Skip to content
This repository has been archived by the owner on Oct 10, 2020. It is now read-only.

data explore using machine learning & deep learning

License

Notifications You must be signed in to change notification settings

konabuta/DataExplore-Workshop

Repository files navigation

DataExplore-Workshop


Dllab Engineers Days 2019 Handson Day1

自動機械学習 AutoML & 要因探索 ハンズオン

機械学習を使って色んな問題を解きたいけど、人材やスキル不足が課題でなかなか前に進まない…そんな状況から脱出しませんか? Azureは機械学習の民主化を実現するプラットフォームを提供しています。本ハンズオンでは、マウス操作による簡単な要因探索、また自動機械学習を利用して高速にプロトタイプを構築するプロセスをご体感頂きます。今回は Azure Machine Learning service の自動機械学習 (Automated Machine Learning) や PowerBI の要因探索機能 (Key Influencers) を中心に利用します。Azure Machine Learning service によるモデルの解釈方法についても言及します。

Agenda

  • 本ワークショップの開発環境の構築作業
  • 一般的な自動機械学習やモデル解釈のアプローチ方法
  • 関連する Microsoft のサービス
  • ハンズオンのコンテンツ

Sample Code

下記の表には、本リポジトリに含まれるサンプルコードのリストです。Environment 列のリンクからアクセスできます。

Algorithm Environment Interpretability Type Description
Decision Tree Azure ML service Python SDK / InterpretML Interpretable Decision Tree (決定木) を用いたモデル開発のサンプルコード
Linear Regression Excel / Azure ML service Python SDK / InterpretML Interpretable Linear Regression (線形回帰) を用いたモデル開発のサンプルコード
Decision Tree & Linear Regressionn Power BI - Key Influencers Interpretable KPI 要因探索ビジュアル機能
AutoML + Model Interpretability Azure ML service Python SDK (Automated ML + Interpretabiliy SDK) Model-Agnostic 自動機械学習 + モデル解釈統合フレームワーク
Microsoft InterpretML Python - Classification Interpretable Microsoft Interpret ML によるモデル開発のサンプルコード
Global Surrogate n/a Model-Agnostic Global Surrogate を用いたモデル解釈
Permutation Feature Importance n/a Model-Agnostic PFI を用いたモデル解釈
LIME external site Model-Agnostic LIME によるモデル解釈
SHAP external site Model-Agnostic SHAP によるモデル解釈

Known Issues

うまくいかない時は、こちらのドキュメント を参照してください。

参考

Releases

No releases published

Packages

No packages published