影響全球不動產投資信託關鍵因素之研究 －灰色關聯分析與類神經網路之應用

A Study of Key Elements on Global Real Estate Investment Trust—An Application of Grey Relational Analysis and Artificial Neural Network

張婷慈(Ting-Tzu Chang)

指導教授 ： 陳若暉

中原大學/商學院/企業管理學系/碩士(2007年)

https://doi.org/10.6840/cycu200700219

摘要

由於不動產投資信託(Real Estate Investment Trust, REIT)的投資門檻低，又同時具備股票市場高報酬和不動產市場抗通膨等特性。近年來已成為熱門的投資商品，投資風潮延燒全球。於此同時，英國經濟學人雜誌卻提出全球房市泡沫化的警告。因此，若能準確預測REIT報酬率，不僅能幫助投資人選擇投資標的，更能避免受到房市泡沫化影響。本研究結合總體經濟變數和不動產投資信託過去績效，將研究範圍擴大至全球17個國家，運用灰色關聯分析與類神經網路，以期能找出影響全球不動產投資信託報酬之關鍵變數。
本研究運用灰色關聯分析結果發現，REIT分別受工業生產指數、借款利率、股利收益率、各國股價指數及其本身過去績效影響最深。再利用類神經網路驗證灰色關聯分析效果，首先測試各國在倒傳遞神經網路(BPN)、回饋式神經網路(RNN)與輻狀基底函數神經網路(RBFNN)的網路學習效果，以選取各國最適神經網路。經參數調整後，將上述關鍵變數代入各國最適神經網路中，研究結果發現，在所有國家中，此五項變數可降低預測誤差，使網路輸出值與實際值具有高度一致性。最後並針對各國篩選出預測能力達到99%以上的變數組合，找出影響各國不動產投資信託之關鍵變數，以幫助投資人進行投資決策。

關鍵字

類神經網路 ； 灰色關聯分析 ； 不動產投資信託

並列摘要

Real Estate Investment Trust (REIT) becomes one of the most popular investment instruments recently due to its low investment limit and the unique characteristics associated with high returns from stock market and anti-inflation function from real estate market. At the same time, The Economist gave a warning about the global real estate price bubble. Forecasting the returns of REIT could not only help investors to choose the investment targets, but also avoid the damage from real estate price bubble. This study uses macroeconomic variables and past performance of REIT to expand the research scope into 17 countries. This work utilizes Grey Relational Analysis (GRA) and Artificial Neural Network (ANN) to analyze the key elements on global real estate investment trust.
The findings of GRA suggests that REIT would be most influenced by industrial production index, lending rate, dividend yield, stock index, and it’s lagged performance. Next, this study used ANN to test the results of GRA, and chose the best fit model through comparing the learning effect of Back-propagation Neural Network (BPN), Recurrent Neural Network (RNN), and Radial Basis Function Neural Network (RBFNN). After adjusting the parameters and putting the key elements into the fitted ANN model, the results show that among all countries, these five key elements could reduce the predicting error, and the network output and the desired output could reach at high consistency. Finally, this paper selects the variable mixes with the predictive ability of 99% level, expecting to help investors to make the best investment decisions.

並列關鍵字

Artificial Neural Network (ANN) ； Grey Relational Analysis (GRA) ； Real Estate Investment Trust (REIT)

參考文獻

蔡瑞煌、高明志、張金鶚，1998，「類神經網路應用於房地產估價之研究」，住宅學報，第八期：頁1-20。

馮正民、王榮祖，2001，「應用灰色關聯分析於航空運輸業營運與財務績效代表性指標之擷取」，民航季刊，第三卷第一期：頁107-126。

紀勝財、尤明偉、饒瑞倫，2004，「應用類神經網路於股票技術指標聚類與預測分析之研究」，科技管理學刊，第九卷第二期，53-81。

Mooradian, R. M., and S. X. Yang, Dividend Policy and Firm Performance: Hotel REITs vs. Non-REIT Hotel Companies, Journal of Real Estate Portfolio Management, 2001, 7:1, 79-87.

Chan, S., Han, W. K. Leung, and K. Wang, Institutional Investment in REITs: Evidence and Implications, Journal of Real Estate Research, 1998, 16:3, 357-374.