Data management

ECoMMFiT has a lot of experience in the management, storage and processing of high volumes of data. Some of the key research activities of the Group revolved around Experiments and Computational Fluid Dynamics which generate by design a large amount of data. This information includes: imaging data generated by image acquisition hardware (e.g. high resolution and speed cameras), hydrodynamic data collections from a wide variety of flow sensors (velocity, temperature, pressure) and structured data generated by massive computational numerical simulations in High-Performance Computing facilities, both in-house and external.

The Group has expertise in the fields of Big Data and Machine Learning. A subset of artificial intelligence, Machine (or Statistical) Learning (ML) refers to a vast set of tools for modeling and understanding complex datasets. This topic has recently gained a lot of attention due to its capacity to perform data reduction on very large data sets. Statistical learning has become a very hot field in many scientific areas as well as marketing, finance, engineering and business disciplines.

Big Data analysis provides a tool for the characterization of the data and the detection of anomalies/deviations. Combined with Machine Learning techniques, this large volumes of data can be used to build more or less complex mathematical models for (1) predicting responses to perturbations on a large number of predictors - Supervised Learning and (2) classify large sets of data according to predefined features - Unsupervised Learning.

Based on this knowledge, the group has developed in the last years different projects for different types of industries and applications. Examples of the projects being developed by the group include development of a platform aimed to enhance the water supply management using remote water meters. The specific goals can be summarized as:
  • model the temporal evolution of water usage and identify consumption trends
  • detection of abnormal behaviors in the distribution network track and identify leaks and/or fraud events
  • implement a customer-tailored warning system for extra and infra-consumption.
  • improvement of the billing procedure and fraud detection in an electrical company, or
  • implementation of an alert systems for accidental extubation events in Intensive Care Unit (ICU) patients.
In collaboration with the University Hospital Joan XXIII of Tarragona, this last project uses monitor data from a database of ~1000 ICU patients to:
  • identify patient state patterns that may potentially lead to unplanned extubations in patients under Mechanical Ventilation.
  • derive a model to assess the validity of the RoX index as a predictor for successful extubation. This index, defined as the ratio of oxygen saturation as measured by pulse oximetry/FiO2 to respiratory rate, has already been used as a predictor for a successful extubation in patients with acute hypoxemic respiratory failure.
The results of this study have been published in the following article:
  • Fabregat, A., Magret, M., Ferreacute, J.A., Vernet, A., Guasch, N., Rodríguez, A., Gómez, J., Bodí, M.; A Machine Learning decision-making tool for extubation in Intensive Care Unit patients, Computer Methods and Programs in Biomedicine, https://doi.org/10.1016/j.cmpb.2020.105869

  • Another study based on Machine Learning was developed to evaluate the impact of cruise ship traffic on the urban air quality of a big city. The results have been published in the following article:
  • Fabregat, A., Vàzquez, Ll., Vernet, A.; Using Machine Learning to estimate the Impact of Ports and Cruise ship traffic on urban air quality: the case of Barcelona, Environmental Modelling and Software 139(1-4):104995 DOI: 10.1016/j.envsoft.2021.104995
  • [top]