OPTIMIZATION OF ARTIFICIAL NEURAL NETWORK PARAMETER USING BACKPROPAGATION ALGORITHM FOR PREDICTING TIME SERIES DATA
Forecasts are important for all decision-making tasks, such as an investment decision. The number of forecasting methods for time series data according to its historical patterns caused difficulty in the prediction process. The presence of Backpropagation Neural Network (BPNN) method is expected to...
Saved in:
| Main Author: | |
|---|---|
| Format: | Theses |
| Language: | Indonesia |
| Online Access: | https://digilib.itb.ac.id/gdl/view/23139 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Institution: | Institut Teknologi Bandung |
| Language: | Indonesia |
| Summary: | Forecasts are important for all decision-making tasks, such as an investment decision. The number of forecasting methods for time series data according to its historical patterns caused difficulty in the prediction process. The presence of Backpropagation Neural Network (BPNN) method is expected to adapt for every pattern of historical data. In the process of creating BPNN network, there are some parameters that must be determined. In this theses, there will be discussed about the optimization of BPNN network and the role of its parameters. After the optimization succeeded, BPNN network will be tested to predict time series data with different patterns. <br />
<br />
<br />
<br />
<br />
The obtained results were quite satisfactory. For predicting stock prices of 9 IT companies with different patterns, BPNN network could predict accurately with an average of MSE 0.3505875. Modifications of BPNN network training process are also done to increase the accuracy of predicted results, one of them was curve smoothing. |
|---|