Decision Support Methods And Models Group

DSMM Group: Expertise

Research on Decision Support Systems and Machine Learning concerns the development of innovative solutions for supporting decision making processes and for building models able to forecast the behavior of complex systems. Two different approaches are followed to pursue this aim: on one hand, decision processes are structured and organized to ensure the respect of basic requirements that improve the quality of their application. On the other hand, specific technologies able to extract knowledge from data in intelligible form are designed and implemented to provide the expert with an automatic and comprehensive tool for provisional purposes and what-if analysis.
Although a variety of different statistical and machine learning techniques is considered, rule generation methods are privileged due to their ability to describe the resulting model in term of intelligible if-then rules. This allows to shed light on the behavior of the system of interest and to automatically determine relevant variables with a quantitative measure of their importance.
The increasing complexity of systems and data to be managed, as well as the need for reasonable processing times and for a timely response to critical events, make it useful and often necessary the resort to parallel and distributed computing systems. To cope with these issues, we develop methodologies and algorithms for high performance computing on multi-core/many-core architectures, which combine high performance with low cost, low power consumption and small size. In particular, optimized computational kernels for machine learning applications and big data analysis are implemented.
The considered methodologies are applied in several different fields, among which, bioinformatics, medical image processing, telecommunication, optimal control and territorial governance (i.e. environmental assessment and sustainable mobility), where the use of DSS is in particular targeted to foster the stakeholders involvement (e-democracy).

Topics:

  • Multi attribute analysis techniques with sensitivity and conflict analysis
  • Planning and management of innovative transport systems
  • Structuring and organizing the public decision process
  • Development of intelligent systems that learn from past experience
  • Realization of artificial models capable of extracting knowledge from data
  • Rule generation and statistical analysis of data
  • Solution of pattern recognition and regression problems
  • Feature selection
  • Control of complex systems and reliability analysis
  • Design of experiments in deterministic learning
  • Implementation of machine learning models in small devices such as FPGA and microcontrollers
  • Integrated tools and methods for the specification and validation of complex, adaptive, or critical systems
  • Parallel and distributed computing support to complex and critical systems
  • Development of methodologies, algorithms and applications for High Performance Computing  on  heterogeneous  platforms
  • General-Purpose Computation on Graphics Processing Units (GPGPU)
  • Using distributed computing platforms (Grids, Clouds) for QoS-demanding applications
  • Big data analysis
  • Model checking techniques
  • Methods and models of parallel computing based on formal language theory

Applications:

  • Design, implementation and application of decision-support software tools
  • Technical support to strategic environmental assessment processes of plans at various scales
  • Assistance to negotiation processes involving multiple actors
  • Development of sustainability frameworks of complex programs
  • Design, test, validation and/or implementation of mobility management measures as bike sharing, dial a ride, carpooling, info traveler systems, road pricing
  • Analysis of biomedical and genetic data
  • Support for the diagnosis of pathologies
  • Selecting relevant genes in DNA microarray experiments
  • Design and development of electronic noses
  • Smart bandwidth management in packet networks
  • Home automation for the elderly and disabled