北台灣不動產價格指標之研究-以混沌理論和類神經預測為例

The Study of Real Estate Price Indicator in Northern Taiwan-An Analysis of Chaos and Artificial Neural Network

周發(Fa Jou)

指導教授 ： 陳若暉

中原大學/商學院/企業管理研究所/碩士(2012年)

https://doi.org/10.6840/cycu201200444

摘要

不動產是國內民眾喜愛投資標的之ㄧ，投資人對於有土斯有財之觀相對重視，另不動產除了居住之需要，也象徵財富地位的榮耀性。因為土地之稀有性便影響著房價之變動，需求與供給是不動產市場二個重要互動因素，如何取捨便要依循著其循環趨勢，對於自住型或投資型若能藉由研究分析之結果，判斷適當投入時機，當可減少許多不必要的損失。
本研究利用BDS分析、R/S分析和相關維度分析法探討領先指標、同時指標、國泰房價指數及信義房價指數是否具有混沌現象。在不動產價格預測方面，使用適合非線性預測的倒傳遞類神經模型及時間延遲遞歸類神經模型，並加入全國人口成長數、經濟成長率、營建股股價指數、國內生產毛額、消費者物價指數、購屋貸款利率、建照執照面積、土地增值稅及貨幣供給額等十項自變數進行模型的預測效果比較，研究樣本分共40組樣本，研究期間為2001年第一季至2011年第四季。
研究結果顯示BDS分析、R/S分析的輸出值皆為顯著，且相關維度分析結果也進行收斂，表示領先指標、同時指標、國泰房價指數及信義房價指數有混沌現象，即表示不動產價格之變動為可預測性。類神經網路預測結果顯示以全國人口成長數、經濟成長率、營建股股價指數、國內生產毛額、消費者物價指數、購屋貸款利率、建照執照面積、土地增值稅及貨幣供給額為投入變數，適用於不動產價格預測。倒傳遞類神經網路進行領先指標、同時指標、國泰房價指數及信義房價指數對房價的預測效果較佳。
全國性領先指標、同時指標、國泰房價指數及信義房價指數與台北市、新北市、桃園縣之國泰房價指數及信義房價樣本檢測結果，以同時指標的預測效果較佳。對於區域性房價之比較亦以台北市之預測較能檢測出房價趨勢。

關鍵字

不動產 ； 混沌現象 ； 倒傳遞類神經 ； 時間延遲遞歸類神經

並列摘要

Real estate is the one of the investment targets for public favorite in Taiwan. Land as wealth was relatively important for investors. And the real estate not only provided demand of living, but also revealed a symbol of wealth and status. The house prices changed caused by the rarity of land. Demand and supply reflecting the the cyclical trend are two important interactive factors of the real estate market to make choose. Through the analysis results provided the appropriate investment opportunities and enhanced investing profit for self-occupied or investment occupied purchases.
This study utilizes BDS test, R/S analysis, and Correlation Dimension Analysis to examine whether Leading Indicators, Coincident Indicators, Cathay Home Price Index, and the Lutheran Home Price Index have the chaos phenomenon. This paper uses a suitable nonlinear prediction of Back-Propagation Neural Network (BPNN) and Time-Delay Recurrent Neural Network (TDRNN), joining ten independent variables (such as Number of the country's population growth, economic growth rate, construction stocks index, gross domestic product, consumer price index, home mortgage rates, the license of construction permits, land value-added tax, and the money supply) and compares the predict performance of the models. The samples are divided into 40 groups. The period was from first quarter of 2001 to the fourth quarter of 2011.
The results of BDS test and R/S analysis showed that are significant, and the outputs of Correlation Dimension Analysis are convergence. This indicated that the Leading Indicators, Coincident Indicators, Cathay Home Price Index, and the Lutheran Home Price Index have the chaos phenomenon, suggesting that the real estate price was predictability. The results of neural network indicated that population growth, economic growth rate, construction stocks index, gross domestic product, consumer price index, home mortgage rates, the license of construction permits, land value-added tax, and the money supply were suitable for real estate price forecasting. The Back-Propagation Neural Network (BPNN) processed the predict ability to have the better performance for Leading indicators, Coincident Indicators, Cathay Home Price Index and the Lutheran Home Price Index .
In comparison of the results for national level of Leading Indicators, Coincident Indicators, Cathay House Price Index and the Lutheran Home Price Index, and Taipei, new Taipei city, Taoyuan County of the Cathay House Price Index with regional level for the Lutheran Home Price Index samples, this paper found that the Coincident indicators have better forecast performance. In comparison regional prices forecasting, Taipei is more suitable for detecting the prices trend.

並列關鍵字

BPNN ； TDRNN ； Chaos phenomenon ； Real estate

參考文獻

1. 中華民國住宅學會，台灣房地產景氣動向季報(1999)，第一卷第一期至第一卷第四期，內政部建築研究所。

6. 白金安（1994），預售屋訂價模式之探討，中華民國住宅學會第三屆年會論文，1994年2月5日。

10. 吳森田（1994)，所得、貨幣與房價-近二十年台北地區之觀察，住宅學報，第二期，49-66。

12. 何肇榮(2006)，景氣對策信號對台灣房地產之研究，中原大學碩士論文。

13. 呂奇傑、李天行、高人龍、陳學群(2008)，結合獨立成份分析與類神經網路於財務時間序列預測模式之建構，交大管理學報，28(2)，187-216