Bidirectional LSTM for Air Quality Forecast

An increase in the number of industries and other human activities have raised the level of contamination of surrounding air. It is imperative to build good prediction models with high accuracy in air quality forecast. In spite of the efforts in research, identifying suitable models with high accuracy is still a challenge especially for locations with less advanced monitoring stations using low cost sensors. Data collected in such situation are prone to missing data in large intervals and the possibility of anomalies is high. This study proposes a bidirectional LSTM (BiLSTM) for the forecast of fine particulate matter pollutant in the air. The performance of a stacked BiLSTM based on the Relu function is compared with two other revised models. The proposed model presented a better performance than the other two models in cases of better learning dependencies and characteristics of the data.