random_forest_regression

Overview

Random forest is a Supervised Learning algorithm which uses ensemble learning method for classification and regression.This example shows you how to use OCI Data Flow to build a Random Forest Regression Model to predict mileage (mpg) for cars.

Prerequisites

Before you begin:

• Ensure your tenant is configured according to the instructions to setup admin
• Know your object store namespace.
• (Optional, strongly recommended): Install Spark to test your code locally before deploying.

Load Required Data

Upload a sample CSV file to OCI object store.

Application Setup

Customize random_forest_regression.py with:

1. Set INPUT_PATH to the OCI path of your CSV data.
2. Customize random_forest_regression.py with the OCI path to your CSV data. The format is oci://<bucket>@<namespace>/path 2a. Don't know what your namespace is? Run oci os ns get 2b. Don't have the OCI CLI installed? See to install it.
3. Customize random_forest_regression.py with the OCI path where you would like to save output data.
4. Recommended: run the sample locally to test it. Refer Develop Oracle Cloud Infrastructure Data Flow Applications Locally, Deploy to The Cloud
5. Upload random_forest_regression.py to an object store bucket.
6. Create a Python Data Flow application pointing to random_forest_regression.py

Testing Locally

Test the Application Locally (recommended):You can test the application locally using :

python3 random_forest_regression.py

If it works you'll see output like :

 RMSE=3.888566643366089 r2=0.7532412722764463

Deploy and Run the Application

• Copy random_forest_regression to object store.
• Create a Data Flow Python application. Refer Create PySpark App for more information.
• Run the application.

Run the Application using OCI Cloud Shell or OCI CLI

Create a bucket. Alternatively you can re-use an existing bucket.

oci os object put --bucket-name <bucket> --file random_forest_regression
oci data-flow application create \
    --compartment-id <compartment_ocid> \
    --display-name "Random Forest Regression" \
    --driver-shape VM.Standard2.1 \
    --executor-shape VM.Standard2.1 \
    --num-executors 1 \
    --spark-version 2.4.4 \
    --file-uri oci://<bucket>@<namespace>/loadadw.py \
    --archive-uri oci://<bucket>@<namespace>/archive.zip \
    --language Python
oci data-flow run create \
    --application-id <application_ocid> \
    --compartment-id <compartment_ocid> \
    --application-id <application_ocid> \
    --display-name "Random Forest Regression"