​ IoT Measurement Kn owledge Graphs: Constructing, Working and Learning with IoT Measureme nt Data as a Knowledge Graph
Promotors: Prof. Dr. F. van Harme len, Dr. V de Boer​
Co-Promotor: Dr. L. Daniele, Dr. R. Siebe s
IoT devices generate substantial amounts of measurements and use many different formats to store this data. In order to make IoT d evices interoperable, a number of ontologies have been developed, suc h as SAREF, to create knowledge graphs that can represent all kinds o f measurements from IoT devices. We call the resulting knowledge grap hs: IoT measurement knowledge graphs. In this thesis, we set out to i nvestigate what sets IoT measurement knowledge graphs apart from regu lar knowledge graphs. In the first half, we describe in detail how to create IoT measurement knowledge graphs from real-world measurement data. In the second half, we investigate what happens when we use the graphs in different scenarios, with different methods and applicatio ns. In order to generate an IoT measurement knowledge graph, we creat e a mapping to transform the measurement data coming from IoT devices . We explore multiple ways to create this mapping and evaluate the re sulting IoT measurement knowledge graph with competency questions. Us ing our mapping, we create OfficeGraph, by transforming IoT measureme nt data recorded over a year from 444 IoT devices located in an offic e building. This IoT measurement knowledge graph is validated with co mpetency questions created with the support of the building owners, s howing that they can now answer questions they were not able to answe r without it. Besides the entities, literals and relations in knowled ge graphs, the combination of these, the context of entities, provide s additional knowledge. Therefore, if we want to use entities in know ledge graphs to train machine learning models, it would be a waste to take only the entities from the graph, because this would leave out information. Representation v models, such as RDF2Vec and GCNs can be used to learn (embedding) representations for entities that take con text into account, creating a representation based on which relations , entities, and literals occur near the entity in the graph. We learn over IoT measurement knowledge graphs using multiple representation models and experiment with the effect of making more knowledge availa ble to the representation models through semantic enrichment. This is done by making implicit information, such as (e.g.) consecutive meas urements, explicitly available, by adding a relation between two meas urements. Results show that the semantic enrichment has a positive ef fect on the learnability of the entity representations. Due to the dy namic nature of IoT measurement knowledge graphs, it is to be expecte d that new measurements are made after the initial IoT measurement kn owledge graphs are created. Therefore, we investigate the possibility of estimating embedding representations for new entities that are ba sed off the existing entity representations. We introduce the embeddi ng estimation method, which used the numerical attributes of entities , the measurement values, to find entities similar to the new entity. Then by averaging the embedding representations of those entities, i t creates a new, estimated, embedding for the new entity. We perform experiments to test the embedding estimation method, and show that th ere is a trade-off between time saved by not re-training the entire p ipeline, but also a loss in accuracy, with the estimated embeddings. This thesis offers the first focused research into IoT measurement kn owledge graph. We provide insight into how IoT measurement knowledge graph can be created and how they differ from regular knowledge graph s, specifically in being numerical, dynamic, and shallow knowledge gr aphs. We suggest and validate remedies to the shallowness of IoT meas urement knowledge graphs, through the semantic enrichments. Furthermo re we show how the numerical aspect can be a strength, through the em bedding estimation method, which can help with the dynamic nature of IoT measurement knowledge graph.
DESCRIPTION: Promotors: Prof. Dr. F. van Harmelen, Dr. V de Boer​ Co -Promotor: Dr. L. Daniele, Dr. R. Siebes IoT devices generate substan tial amounts of measurements and use many different formats to store this data. In order to make IoT devices interoperable, a number of on tologies have been developed, such as SAREF, to create knowledge grap hs that can represent all kinds of measurements from IoT devices. We call the resulting knowledge graphs: IoT measurement knowledge graphs . In this thesis, we set out to investigate what sets IoT measurement knowledge graphs apart from regular knowledge graphs. In the first h alf, we describe in detail how to create IoT measurement knowledge gr aphs from real-world measurement data. In the second half, we investi gate what happens when we use the graphs in different scenarios, with different methods and applications. In order to generate an IoT meas urement knowledge graph, we create a mapping to transform the measure ment data coming from IoT devices. We explore multiple ways to create this mapping and evaluate the resulting IoT measurement knowledge gr aph with competency questions. Using our mapping, we create OfficeGra ph, by transforming IoT measurement data recorded over a year from 44 4 IoT devices located in an office building. This IoT measurement kno wledge graph is validated with competency questions created with the support of the building owners, showing that they can now answer ques tions they were not able to answer without it. Besides the entities, literals and relations in knowledge graphs, the combination of these, the context of entities, provides additional knowledge. Therefore, i f we want to use entities in knowledge graphs to train machine learni ng models, it would be a waste to take only the entities from the gra ph, because this would leave out information. Representation v models , such as RDF2Vec and GCNs can be used to learn (embedding) represent ations for entities that take context into account, creating a repres entation based on which relations, entities, and literals occur near the entity in the graph. We learn over IoT measurement knowledge grap hs using multiple representation models and experiment with the effec t of making more knowledge available to the representation models thr ough semantic enrichment. This is done by making implicit information , such as (e.g.) consecutive measurements, explicitly available, by a dding a relation between two measurements. Results show that the sema ntic enrichment has a positive effect on the learnability of the enti ty representations. Due to the dynamic nature of IoT measurement know ledge graphs, it is to be expected that new measurements are made aft er the initial IoT measurement knowledge graphs are created. Therefor e, we investigate the possibility of estimating embedding representat ions for new entities that are based off the existing entity represen tations. We introduce the embedding estimation method, which used the numerical attributes of entities, the measurement values, to find en tities similar to the new entity. Then by averaging the embedding rep resentations of those entities, it creates a new, estimated, embeddin g for the new entity. We perform experiments to test the embedding es timation method, and show that there is a trade-off between time save d by not re-training the entire pipeline, but also a loss in accuracy , with the estimated embeddings. This thesis offers the first focused research into IoT measurement knowledge graph. We provide insight in to how IoT measurement knowledge graph can be created and how they di ffer from regular knowledge graphs, specifically in being numerical, dynamic, and shallow knowledge graphs. We suggest and validate remedi es to the shallowness of IoT measurement knowledge graphs, through th e semantic enrichments. Furthermore we show how the numerical aspect can be a strength, through the embedding estimation method, which can help with the dynamic nature of IoT measurement knowledge graph. ​ IoT Measurement Knowledge Graphs: Constructing, Working and Learning with IoT Measurement Data as aKnowledge Graph END:VEVENT END:VCALENDAR