Peramalan Konsentrasi Zat Pencemar Udara (PM₁₀, NO₂, Dan SO₂) Di Kota Surabaya Dengan Metode Hybrid SARIMA-LSTM

Kalila, Fathia Zahrani (2026) Peramalan Konsentrasi Zat Pencemar Udara (PM₁₀, NO₂, Dan SO₂) Di Kota Surabaya Dengan Metode Hybrid SARIMA-LSTM. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5003221177-Undergraduate-Thesis.pdf - Accepted Version Restricted to Repository staff only Download (13MB) | Request a copy

Abstract

Kualitas udara di kota besar seperti Surabaya menjadi isu krusial akibat tingginya aktivitas industri dan volume lalu lintas yang menghasilkan polutan berbahaya seperti PM10, NO2, dan SO2. Data konsentrasi ketiga polutan tersebut memiliki karakteristik kompleks berupa pola musiman harian serta dinamika nonlinier yang sulit ditangkap oleh metode statistik konvensional. Penelitian ini bertujuan mengembangkan dan mengevaluasi model peramalan hybrid yang memadukan model linier Seasonal Autoregressive Integrated Moving Average (SARIMA) dengan Long Short-Term Memory (LSTM) untuk meningkatkan akurasi peramalan konsentrasi PM10, NO2, dan SO2 di Kota Surabaya. Data yang digunakan adalah data konsentrasi per jam yang diperoleh melalui Google Air Quality API pada stasiun pemantauan Balongsari periode 8 September-30 November 2025, dengan data 8 September-16 November 2025 sebagai in-sample dan 17-30 November 2025 sebagai out-sample. Pemodelan dilakukan menggunakan model SARIMA sebagai model linier, serta dua pendekatan hybrid, yaitu pendekatan pertama yang memodelkan residual model linier dan pendekatan kedua yang memodelkan konsentrasi aktual menggunakan lag dari model linier sebagai input LSTM. Performa model dievaluasi menggunakan metrik MAE, RMSE, dan MAPE. Hasil penelitian menunjukkan bahwa penambahan komponen LSTM menurunkan nilai MAE dan RMSE dibandingkan model SARIMA tunggal pada seluruh polutan, meskipun berdasarkan MAPE penurunan tersebut tidak terjadi pada pendekatan kedua untuk PM10. Model terbaik berbeda untuk tiap polutan, yaitu SARIMA-LSTM pendekatan pertama untuk PM10 (MAPE 21,2246%) serta SARIMA-LSTM pendekatan kedua untuk NO2 (MAPE 7,2404%) dan SO2 (MAPE 6,9540%), sehingga pendekatan hybrid terbaik bergantung pada karakteristik data masing-masing polutan. Peramalan untuk periode 1-14 Desember 2025 menunjukkan konsentrasi ketiga polutan berada di bawah baku mutu udara ambien PP No. 22 Tahun 2021, sehingga kualitas udara Kota Surabaya pada periode tersebut diprakirakan masih terkendali.
=======================================================================================================================================
Air quality in major cities such as Surabaya is a crucial issue, as high industrial activity and traffic volume generate harmful pollutants such as PM₁₀, NO₂, and SO₂. Concentration data for these pollutants exhibit complex characteristics, namely daily seasonal patterns and nonlinear dynamics that conventional statistical methods struggle to capture. This study aims to develop and evaluate hybrid forecasting models that combine the linear Seasonal Autoregressive Integrated Moving Average (SARIMA) model with Long Short-Term Memory (LSTM) to improve the forecasting accuracy of PM₁₀, NO₂, and SO₂ concentrations in Surabaya. Hourly concentration data were obtained through the Google Air Quality API at the Balongsari monitoring station from 8 September to 30 November 2025, with 8 September–16 November 2025 as the in-sample and 17–30 November 2025 as the out-sample. SARIMA was used as the linear model, together with two hybrid approaches: the first models the residuals of the linear model, while the second models the actual concentration using lags derived from the linear model as LSTM inputs. Model performance was evaluated using MAE, RMSE, and MAPE. The results show that adding the LSTM component reduced MAE and RMSE relative to the single SARIMA model for all pollutants, although based on MAPE this reduction did not occur for the second approach on PM₁₀. The best model differed for each pollutant, namely the first SARIMA-LSTM approach for PM₁₀ (MAPE 21.2246%) and the second SARIMA-LSTM approach for NO₂ (MAPE 7.2404%) and SO₂ (MAPE 6.9540%), so the best hybrid approach depends on the data characteristics of each pollutant. Forecasts for 1-14 December 2025 show that the concentrations of all three pollutants remain below the ambient air quality standards of Government Regulation No. 22 of 2021, so Surabaya's air quality in that period is projected to remain under control.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Kualitas Udara, LSTM, Peramalan, SARIMA, Surabaya, Air Quality, Forecasting, LSTM, SARIMA, Surabaya
Subjects:	Q Science > QA Mathematics > QA276 Mathematical statistics. Time-series analysis. Failure time data analysis. Survival analysis (Biometry) Q Science > QA Mathematics > QA280 Box-Jenkins forecasting Q Science > QA Mathematics > QA76.87 Neural networks (Computer Science)
Divisions:	Faculty of Science and Data Analytics (SCIENTICS) > Statistics > 49201-(S1) Undergraduate Thesis
Depositing User:	Fathia Zahrani Kalila
Date Deposited:	03 Aug 2026 04:50
Last Modified:	03 Aug 2026 04:50
URI:	http://repository.its.ac.id/id/eprint/141673

Actions (login required)

View Item