From bd1fc113b82bbc8dd19c5225375ab0c734180933 Mon Sep 17 00:00:00 2001
From: Uwe Schmitt <uwe.schmitt@id.ethz.ch>
Date: Thu, 4 Mar 2021 12:54:30 +0100
Subject: [PATCH] mention special metrics for GLMs
---
07_regression.ipynb | 5 ++++-
1 file changed, 4 insertions(+), 1 deletion(-)
diff --git a/07_regression.ipynb b/07_regression.ipynb
index 092f4f3..89119f0 100644
--- a/07_regression.ipynb
+++ b/07_regression.ipynb
@@ -910,7 +910,10 @@
"\n",
"- `sklearn.tree.DecisionTreeRegressor` expands the concept of decision trees to regression [is documented here](https://scikit-learn.org/stable/modules/tree.html#regression).\n",
"\n",
- "- `sklearn.linear_model.TweedieRegressor`, `sklearn.linear_model.PoissonRegressor`, `sklearn.linear_model.GammaRegressor` offer so-called *Generalized Linear Models* (**GLM**) which might be of interest if your target values are bounded or discrete or if the error distribution is skewed. This is usually the case when your target values are event-based discrete counts/frequencies, or continuous amounts, durations, costs/prices, or rates/probabilities. The [scikit-learn GLM tutorial](https://scikit-learn.org/stable/modules/linear_model.html#generalized-linear-regression) provides a formal insight as well as tips for choosing GLM with some use case examples. Beyond that the [wikipedia article about generalized linear models](https://en.wikipedia.org/wiki/Generalized_linear_model#Intuition) gives a nice intuition and [this discussion](https://stats.stackexchange.com/questions/190763/how-to-decide-which-glm-family-to-use) provides a guide which GLM should be used when."
+ "- `sklearn.linear_model.TweedieRegressor`, `sklearn.linear_model.PoissonRegressor`, `sklearn.linear_model.GammaRegressor` offer so-called *Generalized Linear Models* (**GLM**) which might be of interest if your target values are bounded or discrete or if the error distribution is skewed. \n",
+ " - This is usually the case when your target values are event-based discrete counts/frequencies, or continuous amounts, durations, costs/prices, or rates/probabilities. The [scikit-learn GLM tutorial](https://scikit-learn.org/stable/modules/linear_model.html#generalized-linear-regression) provides a formal insight as well as tips for choosing GLM with some use case examples. \n",
+ " - Beyond that the [wikipedia article about generalized linear models](https://en.wikipedia.org/wiki/Generalized_linear_model#Intuition) gives a nice intuition and [this discussion](https://stats.stackexchange.com/questions/190763/how-to-decide-which-glm-family-to-use) provides a guide which GLM should be used when.\n",
+ " - For assessing and hyperparameter-optimization of such General Linear Models you should also use [suitable metrics from scikit-learn](https://scikit-learn.org/stable/modules/model_evaluation.html#mean-poisson-gamma-and-tweedie-deviances)."
]
},
{
--
GitLab