&alex-jg3 [DOC] anomaly and changepoint detection notebook from ODSC …

…2024 (#7284) This PR adds the notebook of @Alex-JG3 from ODSC 2024 on anomaly and changepoint detection with `sktime` and `skchange` to the examples. This is without the metrics section, as I expect that the metrics end state will be different (i.e., following `BaseMetric` design). Also makes the following changes: * adds `skchange` to the dependencies installed for binder notebooks so the code runs * adds the data used to the data folder, without loaders - only a few dozen kb, and without loaders for now * changes some of the headers in other notebooks, so the auto-generated thumbnail and link names are sensible
sktime · Oct 17, 2024 · c392695 · c392695
1 parent be4deb4
commit c392695
Show file tree

Hide file tree

Showing 9 changed files with 9,691 additions and 16 deletions.
diff --git a/docs/source/examples.rst b/docs/source/examples.rst
@@ -68,6 +68,7 @@ Changepoint, Anomaly or Outlier Detection
 .. nbgallery::
     :glob:
 
+    examples/07_detection_anomaly_changepoints.ipynb
     examples/annotation/*
 
 Time series Distances, Kernels, Alignment

diff --git a/examples/03_transformers.ipynb b/examples/03_transformers.ipynb
@@ -1,5 +1,12 @@
 {
  "cells": [
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "# Transformers in `sktime` <a class=\"anchor\" id=\"chapter3\"></a>"
+   ]
+  },
   {
    "attachments": {},
    "cell_type": "markdown",
@@ -64,14 +71,6 @@
     "---"
    ]
   },
-  {
-   "attachments": {},
-   "cell_type": "markdown",
-   "metadata": {},
-   "source": [
-    "# 3. Transformers in `sktime` <a class=\"anchor\" id=\"chapter3\"></a>"
-   ]
-  },
   {
    "attachments": {},
    "cell_type": "markdown",
@@ -5721,7 +5720,7 @@
    "name": "python",
    "nbconvert_exporter": "python",
    "pygments_lexer": "ipython3",
-   "version": "3.11.5"
+   "version": "3.11.9"
   },
   "latex_envs": {
    "LaTeX_envs_menu_present": true,

diff --git a/examples/03b_forecasting_transformers_pipelines_tuning.ipynb b/examples/03b_forecasting_transformers_pipelines_tuning.ipynb
@@ -4,7 +4,7 @@
    "cell_type": "markdown",
    "metadata": {},
    "source": [
-    "# 4 Forecasting Pipelines, Tuning, and AutoML\n",
+    "# Forecasting Pipelines, Tuning, and AutoML\n",
     "This notebook is about pipelining and tuning (grid search) for time series forecasting with `sktime`"
    ]
   },
@@ -25,7 +25,7 @@
    "cell_type": "markdown",
    "metadata": {},
    "source": [
-    "## 4.1 Using exogeneous data `X` in forecasting\n",
+    "## 1. Using exogeneous data `X` in forecasting\n",
     "The exogenous data `X` is a `pd.DataFrame` with the same index as `y`. It can be given optionally to a forecaster in the `fit` and `predict` methods. If `X` is given in `fit`, it also has to be given in `predict`, this means the future values of `X` have to be available at the time of prediction. "
    ]
   },
@@ -213,7 +213,7 @@
    "cell_type": "markdown",
    "metadata": {},
    "source": [
-    "## 4.2 Simple Forecasting Pipelines\n",
+    "## 2. Simple Forecasting Pipelines\n",
     "\n",
     "We hae seen: pipelines combine multiple estimators into one.\n",
     "\n",
@@ -432,7 +432,7 @@
    "cell_type": "markdown",
    "metadata": {},
    "source": [
-    "## 4.3 Tuning"
+    "## 3. Tuning"
    ]
   },
   {
@@ -1587,7 +1587,7 @@
    "cell_type": "markdown",
    "metadata": {},
    "source": [
-    "## 4.4 Putting it all together: AutoML-like pipeline and tuning\n",
+    "## 4. Putting it all together: AutoML-like pipeline and tuning\n",
     "\n",
     "Taking all incredients from above examples, we can build a forecaster that comes close to what is usually called [AutoML](https://en.wikipedia.org/wiki/Automated_machine_learning).\n",
     "With AutoML we aim to automate as many steps of an ML model creation as possible. The main compositions from `sktime` that we can use for this are:\n",
@@ -1788,7 +1788,7 @@
    "cell_type": "markdown",
    "metadata": {},
    "source": [
-    "## 4.5 Model backtesting\n",
+    "## 5. Model backtesting\n",
     "\n",
     "After fitting a model, we can evaluate the model error in the past similar to a cross-validation. For this we can use the `evaluate` function. This approach is often also called backtesting as we want to test the forecasters performance in the past."
    ]
@@ -2365,7 +2365,7 @@
    "source": [
     "---\n",
     "\n",
-    "### Credits: notebook 4 - pipelines\n",
+    "### Credits: notebook - pipelines\n",
     "\n",
     "notebook creation: aiwalter\n",
     "\n",

diff --git a/examples/07_detection_anomaly_changepoints.ipynb b/examples/07_detection_anomaly_changepoints.ipynb
diff --git a/examples/img/annotation_tree.png b/examples/img/annotation_tree.png
diff --git a/pyproject.toml b/pyproject.toml
@@ -270,6 +270,7 @@ tests = [
 binder = [
   "jupyter",
   "pandas<2.0.0",
+  "skchange<0.9.0",
 ]
 cython_extras = [
   "mrseql < 0.0.3",