Pasar al contenido principal

PLATAFORMA PARA LA MONITORIZACION Y EVALUACION DE LA EFICIENCIA DE LOS SISTEMAS DE DISTRIBUCION EN SMART CITIES

DPI2013-47450-C2-1-R

•
Nombre agencia financiadora Ministerio de Economía y Competitividad
Acrónimo agencia financiadora MINECO
Programa Programa Estatal de I+D+i Orientado a los Retos de la Sociedad
Convocatoria Proyectos I+D+i Retos de Investigación
Año convocatoria 2013
Plan Estatal 2013-2016 Plan Estatal de Investigación Científica y Técnica y de Innovación
Unidad de gestión Secretaría de Estado de Educación, Formación Profesional y Universidades
Centro beneficiario UNIVERSITAT DE GIRONA
Identificador persistente http://dx.doi.org/10.13039/501100003329

Resultados relacionados

Resultados totales (Incluyendo duplicados): 18
Encontrada(s) 1 página(s)

Recercat. Dipósit de la Recerca de Catalunya
Publicación Artículo científico (JournalArticle)

Multi-dimensional fairness for auction-based resource allocation

  • Pla Planas, Albert
  • López Ibáñez, Beatriz
  • Murillo Espinar, Javier
Multi-attribute resource allocation problems involve the allocation of resources on the basis of several attributes, therefore, the definition of a fairness method for this kind of auctions should be formulated from a multi-dimensional perspective. Under such point of view, fairness should take into account all the attributes involved in the allocation problem, since focusing on just a single attribute may compromise the allocations regarding the remainder attributes (e.g. incurring in delayed or bad quality tasks). In this paper, we present a multi-dimensional fairness approach based on priorities. For that purpose, a recurrent auction scenario is assumed, in which the auctioneer keeps track of winner and losers. From that information, the priority methods are defined based on the lost auctions number, the number of consecutive loses, and the fitness of their loser bids. Moreover, some methods contain a probabilistic parameter that enables handling wealth ranking disorders due to fairness. We test our approach in real-data based simulator which emulates an industrial production environment where several resource providers compete to perform different tasks. The results pointed that multi-dimensional fairness incentives agents to remain in the market whilst it improves the equity of the wealth distribution without compromising the quality of the allocation attributes, This research project has been partially funded through the BR10/18 Scholarship granted to the first author of the paper. Work developed with the support of the research group SITES awarded with distinction by the Generalitat de Catalunya (SGR 2014-2016) and the MESC project funded by the Spanish MINECO (Ref. DPI2013-47450-C2-1-R)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Artículo científico (JournalArticle)

Self-organising energy demand allocation through canons of distributive justice in a microgrid

  • Torrent-Fontbona, Ferran
  • López Ibáñez, Beatriz
  • Busquets, Dídac
  • Pitt, J.
The energy sector is being driven into a new era where considerable portions of electrical demand will be met through distributed energy resources (DERs). Microgrids have been suggested as a tool for integrating and managing DERs. In this context, we formulate the energy demand allocation problem in order to provide service to a given load. We then propose a dynamic method for agreeing and setting the rules to perform the allocation. The methodology is based on self-organisation and the concept of distributive justice which integrates different principles of fairness represented as legitimate claims. Legitimate claims are implemented as voting functions and are used to determine how the DER requests are satisfied. The method is tested by considering different configurations of DERs, mainly of the renewable type, and comparing them with other allocation methods. Results show that this self-organising allocation method provides a better balance amongst all the representations of justice, but also it is more robust for the external authorities that manipulate the allocation process, This research project has been partially funded through BR-UdG Scholarship of the University of Girona granted to Ferran Torrent-Fontbona. Work developed with the support of the research group SITES awarded with distinction by the Generalitat de Catalunya (SGR 2014-2016) and the MESC project funded by the Spanish MINECO (Ref. DPI2013-47450-C2-1-R)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Contribución de congreso (ConferenceOutput)

Case base maintenance of a personalized insulin dose recommender system for Type 1 Diabetes Mellitus

  • Torrent-Fontbona, Ferran
  • Massana i Raurich, Joaquim
  • López Ibáñez, Beatriz
Comunicació presentada al First Joint Workshop on AI in Health (AIH 2018), Stockholm, Sweden, July 13-14, 2018, With the goal of supporting people su ering Type 1 Diabetes Mellitus (T1DM), some mobile applications are being developed based on arti cial intelligence techniques. Some of these applications are based on Case-Based Reasoning methodologies (CBR) due to the advantage regarding a personal, adapted recommendation. However, the amount and quality of the cases in the CBR system will threat the system outcome. Most of the maintenance methods developed deals with classi cation tasks, while recommending an insulin dose (bolus) involves a regression task. In this paper, a new maintenance method presented, with the particularity of dealing with a regression tasks. The method is applied over the Pepper insulin dose recommender system, and tested using the UVA/Padova simulator, exhibiting the improvements of the proposal in terms of both, the person health and the case-base size, This research project has been partially funded through BR-UdG Scholarship of the University of Girona granted to Joaquim Massana Raurich. Work developed with the support of the research group SITES awarded with distinction by the Generalitat de Catalunya (SGR 2014-2016), the MESC project funded by the Spanish MINECO (Ref. DPI2013-47450- C2-1-R) and the European Unions Horizon 2020 research and innovation programme under grant agreement No 680708
Recercat. Dipósit de la Recerca de Catalunya
Publicación Artículo científico (JournalArticle)

Identifying services for short-term load forecasting using data driven models in a Smart City platform

  • Massana i Raurich, Joaquim
  • Pous i Sabadí, Carles
  • Burgas Nadal, Llorenç
  • Meléndez i Frigola, Joaquim
  • Colomer Llinàs, Joan
The paper describes an ongoing work to embed several services in a Smart City architecture with the aim of achieving a sustainable city. In particular, the main goal is to identify services required in such framework to define the requirements and features of a reference architecture to support the data-driven methods for energy efficiency monitoring or load prediction. With this object in mind, a use case of short-term load forecasting in non-residential buildings in the University of Girona is provided, in order to practically explain the services embedded in the described general layers architecture. In the work, classic data-driven models for load forecasting in buildings are used as an example, This research project has been partially funded through BR-UdGScholarship of the University of Girona granted to Joaquim MassanaRaurich. Work developed with the support of the research groupSITES awarded with distinction by the Generalitat de Catalunya(SGR 2014–2016), the MESC project funded by the Spanish MINECO (Ref. DPI2013-47450-C2-1-R) and the European Union’s Horizon2020 Research and Innovation Programme under grant agreementNo 680708
Recercat. Dipósit de la Recerca de Catalunya
Publicación Artículo científico (JournalArticle)

A new perspective of trust through multi-attribute auctions

  • Torrent-Fontbona, Ferran
  • Pla Planas, Albert
  • López Ibáñez, Beatriz
The use of trust in auctions is a well-studied problem; however, most of the works in the literature focus on how to model trust rather on how trust is used in the mechanism. In this paper we propose a method to manage trust in multi-attribute auctions. Complementary, we propose a new trust model and we compare it withmother models in the literature. The proposed methodology is tested using a real data-based simulator, showing the benefits of incorporating trust in multi-attribute auctions, This research project has been partially funded through BRUdG Scholarship of the University of Girona granted to Ferran Torrent-Fontbona. Work developed with the support of the research group SITES awarded with distinction by the Generalitat de Catalunya (SGR 2014-2016) and the MESC project funded by the Spanish MINECO (Ref. DPI2013- 47450-C2-1-R)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Artículo científico (JournalArticle)

Prediction of glucose level conditions from sequential data

  • Mordvanyuk, Natalia
  • Torrent-Fontbona, Ferran
  • López Ibáñez, Beatriz
In type 1 diabetes management, mobile health applications are becoming a cornerstone to empower people to self-manage their disease. There are many applications addressed to calculate insulin doses based on the current information (e.g. carbohydrates intake) and a few of them are accompanied by modules able to supervise postprandial conditions and recommend corrective actions if the user falls in an abnormal state (i.e. hyperglycaemia or hypoglycaemia). On the other hand, mobile apps favour the gathering of historical data from which machine learning techniques can be used to predict if user conditions will worsen. This work presents the application of k-nearest neighbour on the historical data gathered on patients, so that given the information related to a sequence of meals, the method is able to predict if the patient will fall in an abnormal condition. The experimentation has been carried out with the UVA-Padova type 1 diabetes simulator over eleven adult profiles. Results corroborate that the use of sequential data improve significantly the prediction outcome when forecasts distinguish the type of meal (breakfast, lunch and dinner), This work has received funding from the EU Horizon 2020 research and innovation programme under grant agreement No 689810 (PEPPER), and from the University of Girona under the grant MPCUdG2016 (Ajut per a la millora de la productivitat científica dels grups de recerca), and the Spanish MINECO under the grant number DPI2013- 47450-C21-R. This work has been developed with the support of the research group SITES awarded with distinction by the Generalitat de Catalunya (SGR 2014-2016)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Artículo científico (JournalArticle)

Compatibility of municipal services based on service similarity

  • Rusek, Robert
  • Marsal, Maria Lluïsa
  • Torrent-Fontbona, Ferran
  • Colomer Llinàs, Joan
The aim of this article is to propose and examine a quantitative method of determining the degree of compatibility between municipal services. Provision of services and facilities maintenance are usually two biggest expenditures of local governments. Traditionally, facilities host only one service, whereas the challenge and opportunity lies in combining various, compatible services and offering them together under one roof. Such a combination decreases municipal expenditure and has a strong positive impact on the general service quality. For this purpose, we take advantage of the City-block distance formula to calculate the degree of compatibility between municipal services. The method is examined and discussed on a sample of 30 real municipal services. This allows us to find possible combinations of strongly compatible services that should be offered together in Multi-Service Facilities and, at the same time, avoid an unwanted combination of services that are incompatible
Recercat. Dipósit de la Recerca de Catalunya
Publicación Artículo científico (JournalArticle)

Multivariate statistical monitoring of buildings. Case study: Energy monitoring of a social housing building

  • Burgas Nadal, Llorenç
  • Meléndez i Frigola, Joaquim
  • Colomer Llinàs, Joan
  • Massana i Raurich, Joaquim
  • Pous i Sabadí, Carles
A complete methodology for energy building monitoring based on Principal Component Analysis (PCA) is proposed. The method extends the Unfolding or Multiway Principal Component Analysis (MPCA) used in statistical batch process control in terms of building and neighbourhood monitoring. Relationships between energy consumption and independent variables such as weather, occupancy or any other variables that are significant for monitoring can be gathered in a model using the proposed methodology. Historic data are used to obtain a reference model that will be used for monitoring. Two unfolding strategies are proposed (time-wise and entity-based) offering complementary views of the building or of the community under consideration. The first, time-wise unfolding, is suitable for detecting behavioural changes over time, whereas entity-wise unfolding allows the identification of entities, e.g. dwellings in a building, that behave substantially differently from others over a period of time. Two simple statistics, T<sup>2</sup> and SPE, are used to define two monitoring charts capable of detecting abnormal behaviours and, furthermore, the isolation of variables that mainly explain such a situation. The paper presents the theoretical background, followed by the methodological principles. The results are illustrated by a case study, This work has been developed within the project Plataforma para la monitorización y evaluación de la eficiencia de los sistemas de distribución en Smart Cities, ref. DPI2013-47450-C2-1-R and project ACCUS (Adaptive Cooperative Control in Urban (sub) Systems., ART-010000-2013-2 -333020-1), funded by the Spanish Ministry of Industry, Energy and Tourism and by the JTI ARTEMIS Joint Undertaking of the European Commission. Appreciation is given for the data provided by the Patronat de l’habitatge de Barcelona with the collaboration of AITEL. Data fromMeteocat were also used
Recercat. Dipósit de la Recerca de Catalunya
Publicación Artículo científico (JournalArticle)

Power re-allocation for reducing contracted electric power costs

  • Torrent-Fontbona, Ferran
  • López Ibáñez, Beatriz
Electric bills consist of a cost related to the consumed energy and a cost related to the maximum demanded power. This latter part usually accounts for approximately 25-40% of the bill. Demanded power by big consumers is measured in real time and electric companies highly penalise them if the maximum demanded power (along the billing period) exceeds the contracted power by the consumer. In this paper we propose a new method that, given the demanded power of close consumers for a time window (power profile), power costs are reduced by re-allocating the demanded power among consumers in order to keep all of them below or equal to their contracted power. We also propose and analyse some strategies to set a preference when not all power profiles can be kept below the contracted power. We tested this method using real-based simulated power profiles of eight different business buildings located in Girona and the power cost reduction achieved reached approximately 20%, This research project has been partially funded through BR-UdG2012 Scholarship of the University of Girona granted to Ferran Torrent-Fontbona. The work developed with the support of the research group SITES awarded with distinction by the Generalitat de Catalunya (2014 SGR 1052) and the MESC project funded by the Spanish MINECO (Ref. DPI2013-47450-C2-1-R)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Artículo científico (JournalArticle)

Short-term load forecasting in a non-residential building contrasting models and attributes

  • Massana i Raurich, Joaquim
  • Pous i Sabadí, Carles
  • Burgas Nadal, Llorenç
  • Meléndez i Frigola, Joaquim
  • Colomer Llinàs, Joan
The electric grid is evolving. Smart grids and demand response systems will increase the performance of the grid in terms of cost efficiency, resilience and safety. Accurate load forecasting is an important issue in the daily operation and control of a power system. A suitable short term load forecasting will enable a utility provider to plan the resources and also to take control measures to balance the supply and demand of electricity. The aim of this paper is to create a method to forecast the electric load in a non-residential building. Another goal is to analyse what kind of data, as weather, indoor ambient, calendar and building occupancy, is the most relevant in building load forecasting. A simple method, tested with three different models, such as MLR, MLP and SVR, is proposed. The results, from a real case study in the University of Girona, show that the proposed forecast method has high accuracy and low computational cost, This research has been partially supported by the Spanish Government project MESC (Ref. DPI2013-47450-C2-1-R). Also we would like to thank the Department of Physics and acknowledge the technical assistance and maintenance service of the UdG (SOTIM) which provided the weather and consumption data respectively. The authors belong to the ‘Smart IT Engineering and Solutions’ accredited research group (Generalitat de Catalunya, 2014 SGR 1052)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Contribución de congreso (ConferenceOutput)

Prediction of glucose level conditions from sequential data [Pòster]

  • Mordvanyuk, Natalia
  • Torrent-Fontbona, Ferran
  • López Ibáñez, Beatriz
Pòster de congrés presentat a: 20th International Conference of the Catalan Association for Artificial Intelligence, CCIA, October 25-27 th, Deltebre, 2017, Pòster relacionat amb la comunicació presentada al '20th International Conference of the Catalan Association for Artificial Intelligence' i publicada a 'Frontiers in Artificial Intelligence and Applications', 2017, vol. 300, p. 227-232. DOI: https://doi.org/10.3233/978-1-61499-806-8-227, In type 1 diabetes management, mobile health applications are becoming a cornerstone to empower people to self-manage their disease. There are many applications addressed to calculate insulin doses based on the current information (e.g. carbohydrates intake) and a few of them are accompanied by modules able to supervise postprandial conditions and recommend corrective actions if the user falls in an abnormal state (i.e. hyperglycaemia or hypoglycaemia). On the other hand, mobile apps favour the gathering of historical data from which machine learning techniques can be used to predict if user conditions will worsen. This work presents the application of k-nearest neighbour on the historical data gathered on patients, so that given the information related to a sequence of meals, the method is able to predict if the patient will fall in an abnormal condition. The experimentation has been carried out with the UVA-Padova type 1 diabetes simulator over eleven adult profiles. Results corroborate that the use of sequential data improve significantly the prediction outcome when forecasts distinguish the type of meal (breakfast, lunch and dinner)., This work has received funding from the EU Horizon2020 research and innovation programme under grant agreement No689810 (PEPPER), and from the University of Girona under the grant MPCUdG2016 (Ajut per a la millora de la productivitat científica dels grups de recerca), and the Spanish MINECO under the grant number DPI2013-47450-C21-R. This work has been developed with the support of the research group SITES awarded with distinction by the Generalitat de Catalunya (SGR 2014-2016)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Artículo científico (JournalArticle)

N-dimensional extension of unfold-PCA for granular systems monitoring

  • Burgas Nadal, Llorenç
  • Meléndez i Frigola, Joaquim
  • Colomer Llinàs, Joan
  • Massana i Raurich, Joaquim
  • Pous i Sabadí, Carles
This work is focused on the data based modelling and monitoring of a family of modular systems that have multiple replicated structures with the same nominal variables and show temporal behaviour with certain periodicity. These characteristics are present in many systems in numerous fields such as the construction or energy sector or in industry. The challenge for these systems is to be able to exploit the redundancy in both time and the physical structure. In this paper the authors present a method for representing such granular systems using N-dimensional data arrays which are then transformed into the suitable 2-dimensional matrices required to perform statistical processing. Here, the focus is on pre-processing data using a non-unique folding-unfolding algorithm in a way that allows for different statistical models to be built in accordance with the monitoring requirements selected. Principal Component Analysis (PCA) is assumed as the underlying principle to carry out the monitoring. Thus, the method extends the Unfold Principal Component Analysis (Unfold-PCA or Multiway PCA), applied to 3D arrays, to deal with N-dimensional matrices. However, this method is general enough to be applied in other multivariate monitoring strategies. Two of examples in the area of energy efficiency illustrate the application of the method for modelling. Both examples illustrate how when a unique data-set folded and unfolded in different ways, it offers different modelling capabilities. Moreover, one of the examples is extended to exploit real data. In this case, real data collected over a two-year period from a multi-housing social-building located in down town Barcelona (Catalonia) has been used, This work has been carried out by the research group eXIT (http://exit.udg.edu), funded through the following projects: MESC project(Ref. DPI2013-47450-C21-R) and its continuation CROWDSAVING (Ref.TIN2016-79726-C2-2-R), both funded by the Spanish Ministerio de Industria y Competitividad within the Research, Development and Innovation Program oriented towards the Societal Challenges, and also the project Hit2Gap of the Horizon 2020 research and innovation program under grant agreement N680708. The author Llorenç Burgas would also like to thank Girona University for their support through the competitive grant for doctoral formation IFUdG2016
Recercat. Dipósit de la Recerca de Catalunya
Publicación Artículo científico (JournalArticle)

Decision support for grid-connected renewable energy generators planning

  • Torrent-Fontbona, Ferran
  • López Ibáñez, Beatriz
Recent technological advances and the incremental demand for electrical energy are leading a growth in the prevalence of distributed generation. There are some off-the-shelf tools to support grid planners in locating and sizing a given number of Distributed Generators (DGs), but they approach the problem using a single set of the variables (either location, size or number of DGs). This paper reviews the problem and provides a new pathway for supporting grid planning with an integrated view; hence, a new planning problem is formulated to jointly determine how many new DGs are needed, of which type, their location and size, while attempting to maximise the profit of the generators, minimise the system losses and improve the voltage profile. Accompanying the new grid planning problem, solution approaches based on meta-heuristic methods are provided. A detailed performance analysis of the proposed approaches is carried out on 14- and 57-bus systems to illustrate what could be the outcomes of the new problem. In so doing, particle swarm optimisation-based approaches are able to find the best optimised solutions, Work developed with the support of the research group SITES awarded with distinction by the Generalitat de Catalunya (SGR 2014e2016) and the MESC project funded by the Spanish MINECO (Ref. DPI2013-47450-C2-1-R)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Artículo científico (JournalArticle)

eXiT Research Group at the University of Girona: Artificial Intelligence and Machine Learning Applied to Medicine and Healthcare

  • López Ibáñez, Beatriz
  • Mordvanyuk, Natalia
  • Massana i Raurich, Joaquim
  • Torrent-Fontbona, Ferran
  • Cáceres Bartrolí, Gerard
  • Pous i Sabadí, Carles
Comunicació presentada al I Workshop de Grupos de Investigación Españoles de IA en Biomedicina (IABiomed 2018) : XVIII Conferencia de la Asociación Española para la Inteligencia Artificial (CAEPIA 2018) : Granada, 23-26 de octubre de 2018, This paper describes the research activity of the eXiT research group at the Universitat of Girona and the main results achieved. The research is organized in four main groups: prognosis, clinical decision support systems, mHealth, and biosignal processing. The paper describes past, current and future research of the group, This project has received funding from the European Unions Horizon 2020 research and innovation programme under grant agreement No 689810 (PEPPER). Work developed with the support of the awarded distinction by the Generalitat de Catalunya 2017 SGR 1551, the MESC project funded by the Spanish MINECO (Ref. DPI2013-47450-C21-R), and the University of Girona grant to supporting research groups to improve their scientific productivity (MPCUdG2016)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Contribución de congreso (ConferenceOutput)

Towards a Data Driven Platform for Energy Efficiency Monitoring: Two Use Cases

  • Meléndez i Frigola, Joaquim
  • Colomer Llinàs, Joan
  • Pous i Sabadí, Carles
  • Burgas Nadal, Llorenç
  • Massana i Raurich, Joaquim
Comunicació presentada a: 11th International Conference on Artificial Intelligence Applications and Innovations (AIAI 2015) (AIAI 2015), Bayonne, France, September 14-17, 2015, The paper describes an ongoing work to define a framework to support monitoring procedures for distribution systems in smart cities. Specific use cases, involving data driven methods for energy monitoring, benchmarking and forecasting in urban scenarios are analysed. The objective is to identify services required in such framework to define the requirements of a reference architecture to support data-driven methods for energy efficiency monitoring and assessment. Use cases focus on multi-entity buildings monitoring and consumption forecasting, This work is being developed within the project, MESC- Platform for Monitoring and assessing the Efficiency of distribution systems in Smart Cities (Ref. DPI2013-47450-C2-1-R, 2014/16). Authors want to thank the organization of the Workshop SmarTABCD'15 (Workshop on Smart Technologies and Applications on Buildings, Cities and Districts, collocated with AIAI’15) to invite to present the MESC project for discussion in the context of Highly Innovative Building Control Tools Tackling the Energy Performance GAP (HIT2GAP)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Artículo científico (JournalArticle)

Short-term load forecasting for non-residential buildings contrasting artificial occupancy attributes

  • Massana i Raurich, Joaquim
  • Pous i Sabadí, Carles
  • Burgas Nadal, Llorenç
  • Meléndez i Frigola, Joaquim
  • Colomer Llinàs, Joan
An accurate short-term load forecasting system allows an optimum daily operation of the power system and a suitable process of decision-making, such as with regard to control measures, resource planning or initial investment, to be achieved. In a previous work, the authors demonstrated that an SVR model to forecast the electric load in a non-residential building using only the temperature and occupancy of the building as attributes is the one that gives the best balance of accuracy and computational cost for the cases under study. Starting from this conclusion, a simple, low-computational requirements and economical hourly consumption prediction method, based on SVR model and only the calculated occupancy indicator as attribute, is proposed. The method, unlike the others, is able to perform hourly predictions months in advance using only the occupancy indicator. Due to the relevance of the occupancy indicator in the model, this paper provides a complete study of the methods and data sources employed in the creation of the artificial occupancy attributes. Several occupancy indicators are defined, from the simplest one, using general information, to the most complex one, based on very detailed information. Then, a load forecasting performance discrimination between the artificial occupancy attributes is realized demonstrating that using the most complex indicator increases the workload and complexity while not improving the load prediction significantly. A real case study, applying the forecasting method to seve, This research project has been partially funded through BR-UdG Scholarship ofthe University of Girona granted to Joaquim Massana Raurich. Work developed with the support of the research group SITES awarded with distinction by the Generalitat de Catalunya (SGR 2014-2016) and the MESC project funded by the Spanish MINECO (Ref. DPI2013-47450-C2-1-R)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Contribución de congreso (ConferenceOutput)

Prediction of hyperglycaemia and hypoglycaemia events using longitudinal data [Pòster]

  • Mordvanyuk, Natalia
  • Torrent-Fontbona, Ferran
  • López Ibáñez, Beatriz
Pòster de congrés presentat a: 11th International Conference on Advanced Technologies & Treatments for Diabetes (ATTD 2018): 14-17 February, 2018, Vienna, Austria, In type 1 diabetes management, mobile health applications are becoming a cornerstone to empower people to self-manage their disease. There are many applications addressed to calculate insulin doses based on the current information (e.g. carbohydrates intake) and a few of them are accompanied by modules able to supervise postprandial conditions and recommend corrective actions if the user falls in an abnormal state (i.e. hyperglycaemia or hypoglycaemia). On the other hand, mobile apps favour the gathering of historical data from which machine learning techniquescanbeusedtopredictifuserconditionswillworsen. Thisworkpresentstheapplicationofk-nearestneighbouronthehistoricaldatagatheredonpatients,sothatgiven the information related to a sequence of meals, the method is able to predict if the patient will fall in an abnormal condition. The experimentation has been carried out with the UVA-Padova type 1 diabetes simulator over eleven adult profiles. Results corroborate that the use of sequential data improve significantly the prediction outcome when forecastsdistinguishthetypeofmeal(breakfast,lunchanddinner), This work has received funding from the EU Horizon2020 research and innovation programme under grant agreement No689810 (PEPPER), and from the University of Girona under the grant MPCUdG2016 (Ajut per a la millora de la productivitat científica dels grups de recerca), and the Spanish MINECO under the grant number DPI2013-47450-C21-R. This work has been developed with the support of the research group SITES awarded with distinction by the Generalitat de Catalunya (SGR 2014-2016)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Artículo científico (JournalArticle)

Single Nucleotide Polymorphism relevance learning with Random Forests for Type 2 diabetes risk prediction Type 2 diabetes Random Forest Feature learning Predictive model Gini importance

  • López Ibáñez, Beatriz
  • Torrent-Fontbona, Ferran
  • Viñas, Ramon
  • Fernández-Real Lemos, José Manuel
The use of artificial intelligence techniques to find out which Single Nucleotide Polymorphisms (SNPs) promote the development of a disease is one of the features of medical research, as such techniques may potentially aid early diagnosis and help in the prescription of preventive measures. In particular, the aim is to help physicians to identify the relevant SNPs related to Type 2 diabetes, and to build a decision-support tool for risk prediction. Methods: We use the Random Forest (RF) technique in order to search for the most important attributes (SNPs) related to diabetes, giving a weight (degree of importance), ranging between 0 and 1, to each attribute. Support Vector Machines and Logistic Regression have also been used since they are two other machine learning techniques that are well-established in the health community. Their performance has been compared to that achieved by RF. Furthermore, the relevance of the attributes obtained through the use of RF has then been used to perform predictions with k-Nearest Neighbour method weighting attributes in the similarity measure according to the relevance of the attributes with RF. Results: Testing is performed on a set of 677 subjects. RF is able to handle the complexity of features' interactions, overfitting, and unknown attribute values, providing the SNPs' relevance with an up to 0.89 area under the ROC curve in terms of risk prediction. RF outperforms all the other tested machine learning techniques in terms of prediction accuracy, and in terms of the stability of the estimated relevance of the attributes. Conclusions: The Random Forest is a useful method for learning predictive models and the relevance of SNPs without any underlying assumption, This work was supported by the European Unions Horizon 2020 research and innovation programme [grant number 689810, PEPPER]; the University of Girona [grant number MPCUdG2016]; and the Spanish MINECO [grant number DPI2013-47450-C21-R].