woonsan / unomi-predictionio-plugin

This repository contains a prototype integration of Apache Unomi with Apache PredictionIO

Apache Unomi - Apache PredictionIO Integration example

Requirements

• Apache PredictionIO 0.13.0+
• PredictionIO Lead Scoring Template
• Apache Unomi 1.3.0+

Installation

Install & Start PredictionIO (With Docker):

git clone https://github.com/apache/predictionio.git
cd predictionio
cd docker
docker build -t predictionio/pio pio
docker-compose -f docker-compose.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

Install & Start PredictionIO (From Source):

wget https://archive.apache.org/dist/predictionio/0.14.0/apache-predictionio-0.14.0.tar.gz
mkdir apache-predictionio-0.14.0
cd apache-prediction-0.14.0
tar zxvf ../apache-predictionio-0.14.0.tar.gz
./make-distribution.sh
tar zxvf PredictionIO-0.14.0.tar.gz
cd PredictionIO-0.14.0
mkdir vendors
wget https://archive.apache.org/dist/spark/spark-2.1.3/spark-2.1.3-bin-hadoop2.7.tgz
tar zxvfC spark-2.1.3-bin-hadoop2.7.tgz vendors
cd lib 
wget https://jdbc.postgresql.org/download/postgresql-42.2.8.jar
cd ..
edit conf/pio-env.sh to point to proper version of spark and postgres driver. Should look like this:
SPARK_HOME=$PIO_HOME/vendors/spark-2.1.3-bin-hadoop2.7
POSTGRES_JDBC_DRIVER=$PIO_HOME/lib/postgresql-42.2.8.jar
brew install postgresql
createdb pio
pg_ctl -D /usr/local/var/postgres -l /usr/local/var/postgres/server.log start
psql -c "create user pio with password 'pio'"
bin/pio eventserver &
bin/pio status

Build, Train & deploy Lead Scoring PredictionIO template (with Docker):

git clone https://github.com/sergehuber/template-scala-parallel-leadscoring.git MyLeadScoring
cd MyLeadScoring
git checkout unomi-integration
export PATH=**REPLACE_WITH_PATH_TO_PREDICTIONIO_SRC**/docker/bin/:$PATH
pio-docker app new MyLeadScoring    
pio-docker build --verbose
pio-docker app list
ACCESS_KEY=cdelLgZqZxj7CI_2hDM_vy-Q3fhLDxlTQKao_UHe9DgcFLkSVm9Yfq_3ve8BTgzl
pip install predictionio
python data/import_eventserver.py --access_key $ACCESS_KEY
pio-docker train
pio-docker deploy

Build, Train & deploy Lead Scoring PredictionIO template (with Source):

git clone https://github.com/sergehuber/template-scala-parallel-leadscoring.git MyLeadScoring
cd MyLeadScoring
git checkout unomi-integration
export PATH=**REPLACE_WITH_PATH_TO_PREDICTIONIO_SRC**/bin/:$PATH
pio app new MyLeadScoring    
pio build --verbose
pio app list
ACCESS_KEY=cdelLgZqZxj7CI_2hDM_vy-Q3fhLDxlTQKao_UHe9DgcFLkSVm9Yfq_3ve8BTgzl
pip install predictionio
python data/import_eventserver.py --access_key $ACCESS_KEY
pio train
pio deploy

Download, configure and start ElasticSearch 5.6.3

curl -O https://artifacts.elastic.co/downloads/elasticsearch/elasticsearch-5.6.3.tar.gz
tar zxvf elasticsearch-5.6.3.tar.gz
cd elasticsearch-5.6.3
change cluster.name to cluster.name: contextElasticSearch-COMPUTERNAME in config/elasticsearch.yml file
bin/elasticsearch

Compile & start Apache Unomi:

git clone https://github.com/apache/unomi.git
cd unomi
./buildAndRunNoTests.sh
unomi:start (in Karaf shell) 

Build & Deploy this plugin to Apache Unomi (in Karaf shell):

make clean install
feature:repo-add mvn:org.apache.unomi/unomi-predictionio-kar/${project.version}/xml/features
feature:install unomi-predictionio-kar

Testing

You can then open a browser at the following URL to test the integration:

http://localhost:8181/predictionio

First click on 'Page 1' on the top right back, then click Page 2 and go back to Page 1, nothing changes, click Page 3 and go back to Page 1 and you should see the banner change.

Also check the Unomi logs in the Karaf shell using : log:tail to see the probability value.