Show simple item record

dc.contributor.advisorErfiani
dc.contributor.advisorIndahwati
dc.contributor.authorYumna, Fadhilah
dc.date.accessioned2026-08-04T13:00:35Z
dc.date.available2026-08-04T13:00:35Z
dc.date.issued2026
dc.identifier.urihttp://repository.ipb.ac.id/handle/123456789/177066
dc.description.abstractPergerakan harga saham bersifat dinamis. Setiap emiten memiliki karakteristik pola deret waktu yang berbeda. Perbedaan karakteristik tersebut memerlukan metode yang mampu menggerombolkan saham berdasarkan kemiripan pola sebelum dilakukan peramalan. Penelitian ini bertujuan menggerombolkan emiten indeks LQ45 menggunakan metode k-medoids berbasis dynamic time warping (DTW). Penelitian ini juga membandingkan kinerja model autoregressive integrated moving average (ARIMA) dan gated recurrent unit (GRU) pada prototype setiap gerombol. Penelitian ini menggunakan model terbaik untuk meramalkan harga penutupan saham. Penelitian ini menggunakan data harga penutupan mingguan dari 27 emiten indeks LQ45 selama periode 2022–2026. Metode k-medoids berbasis DTW membentuk enam gerombol saham. Prototype median mewakili setiap gerombol. Model ARIMA dan GRU memodelkan setiap prototype. Penelitian ini mengevaluasi kinerja model menggunakan mean absolute percentage error (MAPE). Hasil penelitian menunjukkan bahwa ARIMA memberikan kinerja yang lebih baik pada prototype dengan pola pergerakan yang relatif stabil dan didominasi hubungan linear. Model GRU memberikan kinerja yang lebih baik pada prototype dengan pola pergerakan yang lebih kompleks dan hubungan nonlinier. Seluruh model terbaik menghasilkan nilai MAPE kurang dari 10%. Penggerombolan berbasis DTW mampu mengurangi kompleksitas data. Pemodelan pada prototype menghasilkan peramalan dengan tingkat akurasi yang baik.
dc.description.abstractStock price movements are dynamic. Each listed company exhibits distinct time series characteristics. These differences require a method that can group stocks based on pattern similarity before forecasting. This study aims to cluster companies listed in the LQ45 Index using the k-medoids method based on dynamic time warping (DTW). This study also compares the performance of the autoregressive integrated moving average (ARIMA) and gated recurrent unit (GRU) models on the prototype of each cluster. The best-performing model is then used to forecast stock closing prices. This study uses weekly closing price data from 27 companies consistently included in the LQ45 Index during the 2022–2026 period. The DTW-based k-medoids method produces six stock clusters. The median prototype represents each cluster. The ARIMA and GRU models are fitted to each prototype. Model performance is evaluated using the mean absolute percentage error (MAPE). The results show that ARIMA performs better on prototypes with relatively stable movement patterns dominated by linear relationships. The GRU model performs better on prototypes with more complex movement patterns and nonlinear relationships. All selected models achieve MAPE values below 10%. DTW-based clustering reduces data complexity. Prototype-based modeling also produces forecasts with good predictive accuracy.
dc.description.sponsorship
dc.language.isoid
dc.publisherIPB Universityid
dc.titlePenggerombolan Deret Waktu Harga Saham Indeks LQ45 Berbasis Dynamic Time Warping dengan Pemodelan ARIMA dan GRUid
dc.title.alternativeTime Series Clustering of LQ45 Index Stock Prices Based on Dynamic Time Warping with ARIMA and GRU Modeling
dc.typeSkripsi
dc.subject.keywordARIMAid
dc.subject.keywordDTWid
dc.subject.keywordGated Recurrent Unit (GRU)id
dc.subject.keywordK-Medoidsid
dc.subject.keywordperamalan sahamid
dc.subtypeUndergraduate Theses


Files in this item

Thumbnail
Thumbnail
Thumbnail

This item appears in the following Collection(s)

Show simple item record