Air Pollution Index Prediction Using Multiple Neural Networks

• Main Authors: Ahmad, Zainal, Rahim, Nazira Aniza, Bahadori, Alireza, Jie, Zhang
• Format: Article
• Language: English
• Published: IIUM Press, International Islamic University Malaysia 2017
• Subjects: TP155-156 Chemical engineering
• Online Access: http://eprints.usm.my/36781/1/%28AIR_POLLUITON_INDEX%29__684-3577-2-PB.pdf; http://eprints.usm.my/36781/; http://journals.iium.edu.my/ejournal/index.php/iiumej/article/view/684/451

Air quality monitoring and forecasting tools are necessary for the purpose of taking precautionary measures against air pollution, such as reducing the effect of a predicted air pollution peak on the surrounding population and ecosystem. In this study a single Feed-forward Artificial Neural Network (FANN) is shown to be able to predict the Air Pollution Index (API) with a Mean Squared Error (MSE) and coefficient determination, R2, of 0.1856 and 0.7950 respectively. However, due to the non-robust nature of single FANN, a selective combination of Multiple Neural Networks (MNN) is introduced using backward elimination and a forward selection method. The results show that both selective combination methods can improve the robustness and performance of the API prediction with the MSE and R2 of 0.1614 and 0.8210 respectively. This clearly shows that it is possible to reduce the number of networks combined in MNN for API prediction, without losses of any information in terms of the performance of the final API prediction model. Pemantauan dan ramalan kualiti udara adalah perlu bagi mengambil langkah berjaga-jaga terhadap pencemaran udara, seperti untuk meramalkan mengurangkan kesan puncak pencemaran udara terhadap penduduk sekitar dan ekosistem. Dalam kajian ini rangkaian tiruan tunggal neural suap depan (FANN) ditunjukkan masing-masing dapat meramalkan indek pencemaran udara (IPU) dengan purata ralat kuasa dua (MSE) dan pekali penentuan, R2, daripada 0.1856 dan 0.7950. Namun disebabkan oleh sifat tidak mantap FANN tunggal, gabungan terpilih pelbagai rangkaian neural (MNN) diperkenalkan dengan menggunakan penghapusan ke belakang dan kaedah pemilihan ke hadapan. Keputusan kajian menunjukkan bahawa kedua-dua kaedah gabungan terpilih boleh meningkatkan keteguhan dan prestasi ramalan API masing-masing dengan MSE dan R2 daripada 0.1614 dan 0.8210. Ini jelas menunjukkan bahawa ia adalah mungkin untuk mengurangkan bilangan rangkaian digabungkan dalam MNN untuk ramalan API, tanpa menjejaskan keupayaan mana-mana maklumat dari segi prestasi model ramalan akhir API.