A Confidence-Aware Framework for Semantic Alignment and Temporal Harmonization of Heterogeneous IoT Data in Smart Buildings

MohammadReza VelayatiFar¹ Karamollah Bagheri Fard² Razieh Malek Hosseini³

1) Department of Computer Engineering, Yasouj Branch, Islamic Azad University, Yasouj, Iran Email:
2) Department of Computer Engineering, Yasouj Branch, Islamic Azad University, Yasouj, Iran Email:
3) Department of Computer Engineering, Yasouj Branch, Islamic Azad University, Yasouj, Iran

IOI Code : ACCD-ACABAD
Publication : 10th International Conference on Applied Researches in Science & Engineering (carseconf.com)
Abstract :
    Heterogeneity in sensor naming conventions, measurement units, semantic models, and data acquisition rates makes the integration and unified analysis of Internet of Things (IoT) data challenging. This paper proposes a confidence-aware framework for semantic alignment and temporal harmonization of sensor data in smart buildings. The framework consists of four stages: (1) normalization of terminology and measurement units; (2) calculation of matching scores based on lexical similarity and dimensional, functional, and contextual consistency; (3) application of matching prohibition rules to address physical or semantic conflicts; and (4) alignment of time series on a common temporal grid, with limited gap filling. For preliminary evaluation, a controlled synthetic benchmark comprising 41 sensor concepts and a 14-day simulation involving four variables (temperature, humidity, CO₂ concentration, and power) was developed. Across 30 independent benchmark runs, the hybrid method achieved a mean F1-score of 0.798 with a standard deviation of 0.069, compared with 0.633 and 0.063, respectively, for lexical matching. In the temporal simulation, the proposed method reduced the mean absolute error (MAE) for temperature, humidity, CO₂ concentration, and power compared with a baseline based on unit conversion and grid alignment, while increasing data coverage to approximately 100%. However, these findings are based on synthetic data and simulation assumptions and should not be interpreted as empirical results obtained from RIoT or COLLECTiEF datasets.

Keywords : internet of things semantic alignment temporal harmonization data interoperability smart building