Repository navigation

History

Name		Name	Last commit message	Last commit date
parent directory ..
AI_Kit_XGBoost_Predictive_Modeling.complete		AI_Kit_XGBoost_Predictive_Modeling.complete
License.txt		License.txt
Makefile		Makefile
README.md		README.md
sample.json		sample.json
third-party-programs.txt		third-party-programs.txt

README.md

Predictive Modeling with XGBoost* and the AI Tools

The purpose of this repository is to be the central aggregation, curation, and distribution point for Juypter Notebooks that are developed in support of the AI Tools. These initial hands-on exercises introduce you to predictive modeling using decision trees, bagging, and XGBoost.

The Jupyter Notebooks for the exercises are in the AI_Kit_XGBoost_Predictive_Modeling folder, and the answers to these exercises in the AI_Kit_XGBoost_Predictive_Modeling.complete folder.

Prerequisites

Optimized for	Description
OS	Ubuntu* 20.04 (or newer) Windows Subsystem for Linux (WSL)
Software	Intel® oneAPI Base Toolkit (Base Kit) AI Tools

The Jupyter Notebooks are tested for and can be run on the Intel® Devcloud for oneAPI.

Jupyter Notebook Directories and Descriptions

The referenced folders and Notebooks are in the AI_Kit_XGBoost_Predictive_Modeling folder. The AI_Kit_XGBoost_Predictive_Modeling.complete folder has the same structure.

Notebook Directory and Name	Notebook Focus
00_Local_Setup\Local_Setup.ipynb	- How to setup the environment for running on a local machine - Anaconda setup - Intel® Distribution for Python* programming language - AI Tools - Intel data science workstation kits
01_Decision_Trees\Decision_Trees.ipynb	- Recognize decision trees and how to use them for classification problems - Recognize how to identify the best split and the factors for splitting. - Explain strengths and weaknesses of decision trees - Explain how regression trees help with classifying continuous values - Apply Intel® Extension for Scikit-learn* to leverage underlying compute capabilities of hardware
02_Bagging\Bagging_RF.ipynb	- Determine if stratefiedshuffle split is the best approach - Recognize how to identify the optimal number of trees - Understand the resulting plot of out-of-band errors - Explore Random Forest vs Extra Random Trees and determine which one worked better - Apply Intel® Extension for Scikit-learn* to leverage underlying compute capabilities of hardware
03_XGBoost\XGBoost.ipynb	- Use XGBoost with the AI Tools - Take advantage of Intel® Extension for Scikit-learn* by enabling them with XGBoost - Use Cross Validation technique to find better XGBoost Hyperparameters - Use a learning curve to estimate the ideal number of trees - Improve performance by implementing early stopping
04_oneDal\XGBoost-oneDal.ipynb	- Utilize XGBoost with the AI Tools - Take advantage of Intel® Extension for Scikit-learn* by enabling them with XGBoost - Use Intel® oneAPI Data Analytics Library (oneDAL) to enhance prediction performance

Run the Jupyter Notebooks Locally (on Linux* or WSL)

Update the package manager on your system.
```
sudo apt update && sudo apt upgrade -y
```
After the update, reboot your system.
```
sudo reboot
```
Download and install Intel® oneAPI Base Toolkit (Base Kit) and AI Tools from the Intel® oneAPI Toolkits page.
After you complete the installation, refresh the new environment variables.
```
source .bashrc
```
Initialize the oneAPI environment enter.
```
source /opt/intel/oneapi/setvars.sh
```
Install JupyterLab*. (In this case, we are cloning our base environment so that we can always get back to a clean start.)
```
conda create --clone base --name jupyter
```
Switch to the newly created environment.
```
conda activate jupyter
```
Install Jupyterlab.
```
conda install -c conda-forge jupyterlab
```
Clone the oneAPI-samples GitHub repository.
Note: If Git is not installed, install it now.
```
sudo apt install git
```
```
git clone https://github.com/oneapi-src/oneAPI-samples.git
```
From a terminal, start JupyterLab.
```
jupyter lab
```
Make note of the address printed in the terminal, and paste the address into your browser address bar.

Once Jupyterlab opens, navigate to the following directory.

~/oneAPI-samples/AI-and-Analytics/Jupyter/Predictive_Modeling_Training

From the navigation panel, navigate through the directory structure and select a Notebook to run. (The notebooks have a .ipynb extension.)

Run the Jupyter Notebooks on Intel® Devcloud (Optional)

Use these general steps to access the notebooks on the Intel® Devcloud for oneAPI.

Note: For more information on using Intel® DevCloud, see the Intel® oneAPI Get Started page.

If you do not already have an account, request an Intel® DevCloud account at Create an Intel® DevCloud Account.
Once you get your credentials, open a terminal on a Linux* system
Log in to the Intel® DevCloud.
```
ssh devcloud
```
Note: Alternatively, you can use the Intel JupyterLab to connect with your account credentials.
From a terminal, enter the following command to obtain the latest series of Jupyter Notebooks into your Intel® DevCloud account:
```
/data/oneapi_workshop/get_jupyter_notebooks.sh
```
Note: If you are setting up your account for the first time this script will run automatically.
From the navigation panel, navigate through the directory structure and select a Notebook to run. (The notebooks have a .ipynb extension.)

License

Code samples are licensed under the MIT license. See License.txt for details.

Third-party program Licenses can be found here: third-party-programs.txt.

oneAPI-samples/Training/Predictive_Modeling_Training at development · oneapi-src/oneAPI-samples · GitHub

Repository navigation

Expand file tree

Breadcrumbs

Predictive_Modeling_Training

Directory actions

More options