Jupyter With PredictionIO

Overview

Using Jupyter based docker, you can use Jupyter Notebook with PredictionIO environment. It helps you with your exploratory data analysis (EDA).

Run Jupyter Notebook

First of all, start Jupyter container with PredictionIO environment:

docker-compose -f docker-compose.jupyter.yml \
  -f pgsql/docker-compose.base.yml \
  -f pgsql/docker-compose.meta.yml \
  -f pgsql/docker-compose.event.yml \
  -f pgsql/docker-compose.model.yml \
  up

Open http://127.0.0.1:8888/ and then open a new terminal in Jupyter from New pulldown button.

Getting Started With Scala Based Template

Download Template

Clone a template using Git:

cd templates/
git clone https://github.com/apache/predictionio-template-recommender.git
cd predictionio-template-recommender/

Replace a name with MyApp1.

sed -i "s/INVALID_APP_NAME/MyApp1/" engine.json

Register New Application

Using pio command, register a new application as MyApp1.

pio app new MyApp1

This command prints an access key as below.

[INFO] [Pio$] Access Key: bbe8xRHN1j3Sa8WeAT8TSxt5op3lUqhvXmKY1gLRjg70K-DUhHIJJ0-UzgKumxGm

Set it to an environment variable ACCESS_KEY.

ACCESS_KEY=bbe8xRHN1j3Sa8WeAT8TSxt5op3lUqhvXmKY1gLRjg70K-DUhHIJJ0-UzgKumxGm

Import Training Data

Download trainging data and import them to PredictionIO Event server.

curl https://raw.githubusercontent.com/apache/spark/master/data/mllib/sample_movielens_data.txt --create-dirs -o data/sample_movielens_data.txt
python data/import_eventserver.py --access_key $ACCESS_KEY

Build Template

Build your template by the following command:

pio build --verbose

Create Model

To create a model, run:

pio train

Getting Started With Python Based Template

Download Template

Clone a template using Git:

cd templates/
git clone https://github.com/jpioug/predictionio-template-iris.git
predictionio-template-iris/

Register New Application

Using pio command, register a new application as IrisApp.

pio app new --access-key IRIS_TOKEN IrisApp

Import Training Data

Download trainging data and import them to PredictionIO Event server.

python data/import_eventserver.py

Build Template

Build your template by the following command:

pio build --verbose

EDA

To do data analysis, open templates/predictionio-template-iris/eda.ipynb on Jupyter.

Create Model

You need to clear the following environment variables in the terminal before executing pio train.

unset PYSPARK_PYTHON
unset PYSPARK_DRIVER_PYTHON
unset PYSPARK_DRIVER_PYTHON_OPTS

To create a model, run:

pio train --main-py-file train.py