {
      ...sharedConfig,
      isOpen: false,
      onClose: () => {
        searchButtonWithCustomTrigger.render({
          isOpen: false,
        });
      },
    },
  });

  // When the Mintlify search bar clone is clicked, open the Inkeep search modal
  clonedSearchButtonContainerEl.addEventListener("click", function () {
    searchButtonWithCustomTrigger.render({
      isOpen: true,
    });
  });
});


!(function (t) {
  if (window.oneloop) return;
  window.oneloop = {key: "cus_b57d16380d7240e4a47c9f4a6f0893c6"};
  var n = document.createElement("script");
  (n.async = !0),
    n.setAttribute(
      "src",
      "https://oneloop-website-script.s3.us-west-1.amazonaws.com/oneloop-main.js"
    ),
    (document.body || document.head).appendChild(n);
})();


evaluate

TimeGPT

StatsForecast

MLForecast

NeuralForecast

HierarchicalForecast

UtilsForecast

DatasetsForecast

CoreForecast

Nixtla

Evaluation

About TimeGPT

TimeGPT is a production ready, generative pretrained transformer for time series. It's capable of accurately predicting various domains such as retail, electricity, finance, and IoT with just a few lines of code 🚀.

TimeGPT Quickstart

TimeGPT Quickstart (Polars)

TimeGEN-1 is TimeGPT optimized for the Azure infrastructure. It is a production ready, generative pretrained transformer for time series. It's capable of accurately predicting various domains such as retail, electricity, finance, and IoT with just a few lines of code 🚀.

TimeGEN-1 Quickstart (Azure)

Setting up your API key

This section explains the data requirements for `TimeGPT`.

Data Requirements

These are some key concepts related to time series forecasting, designed to help you better understand and leverage the capabilities of TimeGPT.

Glossary

Quickstart

Add exogenous variables

Add holidays and special dates

Add categorical variables

Long-horizon forecasting

Multiple series forecasting

Fine-tuning

Finetuning with a custom loss function

Cross validation

Predictions intervals

Irregular timestamps

Add date features

Add confidence levels

The foundational models for time series by Nixtla can be deployed on your Azure subscription. This page explains how to easily get started with TimeGEN-1 deployed as an Azure AI endpoint. If you use the `nixtla` library, it should be a drop-in replacement where you only need to change the client parameters (endpoint URL, API key, model name).

AzureAI

Anomaly detection

Exogenous variables

Holidays and special dates

Categorical variables

SHAP Values for TimeGPT and TimeGEN

Fine-tuning with a specific loss function

Cross-validation

Historical forecast

Quantile forecasts

Prediction intervals

Bounded forecasts

Hierarchical forecasting

Missing Values

Improve Forecast Accuracy with TimeGPT

Computing at scale

Run TimeGPT distributedly on top of Spark

Spark

Dask

Forecasting web traffic

Learn how to use TimeGPT for financial time series forecasting

Bitcoin price prediction

Forecasting Energy Demand

What if? Forecasting price effects in retail

SDK Reference

Date Features

TimeGPT Excel Add-in (Beta)

StatsForecast offers a collection of popular univariate time series forecasting models optimized for high performance and scalability.

StatsForecast ⚡️

Install

Quick Start

Model training, evaluation and selection for multiple time series

End to End Walkthrough

End to End Walkthrough with Polars

In this notebook, we'll implement anomaly detection in time series data

Anomaly Detection

In this example, we'll implement conformal prediction

Conformal Prediction

In this example, we'll implement time series cross-validation to evaluate model's performance.

In this example we will show how to perform electricity load forecasting considering a model capable of handling multiple seasonalities (MSTL).

Electricity Load Forecast

In this example we will show how to perform electricity load forecasting on the ERCOT (Texas) market for detecting daily peaks.

Detect Demand Peaks

In this example, we'll forecast the volatility of the S&P 500 and several publicly traded companies using GARCH and ARCH models

Volatility forecasting (GARCH & ARCH)

In this notebook, we'll implement models for intermittent or sparse data

Intermittent or Sparse Data

MLFlow

In this example we will show how to forecast data with multiple seasonalities using an MSTL.

Multiple seasonalities

In this notebook, you will make forecasts for the M5 dataset choosing the best model for each time series using cross validation.

Statistical, Machine Learning and Neural Forecasting methods

In this example, we'll implement prediction intervals

Probabilistic Forecasting

How to do automatic forecasting using `AutoARIMA`, `AutoETS`, `AutoCES` and `AutoTheta`.

Automatic Time Series Forecasting

In this notebook, we'll incorporate exogenous regressors to a StatsForecast model.

Exogenous Regressors

Leverage StatsForecast models to create features

Generating features

Use any scikit-learn model for forecasting

Sklearn models

Migrating from R

Enabling caching for numba functions to reduce cold-starts

Numba caching

Run StatsForecast distributedly on top of Dask.

Run StatsForecast distributedly on top of Ray.

Run StatsForecast distributedly on top of Spark.

Amazon's AutoML vs open source statistical methods

Amazon Forecast vs StatsForecast

AutoARIMA Comparison (Prophet and pmdarima)

Forecasting at Scale using ETS and ray (M5)

This notebook was originally executed using DataBricks

StatsForecast ETS and Facebook Prophet on Spark (M5)

Step-by-step guide on using the `ADIDA Model` with `Statsforecast`.

ADIDA Model

Step-by-step guide on using the `ARCH Model` with `Statsforecast`.

ARCH Model

Step-by-step guide on using the `ARIMA Model` with `Statsforecast`.

ARIMA Model

Step-by-step guide on using the `AutoARIMA Model` with `Statsforecast`.

AutoARIMA Model

Step-by-step guide on using the `AutoCES Model` with `Statsforecast`.

AutoCES Model

Step-by-step guide on using the `AutoETS Model` with `Statsforecast`.

AutoETS Model

Step-by-step guide on using the `AutoRegressive Model` with `Statsforecast`.

AutoRegressive Model

Step-by-step guide on using the `AutoTheta Model` with `Statsforecast`.

AutoTheta Model

Step-by-step guide on using the `CrostonClassic Model` with `Statsforecast`.

CrostonClassic Model

Step-by-step guide on using the `CrostonOptimized Model` with `Statsforecast`.

CrostonOptimized Model

Step-by-step guide on using the `CrostonSBA Model` with `Statsforecast`.

CrostonSBA Model

Step-by-step guide on using the `DynamicOptimizedTheta Model` with `Statsforecast`.

Dynamic Optimized Theta Model

Step-by-step guide on using the `DynamicStandardTheta Model` with `Statsforecast`.

Dynamic Standard Theta Model

Step-by-step guide on using the `GARCH Model` with `Statsforecast`.

GARCH Model

Step-by-step guide on using the `Holt Model` with `Statsforecast`.

Holt Model

Step-by-step guide on using the `Holt Winters Model` with `Statsforecast`.

Holt Winters Model

Step-by-step guide on using the `IMAPA Model` with `Statsforecast`.

IMAPA Model

MFLES is a simple time series method based on gradient boosting time series decomposition.

MFLES

Step-by-step guide on using the `MSTL Model` with `Statsforecast`.

Multiple Seasonal Trend (MSTL)

Step-by-step guide on using the `OptimizedTheta Model` with `Statsforecast`.

Optimized Theta Model

Step-by-step guide on using the `SeasonalExponentialSmoothing Model` with `Statsforecast`.

Seasonal Exponential Smoothing Model

Step-by-step guide on using the `SeasonalExponentialSmoothingOptimized Model` with `Statsforecast`.

Seasonal Exponential Smoothing Optimized Model

Step-by-step guide on using the `SimpleExponentialSmoothingOptimized Model` with `Statsforecast`.

Simple Exponential Smoothing Optimized Model

Step-by-step guide on using the `SimpleExponentialSmoothing Model` with `Statsforecast`.

Simple Exponential Smoothing Model

Step-by-step guide on using the `Standard Theta Model` with `Statsforecast`.

Standard Theta Model

Step-by-step guide on using the `TSB Model` with `Statsforecast`.

TSB Model

Methods for Fit, Predict, Forecast (fast), Cross Validation and plotting

Core Methods

FugueBackend

Models currently supported by StatsForecast

Models

StatsForecast's Models

Feature engineering

How to Contribute

Contribute to Nixtla

Writing Documentation

Nixtla Documentation

Interpreting Issue Labels

Understanding Issue Labels

Submit an Issue

Submit an Issue 📢

mlforecast

Instructions to install the package from different sources.

Quick start (local)

Minimal example of distributed training with MLForecast

Quick start (distributed)

Detailed description of all the functionalities that MLForecast provides.

End to end walkthrough

Use exogenous regressors for training and predicting

Exogenous features

Lag transformations

Hyperparameter optimization

Leverage scikit-learn's composability to define pipelines as models

Using scikit-learn pipelines

Provide a column to pass through to the underlying models as sample weights

Sample weights

Probabilistic forecasting

Target transformations

Access and interpret the models after fitting

Analyzing the trained models

MLflow

Compute transformations on your exogenous features for MLForecast

Transforming exogenous features

Customize the training procedure for your models

	Type	Default	Details
df	AnyDFType		Forecasts to evaluate. Must have `id_col`, `time_col`, `target_col` and models’ predictions.
metrics	List		Functions with arguments `df`, `models`, `id_col`, `target_col` and optionally `train_df`.
models	Optional	None	Names of the models to evaluate. If `None` will use every column in the dataframe after removing id, time and target.
train_df	Optional	None	Training set. Used to evaluate metrics such as `mase`.
level	Optional	None	Prediction interval levels. Used to compute losses that rely on quantiles.
id_col	str	unique_id	Column that identifies each serie.
time_col	str	ds	Column that identifies each timestep, its values can be timestamps or integers.
target_col	str	y	Column that contains the target.
agg_fn	Optional	None	Statistic to compute on the scores by id to reduce them to a single number.
Returns	AnyDFType		Metrics with one row per (id, metric) combination and one column per model. If `agg_fn` is not `None`, there is only one row per metric.

	unique_id	metric	model0	model1
0	0	mae	0.158108	0.163246
1	1	mae	0.160109	0.143805
2	2	mae	0.159815	0.170510
3	3	mae	0.168537	0.161595
4	4	mae	0.170182	0.163329
…	…	…	…	…
175	5	scaled_crps	0.034202	0.035472
176	6	scaled_crps	0.034880	0.033610
177	7	scaled_crps	0.034337	0.034745
178	8	scaled_crps	0.033336	0.032459
179	9	scaled_crps	0.034766	0.035243

	metric	model0	model1
0	calibration_q0.025	0.000000	0.000000
1	calibration_q0.1	0.000000	0.000000
2	calibration_q0.9	0.833993	0.815833
3	calibration_q0.975	0.853991	0.836949
4	coverage_level80	0.833993	0.815833
5	coverage_level95	0.853991	0.836949
6	mae	0.161286	0.162281
7	mape	0.048894	0.049624
8	mase	0.966846	0.975354
9	mqloss	0.056904	0.056216
10	mse	0.048653	0.049198
11	quantile_loss_q0.025	0.019990	0.019474
12	quantile_loss_q0.1	0.067315	0.065781
13	quantile_loss_q0.9	0.095510	0.093841
14	quantile_loss_q0.975	0.044803	0.045767
15	rmse	0.220357	0.221543
16	scaled_crps	0.035003	0.034576
17	smape	0.024475	0.024902

API Reference

​evaluate

evaluate