An application of Artificial Neural Networks to forecast winning price: A comparison of Back-Propagation Networks and Adaptive Network-Based Fuzzy Inference Systems.

摘要

The concept of Fuzzy Set Theory has been widely utilized for decades since Zadeh purposed it in 1965. Artificial Neural Network (ANN) has been applied in various fields especially forecasting as well. Adaptive Network-Based Fuzzy Inference System (ANFIS), a hybrid method purposed by Jang in 1965 owns the advantages of both methods mentioned above. Thus, the purpose of this study is to apply two different models: Back-Propagation Network (BPN) and ANFIS on forecasting the winning price. We took Priceline.com as the subject and collected data from BiddingForTravel.com for training and testing models. The performance of forecasting is based on these four evaluation methods: Mean Square Error (MSE), Mean Absolute Deviation (MAD), Mean Absolute Percentage Error (MAPE) and Coefficient of Determination (R2). The result shows that ANFIS model has higher accuracy than BPN model among overall evaluations. Therefore, we conclude that ANFIS performs better than BPN on forecasting in this research.