Towards Re-identification of Expert Models: MLP-COMET in the Evaluation of Bitcoin Networks

In recent years, the application of complex network techniques has seen substantial progress, driven by enhanced computational capabilities and their efficacy in diverse practical domains. Central to complex network models is the assessment of nodes, often addressed through centrality measures. To face challenges in node evaluation, Multi-Criteria Decision Analysis (MCDA) methods, known for accommodating diverse preferences, are employed. MCDA techniques enable the assessment of decision alternatives against multiple criteria by adjusting criteria weights, reflecting the importance of various decision factors. Despite the advantages of expert knowledge in decision-making models, challenges include inaccessibility, inaccuracy of assessment, and limited reusability. To overcome these issues, this paper focuses on an approach that combines MCDA with a complex network technique, specifically Multi-Layer Perceptron (MLP). The study employs the Characteristic Objects Method (COMET) for evaluation, using MLP as an artificial expert to assess Characteristic Objects in the COMET method. The practical application focuses on assessing the Bitcoin network, showcasing contributions such as re-identifying decision models, an alternative methodology for decision processes without a domain expert, and applying MLP-COMET to analyze the Bitcoin network. Furthermore, the study explores various MLP structures and input parameter values to determine their influence on node evaluation stability within the Bitcoin network problem. By enhancing the stability and reliability of the decision model, this research equips decision-makers with more robust tools, fostering more effective and informed decision-making processes.