PageRank on the DBLP dataset using Spark

Installation requirements:

1. Spark 2.4.x

2. Hadoop 3.x.x

3. Java 8

4. Scala 2.1x.x

5. Sbt version 2.1x.x

I installed hadoop and spark on windows and thus, did not need cloudera or hortonworks for executing my jars.

The algorithm is modified from the implementation at https://github.com/abbas-taher/pagerank-example-spark2.0-deep-dive/blob/master/SparkPageRank.scala

How to run:

1)Get the dblp.xml and dblp.dtd file from https://dblp.uni-trier.de/. Place both in same folder.

2)Go to the project folder and run the following: sbt clean compile

3)Method 1: Without jar

1) `sbt run <input-file-location> <output-file-location>` in terminal(cmd or in Intellij). 

Use absolute paths and '\\' instead of '\'(Windows). 

If running using Intellij, add the argument -Xmx6000m in Run-> Edit Configuration-> VM Options

This increases memory allocation to the VM.

Method 2: Using jar

1) Create the fat jar using
  ~sbt assembly

2) use
  `spark-submit --class prtest pagerank.jar <input-file-location> <output-file-location>`

4. The output folder will have files called part-xxxxx. Open as a text file. This is the required result.

Tests can be run using sbt clean compile test

Output Format:

The output generated will be of the form

(University of Paris-Sud, Orsay, France,0.8154981739701659)

(Elena,0.9850243302878131)

(John Bell,1.4621033282930214)

(University of Nice Sophia Antipolis, France,0.5678480111313514)

(Joseph Fourier University, Grenoble, France,0.8154981739701659)

(Elena Zheleva,1.3690036520596678)

(Acta Inf.,0.9850243302878131)