不動產投資信託型指數股票基金(REIT型ETF)具備REIT可以同時賺取股票價格成長的資本利得以及股票的配息外,又擁有ETF指數化被動式管理之特性,且比投資單一REIT更能分散風險,是投資不動產的最新投資工具。2007年8月美國次貸風暴席捲全世界,導致全球經濟成長疲軟,信用市場嚴重緊縮,房地產市場首當其衝,均出現房地產泡沫化跡象。若能預測出房市價格之走勢,及早察覺資產泡沫化的危機,則投資人可提高投資決策的正確性。 本研究以投資於美國、加拿大、歐洲、亞洲、全球之REIT型ETF共21檔為研究對象,並結合利率、美元指數、股價指數、賣權/買權比率、波動指數、CRB商品指數及西德州原油價格等影響變數,利用灰關聯分析排序影響REIT型ETF之相關程度之變數,再將所選之變數以灰預測GM(1,N)模型與類神經網路模型分別建構出預測REIT型ETF之最適模型。 結果發現REIT型ETF受到各國股價指數、西德州原油價格、CRB商品指數及賣權/買權比率最深。在預測結果方面,以倒傳遞神經網路擁有較小之平均RMSE值,優於灰預測GM(1,N)模型。另灰預測GM(1,N)模型與倒傳遞神經網路模型皆以投資於加拿大之XRE預測效果最好,代表由各國股價指數、西德州原油價格、CRB商品指數及賣權/買權比率所建構之預測模型有良好的預測績效。
The REIT-ETF possesses the features of making profit on stock from capital gain and its dividens, as well as the passive management of ETF index. Further, it may spread risks more effectively than single REIT, and thus, it becomes the newest investing tool in the field of real estate. In August of 2007, the Subprime Mortgage Crisis happened in the USA and rolled up worldwide, affecting to the weak economic growth globally. Therefore, the credit markets shrank seriously back then, and the real estate market was in the front of this tide. Causing a bubbling trend. If the trends in house prices could be precisely estimated with early, detection of asset bubbles, thereby offering investors with crucial reference. This research uses the 21 REIT-ETFs among the USA, Canada, Europe, Asia and global countries, exaiming the effects of interest rate, dollar index, stock index, Put/Call ratio, fluctuation index, Commodity Research Bureau index (CRB) index and oil price of West Texas. etc. To ranking the the most crucial factor on the REIT-ETF by means of Gray Relation Analysis, the determining factors will be used to forecaste REIT-ETF, through the cross-match of Grey Prediction GM(1,N) and Artificial Neural Network System of Back-propagation Network. The study results show that the REIT-ETF is affected the most notably by stock indexes, oil price of West Texas, CRB index, Put/Call ratio. In relation to forecasting results, the Back-propagation Network is superior than GM (1,N) model in terms of smaller average Root Mean Square Error (RMSE). Among them, the best estimated performance is for XRE in Canada investment, representing that the forecast model established by stock indexes, oil price of West Texas, CRB index and Put/Call ratio has excellent predicting performance.