中文摘要 房地產是國人喜愛的投資標的之一,因具有滿足住的需求性與象徵財富地位的榮耀性外,亦因土地不可移動與供給量限制之特性,致使土地擁有長期上升趨勢的保值性。需求與供給是房地產市場的兩個影響因素,在需求量大時會導引資金流入房地產,影響租賃的成長率和房地產價格。故房地產景氣循環為必然趨勢,惟其循環高點與低點之房屋價差,有達30%以上。若能掌握其循環趨勢,對於自住型者,可判斷先行租賃或直接購屋較適當。對於投資型者,則可判斷適當投入時機。 本研究欲從景氣對策信號、國內生產毛額(GDP)、貨幣供給額-M2、核發建造執照樓地板面積、買賣契稅件數、購屋貸款利率、營建業股價指數、經濟成長率、土地增值稅、消費者物價指數變動率等10種總體經濟變數,並採用房地產景氣綜合指標(領先、同時二種指標)、國泰房價指數、信義房價指數等4個房地產指標為應變數。使用能對事物不確定性資料,進行有效處理的灰色關聯分析(Grey Relational Analysis),依變數相關性進行排序後,以瞭解其灰關聯程度高低。再利用類神經網路驗證其預測效果,並期預測房地產波動之趨勢,使房地產生產者(建築商)、消費者(購屋者)、政策執行機構(政府),皆能瞭解房地產的趨勢,以預作策略因應。 本研究以臺灣地區為樣本,採取1999年第一季至2008年第四季,共計40季的季資料進行研究。由灰色關聯分析發現,十個自變數之灰關聯排序,以國內生產毛額(GDP)、貨幣供給額-M2、核發建造執照樓地板面積,與房地產景氣綜合指標等四個指標之灰關聯度較高。而類神經網路分析發現,以倒傳遞神經網路(BPN)的表現最好。將灰色關聯分析結果,代入倒傳遞神經網路(BPN)中發現,全部變數組代入神經網路後,均可將誤差值(MAE)降至0.076以下,顯示以此10個變數預測房地產走勢,準確度頗高。然灰關聯度高之變數組的誤差值,並未優於灰關聯度低的變數組。
Abstract Real Estate is one of the favorite investments in Taiwan. Besides of fulfilling the needs of residence and the dignity of wealth status, the features of its immobility and limited amount caused the long-term trend of an increase in value. Demand and supply are the two major influential factors in the real estate market. When the demand is high, funds would be drawn into the real estate market and thus influencing the growth rate of rents as well as the price of properties. This demonstrates that the real estate market presents a circular cycle in prosperity. However, the high point and low point of this cycle have an over 30% difference in price. If this cycle can be predicted, the potential resident buyers can decide whether it is more economical to rent or to buy, while the investors can judge a better timing for entering the market. This study focuses on ten macroeconomic variables, including Monitoring Indicator, gross domestic product (GDP), money supply balance – M2, the issuance of building license for total floor coverage, the number of contracts dealt, the mortgage interest rate, the stock index of constructing sector, the economic growth rate, land value-added tax, and consumer price index. It also refers to four dependent variables: real estate indicator (both leading and current indicators), Cathay housing index, and Sinyi housing index. The application chosen is the Gray Relational Analysis that can effectively process uncertain data, sorted in accordance to variables relativity, to verify the level of the gray relational correlation. Then applying artificial neural network analysis to examine the predicted effects and to foresee the fluctuation trend, it will allow the real estate producers (the constructors), consumers (the buyers) and policy makers (the government) to fully understand the trend of the real estate market, and react accordingly. The samples used the quarterly data of 40 quarters in Taiwan, namely those from the first quarter of 1999 to the fourth quarter of 2008. From the Gray Relation Analysis, in accordance to the sorting of gray relativity of the ten variables, the gray relations of GDP, money supply balance – M2, issuance of building license of total floor coverage and the real estate indicator have the higher relativity. The results of artificial neural network analysis showed that Back-propagation Neural Network (BPN) performed best. When applying the gray relation analysis results into the BPN, the Mean Absolute Error (MAE) value connected with all variables could be lowered under 0.076, meaning high accuracy of predicting the developmental trend of real estate. However, the MAE of higher gray relation variables is not superior to the lower gray relation variables. Key Words:Real Estate, Grey Relational Analysis, Artificial Neural Network, Back-propagation Neural Network,Mean Absolute Error