IW-NET BDA: A Big Data Infrastructure for Predictive and Geotemporal Analytics of Inland Waterways

Abstract

The recent shift towards digitalization in traditional sectors like logistics and transportation has unlocked new avenues for gaining valuable insights and streamlining operations. This transformation is facilitated by the abundance and specificity of data now available, including fleet IoT data, transactional documents, and event notifications. These businesses leave a substantial digital footprint, ripe for analysis when combined with external data sources. However, harnessing this information requires robust computing infrastructure and adaptable software capable of handling vast amounts of data. In this paper, we introduce IW-NET BDA, a big-data analytics framework built on open-source technologies to address the storage and processing demands of massive datasets from various origins. Developed within the framework of the EU-funded research and innovation project IW-NET (Innovation driven Collaborative European Inland Waterways Transport Network), our system caters to the logistics domain but offers a versatile IT service backbone due to its agnostic design, focusing on infrastructure-as-a-service provision. Furthermore, it allows for the development and deployment of applications that encapsulate business logic, thus tailored to specific business needs. In the subsequent sections, we delve into the design principles, architectural components, and deployment possibilities of IW-NET BDA. Additionally, we present two illustrative use cases: firstly, the automated detection of areas of interest and vessel activity tracking for insightful geo-temporal data analytics along the River Weser corridor; secondly, the utilization of recurrent neural networks to forecast water levels in the Danube River corridor. These examples highlight the adaptability and efficacy of IW-NET BDA in tackling diverse challenges across different contexts, underscoring its versatility and utility.

Keywords

• Big data
• cloud computing
• DBSCAN
• machine learning
• recurrent neural networks