Optimasi Model Peramalan Fuzzy Time Series dengan Integrasi Fuzzy Clustering, Prinsip Justifiable Granularity, dan Particle Swarm Optimization (Studi Kasus Harga Penutupan Saham Sektor Energi)
Syeikha Kirana Dewi, Prof. Dr. Abdurakhman, S.Si., M.Si.
2026 | Skripsi | STATISTIKA
Pergerakan harga saham sektor energi bersifat fluktuatif, tidak linier, dan sensitif terhadap perubahan harga komoditas, kebijakan energi, kondisi ekonomi global, serta sentimen investor. Karakteristik tersebut memerlukan metode peramalan yang mampu menangani ketidakpastian sekaligus membentuk interval fuzzy secara adaptif. Penelitian ini bertujuan mengoptimalkan model Fuzzy Time Series (FTS) melalui integrasi Fuzzy C-Means Clustering (FCM), Prinsip Justifiable Granularity (PJG), dan Particle Swarm Optimization (PSO) untuk meramalkan harga penutupan saham batu bara Indonesia yang tergabung dalam LQ45. Data yang digunakan berupa harga penutupan harian saham ADRO, ITMG, dan PTBA periode 2023 – 2025. Data ditransformasikan menjadi return, dipisahkan berdasarkan saham dan tahun, lalu dimodelkan menggunakan FTS Tsaur, FTS-FCM, FTS-FCM-PSO, dan FTS-FCM-PJG-PSO. Kinerja model dievaluasi melalui RMSE, MAE, dan MAPE. Hasil penelitian menunjukkan bahwa seluruh model menghasilkan MAPE di bawah 5%. FTS-FCM-PJG-PSO memberikan kinerja terbaik pada lima dari sembilan dataset dengan rata-rata MAPE terendah sebesar 1,4603%. Temuan ini menunjukkan bahwa integrasi FCM, PJG, dan PSO mampu meningkatkan konsistensi akurasi, meskipun model yang lebih kompleks tidak selalu menjadi pilihan terbaik pada setiap dataset.
Energy-sector stock prices are volatile, nonlinear, and sensitive to changes in commodity prices, energy policies, global economic conditions, and investor sentiment. These characteristics require a forecasting method capable of addressing uncertainty while constructing fuzzy intervals adaptively. This study aims to optimize the Fuzzy Time Series (FTS) model by integrating Fuzzy C-Means Clustering (FCM), the Principle of Justifiable Granularity (PJG), and Particle Swarm Optimization (PSO) to forecast the closing prices of Indonesian coal stocks listed in the LQ45 index. The data consist of daily closing prices of ADRO, ITMG, and PTBA from 2023 to 2025. The data were transformed into returns, separated by stock and year, and modeled using FTS Tsaur, FTS-FCM, FTS-FCM-PSO, and FTS-FCM-PJG-PSO. Model performance was evaluated using RMSE, MAE, and MAPE. The results show that all models produced MAPE values below 5%. FTS-FCM-PJG-PSO achieved the best performance in five of the nine datasets and recorded the lowest average MAPE of 1.4603%. These findings indicate that integrating FCM, PJG, and PSO can improve forecasting consistency, although a more complex model does not always provide the best performance for every dataset.
Kata Kunci : Peramalan, Fuzzy Time Series, Fuzzy Clustering, Justifiable Granularity, Particle Swarm Optimization