Text Classification using Apache Spark

The goal of that project was the implementation of a multi-layer perceptron classifier, using Spark and HDFS, in order, a huge collection of complaints about consumer financial products and services, to be categorized (Customer Complaint Database).

File Format

The dataset includes complains from customers from 2011 untill today and it is a comma-delimited CSV file with the following format:

0 <- date %Y-%m-%d
1 <- category
2 <- comment

Usage

Given that Spark and HDFS are properly installed and working on our system:

  • Upload data file in HDFS
hadoop fs -put ./customer_complaints.csv hdfs://master:9000/customer_complaints.csv
  • Submit job in a Spark environment
spark-submit ml.py

Libraries used

nltk, pyspark