Institutional research project

An efficient multimodal environment for data analysis and anomaly detection using machine learning methods in time-critical applications (ANO-ML)

Area

STEM

Duration

48 months

Total funds

137,220.00 EUR

Abstract

The research within the project aims to develop an efficient multimodal computing environment for data analysis and anomaly detection using scientifically innovative approaches using machine learning and deep learning procedures and models adapted for work in time-critical applications.

A custom dataset from various sources will be created and adapted, and a procedure for solving the problem of data balance and partial labeling will be proposed.

It will be used to train our own and improve pre-trained machine and deep learning models that will be used in a multimodal system for data analysis, detection and visualization of anomalies based on data of different levels of structure coming from sensors and visual cameras.

System improvements will be implemented based on advanced data fusion procedures, efficient resource management in data analysis, and the application of energy-efficient data processing principles at the edge of the network based on clusters of single-board computers and a high-performance computing environment.

A blockchain architecture solution will be developed to improve data security, integrity and traceability. Such solutions have applications in a range of areas, from fire protection, environmental hazard recognition in general, medicine and various industries.

The entire system will be implemented in software and hardware in the form of a prototype, and tested and analyzed.

Main objectives

The main goals of the project are to develop an efficient multimodal environment for data analysis and anomaly detection using machine learning with an emphasis on deep learning, approaches for analyzing data of various structures, and data flow analysis, task handling procedures for real-time work, and energy-aware approaches for data processing in parallel and distributed environments.

Specific objectives

  1. Create and adapt datasets and develop efficient machine learning models for processing spatial and temporal data from different sources
  2. Model and develop a multimodal system for anomaly detection and visualization on heterogeneous data using structured data, image and data stream processing methods
  3. Improve and implement system improvements with the aim of applicability in time-critical applications, and increasing security and privacy and accuracy of machine learning models
  4. Create and evaluate data analysis architecture based on proposed approaches and solutions with the aim of improving computational, data and energy efficiency and adapting to execution in parallel and distributed computing environments
  5. Implement blockchain-based solutions to ensure confidentiality, integrity and transparency of data in the learning process
  6. Increase scientific excellence and international recognition through publication of results in WoS/SCOPUS journals and active involvement of young researchers in all phases of the project

Istraživački tim

izv. prof. dr. sc. Josip Balen

Leader on work packages 2 and 3, co-leader on 1

izv. prof. dr. sc. Alfonzo Baumgartner

Co-leader on work package 4

prof. dr. sc. Josip Job

Co-leader on work package 3

prof. dr. sc. Ivica Lukić

Co-leader on work package 6

Funded by the European Union - Next Generation EU. However, the views and opinions expressed are solely those of the author(s) and do not necessarily reflect those of the European Union or the European Commission. Neither the European Union nor the European Commission can be held responsible for them.