This research is focused on applications of semantic web and data mining techniques to the field of AAL. In particular the activities concern two major aspects. The first one is related to solutions aimed at supporting the design of AAL systems. Indeed, the introduced methodology that, starting from designer's requirements, provides a set of sensors which allow to achieve the project objectives. In order to fulfill such a task, a goal-driven ontology has been developed and on its top a reasoning framework is introduced. The ontology includes three different modules. The Goal module which describes the decomposition of a goal. The Measure module which is aimed at describing the measures and their relations with all the objects belonging to the modeled domain. Finally the Sensor module which defines the sensors and their characteristics. The second aspect which is the objective of this research, is related to Process Mining techniques aimed to study the behavior of a user that lives in an AAL environment. In particular, taking advantage of process discovery algorithms which are able to extract the model of a process, starting from process event logs. Such algorithms can be adopted in contexts in which the process model is unknown or they can be helpful to check the conformance of the actual model to the design-time model. In the AAL field such algorithms have been used on data collected through sensors of a smart environment. After a proper pre-processing stage, data have been exploited to extract two kinds of different models. The daily behavioral model which represents the behavior of a standard day of the user, whereas the macro activity models which are a set of models representing the flow of sensors activations when the user perform a given daily activity (e.g. watching the tv, cooking, reading and so forth). The obtained models have been verified with interesting results through conformance checking algorithms.
La ricerca è focalizzata sull’applicazione di tecniche, sviluppate nell’ambito del semantic web e del data mining, al settore dell'Ambient Assisted Living (AAL). La ricerca ha riguardato due filoni principali. Il primo relativo a soluzioni di supporto alla progettazione di sistemi AAL, in cui si è sviluppato un sistema che a partire dai requisiti definiti dal progettista restituisce un set di sensori che permettono il raggiungimento di tali obiettivi. Per fare ciò è stata sviluppata un’ontologia goal-driven e relativi meccanismi di reasoning. L’ontologia è formata da tre moduli. Il modulo Goal che descrive la decomposizione di goal, il modulo Measure che descrive le misure e le loro relazioni con gli oggetti del dominio, ed infine il modulo Sensor che definisce i sensori e le loro caratteristiche. Sulla base di questa ontologia sono stati sviluppati meccanismi di reasoning che permettono il passaggio dai requisiti di progetto ai sensori necessari per soddisfarli. Un secondo filone di ricerca che è stato portato avanti è legato a tecniche di Process Mining, finalizzate allo studio del comportamento di un utente in un ambiente AAL. Sono stati usati algoritmi di process discovery che a partire da event log di un processo sono in grado di indurre il modello del processo stesso. Tali algoritmi sono molto utili in contesti in cui il modello di processo è sconosciuto oppure a verificare che il modello effettivamente eseguito dagli utenti sia conforme al modello definito a design-time. Nel contesto dell’AAL sono stati usati tali algoritmi su dati generati dai sensori presenti in un ambiente domotico. Dopo una fase di pre-processing, i dati sono stati usati per estrarre due tipi di modelli. Il daily behavioral model che rappresenta il comportamento della giornata tipo dell’utente ed i macro activity model, una serie di modelli che rappresentano la sequenza di attivazione dei sensori quando l'utente esegue una determinata attività quotidiana. I modelli ottenuti, sono stati testati con buoni risultati tramite algoritmi di conformance checking.
Smart Environment Design and User's Behaviors Mining in AAL Domain / Cameranesi, Marco. - (2018 Mar 27).
Smart Environment Design and User's Behaviors Mining in AAL Domain
CAMERANESI, MARCO
2018-03-27
Abstract
This research is focused on applications of semantic web and data mining techniques to the field of AAL. In particular the activities concern two major aspects. The first one is related to solutions aimed at supporting the design of AAL systems. Indeed, the introduced methodology that, starting from designer's requirements, provides a set of sensors which allow to achieve the project objectives. In order to fulfill such a task, a goal-driven ontology has been developed and on its top a reasoning framework is introduced. The ontology includes three different modules. The Goal module which describes the decomposition of a goal. The Measure module which is aimed at describing the measures and their relations with all the objects belonging to the modeled domain. Finally the Sensor module which defines the sensors and their characteristics. The second aspect which is the objective of this research, is related to Process Mining techniques aimed to study the behavior of a user that lives in an AAL environment. In particular, taking advantage of process discovery algorithms which are able to extract the model of a process, starting from process event logs. Such algorithms can be adopted in contexts in which the process model is unknown or they can be helpful to check the conformance of the actual model to the design-time model. In the AAL field such algorithms have been used on data collected through sensors of a smart environment. After a proper pre-processing stage, data have been exploited to extract two kinds of different models. The daily behavioral model which represents the behavior of a standard day of the user, whereas the macro activity models which are a set of models representing the flow of sensors activations when the user perform a given daily activity (e.g. watching the tv, cooking, reading and so forth). The obtained models have been verified with interesting results through conformance checking algorithms.File | Dimensione | Formato | |
---|---|---|---|
Tesi_Cameranesi.pdf
accesso aperto
Descrizione: Tesi_Cameranesi.pdf
Tipologia:
Tesi di dottorato
Licenza d'uso:
Creative commons
Dimensione
4.45 MB
Formato
Adobe PDF
|
4.45 MB | Adobe PDF | Visualizza/Apri |
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.