Crystal Discoball

Authors: Biswajeeth Dey, Rashmi Dwaraka

This project aims to predict the number of downloads for a given song primarily through the attributes provided by the million song dataset. More details of the project can be found at http://janvitek.org/pdpmr/f17/project.html

Pre-requisite packages:

1. Java (v1.8+)
2. Spark (v2.2.0 compiled with Scala 2.11.X and hadoop-2.7)
3. Scala (v2.11.X)
4. .Renviron should exist to generate the report in R

Build the project by typing command make build. Run the project locally by running make run

Pre-requisites:

1. Update the variables SCALA_HOME and SPARK_HOME as per your configuration in Makefile.
2. The model can be loaded in memory by running the make run command. It assumes that you have a partition /mnt/pdpmr of size 100 MB or uses the /tmp as the scratch directory for loading the model.

Project structure

1. The project has both java and scala files. sources/bishwajeet_rashmi/src/main/java contains the java files. sources/bishwajeet_rashmi/src/main/scala contains the scala files.
2. The RandomForest model is built using RFEngine.
3. The loading/deployment of the model is part of the Model class.