Tentang SmartLib Bantuan Pengguna

Skripsi S1

Record Data Skripsi S1

MhswID :

210401181

Judul :

INTEGRASI SHAP – BASED EXPLAINABLE AI PADA MODEL HYBRID HHT-LSTM UNTUK INTERPRETASI PERAMALAN KEDATANGAN WISATAWAN

Penulis :

SYAHRUL NIZAM

Abstrak :

Abstrak. Peramalan kedatangan wisatawan merupakan aspek penting dalam mendukung pengambilan keputusan pada sektor pariwisata. Model berbasis deep learning, seperti Long Short-Term Memory (LSTM), memiliki kemampuan yang baik dalam memprediksi data deret waktu, tetapi masih memiliki keterbatasan dalam interpretasi hasil prediksi karena bersifat black-box. Penelitian ini bertujuan mengintegrasikan SHapley Additive exPlanations (SHAP) pada model Hybrid Hilbert-Huang Transform–Long Short-Term Memory (HHT-LSTM) untuk meningkatkan interpretabilitas peramalan kedatangan wisatawan mancanegara ke Indonesia. Data yang digunakan berupa data kedatangan wisatawan dari Badan Pusat Statistik (BPS) dan data Google Trends periode Januari 2008 hingga Desember 2025. Metode penelitian meliputi pra-pemrosesan data, pembentukan variabel delay, dekomposisi sinyal menggunakan Hilbert-Huang Transform (HHT), pemodelan menggunakan LSTM, serta interpretasi model menggunakan SHAP. Hasil penelitian menunjukkan bahwa model Hybrid HHT-LSTM memperoleh nilai MAE sebesar 157.717,04, RMSE sebesar 191.152,97, dan MAPE sebesar 15,43%, lebih baik dibandingkan model LSTM standar yang memiliki MAPE sebesar 18,97%. Integrasi SHAP juga mampu memberikan interpretasi global dan lokal sehingga hasil prediksi menjadi lebih transparan dan mudah dipahami. Kata kunci: HHT-LSTM, SHAP, peramalan, wisatawan, interpretabilitas.

Abstrak Inggris :

Abstract. Forecasting tourist arrivals is an important aspect in supporting decision-making in the tourism sector. Deep learning models, such as Long Short-Term Memory (LSTM), have shown good performance in time-series forecasting; however, they still suffer from limited interpretability due to their black-box nature. This study aims to integrate SHapley Additive exPlanations (SHAP) into a Hybrid Hilbert-Huang Transform–Long Short-Term Memory (HHT-LSTM) model to improve the interpretability of international tourist arrival forecasting in Indonesia. The data used in this study consist of tourist arrival data obtained from Statistics Indonesia (BPS) and Google Trends data from January 2008 to December 2025. The research methodology includes data preprocessing, delay variable formulation, signal decomposition using Hilbert-Huang Transform (HHT), forecasting using LSTM, and model interpretation using SHAP. The results show that the Hybrid HHT-LSTM model achieved an MAE of 157,717.04, an RMSE of 191,152.97, and a MAPE of 15.43%, outperforming the standard LSTM model, which obtained a MAPE of 18.97%. Furthermore, the integration of SHAP provides both global and local interpretations, making the forecasting results more transparent and easier to understand. Keywords: HHT-LSTM, SHAP, forecasting, tourist arrivals, interpretability.

Tahun :

2026

Cover :

210401181-cover-20261015154738636.pdf

Bab 1 :

210401181-bab1-20261015154738907.pdf

Bab 2 :

210401181-bab2-20261015154738957.pdf

Bab 3 :

210401181-bab3-20261015154739013.pdf

Bab 4 :

210401181-bab4-20261015154739066.pdf