AI-based university KPI system: React.js, ASP.NET Core, and PostgreSQL architecture

This article presents a layered architecture for a system designed to collect, analyze, and generate recommendations based on university employees’ Key Performance Indicators (KPIs). The technology stack includes React.js, ASP.NET Core 8, PostgreSQL 16, and a Python-based AI service. The ETL process operates on a star-schema data model under the condition Q_period ≥ 0.90. Anomaly detection is performed using Isolation Forest combined with forecasting residuals (AUC = 0.94), while recommendations are constrained by a weighted gap-based rule. The prototype achieved a p95 response time of 0.42 s and a System Usability Scale (SUS) score of 78, while reducing report preparation time from 36 hours to 9 hours. A human-in-the-loop approach is maintained: the AI system provides recommendations but does not issue commands.

Hali tarjima qilinmagan