Architecture
Prerequisites
To implement this solution, you need the following resources:
- Amazon S3 to stage a ‘ribbon’ call detail record sample in a CSV format.
- AWS Glue running an ETL job in PySpark.
- AWS Glue crawlers to discover the schema of the tables and update the AWS Glue Data Catalog.
- Amazon Athena to query the Amazon QuickSight dataset.
- Amazon QuickSight to build visualizations and perform anomaly detection using ML Insights.
The following is a diagram of a sample architecture for a fraudulent
call-detection solution.
It makes use of a PySpark script to prepare the data and transform it into
Parquet;
An AWS Glue Crawler is the used to build the AWS Glue Data Catalog;
Amazon Athena is used to run SQL queries on the result data set and QuickSight
can be plugged in to create analysis and dashboards.
