ENBIS-17 in Naples

9 – 14 September 2017; Naples (Italy) Abstract submission: 21 November 2016 – 10 May 2017

My abstracts


The following abstracts have been accepted for this event:

  • Functional Data Analysis for Business and Industry

    Authors: Simone Vantini (MOX - Dept. of Mathematics, Politecnico di Milano)
    Primary area of focus / application: Other: Italian SIS session on Statistics in Data Science
    Keywords: Functional data analysis, Functional time series, Local inference, Curve prediction, Profile monitoring
    Submitted at 3-Apr-2017 16:17 by Simone Vantini
    12-Sep-2017 12:40 Functional Data Analysis for Business and Industry
    The continuous and outstanding advances of measurement technologies have enabled the collection and storage of high-resolution data which can often be modeled as smooth functions (e.g., curves or surfaces). This kind of data are at the basis of functional data analysis (FDA, Ramsay and Silverman, 2005) which is a well-known lively and expanding research area of modern statistics. In FDA, the classical concept for scalar or multivariate random variable is indeed replaced by the concept of functional random variable. Consequently, in FDA the typical data set is not made of numbers or Euclidean vectors but a collection of functions embedded in a suitable separable functional Hilbert space meant to formalize application-specific relations between sample units.
    Recent applications of FDA techniques in different and many fields of science are countless. Nevertheless, very few business and industrial applications can be found, thus pointing out the existence of an unexploited potential of this type of techniques in these two fields.
    With respect to this discrepancy, this talk will showcase two recent business and industrial applications in which state-of-the-art FDA techniques are fruitfully used. The first application (i.e., Canale and Vantini, 2016) pertains to the one-day-ahead prediction of natural gas demand and supply curves in the Italian gas balancing platform (i.e., Mercato del Bilanciamento). In this former application, the concept of constrained functional auto-regressive model is introduced and then used for predictive purposes. The latter application (Pini et al., 2017) pertains instead to the real-time monitoring of a laser-based welding process based on the analysis of plasma/metal emission spectrum. In this application, a local non-parametric functional ANOVA is performed such to build ad-hoc monitoring tools for the early detection of out-of-control dynamics.

    Canale, A. and Vantini, S. (2016): " Constrained functional time series: Applications to the Italian gas market”, International Journal of Forecasting, Vol. 32(4), pp. 1340-1351.
    Pini, A., Vantini, S., Colosimo, B. M., Grasso, M. (2017): “Domain-Selective Functional ANOVA for Supervised Statistical Profile Monitoring of Signal Data”, Journal of the Royal Statistical Society – Series C (to appear).
    Ramsay, J. O. and Silverman, B. W. (2005): Functional data analysis, Springer, New York.
  • Analysis of High Frequency Acoustic Data

    Authors: Tim Park (Shell)
    Primary area of focus / application: Other: Novel methods for industrial data streams
    Keywords: Time Series, Spectral analysis, Industrial statistics, Spatial temporal modelling
    Submitted at 4-Apr-2017 15:07 by Tim Park
    Accepted (view paper)
    12-Sep-2017 15:40 Analysis of High Frequency Acoustic Data
    Distributed Acoustic Sensing (DAS) is a recently introduced sensing technique which has opened up many new possibilities in the oil and gas industry. It involves using a fibre optic cable to measure acoustic signals at different spatial locations along the length of a pipe. The sound is collected continuously at very high temporal frequency and at multiple locations simultaneously. The fibre is easy to deploy both for surface pipes and downhole wells. This has opened up a lot of opportunities to collect data where it was not possible before. Applications include real time estimation of quantities such as flow rates, flow composition and the detection of events such as leaks and water breakthroughs. The downside, as compared to standard measurement devices such as pressure gauges and flow meters, is that it is not a direct measurement of the quantity of interest and is susceptible to noise and corruption. In this talk I will introduce some of the statistical challenges this data presents and the ways in which we use time series, alongside other areas of statistics, to extract relevant information from the acoustic signal.
  • Online Classification of Non-Stationary Industrial Data Streams

    Authors: Idris Eckley (Lancaster University)
    Primary area of focus / application: Other: Novel methods for industrial data streams
    Keywords: Wavelets, Non-stationary time series, Dynamic classification, Streaming data
    Submitted at 4-Apr-2017 18:27 by Idris Eckley
    12-Sep-2017 16:20 Online Classification of Non-Stationary Industrial Data Streams
    The long-established need to be able to accurately detect, diagnose and act in (or close to) real-time on process data has become ever more apparent, and computationally challenging, with the advent of high-resolution sensors capable of recording many variables at kHz or even GHz. Such signals are typically non-stationary in structure, with potentially time-varying dependence between the various components. In this talk I will motivate and introduce some recent work in this area that focuses on the challenge of online estimation of the time-dependent coherence and its use within a novel, dynamic classification approach for data streams arising from a collaboration with an industrial partner.
  • Bayesian Networks in Survey Data: Robustness and Sensitivity Issues

    Authors: Silvia Salini (University of Milan), Ron Kenett (KPA Ltd., Raanana, Israel and University of Turin, Italy), Federica Cugnata (University Centre for Statistics in the Biomedical Sciences, Vita-Salute San Raffaele University)
    Primary area of focus / application: Other: ASQ international journal session
    Keywords: Do calculus, Hard and soft evidence, Importance performance analysis, Information quality (InfoQ), Survey data, What-if scenario
    Submitted at 5-Apr-2017 14:49 by SALINI SILVIA
    11-Sep-2017 13:00 Bayesian Networks in Survey Data: Robustness and Sensitivity Issues
    Bayesian networks (BN) implement a graphical model structure known as a directed acyclic graph (DAG) that is popular in statistics, machine learning, and artificial intelligence. They enable an effective representation and computation of a joint probability distribution (JPD) over a set of random variables.
    The paper focuses on the selection of a robust network structure according to different learning algorithms and the measure of arc strength using resampling techniques. Moreover, it shows how ‘what-if’ sensitivity scenarios are generated with BN using hard and soft evidence in the framework of predictive inference. Establishing a robust network structure and using it for decision support are two essential enablers for efficient and effective applications of BN to improvements of products and processes. A customer-satisfaction survey example is presented and R scripts are provided.
  • Two-Player Zero Sum Games with Fuzzy Payoffs for Different Risk Levels

    Authors: Yesim Koca (Hacettepe University), Ozlem Muge Testik (Hacettepe University)
    Primary area of focus / application: Business
    Keywords: Multi objective linear programming, Game theory, Two-player zero sum games, Fuzzy logic
    Submitted at 5-Apr-2017 22:52 by Ozlem Testik
    12-Sep-2017 14:50 Two-Player Zero Sum Games with Fuzzy Payoffs for Different Risk Levels
    This study aims to define best strategies for competing two online shopping sites by using a zero sum game with fuzzy payoffs. Since the importance of the game mostly varies for the stores in this competitive environment, different risk levels are considered for the players. The shopping sites are assumed offering the same products for the same prices; in order to focus only on determined marketing activities. Marketing activities are regarded as game strategies and payoffs for a website are determined in terms of the number of people preferring that website among the constant number of customers which are known to make the purchase from one of these sites. Potential customers’ opinions are investigated by asking them to state which site they would prefer for each strategy combination and results are modeled by fuzzy numbers due to subjectivity and uncertainty of the data. After that the game is converted to a multi-objective linear programming problem and optimal solution of this problem is obtained. This optimal solution refers to the optimal strategies for players. The results are discussed for different customer characteristics (age, gender, profession) and suggestions for online shopping websites are presented.
  • A Non-Standard Approach to the Calibration of Selected Dynamic Factor Models in Macroeconomic Forecasting

    Authors: Fabio Della Marra (University of Parma)
    Primary area of focus / application: Economics
    Keywords: Macroeconometrics, Time series forecasting, Dynamic factor models, Genetic algorithms
    Submitted at 6-Apr-2017 14:43 by Fabio Della Marra
    Accepted (view paper)
    13-Sep-2017 09:20 A Non-Standard Approach to the Calibration of Selected Dynamic Factor Models in Macroeconomic Forecasting
    In this paper, we present a comparison of the forecasting performance of selected factor models on two large monthly data panels. The first dataset contains EU variables, whereas the other contains US variables. These data panels are split into two parts: the first subsample is used to select the most performing specification for each class of models in a in-sample environment, and the second subsample is used to compare the performances of the selected models in an out-of-sample environment. In the first subsample, genetic algorithms are employed to achieve an efficient exploration of the parameter space. We find that selected dynamic factor models are globally the most performing methods on the second subsamples of both data panels.