Monitoring air quality has become one of the most essential activities in IoT industrial and urban areas. The quality of air is adversely affected due to various forms of pollution caused by transportation, electricity, fuel uses. Air quality monitoring enables early detection of harmful pollutants, allowing for timely interventions to protect public health and improves the quality of life. However, traditional way of using fixed sensors cannot effectively provide a comprehensive view of air pollution in people’s surroundings. To overcome these issues, a novel GNN Model Based Air quality monitoring (GNN-MBA) has been proposed to enable real-time prediction of air quality and displays the quality results to user’s mobile devices. The proposed method utilizing the Variational Autoencoder (VAE) for feature extraction to efficiently capture features for accurate classification. The extracted features are fed into Graph Neural Network (GNN) deep learning model, which categorizes the data into three classes such as Good, Moderate and Poor, the result is then forwards to gateway. The gateway sends the processed output to WiFi network, making the air quality information accessible to the user’s mobile device. Measures including F1score (F1S), accuracy, precision and recall are utilized to assess the suggested approach. The GNN-MBA accuracy in the Beijing Air Quality dataset is 0.7%, 2.5%, and 3.2% greater than the current ICEEMDAN-WOA-ELM and SARIMA approaches, while its RMSE is decreased by 4.5%, 1.5%, and 2.0% respectively.