Prediction of Low Cost Housing Demand in Malaysia Using ARIMA Model

Among the key challenges in construction industry sector faces are matching supply of and demand for affordable housing. It is very crucial to predict low-cost housing demand to match the demand and supply so that the government can plan the allocation of low cost housing based on the demand. In Johor, housing provision is very crucial due to urbanization. The supply of houses seems to be swamping the demand for luxury condos and houses especially in Johor Bharu. Thus the aim of this study is to predict low-cost housing demand in Johor, Malaysia using ARIMA model. Time series data on low-cost housing demand have been converted to Ln before develop the model. The actual data and forecasted data will be compared and validate using Mean Absolute Percentage Error (MAPE). After that, the results using ARIMA method will be compared with ANN method. The results show that MAPE analysis for ARIMA is 15.39% while ANN is 18.27%. It can be conclude that ARIMA model can forecast low cost housing demand in Johor quite good.


Introduction
One of Malaysia's longstanding development objectives is the provision of affordable housing for Malaysian, with a focus on lower-income groups [1].Low cost housing can be defined as a development projects sold at the price set by the government that is between RM25, 000 to RM42, 000 [2].Low cost housing built is intended to provide housing that is affordable for low earners in rural and suburban areas.The target groups for this project are households with monthly income of between RM500 to RM750 [3].
To provide adequate housing and affordable for Malaysians, especially for those with low incomes has become the main agenda through Malaysia plans prior to now [4].However, there is mismatched data between the supply and demand for low-cost housing in Malaysia [5].In some places, the supplies of low cost housing are exceeding compare to demand and lead to wastage of construction and of course has an impact on the cost and economic aspects.While in other areas the demand is exceeding supply provided, which supplies low-cost houses are insufficient, especially in urban areas [6].Therefore, an alternative approaches need to be done to resolve these issues.
There are many series of forecasting methods can be used to predict the housing demand such as Artificial Neural Network (ANN), Autoregressive Integrated Moving Average (ARIMA), Power a Corresponding author : nryasmin@uthm.edu.myModel and Multiple Log linear Regression [7].In this study ARIMA model known as the Box-Jenkins time series is used because it has good accuracy for the short term forecasting.

Scope and Limitation of Study
This study will focuses on forecasting low cost housing demand in Johor, Malaysia only.Previous time series data from [6] will be used to forecast low-cost housing demand in Johor using ARIMA model.

Methodology
The time series data were changed to Ln then analyzed using ARIMA software adopted from SPSS 20.0.Results were validated using MAPE where actual and forecasted data were compared to determine the accuracy of the model.Finally, the MAPE value will be compared to establish the performance of model.

Results and Discussion
Table 1 shows the monthly time series data on low cost housing demand in Johor from January 2000 to January 2007.Data were change to Ln to get idea for p, d, q value for a non seasonal ARIMA model ARIMA (p, d, q), where p is the number of autoregressive terms, d is the number of nonseasonal differences needed for stationarity, and q is the number of lagged forecast errors in the prediction equation.p and d were determined using coefficient Auto Correlation (AC) and Partial Auto Correlation (PAC) where AR(p) and MA(q) are the components for the time series.

Conclusion
Since the MAPE value is less than 20%, it can be conclude that ARIMA model can predict low-cost housing demand in Johor quite good.It is recommend further study should be done to reduce the error of performance since the results generated are able to assist the construction of low-cost housing scheme in terms of the accuracy of necessity based on actual demand.Subsequently there would be a minimal possibility of the procurement of either under-construction or over construction of low cost houses particularly in the state of Johor.

Figure 1
Figure1are time series housing demand in Ln.Data were change to Ln to get idea for p, d, q value for a non seasonal ARIMA model ARIMA (p, d, q), where p is the number of autoregressive terms, d is the number of nonseasonal differences needed for stationarity, and q is the number of lagged forecast errors in the prediction equation.p and d were determined using coefficient Auto Correlation (AC) and Partial Auto Correlation (PAC) where AR(p) and MA(q) are the components for the time series.Figure2views the ACF and PACF of housing demand using ARIMA (0,0,0).It can be seen that there are a few extrusions and improvements were done to create stationary data.From the p,d,q value, significant calculations for each parameters were done to determine the best model.In this study, three model were used; ARIMA (1,0,1); ARIMA (1,0,0) and ARIMA (2,0,0).ARIMA model produced the lowest value Akaike Information Criterion (AIC) and Schwarz Criterion (SC) is the best model.

Table 1 .
Time series data on low cost housing demand.

Table 2 .
Value for AIC and SC for each model.

Table 4 .
Comparison forecasting value between tentative models.

Table 4
is the comparison forecasting value between tentative models ARIMA.It shows that ARIMA (1,0,1) have the lowest MAPE with 3.9%.04008-p.4

Table 5
[8]ws the actual and forecasted data from June 2000 to September 2006 using ARIMA (1,0,1) model.From the calculations, MAPE value obtained was 15.39%.Predictive ability is very good if the MAPE is less than 10% while MAPE less than 20% is good[8].The results show that MAPE value for ARIMA less than 20%.