Partial Dependence Plot


Overview

Partial Dependence Plot (PDP) is one of the effective methods when interpreting why a created model makes the predictions it does, and it helps to understand how arbitrary explanatory variables are affecting the inference for a trained model.
When using PDP, the condition is that the arbitrary explanatory variables are not correlated with any other explanatory variables. Note that if there is a correlation, it cannot be displayed correctly.
In addition, it is common to target one or two explanatory variables for the target variable.
The target explanatory variables are selected from the most important features to be checked.

Partial dependency plot in VARISTA

Since VARISTA automatically generates PDPs, you can check them from the resulting model from the feature ' Correlation ' Partial Dependence.
Parcial Dependence Plot - VARISTA AI Machine Learning Knowledge

This example shows the PDP of SalePrice and 1stFlrSF.
We can see that the SalePrice has an upward trend as the value of 1stFlrSF increases.


VARISTAは機械学習モデルの開発、管理をノーコードで効率的に行うことができる新しいプラットフォームです。
データをお持ちでしたらすぐに始められますので、是非ともお試しください。

or

Made with
by VARISTA Team.
© COLLESTA, Inc. 2021. All rights reserved.