부동산 분석을 위한 뉴럴네트웍의 응용 = Applications of Neural Networks for Real Estate Analysis

다국어 초록 (Multilingual Abstract)

This study shows the potential of neural networks for real estate analysis. Also, the study compares the performance of neural networks with that of time series regression analysis as illustration of the appropriate use of real estate analysis. Neural networks are computer-based simulations of living nervous systems and have a mathematical basis. Neural networks learn from experiences, generalize from previous examples to new ones, and abstract essential characteristics from noisy and incomplete inputs. It is required that a user have sufficient expertise to decide the number of neurons in the hidden layers, the learning rate, and the momentum. The selection of these factors is problem-dependent and, at this time, there are no established methods for identifying the appropriate values. Generally, using a small number of neurons in the hidden layer increases the number of iterations required to train the neural network and reduces the predictive ability of the neural network. On the other hand, the use of too many neurons in the hidden layer extends the training time and allows the neural network to memorize rather than generalize the training data. The performance of the neural network is affected by its training time. During the first part of training, the performance of the neural network on the training data and testing data improves. During the next part of training, the performance of the neural network on the training data improves continuously. However, the performance of the neural network on the testing data may become worse because the neural network may memorize the training data, and it may cause nongeneralization of predictive ability of the neural network on the testing data. The empirical results for the data of housing price index are that the neural networks performed better than time series regression on the basis of MAD(mean absolute deviations). Also, the F-tests showed that there were statistical differences between them at 0.05 significance level.

목차 (Table of Contents)

• Ⅰ. 서론 Ⅱ. 뉴럴네트웍의 기본모델 Ⅲ. 주택가격지수 예측을 위한 뉴럴네트웍의 응용 Ⅳ. 결론 참고문헌 Abstract