Development of a business intelligence solution for fraud detection in CEP logistics
Project duration: 2 years
Brief description
In cooperation with the customer, PTA develops a Business Intelligence Solution (BI Solution) in the context of the existing data warehouse with the aim of supporting the specialist department in the investigation and prevention of fraud. PTA advises and supports the customer in the design and implementation of the database model as well as in the realization of the ELT/ETL processes (Extract-Load-Transform) and reporting. To this end, PTA designs the logical data model together with the customer and implements the physical data model based on this, with a focus on optimizing query speed. This is crucial in order to access the detailed data in a user-friendly and clear way via drill-down. For regular data management, PTA also supports the customer in structuring and designing the data pipelines (ELT processes).
Supplement
The conception of a logical data model and the implementation of the physical model in a cloud DWH based on Snowflake is designed to decouple the business logic for recognizing anomalies in shipment processing as far as possible. This is achieved by only calculating predefined metrics at the relevant process stages. By setting threshold values in the Tableau reports, the user is given the opportunity to individually configure the identification of anomalies. The development of ETL/ELT processes and data pipelines for regular data management is based on Python and SQL scripts. These are orchestrated and scheduled using the Argo Workflow Engine in the form of a directed acyclic graph (DAG). Various Tableau reports serve as an interface to the data with the aim of accessing the detailed data in top-down semantics from aggregated data via drill-down.
Subject description
As part of fraud detection and prevention, the customer's responsible department requires extensive evaluation options in order to be able to recognize anomalies in the processing of shipments both at the logistics locations and on the so-called last mile (delivery routes). Furthermore, the application helps the fraud investigation specialists, in cooperation with the location managers, to investigate potential shipment losses. By linking data from the shipment processing systems with data from customer information systems, additional opportunities are created to identify fraud patterns.