Serve Batch Recommendations

Deploy Recommendations to Production ML Serving/Predicting Cluster

The notebook below does the following:

• Generates recommendations from input data using Spark ML ALS Matrix Factorization
• Transform the recommendations into multiple Servables (user-to-item recs, item-to-item sims)
• Deploy the Servables to a production ML serving/predicting cluster

http://<your-cloud-ip>:3123/#/notebook/2AUYFSKXN

Start Prediction Services

start-all-prediction-services.sh

Verify Prediction Service Startup

tail-prediction-service.sh

Verify User-to-Item Predictions

<your-cloud-ip>:9040/prediction/<userId>/<itemId>

Verify User-to-Item Recommendations

<your-cloud-ip>:9040/recommendations/<userId>/0/9

Verify Item-to-Item Similarities

<your-cloud-ip>:9040/similars/<itemId>/0/9

Monitor Hystrix Circuit Breaker Dashboard

http://<your-cloud-ip>:7979/hystrix-dashboard/

• Type the following into the text box: http://<your-cloud-ip>:8989/turbine.stream
• Click Add Stream
• Click Monitor Streams

• Re-run some of the prediction/recommendation/similar urls above
• Verify the Hystrix Dashboard is showing activity

Stop Prediction Services

stop-all-prediction-services.sh