Bernhard, David (2019) Prediksi Harga Saham Menggunakan Metode Backpropagation Dengan Optimasi Ant Colony Optimization. Sarjana thesis, Universitas Brawijaya.
Abstract
Saham merupakan tanda kontribusi penanaman modal seseorang atau pihak terhadap suatu perusahaan atau perseroan terbatas. Pergerakan harga saham berpengaruh terhadap keuntungan dan kerugian yang akan diperoleh investor. Kendalanya, harga saham dapat berubah dalam setiap menit pada hari kerja. Dibutuhkan metode yang mampu memprediksi harga saham dengan akurat dan konsisten, sehingga dapat meminimalkan risiko investasi saham. Disamping kelebihannya, BPNN memiliki kekurangan, seperti lambatnya waktu konvergensi, mudah konvergen ke titik minimum lokal, dan buruknya kemampuan generalisasi. ACO memiliki kelebihan dalam komputasi terdistribusi, umpan balik positif, dan sifat metaheuristik yang dapat memperbaiki kelemahan BPNN. Oleh karena itu, penelitian ini menerapkan ACO-BPNN untuk memprediksi harga saham. Penelitian ini menggunakan data time series harga saham Bank Rakyat Indonesia (Persero) Tbk. periode 1 Januari 2018 sampai 31 Desember 2018. ACO berfungsi untuk mengoptimalkan kombinasi nilai learning rate, momentum, dan jumlah hidden node bagi fase pelatihan BPNN. Lalu bobot, threshold, dan jumlah hidden node dari fase pelatihan digunakan untuk memprediksi harga saham harian pada suatu periode. Diperoleh kombinasi nilai parameter ACO terbaik, yaitu tetapan siklus semut sebesar 0,8, tetapan pengendali intensitas feromon sebesar 0,1, tetapan pengendali visibilitas sebesar 0,1, tetapan penguapan feromon lokal sebesar 0,5, tetapan penguapan feromon global sebesar 0,1, jumlah semut 5, dan jumlah iterasi 7. Kombinasi tersebut menghasilkan rata-rata MAPE 1,745, sedangkan BPNN hanya mencapai 3,024.
English Abstract
Stocks are a sign of a person's or party's investment contribution to a company or limited liability company. Movement of stock prices affects the profits and losses that will be obtained by the investor. The obstacle is stock prices can change in every minute on weekdays. It takes a method that is able to predict stock prices accurately and consistently, so that it can minimize the risk of stock investment. Besides its advantages, BPNN has shortcoming, such as slow convergence time, easy convergence to local minimum points, and poor generalization capabilities. ACO has advantages in distributed computing, positive feedback, and metaheuristic properties that can improve the weaknesses of BPNN. Therefore, this study applies ACO-BPNN to predict stock prices. This study uses time series data from the stock price of Bank Rakyat Indonesia (Persero) Tbk. period 1 January 2018 until 31 December 2018. ACO serves to optimize the value combination of learning rate, momentum, and number of hidden nodes for BPNN training phase. Then weights, thresholds, and number of hidden nodes from the training phase are used to predict daily stock prices for a period. Best combination of ACO parameter values was obtained, namely the ant cycle constant worth 0.8, the control constant of pheromone intensity worth 0.1, the visibility control constant worth 0.1, the local pheromone evaporation constant worth 0.5, global pheromone evaporation constant worth 0.1, number of ants 5, and number of iterations 7. That combination produces an average of MAPE 1,745, while BPNN only reached 3,024
Other obstract
-
| Item Type: | Thesis (Sarjana) |
|---|---|
| Identification Number: | SKR/FILKOM/2019/223/051902973 |
| Uncontrolled Keywords: | prediksi, harga saham, optimasi, ant colony optimization, backpropagation, ACO-BPNN, prediction, stock prices, optimization, ant colony optimization, backpropagation, ACO-BPNN |
| Subjects: | 300 Social sciences > 332 Financial economics > 332.6 Investment > 332.63 Specific forms of investment > 332.632 Securities, real estate, commodities > 332.632 2 Stocks (Shares) > 332.632 22 Prices |
| Divisions: | Fakultas Ilmu Komputer > Teknik Informatika |
| Depositing User: | Nur Cholis |
| Date Deposited: | 30 Jun 2020 04:35 |
| Last Modified: | 24 Oct 2021 05:22 |
| URI: | http://repository.ub.ac.id/id/eprint/169522 |
Preview |
Text
SKRIPSI_David Bernhard (2).pdf Download (4MB) | Preview |
Actions (login required)
 |
View Item |