Before you use this demo, please make sure that you have Hadoop, Spark and Cassandra up and running. You can download both from the internet:
- Spark: spark.apache.org
- Hadoop: hadoop.apache.org
- Cassandra: cassandra.apache.org
- Akka: com.typesafe.akka
After you have Spark, Hadoop and cassandra installed run the following scripts on the Cassandra server using CQLSH in order to get the database created and filled with sample data:
create keyspace recommendations with replication = { 'class': 'SimpleStrategy', 'replication_factor': 1 };
create table recommendations.user_item_rating (
user_id int,
item_id int,
rating double,
primary key(user_id, item_id)
);
After you created the cassandra database, create a CSV file with the following content: user_id, item_id, rating where each is an integer value. Load this new file into Cassandra using the command below in CQLSH:
copy user_item_rating (user_id, item_id, rating) from 'user_item_ratings.csv' with HEADER=true;
I've created the user_item_rating.csv file which you can readily use, Don't forget to fireup your cassandra; Next start the application using the follow command:
sbt run
Finally use CURL to train the first version of the recommendation system model
curl -XPOST http://localhost:8080/train
The Rest API is created using Akka;
You can get recommendations by opening http://localhost:8080/recommendations/[id] in a browser.
This can be fine tuned to be consumed by a mobile Application