Published Paper Details:

This study investigates the application of contrastive learning to Key Performance Indicator (KPI) analysis and dashboard optimization by integrating human psychology and user attention span modeling. Traditional KPI dashboards often fail to consider the cognitive load and interaction behavior of users, leading to inefficient decision-making. This research proposes a novel approach that leverages contrastive learning to derive KPI relationships based on time spent and user interaction patterns. By analyzing real-time engagement data, the study identifies high-focus and low-focus dashboard regions and dynamically reconfigures KPI layouts to enhance usability and minimize cognitive overload. Using a dataset of recorded user interactions, we train a contrastive learning model to cluster KPIs according to interaction relationships. Positive relationships are described as KPIs that are frequently accessed together, whilst negative pairs are KPIs that are rarely engaged together. The learned representations enable a dynamic dashboard adaptation architecture that optimizes KPI placement based on attention-driven heatmaps. The effectiveness of this technique is assessed via A/B testing, which measures gains in decision speed, accuracy, and user engagement. The findings show that contrastive learning-based KPI optimization significantly improves dashboard usability by prioritizing relevant information, reducing unnecessary cognitive load, and increasing user efficiency. This research advances the fields of business intelligence, human-computer interaction, and machine learning, paving the way for more intuitive and adaptive analytical dashboards.