Restoration of missing environmental data in an air quality monitoring system based on a naive Bayes classifier

• Date: 2026
• Main Authors: Shevchenko, Dmytro, Holub, Bella, Trysnyuk, Taras
• Format: Article
• Language: Ukrainian
• Published: Kyiv National University of Construction and Architecture 2026
• Subjects: Data Mining; K-Means; data restoration; information and analytical system; intelligent technology; information technologies; data quality; data reliability and validity
• Online Access: https://es-journal.in.ua/article/view/365072
• Journal Title: Environmental safety and natural resources

The article examines an approach to restoring missing environmental data in an air quality monitoring system based on a Naive Bayes classifier. The relevance of the study is determined by the fact that missing values in observational time series reduce the reliability of air quality index calculations, complicate the interpretation of environmental conditions, and weaken the analytical support of managerial decision-making. The study is a logical continuation of previous research in which stable and representative monitoring stations suitable for forming a high-quality training dataset were identified using cluster analysis. In contrast to approaches that use the entire set of available measurements without considering their reliability, the proposed method involves training the model only on data from selected stations characterized by higher completeness, stability, and credibility of time series.The study forms a feature space based on concentrations of major pollutants and accompanying meteorological parameters, uses CAQI categories as the target variable, and implements a procedure for restoring missing values according to the most probable air quality class. The obtained results confirmed the acceptable quality of the constructed model: the overall classification accuracy reached 0.71, which indicates the suitability of the approach for basic air quality assessment and its further use in intelligent data restoration tasks.The practical value of the proposed approach lies in its potential integration into environmental information and analytical systems in order to improve data completeness, enhance the quality of air quality index calculations, and provide more reliable analytical support for decision-making.