AIR QUALITY INDEX PREDICTION IN BESIKTAS DISTRICT BY ARTIFICIAL NEURAL NETWORKS AND K NEAREST NEIGHBORS

Abstract

In this study, Air Quality Index (AQI) in Besiktas was intended to be predicted by Artificial Neural Networks (ANN) and k Nearest Neighbors (kNN) algorithms. For this purpose, eight parameters have been selected which may affect the AQI. These parameters are PM10, SO2, CO, O3, temperature, humidity, pressure and wind speed, respectively. 124 data for 2015, 2016, 2017 and 2018 January, which includes eight parameters, were determined as training data. The first 14-day data of January 2019 were determined as test data. Similarly, the first 14-day data of January, March and December of 2018 were used as test data. In addition, The first 14-day data for January 2019 were normalized and set as test data. The success of ANN and kNN were measured by comparing. Performance rate of ANN with raw data for January 2018 was 85.71%, for March 2018 was 71.43%, for December 2018 was 78.57%. Both with raw and with normalized data for January 2019 was 85.71% performance rate. Performance rate of kNN with raw data for January 2019 was 92.86%, for March 2018 was 28.57%, for December 2018 was 71.43%. Performance rate of kNN with normalized data for January 2019 was 92,86%.

Keywords

• Air Quality ındex,Artificial Neural Networks,k Nearest Neighbours,Classification

BEŞİKTAŞ’TAKİ HAVA KALİTESİ İNDEKSİNİN YAPAY SİNİR AĞLARI VE K EN YAKIN KOMŞULUK ALGORİTMALARI İLE TAHMİNİ

Abstract

Bu çalışmada, Beşiktaş’taki Hava Kalitesi İndeksinin (HKİ) Yapay Sinir Ağları (YSA) ve k En Yakın Komşuluk (kNN) algoritmaları ile tahmin edilmesi amaçlanmıştır. Bu amaçla HKİ’yi etkileyebilecek 8 adet parametre seçilmiştir. Bu parametreler sırasıyla PM10, SO2, CO, O3, sıcaklık, nem, basınç ve rüzgar hızı’dır. Bu parametreleri içeren 2015, 2016, 2017 ve 2018 yıllarının Ocak aylarına ait 124 adet veri eğitim verisi olarak belirlenmiştir. 2019 yılı Ocak ayına ait ilk 14 günlük veriler ile 2018 yılının Ocak, Mart ve Aralık aylarının ilk 14 günlük verileri ise test verisi olarak kullanılmıştır. 2019 yılı Ocak ayının ilk 14 günlük verisi normalize edililerek, ayrıca test verisi olarak ta kullanılmıştır. YSA ve kNN’nin sonuçları karşılaştırılarak başarıları ölçülmüştür. YSA’nın Ocak 2018 ham verileri ile başarım oranı % 85.71, Mart 2018 için % 71.43, Aralık 2018 için % 78.57 olmuştur. Hem ham hem de normalize edilmiş Ocak 2019 verileri için ise başarım oranı % 85,71 olmuştur. kNN nin Ocak 2019 ham verileri ile başarım oranı % 92.86, Mart 2018 için % 28.57, Aralık 2018 için % 71.43 olmuştur. kNN’nin normalize edilmiş Ocak 2019 verileri ile başarım oranı ise % 92.86 olmuştur.

Keywords

• Hava Kalitesi İndeksi,Yapay Sinir Ağları,k en Yakın Komşuluk,Sınıflandırma

References

1. Alkasassbeh, M., Sheta, A.F., Faris, H.,Turabieh, H., 2013. Prediction of PM10 and TSP Air Pollution Parameters Using Artificial Neural Network Autoregressive, External Input Models: A Case Study in Salt, Jordan. Middle-East Journal of Scientific Research. 14, 999-1009.
2. Artificial Neural Network, http://kod5.org/yapay-sinir-aglari-ysa-nedir/ [last access date: 02.02.2019].
3. AQI Calculator: [Online]. Available:https://app.cpcbccr.com/ccr_docs/AQI%20-Calculator.xls. [last access date: 01.12.2018].
4. Back propagation Algorithm: [Online]. Available:https://ab.org.tr/ab06/sunum/8.ppt. [last access date: 03.12.2018].
5. Baran, B., 2017. Yenilenebilir Enerji Kaynaklarını İçeren Mikro-şebeke Sistemlerin Akıllı Yönetimi, İnönü Üniversitesi, Fen Bilimleri Enstitüsü, Doktora Tezi, Malatya, Türkiye.
6. Baran B., 2019. Prediction of Air Quality Index by Extreme Learning Machines. International Conference on Artificial Intelligence and Data Processing (IDAP), Malatya, Turkey.
7. Barrero, M.A., Orza, J.A.G., Cabello, M., Cantón, L., 2015. Categorisation of air quality monitoring stations by evaluation of PM10 variability.Science of The Total Environment. 524–525: 225-236.
8. Besiktas, 2015, 2016, 2017, 2018, 2019 temperature, pressure, wind speed, humidity values. https://www.timeanddate.com/weather/turkey/besiktas/historic?month=1&year=2019. [Accessed 23 January 2019].

9. CO, O3. [Online]. Available:https://www.epa.gov. [last access date: 12.01.2019].
10. Directorate of Environment and Urbanization National Air Quality Monitoring Network. [Online]. Available:www.havaizleme.gov.tr. [last access date: 27.11.2019].
11. Dursun, Ş., İbrahimova İ., 2014. Bakü Hava Kirlenmesinde SO2’nin Rolü ve Meteorolojik Olaylarla İlişkisinin Araştırılması. Avrupa Bilim ve Teknoloji Dergisi. 1(3), 84-91.
12. Feed-forward Artificial Neural Network Figure. [Online]. Available:https://www.saedsayad.com/artificial_neural_network.htm.[last access date: 18.01.2019].
13. Feed-forward Back Propagation Neural Network. [Online]. Available:https://stackoverflow.com/questions/28403782/what-is-the-difference-between-back-propagation-and-feed-forward-neural-network.[last access date: 01.02.2019].
14. Kaplan, Y., Saray, U., Azkeskin, E., 2014. Hava Kirliliğine Neden olan PM10 ve SO2 maddesinin Yapay Sinir Ağı Kullanılarak Tahmininin Yapılması ve Hata Oranının Hesaplanması. Afyon Kocatepe Üniversitesi Fen ve Mühendislik Bilimleri Dergisi. 14(025201), 1-6.
15. Karacı, A., 2018. Development of PM2.5 Concentration Measurement Device for Intelligent City Air Tracking System And Asthma Diseases, Journal of Engineering Sciences and Design, 6(3), 418 – 425.
16. Karatzas, K., Katsifarakis, N., Orlowski, C., Sarzyński, A., 2017. Urban Air Quality Forecasting: A Regression and a Classification Approach.Asian Conference on Intelligent Information and Database Systems (ACIIDS 2017). 539-548.
17. kNN1 - k Nearest Neighbors Algorithm. [Online]. Available:https://medium.com/@ekrem.hatipoglu/machine-learning-classification-k-nn-k-en-yak%C4%B1n-kom%C5%9Fu-part-9-6f18cd6185d. [last access date: 15.11.2019].
18. kNN2 - k Nearest Neighbors Algorithm. [Online]. Available:https://www.saedsayad.com/k_nearest_neighbors.htm. [last access date:15.11.2019].
19. Koçak, E., 2018. Temporal Variation of PM10 and SO2 Concentrations of Aksaray Atmosphere: Conditional Bivariate Probability Function and Kmeans Clustering. Journal of Engineering Sciences and Design, 6(3), 471 – 478.
20. Kunt, F., Dursun, Ş., 2018. Konya Merkezinde Hava Kirliliğine Bazı Meteorolojik Faktörlerin Etkisi. Ulusal Çevre Bilimleri Araştırma Dergisi. 1(1), 54-61.
21. Menteşe, S., Tağıl, Ş., 2012. Bilecik’te İklim Elemanlarının Hava Kirliliği Üzerine Etkisi. Balikesir University The Journal of Social Sciences Institute. 15(28), 3-16.
22. Nejadkoorki, F., Baroutian, S., 2012. Forecasting Extreme PM10 Concentrations Using Artificial Neural Networks. Int. J. Environ. Res. 6(1), 277-284.
23. Rahman, N.H.A., Lee, M.H., Suhartono, L.M.T., 2015. Artificial neural networks and fuzzy time series forecasting: an application to air quality. Quality&Quantity International Journal of Methodology. 49(6), 2633–2647.
24. Neural Network Figure. [Online]. Available: https://blogs.mathworks.com/loren/2015/08/04/artificial-neural-networks-for-beginners/. [last access date:23.01.2019].
25. Saxena, A., Shekhawat, S., 2017. Ambient Air Quality Classification by Grey Wolf Optimizer Based Support Vector Machine. Journal of Environmental and Public Health. 12.
26. Sulfur Dioxide. [Online]. Available: www.environment.gov.au/protection/publications/factsheet-sulfur-dioxide-so2. [last access date:12.01.2019].
27. Tepe, A.M., Doğan, G., 2019. Investigation of Air Qualities of Four Cities Located on Southern Coast of Turkey. Journal of Engineering Sciences and Design, 7(3), 585 – 595.
28. Turgut, D., Temiz, İ., 2015. Time Series Analysis And Forecasting For Air Pollution in Ankara: A Box-Jenkıns Approach. alphanumericjournal. 3(2), 131-138.
29. XieY, Zhao B, Zhang L., Luo R., 2015. Spatiotemporal variations of PM2.5 and PM10 concentrations between 31 Chinese cities and their relationships with SO2, NO2, CO and O3.Particuology. 20, 141-149.
30. Yavuz, S., Deveci, M., 2012. İstatiksel Normalizasyon Tekniklerinin Yapay Sinir Ağın Performansına Etkisi. ErciyesÜniversitesi İktisadi ve İdari Bilimler Fakültesi Dergisi. 40(2), 167-187.