Peramalan Spatio-temporal Kualitas Udara di DKI Jakarta Menggunakan Hybrid Convoutional Neural Network dengan Long Short-Term Memory

Nareswari, Ina Tantri (2024) Peramalan Spatio-temporal Kualitas Udara di DKI Jakarta Menggunakan Hybrid Convoutional Neural Network dengan Long Short-Term Memory. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5003201045-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only until 1 October 2026. Download (6MB) | Request a copy

Abstract

Pertumbuhan ekonomi yang pesat dan tingkat aktivitas industri yang tinggi di DKI Jakarta meningkatkan kadar polusi udara. Oleh karena itu, peramalan kualitas udara yang akurat diperlukan dalam pengelolaan polusi yang efektif. Data polusi yang memiliki pola abstrak memerlukan metode peramalan yang mampu menangani ketidaklinieran data, contohnya seperti metode machine learning. Penelitian mempertimbangkan metode hybrid Convolutional Neural Network dengan Long Short-Term Memory (CNN-LSTM) untuk meramalkan lima parameter pembentuk kualitas udara, yakni PM10, NO2, SO2, CO, dan O3. Model LSTM digunakan untuk mengatasi dependensi temporal, sementara CNN dimanfaatkan untuk mengatasi dependensi spasial. Variasi hanya ditangkap dari data historis tanpa penambahan variabel eksogen. Data yang digunakan diambil dari lima stasiun pemantauan kualitas udara di Provinsi DKI Jakarta yang tercatat harian dalam rentang 1 Januari 2013 hingga 31 Desember 2021. Pemilihan model optimal didasarkan pada nilai Root Mean Square Error (RMSE) dan Mean Average Percentage Error (MAPE) terkecil. Nilai yang hilang diatasi dengan Kalman Smoothing. Analisis karakteristik data awal menunjukkan bahwa PM10 dan O3 merupakan parameter yang mendominasi sebagai parameter kritis harian. Dari hasil penelitian, model CNN-LSTM mampu untuk meramalkan data kualitas udara di DKI Jakarta sepanjang 90 langkah. Akan tetapi, diperoleh kesimpulan bahwa model yang lebih kompleks tidak selalu memberikan akurasi yang lebih tinggi jika dibandingkan dengan model yang lebih sederhana. Menggunakan epochs = 100, peramalan menunjukkan kualitas udara di Jakarta pada rentang 3 Oktober 2021 hingga 31 Desember 2021 cenderung masuk ke dalam kategori sehat atau sedang. Pada kelima stasiun, nilai-nilai polutan tertinggi harian didominasi oleh PM10. Peramalan ini diharapkan dapat mendukung pengambilan keputusan terkait manajemen polusi udara di Provinsi DKI Jakarta.
==============================================================================================================================
The rapid economic growth and high level of industrial activity in DKI Jakarta have increased air pollution levels. Therefore, accurate forecasting air quality is necessary for effective pollution management. Air pollution data, which often have abstract patterns, require forecasting methods capable of handling data nonlinearity, such as machine learning methods. This research considers a hybrid Convolution Neural Network with Long Short-Term Memory (CNN-LSTM) method to forecast five air quality parameters, namely PM10, NO2, SO2, CO, and O3. The LSTM model is used to address temporal dependencies, while CNN is utilized to handle spatial dependencies. Variations are captured only from historical data without adding exogenous variables. The data used is taken from five air quality monitoring stations in DKI Jakarta Province, recorded daily from January 1, 2013, to December 31, 2021. The optimal model selection is based on the smallest Root Mean Square Error (RMSE) and Mean Average Percentage Error (MAPE) values. Missing values were addressed using Kalman Smoothing. Initial data characteristic analysis showed that PM10 and O3 are dominant as critical daily parameters. The research results indicate that the CNN-LSTM model can forecast air quality data in DKI Jakarta for up to 90 steps. However, it was concluded that more complex models do not always provide higher accuracy compared to simpler models. Using 100 epochs, the forecast showed that air quality in Jakarta from October 3, 2021, to December 31, 2021, tended to fall into the healthy or moderate category. At the five stations, the highest daily pollutant values were dominated by PM10. This forecast is expected to support decision-making related to air pollution management in DKI Jakarta Province.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	hybrid CNN-LSTM, peramalan nonlinier, polusi udara, spatio-temporal, time series, air pollution, nonlinear forecasting
Subjects:	Q Science > QA Mathematics > QA276 Mathematical statistics. Time-series analysis. Failure time data analysis. Survival analysis (Biometry) Q Science > QA Mathematics > QA76.87 Neural networks (Computer Science) Q Science > QC Physics > QC866.5 Climatology--Forecasting.
Divisions:	Faculty of Science and Data Analytics (SCIENTICS) > Statistics > 49201-(S1) Undergraduate Thesis
Depositing User:	Ina Tantri Nareswari
Date Deposited:	27 Aug 2024 05:12
Last Modified:	27 Aug 2024 05:12
URI:	http://repository.its.ac.id/id/eprint/114980

Actions (login required)

View Item