ENBIS-18 in Nancy2 – 25 September 2018; Ecoles des Mines, Nancy (France) Abstract submission: 20 December 2017 – 4 June 2018
Best Manager Award (Sotiris Bersimis) and Young Statistician Awards (Abel Folch-Fortuny & Alessandra Menafoglio)4 September 2018, 17:20 – 18:50
Statistical Tools for Quantitative Management and Fraud Detection
Best Manager Award
Sotiris Bersimis (Hellenic Organization for Health Care Services Provision (EOPYY), University of Piraeus)
A challenge, in the era of economic crisis and uncertainty, is to provide health care services in an efficient and effective manner. The location of the service centers, the geographical distribution of patients, and the provision of specialist services are some of the topics an Organization responsible for health care services provision has to arrange. Other topics are the assessment of quality, safety and effectiveness of healthcare services provided by private and public health care providers. A prominent place among these topics hold expenditure monitoring and control. A solution to this is the use of statistical models and analytical tools. In this presentation, we give a brief review of the use of statistics to improve health care decision-making under uncertainty. Case studies from the Hellenic National Organization for Healthcare Services Provision (NOHSP) will also be presented.
An Object Oriented Approach to the Analysis of Complex Data
Young statistician award
Alessandra Menafoglio (Politecnico di Milano)
Modern business and industrial applications often involve Big Data. I will focus on a few methodologies that allow analyzing Big Data when “Big” refers more to the data complexity than to the sample size. I will consider an object oriented approach, whose foundational idea is to interpret the data point – regardless of their complexity – as the “atom” of the analysis. The statistical analysis of these atoms is then performed by considering them as points within an appropriate mathematical space. I will illustrate a few object oriented methods addressing the problems of analyzing Big Data from business and industrial application, with a particular attention to the possible (spatial) dependence among data and to their constraints. I will finally discuss a recent object-oriented method for statistical meta-modeling.
Missing Data Imputation in Multivariate Data Sets
Young statistician award
Abel Folch-Fortuny (DSM Food Specialties B.V.)
One of the big challenges in data analysis in any research and industrial field is how to deal with missing values. In multivariate model building stages, when dealing with biological, economics or business databases, practitioners usually deal with 10-30% of missing values. In complex industrial processes, where hundreds of variables are collected per batch, 30–60% of missing data can appear in their historical data sets. Finally, with the paradigm of big data, thousands of variables are collected for large sets of individuals, having sometimes more than 70% of missing values in their databases.
A framework for missing data imputation in exploratory and predictive models will be given in this talk, together with a comparative study with state-of-the art techniques. These new algorithms have been included in a freely available software for missing data imputation, including a graphical user-friendly interface. Furthermore, the use of these new algorithms has been extended to the reconstruction of biological networks, curating data for missing values and faulty observations; and to calibration transfer between near infrared spectrometers, reconstructing unmeasured spectra using imputation methods.