Enhancing Decision Support Systems in Smart Systems by Using Advanced AI Tools and Blockchain

Abstract

Decision Support Systems (DSSs) are computer-based systems used to enhance decision-making in various fields, including transportation and education. These systems take into consideration all alternatives involved in decision-making, ensuring that decisions are made accurately and effectively. However, decision-making can be challenging due to the involvement of multiple criteria and large amounts of data, which must be carefully considered. These factors can negatively impact decisions, affecting business or company goals.

This dissertation introduces advanced Artificial Intelligence (AI) tools and blockchain to address decision-making issues based on two case studies: one in smart transportation and the other in smart education. This dissertation demonstrates the effectiveness of MTL, sentiment analysis, and blockchain in enhancing decision-making in education and transportation smart systems.

The first case study examines how Multi-Task Learning (MTL) enhances smart systems, particularly in the context of smart transportation, across various categories. Numerous related tasks in smart transportation require efficient handling. MTL can help train these multiple associated tasks together, such as traffic flow and speed, sharing their features within one model, making the model more robust, efficient, and scalable. This means that it considers all possible patterns and representations among these tasks. Additionally, MTL helps train new tasks with a pre-trained model instead of training them from scratch. This leads to more accurate predictions and decision-making.

The second case study aims to illustrate how sentiment analysis and blockchain with smart contracts can enhance the accuracy of decision-making in smart systems, particularly in improving the scholarship approval process. There are multiple factors in the scholarship approval process, so using smart contracts with blockchain can enhance decision-making by automating the process. Blockchain, based on the IBFT 2.0 consensus, is used to manage and record student transcripts, supplementary documents, scholarship decisions, and tuition and salary distribution in a transparent, immutable, and auditable way. Additionally, the DistilRoBERTa transformer learning model is utilized for sentiment analysis to address the limitation of unstructured data and enhance its detection (e.g., students supplementary documents) to determine whether the tone is positive, negative, or neutral. This enhances prediction accuracy and, consequently, decision-making in the scholarship process by assigning a confidence score and label to each student document.