Electronic Theses and Dissertation

Pasar modal memiliki peran strategis dalam pembangunan ekonomi Indonesia. Indeks LQ45 menjadi indikator utama bagi kinerja saham-saham dengan likuiditas dan kapitalisasi pasar tinggi. Namun, analisis pergerakan harga saham menghadapi kompleksitas akibat hubungan antarsaham yang dinamis dan heterogen, terutama pada periode ketidakpastian pasar 2022–2026. Penelitian ini bertujuan untuk mengelompokkan saham-saham anggota Indeks LQ45 berdasarkan kemiripan pola deret waktu harga penutupan harian yang disesuaikan. Metode yang diterapkan adalah Agglomerative Hierarchical Clustering dengan empat metode linkage, yaitu single, complete, average, dan Ward’s method, serta ukuran jarak Constrained Dynamic Time Warping (DTW) menggunakan pembatas Sakoe-Chiba band dengan variasi window size 5%, 10%, dan 15% untuk mitigasi risiko pathological alignment. Data yang digunakan berupa harga harian dari Yahoo Finance untuk periode Agustus 2022 hingga Januari 2026. Hasil penelitian menunjukkan bahwa kombinasi average linkage dengan window size 5% menghasilkan representasi struktur jarak antarsaham terbaik berdasarkan nilai Cophenetic Correlation Coefficient. Evaluasi menggunakan indeks Silhouette dan Calinski–Harabasz menunjukkan bahwa tiga klaster (K= 3) merupakan jumlah optimal. Indeks LQ45 terbagi menjadi tiga klaster berdasarkan karakteristik trennya. Klaster 1 berisi 10 saham dengan tren menurun disertai sedikit pemulihan di Januari 2026, klaster 2 berisi 7 saham yang membentuk pola double top dengan dua puncak pada awal hingga pertengahan tahun 2024, serta klaster 3 berisi 11 saham dengan tren meningkat. Temuan ini memberikan implikasi praktis bagi investor sebagai referensi teknis dalam mengidentifikasi pola harga dan menyusun strategi diversifikasi portofolio untuk secara efektif mengelola risiko investasi.

The capital market plays a strategic role in Indonesia’s economic development. The LQ45 Index serves as a key indicator of the performance of stocks with high liquidity and market capitalization. However, stock price movement analysis is inherently complex due to the dynamic and heterogeneous relationships among stocks, particularly during periods of market uncertainty from 2022 to 2026. This study aims to cluster stocks listed in the LQ45 Index based on the similarity of their adjusted daily closing price time series patterns. The method employed is Agglomerative Hierarchical Clustering with four linkage criteria: single, complete, average, and Ward’s method, combined with the Constrained Dynamic Time Warping (DTW) distance measure using the Sakoe–Chiba band constraint with window sizes of 5%, 10%, and 15% to mitigate the risk of pathological alignment. The dataset consists of daily stock prices obtained from Yahoo Finance for the period of August 2022 to January 2026. The results indicate that the combination of average linkage and a 5% window size provides the best representation of the inter-stock distance structure based on the Cophenetic Correlation Coefficient. Evaluation using the Silhouette and Calinski–Harabasz indices suggests that three clusters (K= 3) represent the optimal grouping. The LQ45 stocks are divided into three clusters based on their trend characteristics: cluster 1 consists of 10 stocks exhibiting a downward trend with slight recovery on January 2026, cluster 2 consists of 7 stocks forming a double top pattern, with two peaks occurring from early to mid-2024, and cluster 3 consists of 11 stocks showing an upward trend. These findings provide practical implications for investors as a technical reference for identifying price patterns and developing portfolio diversification strategies to effectively manage investment risk.