Evidence record · UDINUS · SDGs
- Unit
- Rektorat UDINUS
- Period
- 2024
- Goals
- Record
- 1152b71a-59fa-4140-8856-f29cce97b1d8
- Published
- 2026-10-30
Universitas Dian Nuswantoro (UDINUS), through participating researchers Anggraini, Lilis; Noersasongko, Edi; Purwanto; and Marjuni, Aris, conducted an evaluation of time series forecasting techniques for predicting nitrogen dioxide (NO2) concentration. The activity addressed air quality as an important factor for human life because a decrease in air quality has a major impact on human health. NO2 was examined as one of the main pollutants that has a direct effect on human health.
The activity assessed the performance of three models for time series forecasting: AutoRegressive Integrated Moving Average (ARIMA), eXtreme Gradient Boosting (XGBoost), and Long Short-Term Memory (LSTM). The data used consisted of air quality data from Beijing over the period 2013 to 2024. The research methodology included data pre-processing to address missing values and reduce noise. After the pre-processing stage, ARIMA, XGBoost, and LSTM were applied to predict NO2 concentration. The parameters were selected through a cross-validation process, using the most maximized parameters identified through that process.
Based on the test results, XGBoost had the best performance among the evaluated models. This performance was evidenced by the lowest prediction error values for RMSE, MAPE, and MAE. ARIMA and LSTM also provided good results, but their performance remained below that of XGBoost. The results further indicated that XGBoost was able to capture fluctuations in NO2 concentration more accurately. The findings therefore described XGBoost as superior in modelling the complexity of air quality time series data.
The activity is connected with a health-related research focus because it examined a pollutant stated to have a direct effect on human health. The results are expected to contribute to the development of better air quality prediction strategies and to support more informed decision making for public health improvement. Through this research activity, UDINUS presents an institutional contribution based on the evaluation of forecasting methods for NO2 concentration and the comparison of their prediction performance.