Detecting Rainfall Events Leveraging Climate Knowledge Graphs

Rainfall detection nowadays requires vast amounts of diverse climate data, such as temperature, precipitation, and wind speed, to be coupled with machine learning models in order to get a better understanding of meteorological effects. Meanwhile, knowledge graphs excel in integrating disparate data sources and providing an interoperable framework for web-wide data integration, referred to as Linked Data Principles. Despite this, today’s knowledge graph data is often incompatible with the input of many machine learning pipelines, such as the Python package “scikit-learn”, which takes tabular data as input and is frequently used by studies to assist in the creation of data intelligence. In this paper, we utilize our recently created knowledge graph about climate observations as a use case to create a work ow for scholars to facilitate the extraction of climate data from any knowledge graph for automated conversion from graph-form data to tabular-form data suitable for feeding into machine learning models. As a result, our process enables the integration of diverse climate data while also smoothing the introduction of diverse climate graph data into machine learning techniques for rainfall detection.