Malliaridis Konstantinos (Phd Candidate)

Thesis title: AutoML tools for effortless data analysis
Supervisor: Ougiaroglou Stefanos
Advisory Committee Members:
Konstantinos Diamantaras, Professor, Dept. of Information and Electronic Engineering, IHU
Antonis Sidiropoulos, Associate Professor, Dept. of Information and Electronic Engineering, IHU

The present doctoral research proposal aims to promote the field of Knowledge Discovery and Machine Learning through the development of practical, web-based applications and services that automate the processes of Knowledge Discovery and Machine Learning model training. The lack of specialized human resources in data analysis compared to the increasing demand for such expertise poses a challenge. To address this gap, the creation of web-based tools that are easily accessible, eliminating the need for specialized knowledge in machine learning, programming, and specialized software, is proposed. The main focus is on automating the learning of association rules and frequent itemset mining, which will be accessible to non-expert users. Objectives include the implementation of algorithms, evaluating their performance and comparing them on different datasets, automatic characterization of datasets, automatic algorithm selection for each dataset, automatic parameter tuning, automatic elimination of redundant association rules, and ultimately, developing web applications and services through which non-expert users can perform association rule mining on their data. Additionally, the possibility of developing innovative web-based tools for automatic unsupervised learning and automated data preprocessing techniques will be explored. Overall, the research aims to “democratize” the field of data mining and machine learning, promoting its accessibility to a wide audience.