Decisions based on data-driven insights can be vital and have a phenomenal impact on the environment, society, and business. Many critical domains such as crisis management, predictive maintenance, mobility, public safety,and cyber-security have become increasingly disrupted by new means to harness the extreme proliferation of data for effective decision making.
However, extreme big data challenges data-driven analytics and decision-making methods due to the complexity of such data.
ExtremeXP aims to provide accurate, precise, fit-for-purpose, and trustworthy data-driven insights via evaluating different complex analytics variants, considering end users’ preferences and feedback in an automated way.
ExtremeXP will integrate cutting-edge research results from the domains of data integration, machine learning, visual analytics, explainable AI, decentralized trust, knowledge engineering, and model-driven engineering into a common framework.
The €10 million project has been funded by the European Union Horizon Program and will aim at handling the complexity of extreme big data by designing user-friendly tailored processes that place the end user at the centre of complex analytics processes, considering user intents and constraints on accuracy, time-to-answer, specificity, energy consumption etc.
These processes are packaged in the ExtremeXP’s framework that is envisioned as modular and extensible, orchestrating different services around an experimentation engine that brings together data integration, data & knowledge management, and transparent & interactive decision making.>
The framework will be validated on five pilot demonstrators related to critical domains such as crisis management, predictive maintenance, mobility, public safety, and cyber-security.
Developing trustworthy and continual machine learning
Bournemouth University, the only partner in the consortium with machine learning expertise, under the lead of Professor Hamid Bouchachia, will contribute to three major aspects:
- Original constraint-aware machine learning, i.e., new algorithms that accommodate various types of constraints (e.g., high sensitivity and specificity, high precision, physics-informed ML, etc.) addressing the needs of the pilots.
- Continual machine learning where algorithms can learn continuously over time in an incremental manner, as new data becomes available, while avoiding forgetting (retroactive interference with old data).
- Trustworthy machine learning algorithms by implementing explainability mechanisms for the complete workflow.
BU will develop these tools for at least two pilots: cybersecurity, where threat detection algorithms are proposed; and for natural hazards (crisis management) where physics and data-driven are reconciled to develop risk models.