Research on air quality data optimization based on SW-KNN algorithm

The rapid industrialization in China has led to severe environmental issues, with air pollution—particularly PM2.5—attracting significant public and scientific attention due to its harmful effects on human health. Accurate prediction of PM2.5 concentrations plays a vital role in supporting sustainable development and public well-being. However, the reliability of such predictions largely depends on the quality of the input data. To address this, we propose a data optimization framework that enhances air quality time series data for improved predictive performance. First, the one-dimensional time series is transformed into a two-dimensional representation to retain temporal correlations, and the Isolation Forest algorithm is applied to detect and remove outliers. Then, considering the periodic characteristics of air quality data, we introduce a Sliding Window-based K-Nearest Neighbors (SW-KNN) approach to impute missing values by leveraging the average of similar sequence segments. Experimental results confirm the effectiveness of the proposed method in improving data quality and enhancing the accuracy of subsequent PM2.5 predictions.