In the information age, data-driven decision-making has emerged as a critical paradigm for enhancing the intelligence and efficiency of modern information systems. This thesis explores the use of machine learning and its computational capabilities to ...
In the information age, data-driven decision-making has emerged as a critical paradigm for enhancing the intelligence and efficiency of modern information systems. This thesis explores the use of machine learning and its computational capabilities to analyze complex real-world data structures, aiming to support more informed and effective decision-making. Traditional decision-making methods, which are often based on static rules, human intuition, or technical experts, are increasingly struggling to cope with the scale, complexity, and dynamism of contemporary data environments. Machine learning (ML) offers a robust solution by enabling systems to identify patterns, extract meaningful insights, and adapt to non-stationary environments through the analysis of large-scale, heterogeneous datasets collected from real-world operational settings, thereby enhancing generalization, robustness, and real-world adaptability.
The thesis explores the application of machine learning techniques—including text classification, linear lasso regression, and multimodal models that integrate image clas-
sification and multilayer perceptron—to real-world scenarios in order to support system optimization, predictive analytics, and automated decision-making. The first work proposes an intelligent marketing platform based on media text classification to better match influcers with advertisers. The second work performs linear lasso regression to predict kidney function of deceased donor kidney transplant recipients to help understand disease outcomes. The third work also develops a multimodal model that combines convolutional neural network and multilayer perceptron for the auxiliary diagnosis of brain atrophy, which helps clinicians automatically detect brain atrophy in the young population.
By integrating ML, information systems can evolve from passive, rule-based tools into intelligent platforms capable of proactive, adaptive, and autonomous decision-making. This transformation is driven by the ability of ML algorithms to learn from complex, real-world data and respond dynamically to changing conditions. The study highlights the challenges, and strategies of ML-driven decision-making in various domains, such as marketing, and healthcare, ultimately demonstrating its transformative potential for data-intensive environments.