PREDIKSI HARGA SAHAM DENGAN HYBRID GROWING HIERARCHIAL SELF ORGANIZING MAP (GH-SOM) DAN BACKPROPAGATION NEURAL NETWORK (BPNN) (Studi Kasus: Saham Index LQ 45 pada Bursa Efek Indonesia)

• Main Authors: , putu sugiartawan, , Prof. Dra. Sri Hartati, M.Sc., Ph.D.
• Format: Theses and Dissertations NonPeerReviewed
• Published: [Yogyakarta] : Universitas Gadjah Mada 2014
• Subjects: ETD
• Online Access: https://repository.ugm.ac.id/133309/; http://etd.ugm.ac.id/index.php?mod=penelitian_detail&sub=PenelitianDetail&act=view&typ=html&buku_id=73890
• Institution: Universitas Gadjah Mada

Predictions of stock prices are very helpful for investors, especially stockbrokers in deciding the direction of investment from the fluctuation of stock prices in the Stock Exchange. High volatility in the stock markets led to the emergence of the need to understand the patterns and behavior of stock prices and indexes in the stock market. In general, stock predictions are done using technical and fundamental analysis, but these approaches produce a low level of accuracy when compared with the approach of Artificial Neural Network (ANN). Backpropagation neural network algorithm is one of the ANN approaches, which is able to recognize patterns and predict the stock price with a high degree of accuracy, when compared to both approaches. ANN models do not rely on mathematical calculations, but rather to the data of the solved problem, the information presented by the data is obtained from the training process of ANN. The range of training time depends on the amount of data used, the greater the data amount the longer the training process and vice versa. To reduce the training time, this study used a fusion or hybrid algorithms of Growing Hierarchical Self-Organizing Map (GH-SOM) and Backpropagation Neural Network (BPNN). GH-SOM algorithm groups the training data into clusters and then each cluster of data is predicted by BPNN algorithm. The result of this study showed that the clustering of the data with GH-SOM algorithm generated multiple clusters of data and it can reduce training time, but the resulting accuracy of the hybrid algorithm is lower by 53%, when compared with BPNN algorithm that produced an accuracy degree of 92.58%. One of the reasons causing low level of accuracy is the stock split in the stocks researched. Stock that was used in the research is Kalbe Farma stock (KLBF), which fundamentally has a not fluctuated profit level.