Time-Series Prediction of Environmental Noise for Urban IoT Based on Long Short-Term Memory Recurrent Neural Network

Noise pollution is one of the major urban environmental pollutions, and it is increasingly becoming a matter of crucial public concern. Monitoring and predicting environmental noise are of great significance for the prevention and control of noise pollution. With the advent of the Internet of Things (IoT) technology, urban noise monitoring is emerging in the direction of a small interval, long time, and large data amount, which is difficult to model and predict with traditional methods. In this study, an IoT-based noise monitoring system was deployed to acquire the environmental noise data, and a two-layer long short-term memory (LSTM) network was proposed for the prediction of environmental noise under the condition of large data volume. The optimal hyperparameters were selected through testing, and the raw data sets were processed. The urban environmental noise was predicted at time intervals of 1 s, 1 min, 10 min, and 30 min, and their performances were compared with three classic predictive models: random walk (RW), stacked autoencoder (SAE), and support vector machine (SVM). The proposed model outperforms the other three existing classic methods. The time interval of the data set has a close connection with the performance of all models. The results revealed that the LSTM network could reflect changes in noise levels within one day and has good prediction accuracy. Impacts of monitoring point location on prediction results and recommendations for environmental noise management were also discussed in this paper.