Penerapan Model Hybrid ARIMA–Temporal Convolutional Network (ARIMA–TCN) dan ARIMA–Long Short-Term Memory (ARIMA–LSTM) dalam Peramalan Anomali Temperatur Global

Satyatama, Rasendra Akbar (2026) Penerapan Model Hybrid ARIMA–Temporal Convolutional Network (ARIMA–TCN) dan ARIMA–Long Short-Term Memory (ARIMA–LSTM) dalam Peramalan Anomali Temperatur Global. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5003221120-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (4MB) | Request a copy

Abstract

Peramalan anomali temperatur global merupakan aspek penting dalam memahami dinamika perubahan iklim serta mendukung upaya antisipasi terhadap kecenderungan peningkatan suhu global. Karakteristik data anomali yang bersifat kompleks dan non-linear mendorong penggunaan pendekatan hybrid yang mengombinasikan metode statistik klasik dan deep learning. Penelitian ini bertujuan untuk mendapatkan serta membandingkan kinerja model terbaik Autoregressive Integrated Moving Average–Long Short-Term Memory (ARIMA–LSTM) dan Autoregressive Integrated Moving Average–Temporal Convolutional Network (ARIMA–TCN) dalam peramalan anomali temperatur global. Data yang digunakan berupa deret waktu anomali temperatur global yang bersumber dari dataset iklim internasional. Model ARIMA digunakan untuk menangkap pola linier, sedangkan LSTM dan TCN diterapkan pada residual ARIMA untuk memodelkan hubungan nonlinier. Evaluasi kinerja peramalan dilakukan pada data out-sample menggunakan metrik Root Mean Square Error (RMSE) dan Mean Absolute Percentage Error (MAPE). Hasil penelitian menunjukkan bahwa model hybrid ARIMA–TCN memberikan performa terbaik dengan nilai RMSE sebesar 0,14347 dan MAPE sebesar 11,914, lebih rendah dibandingkan model ARIMA–LSTM yang menghasilkan RMSE sebesar 0,19609 dan MAPE sebesar 14,201. Hasil peramalan menunjukkan bahwa nilai prediksi berada pada kisaran yang sejalan dengan pola anomali temperatur global pada akhir periode pengamatan dan tidak menunjukkan lonjakan ekstrem selama horizon peramalan, sehingga model hybrid ARIMA–TCN memberikan gambaran awal yang wajar mengenai perilaku anomali temperatur global pada periode mendatang.
====================================================================================================================================
Forecasting global temperature anomalies is an important aspect of understanding climate change dynamics and supporting anticipatory efforts toward increasing global temperatures. The complex and non-linear characteristics of anomaly data motivate the use of hybrid approaches that combine classical statistical methods with deep learning techniques. This study aims to get and compare the best performance of the Autoregressive Integrated Moving Average–Long Short-Term Memory (ARIMA–LSTM) and Autoregressive Integrated Moving Average–Temporal Convolutional Network (ARIMA–TCN) models in forecasting global temperature anomalies. The data used consist of a global temperature anomaly time series obtained from international climate datasets. The ARIMA model is employed to capture linear patterns, while LSTM and TCN are applied to the ARIMA residuals to model non-linear relationships. Forecasting performance is evaluated on out-of-sample data using the Root Mean Square Error (RMSE) and Mean Absolute Percentage Error (MAPE) metrics. The results indicate that the hybrid ARIMA–TCN model achieves the best performance, with an RMSE of 0.14347 and a MAPE of 11.914, which are lower than those of the ARIMA–LSTM model, which yields an RMSE of 0.19609 and a MAPE of 14.201. The forecasting results show that the predicted values remain within a range consistent with the pattern observed at the end of the historical period and do not exhibit extreme fluctuations over the forecasting horizon. Therefore, the hybrid ARIMA–TCN model provides a reasonable initial representation of global temperature anomaly behavior in future periods.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	anomali temperatur global, ARIMA–LSTM, ARIMA–TCN, model hybrid, peramalan, global temperature anomaly, ARIMA–LSTM, ARIMA–TCN, hybrid model, forecasting
Subjects:	Q Science Q Science > QA Mathematics Q Science > QA Mathematics > QA276 Mathematical statistics. Time-series analysis. Failure time data analysis. Survival analysis (Biometry)
Divisions:	Faculty of Science and Data Analytics (SCIENTICS) > Statistics > 49201-(S1) Undergraduate Thesis
Depositing User:	Rasendra Akbar Satyatama
Date Deposited:	29 Jan 2026 08:08
Last Modified:	29 Jan 2026 08:08
URI:	http://repository.its.ac.id/id/eprint/130967

Actions (login required)

View Item