Abstract
JOURNAL OF ASIAN ARCHITECTURE AND BUILDING ENGINEERING 2020, VOL. 19, NO. 3, 230–241 https://doi.org/10.1080/13467581.2020.1723595 CONSTRUCTION MANAGEMENT Model for predicting price change patterns in multi-family houses post renovation work in South Korea a a a b Kyuman Cho , Jaesung Kim , Taehoon Kim and Taehoon Hong a b School of Architecture, Chosun University, Gwangju, Republic of Korea; Department of Architecture and Architectural Engineering, Yonsei University, Seoul, Republic of Korea ABSTRACT ARTICLE HISTORY Received 5 June 2019 Renovation work on deteriorated multi-family houses (MFHs) is often undertaken to improve Accepted 18 January 2020 their physical performance. However, due to uncertainties in economic benefits from renova- tion, many MFHs frequently withdraw their renovation plans in South Korea. Despite this KEYWORDS problem, there has been very little research on countering this issue. With this background, Renovation work; this study aims to develop a model for predicting the price change patterns (MPPCP) of multi-family house; artificial deteriorated MFHs upon renovation in South Korea. An artificial neural network (ANN)-based neural network; economic MPPCP was developed to detect the relationship between project attributes and price change benefit due to renovation patterns due to renovations. By combining the parameters of the ANN method, 108 candidate models were identified and a final MPPCP was proposed after conducting simulation tests to verify the level of correct for the candidate`s models. The results of model application to actual MFH renovation cases show that the developed model can facilitate a project owner’s decision- making by estimating price change patterns for the deteriorated MFH in the project planning stage itself. 1. Introduction Kim, and Kim 2019). However, most of the existing studies on the economic efficiency of MFH renovation 1.1. Research background and objectives projects analyzed economic efficiency only in terms of According to a report by the Construction & Economy input cost, such as construction costs (Lee 2005; Lee et Research Institute of Korea (CERIK), in addition to phy- al. 2007; Han and Shin 2012; Yeon et al. 2014; Kim and sical deterioration of multi-family houses (MFHs), their Baik 2015). Some existing studies have been con- social performance is experiencing rapid changes due ducted for typical MFHs in estimating their price or in to changes in lifestyle and population structure in identifying the factors influencing the price. However, South Korea (Yoon and Lee, 2012). Renovation and most of these studies have attempted to estimate the reconstruction are considered attractive solutions to price for new and existing MFHs, mainly using the counter this physical and social deterioration. From method based on the typical building appraisal meth- the point of view of low-carbon green construction, ods compensated by referring to the transaction cases renovation is gaining much attention as it imposes less (Kim, Cho, and Kim 2016). environmental burden. Therefore, due to the above limitation, it is highly However, in spite of the positive effect of less envir- difficult to evaluate the economic effect of renova- onmental pollution, the current status of renovations tions, which is one of the main concerns of property for deteriorated MFHs is poor. According to the Korea owners considering renovating their deteriorated Remodeling Association (KRA), in 2011, only 3% MFHs; this is also a major reason for owners cancelling (104,803 households) of 3,177,000 deteriorated house- their renovation plans in South Korea. In this study, we holds considered renovation and of these, 77% later aim to develop a method for evaluating the economic cancelled their renovation plans (KRA 2011). According benefit, represented by changes in the monetary price to Kim, Choi, and Kim (2010), Cho et al. (2012), and Kim of MFHs in South Korea, post renovation. It is expected et al. (2013), the reasons for abandoning MFH renova- that the results of this study would help facilitate tions in South Korea are primarily the uncertainty in decision-making with respect to renovation projects. economic profit after renovation and difficulty in securing the feasibility due to high building costs. 1.2. Research methodologies Economic benefit from renovation is very important to not only evaluate the success or failure of a project The implementation procedure of the economic but also decide whether or not to undertake renova- effect analysis model developed in this study is tion work (Cho and Yoon 2016; Kim et al. 2018; Cho, depicted in Figure 1. One of the main aspects in CONTACT Taehoon Kim thoonkim@chosun.ac.kr School of Architecture, Chosun University, Gwangju, Republic of Korea © 2020 The Author(s). Published by Informa UK Limited, trading as Taylor & Francis Group on behalf of the Architectural Institute of Japan, Architectural Institute of Korea and Architectural Society of China. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. JOURNAL OF ASIAN ARCHITECTURE AND BUILDING ENGINEERING 231 Figure 1. Model framework. developing the model lies in analyzing the relation- 2. Analysis of price changes due to ship between attribute changes and price changes of renovations a given MFH upon renovation. In other words, reno- It is possible to analyse the economic effect of renova- vation work on a deteriorated MFH causes changes in tions on MFHs by evaluating trends in price changes of its physical attributes and monetary value. Moreover, the MFH before and after renovation. if the relationship between the two sets of changes can be structured using scientific and objective meth- odologies, it is possible to predict price changes in 2.1. Analysis of price changes due to renovations the MFH post renovation according to variations in the project’s attributes corresponding to the renova- Generally, the price of an MFH is influenced by various tion plan. With this objective in mind, 120 MFH cases factors, including location, convenience of transporta- were selected to gather data required for analyzing tion, and educational environment (Choi and Song changes in their attributes and monetary values after 2006; Jin et al. 2012; Kim, Cho, and Kim 2016). As renovation; an artificial neural network (ANN) method price changes due to renovation must be measured, was adopted to structure the relationship between the factors above should be controlled while assessing the two changes objectively. As shown in Figure 1, price changes. In other words, price change assess- the attribute change rate (△1) and corresponding ment due to renovations may follow the concept of price change rate (△2) were utilized as inputs and “relative price” for the renovated MFH on the similar out variables for implementing ANN method. comparative cases, which aims at excluding other Through the above process, this research could pro- price-influencing factors. pose an ANN-based model predicting price change of Therefore, a comparative case with location and size MFH post renovation works. similar to those of the renovation case should be 232 K. CHO ET AL. selected; the price of the renovation case should be rate of RC on CC at present, RC = renovation case k, k k assessed relative to the comparison case at three time j and CC = comparative case j for RC . points (before renovation, after renovation, and pre- Using the above method, it is possible to analyse sent). From the obtained results, one can be sure that price change patterns for an MFH upon renovation; the relative price change between the price of the furthermore, the economic effect of renovation can renovated MFH and that of the comparative MFH is be grasped objectively. As shown in Figure 2, if the due to renovations. The relative price ratio can be price after renovation is higher than that before reno- calculated using Equations (1–1), (1–2), and (1–3). RC vation, it can be considered that renovations have a and CC denote renovation case k and comparative positive influence on the monetary value of the MFH. case j for RC , respectively. RC BR 2.2. Collection of renovation cases and R P ¼ 100 (1 1) BR j CC comparative cases BR A total of 17 renovated MFH cases, all of which were RC j located in Seoul, were considered in this study. As AR R P ¼ 100 (1 2) AR CC shown in Table 1, in most of these cases, the house- AR hold unit area increased upon renovation, which took about 1 to 2 years. In some cases, vertical expansion RC was conducted by increasing the number of floors; R P ¼ 100 (1 3) CC most of these structures were constructed around 1970 and about 25 years later, they required renova- Here, R P = price rate of renovation case (RC) k on BR tions for improving building performance in terms of comparative case (CC) j before renovation, R P = AR k energy, physical appearance, and economic perfor- j j price rate of RC on CC after renovation, R P = price mance. Furthermore, as described above, comparative k P k k Proportion of RC`s price on CC`s price at each time (%) R P k P R P Before k AR renovation 100 % After Present renovation R P k BR Figure 2. Example of positive price changes due to renovation. Table 1. Renovation cases. Time line Area of Renovation Number of floors unit households (m ) Year of Renovation cases (before renovation) Location after renovation construction period RC 12(12) SC. gu, Seoul 89.84 1978 2005. 07 ∼ 2006. 12 RC 12(12) SC. gu, Seoul 121.07 1978 2005. 07 ∼ 2006. 12 RC 12(12) SC. gu, Seoul 155.44 1978 2005. 07 ∼ 2006. 12 RC 5(5) MP. gu, Seoul 57.86 1971 2006. 02 ∼ 2007. 03 RC 12(10) MP. gu, Seoul 89.79 1989 2011. 03 ∼ 2012. 12 RC 12(12) YS. gu, Seoul 147.74 1972 2004. 07 ∼ 2005. 12 RC 12(12) YS. gu, Seoul 186.12 1972 2004. 07 ∼ 2005. 12 RC 13(12) YS. gu, Seoul 89.09 1976 2007. 06 ∼ 2008. 08 RC 11(10) YD. gu, Seoul 88.03 1978 2008. 07 ∼ 2010. 07 RC 13(12) YD. gu, Seoul 68.27 1978 2008. 07 ∼ 2010. 07 RC 13(12) YD. gu, Seoul 106.56 1978 2008. 07 ∼ 2010. 07 RC 5(5) MP. gu, Seoul 59.50 1971 2002. 06 ∼ 2003. 07 RC 16 (15) GN. gu, Seoul 110.39 1989 2011. 12 ∼ 2014. 02 RC 15(12) GN. gu, Seoul 110.20 1991 2011. 06 ∼ 2014. 01 RC 19(18) GN. gu, Seoul 110.14 1992 2012. 02 ∼ 2014. 03 RC 11(10) GJ. gu, Seoul 92.40 1987 2010. 08 ∼ 2013. 08 RC 11(10) GJ. gu, Seoul 65.65 1987 2010. 08 ∼ 2013. 08 17 JOURNAL OF ASIAN ARCHITECTURE AND BUILDING ENGINEERING 233 7 2 8 MFH cases were selected for each renovated MFH case CC ,CC , and CC ) exhibited significant differences 3 11 16 selected using a 2-step criterion as follows (see in a unit area. Despite the area difference, because they Figure 3). In Step 1, an MFH located in the same were located very close to the renovation case, it is administrative district as the renovation case was meaningful to compare their price data with that of selected to control the effect on price by location, renovation cases. Price data at the three selected time which has a highly significant effect on MFH price points were collected through the Declared price of (see ⓐ in Figure 3). If the area of the administrative real estate in Korea inquiry system (Ministry of Land, district is too wide, MFHs located within a radius of Infrastructure, and Transport, 2016). 1 km from the renovated MFH are selected as com- parative candidates (see ⓑ in Figure 3). In the second 2.3. Analysis of price change patterns due to step, because price is also influenced by the household renovation work unit area of the MFH, a case with similar unit area was selected from among the candidates screened in the In this study, we collected price data at three time first step as the comparative case (see ⓒ in Figure 3). points (before renovation, after renovation, and pre- In accordance with the above criteria, 103 compara- sent) for selected renovation and comparison cases. tive MFH cases were collected for 17 renovated MFH Using this data, the impact of renovation on MFH cases (see Table 2). As shown in the table, each com- prices can be assessed. Table 3 shows an example of parative case had an area similar to that of the house- such evaluation for Renovation Case 1. The table hold unit area of the renovation case. For example, the shows price information on renovation case RC and 2 1 4 household unit area of RC was 186.12 m while that of its comparative cases CC to CC at the three selected 7 1 1 1 4 comparative cases CC to CC was 177.62, 147.48, time points as well as the price ratio of RC to CC, which 7 7 180.20, and 196.71 m , respectively. A minimum of can be used to analyse relative price change trends two comparison cases (for RC and RC ) were upon renovation. As shown in the table, a total of four 11 12 selected for each renovation case while the maximum relative price changes could be observed (i.e., changes was 10 (for RC ). in the prices of the renovation and comparative cases. 1 4 Although comparative cases in similar areas were The price change of RC as compared to CC to CC 1 1 1 selected in most cases, some comparative cases (i.e., shows that its current relative price is greater than its Within a 1 km Radius < Used legend > : SC-gu, B-dong (Administrative district) : Renovated MFH case : Comparative MFH cases Figure 3. Selection example of comparative cases. 234 K. CHO ET AL. relative price before renovations, which means that renovations exerted a positive influence on its price. To improve conveyance of the concept “the relative price” of each renovation case at three time points, the combination of the renovation case and the equivalent comparative case is expressed as a union (U ) as shown in Equation 2. As described earlier, using 17 renovation cases and 103 comparison cases, 103 com- binations were created and these combinations can be expressed as follows. j j U ¼ Union of RC and CC ; for k ¼ 1 to 17; k k j ¼ 1 to 10 (2) For example, as shown in Table 3, the four combina- 1 2 tions referred to for RC can be expressed as U ,U , 1 1 1 3 4 U , and U (see Union cases in Table 3). 1 1 After studying the price change trends of these 103 unions, three price change patterns could be identified (see Figure 4). The first price change pattern was the constantly increasing pattern, which means that the price of the renovation case before renovation was lower than that of comparative cases but it was higher than that of comparative cases after renovations. For example, the price change trends of three union cases 2 3 4 (U ,U , and U ) described in Table 3 follow this 1 1 1 pattern. These patterns can be diagrammatically repre- sented as shown in Figure 4(a); 30 cases were found to follow this pattern. In the second price change pattern, the price of the renovation case before renovation was lower than that of comparative cases, but it rose sig- nificantly after renovation and this tendency were sus- tained (Figure 4(b), 49 cases). In the third pattern, the price of the renovation case before renovation was lower than that of comparative cases and it increased slightly after renovation, but this increase was not sustained (Figure 4(c), 21 cases). Meanwhile, the remaining 3 cases of 103 cases showed patterns differ- ent from those described above and these trends were not classified as they were few in number. The three meaningful price change patterns can be expressed as Pattern I (Equation 3–1), Pattern II (Equation 3–2), and Pattern III (Equation 3–3). Pattern I and Pattern II have positive economic effects upon renovation while Pattern III shows that the economic benefit due to renovation is insufficient. no j j j j Pattern I ¼ R P R P and R P R P BR AR AR P k k k k (3 1) no j j j j Pattern II ¼ R P R P and R P R P BR AR AR P k k k k (3 2) no j j j j Pattern III ¼ R P R P and R P R P BR AR AR P k k k k (3 3) Table 2. Comparative cases for each renovation case. Ren. Cases, RC RC RC RC RC RC RC RC RC RC RC RC RC RC RC RC RC RC k 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 (Household unit area, m2) (89.84) (121.07) (155.44) (57.86) (89.79) (147.74) (186.12) (89.09) (88.03) (68.27) (106.56) (59.50) (110.39) (110.20) (110.14) (92.40) (65.65) j 1 Household unit area of each CC CC 80.33 133.31 159.47 59.94 122.68 145.19 177.62 84.65 84.93 73.86 114.96 59.99 114.59 89.36 89.36 84.92 75.55 k k CC 105.04 116.13 177.52 59.65 84.84 114.69 174.48 84.88 82.65 60.83 162.21 59.25 128.03 138.51 138.51 84.55 59.94 CC 84.96 128.50 164.55 59.98 114.99 114.96 180.20 104.86 94.75 59.40 115.59 109.53 109.53 162.41 66.03 CC 84.94 129.76 134.76 84.97 140.07 196.71 84.78 84.96 59.40 109.04 107.22 107.22 84.92 59.67 CC 124.81 147.26 59.21 134.90 84.96 82.77 68.88 128.01 107.22 107.22 143.86 59.93 CC 114.74 164.73 84.98 149.09 84.86 84.87 84.43 113.57 113.57 84.46 CC 134.04 239.33 84.94 84.96 121.08 84.59 84.59 84.56 CC 157.56 84.82 139.54 113.21 113.21 226.75 CC 163.36 119.58 84.91 CC 117.70 k JOURNAL OF ASIAN ARCHITECTURE AND BUILDING ENGINEERING 235 Table 3. An example showing the calculation of price ratio for each union case. Price information (unit: 1,000 US$) Price data of RC Price data of CC Price ratio of RC on CC (%) RC RC RC CC CC CC j j j RC P P P CC P P P R P R P R P Union cases BR AR P BR AR P BR AR P k k k 1 1 RC 308.5 584.0 536.0 CC 222.0 442.0 355.0 139 132 151 U 1 1 1 2 2 CC 394.0 672.0 586.0 78 87 91 U 1 1 3 3 CC 367.5 625.0 481.0 84 93 111 U 1 1 4 4 CC 381.0 649.0 523.0 81 90 102 U 1 1 Patterns Union cases classified under each pattern Proportion of RC`s price 2 3 4 on CC`s price at each U , U , U , 1 1 1 1 4 5 6 7 time (%) U2 , U2 , U2 , U2 , U2 , 1 2 3 4 5 8 9 R P k P U , U ,U ,U ,U ,U ,U , 3 3 3 3 3 3 3 1 3 4 5 6 U ,U ,U ,U ,U , R P 4 4 4 4 4 Before k AR 1 3 renovation U ,U , 100 % 5 5 After Present renovation U , R P k BR 3 U , 7 8 9 U ,U ,U , 13 13 13 3 5 8 U16 ,U16 ,U16 (a) U , Proportion of RC`s price U , on CC`s price at each 4 6 time (%) U ,U , 6 6 U , j j 1 2 4 5 6 7 R P R P k AR k P U ,U ,U ,U ,U ,U , 9 9 9 9 9 9 Before 1 2 3 4 5 U ,U ,U ,U ,U , 10 10 10 10 10 renovation 100 % After Present U , renovation 1 2 3 4 6 R P U ,U ,U ,U ,U , k BR 13 13 13 13 13 1 2 3 4 5 6 7 8 U ,U ,U ,U ,U ,U ,U ,U , 14 14 14 14 14 14 14 14 1 2 3 4 5 6 7 8 U15 ,U15 ,U15 ,U15 ,U15 ,U15 ,U15 ,U15 , 1 2 4 6 7 9 U ,U ,U ,U ,U ,U , 16 16 16 16 16 16 (b) 1 2 3 4 5 U ,U ,U ,U ,U 17 17 17 17 17 Proportion of RC`s price on CC`s price at each U , time (%) U , 1 2 3 5 U ,U ,U ,U , 6 6 6 6 Before 3 4 After Present U ,U , renovation 7 7 renovation 100 % 1 2 3 4 5 6 7 8 U ,U ,U ,U ,U ,U ,U ,U , 8 8 8 8 8 8 8 8 R P k AR j 2 R P j k P U , R P k BR 1 2 U ,U , 12 12 5 10 U13 ,U13 (c) Figure 4. Three deduced price change patterns and union cases following each pattern. (a). Pattern I. (b). Pattern II. (c). Pattern III. 3. Development of a model for predicting price 3.1. Network architecture change patterns after renovation on MFHs The model for predicting price change patterns (MPPCP) The derived patterns provide information that helps in in MFHs upon renovation includes the following pro- deciding whether or not to perform renovations on cesses, as described earlier in Figure 1 – (1) inputting deteriorated MFHs. If an MFH for which renovation is attribute values of the target MFH corresponding to the being considered follows Pattern I or II in terms of price renovation plan and (2) predicting price change pat- change, renovations may be undertaken. However, if terns of the MFH due to renovation. To yield this output, Pattern III is predicted, it may be better not to proceed the MPPCP model is required to analyse the relationship with renovation for that MFH. Keeping these aspects in between property value (i.e., attributes) due to renova- mind, this study intends to develop a model that can tion and price change patterns. During the process of predict price change patterns after renovating a dete- identifying the relationship, this study first attempted to riorated MFH. Using this model, MFH owners will be analyze it using multivariate data analysis methods (e.g., able to evaluate the pattern of price changes in the regression analysis, decision tree method, etc.) that are renovation planning stage itself, which would help in implemented based on mathematical analysis, but there better decision-making. was a limitation to mathematically generalizing the 236 K. CHO ET AL. relationship. Therefore, the authors focused only on station, (15) educational environment, (16) distance developing a method to predict the price change pat- to the neighbourhood park, (17) distance to the bus terns of renovated MFHs using 19 attributes. In this stop, (18) distance to the general hospital, and (19) context, a methodology for defining complex nonlinear view from each unit. relationships with 19 input variables (attributes) and 3 To use the above-stated 19 attributes as model output variables (price change patterns) was required to input variables, the ratio of relative attribute values achieve the research goal. was calculated in the same way the relative price According to Patel and Jha (2016) and Kim (2004), ratio was calculated for price change pattern analysis. the artificial neural network (ANN) theory can be effec- In other words, if an arbitrary attribute i chosen from tively applied when a complex nonlinear relationship the list of 19 attributes given above is referred to as m , exists between several input and output variables. In the attribute value of the union case m U (i.e., the addition, it is not necessary to identify a mathematical attribute value of renovation case RC with the com- relationship between variables that affect the output parative case CC ) can be estimated using Equation (4). variable, and there is no limit to the number of input Table 4 shows an example of relative attribute values variables; thus, there is no need to select appropriate for 30 cases belonging to Pattern I. variables (Kim 2004). For these reasons, the ANN RC method is widely used in the field of construction m U ¼ 100 (4) CC management, and this research developed the m MPPCP model using the ANN method. To develop the ANN-based MPPCP model, input and output variables should be defined. As described 3.2. Design parameters and training criteria earlier, the input variable is set to the attribute value of To develop MPPCP using ANN, the structure of ANN MFH due to renovation and the output variable is set and various parameters that determine said structure to the price change pattern described in Equations 3– are defined, as shown in Table 5. To determine the 1, and 3–3. To set the input variables, it is necessary to values of these parameters, empirical methods and define the attributes that affect the price of the reno- genetic algorithms can be used (Hegazy, Fazio, and vated MFH. Kim, Cho, and Kim (2016), which is a pre- Moselhi 1994). In this study, we used empirical meth- vious study of this research, proposed 19 parameters ods. In addition, to operate the ANN, the activation that affect the price of the renovated MFH in two function must be determined. The activation function stages. First, dozens of candidate influencers were serves to activate the input signal value in a certain derived through an analysis of 23 previous studies dealing with factors influencing the price of general Table 5. ANN model parameters and implementation design. MFHs. Then, a correlation analysis was conducted to Parameters Setting References identify the relationship between these candidates Transfer function Sigmoid transfer Kim 2004 and price of the renovated MFH; 19 parameters were (Activation function) function found to have a significant influence on the price of Number of Hidden 1or 2 - layers the renovated MFH. By adopting these parameters, 19 Number of units in the 17, 19, 39 Hegazy, Fazio, and factors that influence the price of renovated MFHs hidden layers Moselhi 1994 Connectivity All units connected were selected as input variables; these include (1) Learning algorithm Error back- household unit area, (2) gross floor area, (3) building propagation footprint, (4) number of unit houses, (5) number of algorithm Learning rate 0.3, 0.6, 0.9 Cho, Seo, and Kang floors, (6) years elapsed since construction, (7) type of Momentum 0.7, 0.8, 0.9 2002 heating system, (8) reputation of the construction Shim, Cho, and Lee 2007 company, (9) number of parking lots, (10) unit plan, Training options Auto setting value SPSS ver. 23 (11) administrative district, (12) number of rooms in a (SPSS 23) Stopping rules Auto setting value SPSS ver. 23 household unit, (13) number of convenience facilities (SPSS 23) near the MFH, (14) distance to the nearest metro Table 4. Example of attribute ratios for each union case. Ratio of RC’s attributes on each comparative case (%) 1 2 3 4 5 18 19 Patterns Union cases m m m m m ... . .. m m Pattern U 72 29 121 45 92 ... . .. 110 100 Ⅰ U 89 26 80 34 60 65 0 (30) U 89 40 73 56 75 69 0 ¦ ¦ ¦¦¦¦ ...... ¦ ¦ U 49 13 214 35 77 ... . .. 110 100 U 56 52 71 168 48 157 100 U 35 268 130 526 100 110 100 16 JOURNAL OF ASIAN ARCHITECTURE AND BUILDING ENGINEERING 237 neuron to the input value of the neuron in the hidden parameters required for implementing the ANN model, layer or the output layer. We found the sigmoid func- includingtrainingoptions andstoppingrules,were tion to be suitable for our study as the input and out- adopted as the setting values on the SPSS software. put variables are nonlinear. In addition, because the sigmoid function is in the range of 0 to 1, it is possible 3.3. ANN model building to solve the problem of one neuron dominating the entire operation of the neural network (Kim 2004). Based on the set parameters, an ANN-based MPPCP Next, it is necessary to determine the number of was constructed. In general, to operate a prediction hidden layers. In this study, as the input and output model using ANN, it is necessary to divide the collected variables are not complicated, it is considered that data into two groups – (i) a set of learning data for sufficient learning can be achieved by determining constructing a neural network structure and (ii) a data the number of hidden layers by one or two. set for model verification. At this time, a data set must According to the criteria laid down by Hegazy, Fazio, be formulated to prevent excessive learning along and Moselhi (1994) for determining the number of with a cross-validation data set (Kim 2004). Therefore, nodes in hidden layers, in this investigation, we devel- we classified 100 union cases into case sets for learn- oped three alternatives in the number of nodes as ing, cross-validation, and testing. In this study, because follows – (i) the number of input variables (i.e., 19 there are many types of input variables (i.e., 19 MFH nodes), (ii) the number corresponding to 75% of the attributes) but the amount of data is relatively small, number of input variables (i.e., 17 nodes), and (iii) twice (i) the proportion of the learning case set was relatively the number of input variables and then plus one (i.e., high for normalization of learning and (ii) the propor- 39 nodes). tion of verification case sets was also increased as there Finally, it is necessary to set the learning rate and are three types of output variables (i.e., Patterns I, II, momentum constant. In general, it is not known and III). Accordingly, the collected union cases for whether certain values of learning rate and momen- learning, cross-validation, and verification were set at tum constant guarantee proper neural network learn- 73%, 11%, and 16%, respectively. Meanwhile, for the ing (Cho, Seo, and Kang 2002). Therefore, to generate operation of the model, Patterns I, II, and III are coded various alternatives to derive the optimal ANN model, nominally as 1, 2, and 3, respectively. the learning rate and momentum were determined according to the criteria applied by Cho, Seo, and Kang (2002) and Shim, Cho, and Lee (2007). The learn- 4. Model validation and discussion ing rate was set at 0.3, 0.6, and 0.9 while the momen- 4.1. Evaluation of the ANN model tum constant was set at 0.7, 0.8, and 0.9. In addition, the error back-propagation algorithm was applied to The input and output variables of case learning and the learning of the ANN model developed in this study. cross-validation sets were input to the ANN model. By This algorithm is based on the Delta rule that learns in controlling each parameter, a total of 108 model candi- the direction of decreasing the error value using the dates were constructed. After stopping the learning of difference between the user-specified target value and the model, based on 16 union cases for model verifica- learning output value. tion, the level of correct between actual patterns and the Meanwhile, the ANN model was developed and ana- model-predicted patterns were calculated. As shown in lysed using the SPSS V23® package (SPSS V 23), a Table 6, 108 model candidates were developed through a commercially available neural network analysis and trial and error method for each parameter. Of these, modelling software. Thus, the settings for the remained candidates 8, 10, 31, 37, and 51 exhibited a 93.80% of Table 6. Level of correct analysis for each trial depending on parameter variation. # of units in the hidden layers Lev. of correct Model candidates # of Hidden layers Layer 1 Layer 2 Learning rate Momentum (%) Remark 1 1 17 0 0.3 0.7 87.50 ¦¦ ¦ ¦ ¦ ¦ ¦ 8 1 17 0 0.6 0.9 93.80 Model 1 9 1 17 0 0.9 0.9 75.00 10 1 19 0 0.3 0.7 93.80 Model 2 ¦¦ ¦ ¦ ¦ ¦ ¦ 31 2 17 17 0.3 0.8 93.80 Model 3 ¦¦ ¦ ¦ ¦ ¦ ¦ 37 2 17 19 0.3 0.9 93.80 Model 4 ¦¦ ¦ ¦ ¦ ¦ ¦ 51 2 17 39 0.9 0.8 93.80 Model 5 ¦¦ ¦ ¦ ¦ ¦ ¦ 107 2 39 39 0.6 0.9 50.00 108 2 39 39 0.9 0.9 81.30 238 K. CHO ET AL. the level of correct and thus these five candidates were corresponding AUC under the seventh simulation condi- namedasModels1to 5. Subsequently, additional analy- tion in Table 7.Asshown in the figure, under the seventh sis was performed to find the best prediction model simulation condition, the AUC value was 0.993 for pattern among the five chosen models. In order to find the I, 0.998 for pattern II, and 0.867 for pattern III, which optimal model among these, in this study, the receiver confirms that classification by pattern using Model 2 is operating characteristic (ROC) and level of correct accord- preferable. Using the same process, it was found that the ing to variation of data set number were further evalu- average AUC value for each model during 10 simulations ated. Recently, in many previous studies adopting ANN was over 0.9 in all the models as shown in the table. method, ROC curve analysis can be used to fine a model According to the testing criterion set by Tserng et al. with best classifier performance. (2011)(if AUC ≥ 0.9, it implies an outstanding discrimina- Figure 5(a) shows an example ROC curve. These curves tion), this result implies that pattern classification by all are commonly used to analyse the trade-off between the tested models is highly desirable. sensitivity and specificity of classifiersacrossdifferent In addition to AUC analysis, based on randomly classification thresholds. In general, the more accurate extracted cases for model verification on each simula- an ANN-based model is, the farther away the ROC curve tion, the level of correct between actual patterns and is from the 45° diagonal, as shown in Figure 5(a). Thus, for model-predicted patterns was calculated. As shown in a very good model, the ROC curve appears close to a Table 7, for example, Model 1 predicted patterns with square. The area under the ROC curve (AUC) can be used 76.5% accuracy for simulation test 1 wherein the to characterise the overall discrimination of a classifica- model was implemented 17 times based on 17 random tion model. The closer the value of AUC is to 1, the better verification cases. Furthermore, Model 1 shows an is the distinction between different classes (Shenfield, average accuracy of 69.1% during 10 simulations. Day, and Ayesh 2018;Yu,Ye,andXiang 2016). With this Similarly, the average level of correct was 69.1% for background, with cases extracted randomly from 100 Model 1, 72.2% for Model 2, 66.6% for Model 3, 69.9% union cases, this study evaluated how accurately the for Model 4, and 71.1% for Model 5. five models classified the patterns of the extracted cases Because Model 2 shows outstanding pattern discri- and compared them to their actual patterns using mination ability (AUC analysis) and its accuracy of pre- the AUC analysis method. Based on the composition diction for random cases is superior to that of other ratio (i.e., 73% for learning, 11% for cross-validation, and models (72.2% of accuracy), Model 2 was finally 16% for verification) of the 100 union cases used for selected as the MPPCP. Table 8 shows the parameters developing the models, as shown in Table 7,10 simula- and performances of the proposed model. tions were carried out to verify how the developed five models distinguish the extracted cases according to their 4.2. Model application and discussion actual patterns; in these simulations, the number of data sets for each ANN implementation was varied randomly. To evaluate the applicability of the MPPCP, this study Figure 5(b)shows theROCcurveofModel 2and its applied the chosen model on an MFH planning actual (a) (b) 1.0 1.0 0.8 0.8 0.6 0.6 0.4 0.4 0.2 0.2 0 0.2 0.4 0.6 0.8 1.0 0 0.2 0.4 0.6 0.8 1.0 1-Specificity 1-Specificity th The example of ROC curve and AUC ROC curve and AUC (7 test of the Model 2) : ROC curve : AUC of Pattern : AUC of Pattern : AUC of Pattern : Area Under the Curve(AUC) (Dependent variable) (0.993) (0.998) (0.867) th Figure 5. Example of ROC and AUC analysis. (a). ROC and AUC. (b). ROC and AUC in the 7 test using Model 2. Sensitivity Sensitivity JOURNAL OF ASIAN ARCHITECTURE AND BUILDING ENGINEERING 239 Table 8. MPPCP properties. Parameter of Model 2 Properties Number of Hidden layers 1 Number of units in the hidden layers Layer #1 19 Layer #2 0 Learning rate 0.3 Momentum 0.7 Level of correct (Training, %) 100 Level of correct (Predicting, %) 93.8 Error rate (%) 6.2 renovation works. The MFH considering renovation work was built in 1992 in Seoul, South Korea. To apply the MPPCP model, 10 comparative cases of simi- lar size were selected. To implement MPPCP, it is necessary to calculate various attribute values of the renovation case and use them against those of comparison cases as model input values. Using Equation 4, the relative ratio values for 19 attributes of 10 union cases, which combine the renovation and comparative cases, are derived (Table 9). After providing the input values of each case to the MPPCP, it predicted price change patterns, as listed in Table 9. As can be seen in the table, the model predicted that Cases 1, 3, and 9 follow Pattern I and the remaining 7 cases follow Pattern II. Therefore, it can be inferred that (i) the MFH undergoing renova- tions is currently lower in price as compared to the 10 comparative cases, (ii) renovations on the MFH will result in a higher price as compared to the prices of the comparative cases, and (iii) the price will continue to increase in the future (Pattern I) or the increased price will be maintained (Pattern II). Meanwhile, 10 cases could be divided into two groups: a group including cases 1, 3, and 9 that pre- dicted Pattern I and another group predicting Pattern II. In order to identify which attributes can significantly influence the pattern prediction, the average value of each attribute in the two groups was evaluated. As a 6 1 result, “m (years elapsed since construction),”“m (household unit area),” and “m (number of parking lots)” were detected as the critical attributes, because the difference in the average values for the three attri- butes between the two groups were evaluated as sig- nificant. For example, the average value of m for group 1 was about 141.33% (i.e., (192 +156 +76)/3), while the average value of m6 for group 2 was 79.14% (i.e., (83 +78 +69 +74 +100 +76 +74)/7). The gap for this attribute was remarkable, and it could be interpreted that the years elapsed since the MFH’s construction increased; the renovation of such an MFH could be expected to yield economic benefit. Similarly, if the MFH has a relatively small house unit area (m , 47.00% of group 1 and 75.57% of group 2) and number of parking lots (m , 77.33% of group 1 and 133.71% of group 2) compared to the surrounding cases, such an MFH could be regarded a good candidate for renova- tion in terms of economic value increase. Table 7. ROC analysis results and level of correct for each model. Division Number of data set Model 1 Model 2 Model 3 Model 4 Model 5 Area under Area under Area under Area under Area under Lev. Lev. Lev. Lev. Lev. ROC* curve (AUC) ROC curve (AUC) ROC curve (AUC) ROC curve (AUC) ROC curve (AUC) Simulation For of of of of of No. For training cross- validation For verification corr. P1 P2 P3 corr. P1 P2 P3 corr. P1 P2 P3 corr. P1 P2 P3 corr. P1 P2 P3 1 69 14 17 76.5 0.998 0.974 0.969 82.4 0.857 0.848 0.859 70.6 0.886 0.83 0.844 76.5 0.901 0.869 0.787 76.5 0.929 0.939 0.868 2 70 12 18 55.6 0.951 0.983 0.903 61.1 0.948 0.967 0.907 50.0 0.940 0.984 0.928 55.6 0.958 0.981 0.885 61.1 0.922 0.934 0.888 3 74 9 17 70.6 1.000 0.998 0.957 82.4 0.999 0.999 0.944 76.5 1.000 0.999 0.993 70.6 1.000 1.000 1.000 82.4 1.000 1.000 0.987 4 72 6 22 68.2 0.987 0.968 0.960 68.2 0.980 0.947 0.948 68.2 0.963 0.956 0.930 63.6 1.000 0.989 0.994 72.7 0.977 0.956 0.917 5 76 5 19 63.2 0.885 0.885 0.862 78.9 0.994 0.989 0.956 78.9 0.958 0.952 0.886 73.7 0.987 0.994 0.864 73.7 0.941 0.935 0.894 6 74 11 15 73.3 0.985 1.000 0.976 66.7 0.995 0.996 0.965 66.7 0.948 0.931 0.871 66.7 0.895 0.893 0.841 60.0 1.000 1.000 1.000 7 71 10 19 73.7 0.989 0.996 0.906 84.2 0.993 0.998 0.867 73.7 0.990 0.999 0.885 78.9 0.987 0.998 0.899 73.7 0.989 0.999 0.887 8 76 6 18 61.1 0.987 0.939 0.915 55.6 0.989 0.938 0.916 55.6 0.987 0.945 0.925 55.6 0.992 0.971 0.939 55.6 0.992 0.969 0.934 9 73 9 18 77.8 0.925 0.949 0.926 77.8 0.935 0.918 0.907 61.1 0.964 1.000 0.993 72.2 0.973 0.951 0.948 83.8 0.933 0.887 0.904 10 74 12 14 71.4 0.874 0.802 0.851 64.3 0.917 0.960 0.961 64.3 0.892 0.842 0.850 85.7 0.861 0.927 0.903 71.4 0.865 0.861 0.877 Average 69.1 0.958 0.949 0.922 72.2 0.960 0.956 0.923 66.6 0.952 0.943 0.910 69.9 0.955 0.957 0.906 71.1 0.954 0.948 0.915 * ROC = Receiver operating characteristic 240 K. CHO ET AL. Table 9. Model application & predicted patterns. Rate of each attribute for each test case Predicted Patterns 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 Union cases m m m m m m m m m m m m m m m m m m m By the model U 37 161 81 829 68 192 0 20 42 67 100 50 100 100 100 125 50 150 100 1 U 73 118 128 219 107 83 0 100 127 100 100 100 100 100 100 100 100 100 100 2 U 52 105 119 279 100 156 0 50 94 100 100 67 80 114 72 91 167 100 100 1 U 59 49 117 172 107 78 0 100 125 100 0 67 95 80 60 67 71 90 100 2 U 65 18 99 40 107 69 0 100 89 100 0 67 83 89 58 67 71 90 100 2 U 93 110 136 195 100 74 0 100 227 100 0 67 118 114 64 111 100 100 0 2 U 126 171 98 214 100 100 0 100 118 100 0 67 100 114 95 143 125 113 100 2 U 61 230 113 602 115 76 0 33 125 100 0 67 118 160 90 77 250 225 0 2 U 52 167 113 369 107 76 0 100 96 100 100 67 80 89 75 200 167 75 0 1 U 52 986 218 1598 300 74 0 20 125 100 100 100 118 100 100 71 125 225 100 2 7 11 19 *m ,m , and m are nominal variables; thus, their values are coded as 1 or 0. Based on the above application results, this research After applying the selected model to the deterio- enables decision-makers to grasp the economic effect rated MFH which is currently planning a renovation, it of MFH renovation and it is expected that they would is shown that the price change pattern of the MFH be able to decide whether or not to carry out after the renovation is well predicted positively. Using renovations. the MPPCP presented in this study, it is possible to predict price change patterns that would be experi- enced by MFHs upon renovation. Such prediction can 5. Conclusion be done at the renovation planning stage, which would be of great help to the MFH owner in coming Interest in the renovation of deteriorated MFHs has to a decision. In future, based on additional collections been growing in recent years. However, many MFH of renovation cases and their comparative cases, a owners cancel their renovation plans as economic model for predicting price changes, and not only advantages after renovation are often not guaranteed. price change patterns, due to renovations should be Although the economic effects of a renovation project developed. are very important for its success or failure and in making decisions on the fate of the project, existing literature on this topic is minimal. Therefore, in this Acknowledgments study, we developed a model (MPPCP) that can predict This work was supported by the research funds from Chosun the economic effects of renovation on older MFHs. The University, 2018. effects of renovation were evaluated based on price changes at three time points (before renovation, after renovation, and present) and price changes due to Disclosure statement renovation were measured by measuring relative No potential conflict of interest was reported by the authors. price changes with respect to surrounding comparison cases. The MPPCP developed in this study was based on Notes on contributors the ANN technique. To develop the model, 17 renova- tion cases and 103 comparison cases were considered Dr. Kyuman Cho is an Associate Professor in the School of Architecture at Chosun University. Before joining Chosun for retrieving attribute data and price data. From the University, he was a Post-doctoral research fellow at Purdue collected data, (i) the price change pattern of renova- University in USA and worked as a Practicing engineer in tion cases at three time points was analysed and (ii) construction industry. He has published about 80 journal relative attribute values corresponding to 19 para- papers and peer-reviewed conference papers, related to meters after renovation were calculated. Furthermore, Construction Engineering and Management. His primary the relationship between price change patterns and research areas include Economic analysis on construction systems, Building renovation techniques, and Automation these attributes was derived using the ANN method. in construction operation. During the process of applying the ANN method, para- Jaesung Kim is a Graduate research assistant in the School of meters required to apply the ANN were designed to Architectural Engineering at Chosun University. His primary generate model candidates and finally 108 candidate research areas include construction engineering and man- models were detected by combining various para- agement, building renovation, and cost estimation in the meters that define the ANN method. From these can- project early phase. didates, five models with relatively high prediction Taehoon Kim received Ph.D. in Civil, Environmental and accuracy were selected. Subsequently, by analysing Architectural Engineering from the Korea University. Since, AUC and the level of correct for randomly retrieved 2015, he is an assistant professor at the School of verification cases, the final model MPPCP with the Architecture of Chosun University. His research interests are highest prediction accuracy was derived. focused on introducing advanced construction methods and JOURNAL OF ASIAN ARCHITECTURE AND BUILDING ENGINEERING 241 technologies for improving construction productivity. He Kim, J., K. Cho, and T. Kim. 2016. “Factors Determining the recently conducted a research project on developing a pro- Price of Remodeled Multi-family Housing.” Korea Journal of cess management technique for efficient construction auto- Construction Engineering and Management 17 (3): 13–22. mation system. doi:10.6106/KJCEM.2016.17.3.013. Kim, J., K. Cho, T. Kim, and Y. Yoon. 2018. “Predicting the Dr. Taehoon Hong is Underwood Distinguished Professor of Monetary Value of Office Property Post Renovation Work.” Department of Architecture and Architectural Engineering at Journal of Urban Planning and Development, ASCE 144 (2): Yonsei University. He is the Associate Editor of Renewable & 04018007. doi:10.1061/(ASCE)UP.1943-5444.0000434. Sustainable Energy Reviews, Elsevier and Journal of Kim, K. H., and H. S. Baik. 2015. “Comparative Analysis between Management in Engineering, ASCE. He is an editorial board RSMeans and Standard Estimating System for Selective member of Applied Energy, Elsevier as well as Energy and Finish Demolition in Remodeling Projects.” Journal of the Buildings, Elsevier. He is also an Associate Editor of Architectural Institute of Korea Structure & Construction 31 (3): Sustainable Design and Construction, Frontiers in Built 73–80. doi:10.5659/JAIK_SC.2015.31.3.73. Environment. Dr. Hong’s current research topics include Kim, W. H., H. S. Cha, D. W. Shin, and K. R. Kim. 2013. “Profit/ renewable energy system, energy performance, solar photo- Cost Prediction Model for Economic Feasibility voltaic system, indoor environmental quality, and sustain- Assessment on Aged-housing Remodeling Projects.” able building. Journal of the Architectural Institute of Korea Structure & Construction 29 (4): 85–92. http://www.dbpia.co.kr/Article/ NODE02148676 Korea remodeling association. 2011. The References Institutional Improvement Committee Seminar. Seoul: Republic of Korea. Cho, H. H., D. S. Seo, and K. I. Kang. 2002. “Development of Lee, J. B. 2005. “Economic Analysis and Preference Study on Building Construction Cost Forecasting Model Using Apartments according to Business Method.” Journal of the Artificial Neural Networks.” Journal of the Architectural Architectural Institute of Korea Structure & Construction 21 Institute of Korea Structure & Construction 18 (8): 107–114. (8): 137 –144. http://www.dbpia.co.kr/Article/ http://www.dbpia.co.kr/Article/NODE00365731 NODE00618931 Cho,J.Y.,K.W.Lee,H.S.Yoo,andY.S.Kim. 2012. “AStudy on Lee, K. S., K. H. Kim., K. R. Kim., and D. W. Shin. 2007. Vitalizing Plans of Extension Remodeling for Apartment “Profitability Influence Factor Analysis by Apartment Housings.” Korea Journal of Construction Engineering and Remodeling Case Study.” Korea Journal of Construction Management 13 (4): 33–47. doi:10.6106/KJCEM.2012.13.4.033. Engineering and Management 8(2):127–135. http://www. Cho, K., J. Kim, and T. Kim. 2019. “Decision Support Method ndsl.kr/ndsl/search/detail/article/articleSearchResultDetail. for Estimating Monetary Value of Post-renovation Office do?cn=JAKO200734515087128 Buildings.” Canadian Journal of Civil Engineering online Ministry of land, infrastructure and transport. 2016. “Declared published 46: 1103–1113. doi:10.1139/cjce-2017-0692. Price of Real Estate in Korea Inquiry System.” Accessed Cho, K., and Y. Yoon. 2016. “Decision Support Model for January 2016. http://www.realtyprice.kr/notice/ Determining Cost-effective Renovation Time.” Journal of Patel, D. A., and K. N. Jha. 2016. “Evaluation of Construction Management in Engineering, ASCE 32 (3): 04015051. Projects Based on the Safe Work Behavior of Co-employ- doi:10.1061/(ASCE)ME.1943-5479.0000418. ees through a Neural Network Model.” Journal of the Safety Choi, Y. A., and B. H. Song. 2006. “A Study on the Science 89: 240–248. doi:10.1016/j.ssci.2016.06.020. Characteristics of the Price Factors in Apartment Houses.” Shenfield, A., D. Day, and A. Ayesh. 2018. “Intelligent Intrusion Journal of the Korean Housing Association 18 (2): 75–82. Detection Systems Using Artificial Neural Networks.” ICT http://www.dbpia.co.kr/Article/NODE00780984 Express 4(2):95–99. doi:10.1016/j.icte.2018.04.003. Han, J. Y., and D. W. Shin. 2012. “A Study on the Story Shim, J. H., C. H. Cho, and S. H. Lee. 2007. “The Industrial Land Increase for Securing the Feasibility of Aged-housing Price Appraisal Based on Artificial Neural Network.” Journal Remodeling.” Korea Journal of Construction Engineering of the Korea Planning Association 42 (5): 223–232. http:// and Management 13 (3): 152–159. doi:10.6106/KJCEM. www.dbpia.co.kr/Article/NODE00914539 2012.13.3.152. Tserng, H. P., G. F. Lin, L. K. Tsai, and P. C. Chen. 2011. “An Hegazy, T., P. Fazio, and O. Moselhi. 1994. “Developing Enforced Support Vector Machine Model for Construction Practical Neural Network Applications Using Back-propaga- Contractor Default Prediction.” Automation in Construction tion.” Computer-Aided Civil and Infrastructure Engineering 9 20 (7): 837–863. doi:10.1016/j.autcon.2011.03.002. (2): 145–159. doi:10.1111/j.1467-8667.1994.tb00369.x. Yeon,J.H.,H.S. Lee, M.S. Park, S.Y.Kim, and J.S. Ahn. 2014. Jin, R., K. Cho, C. Hyun, and M. Son. 2012. “MRA-based CBR “Feasibility Study on Remodeling Project by Using Real Model for Cost Prediction in the Early Stage of Option Model: Focusing on Apartment House Remodeling.” Construction Project.” Expert Systems with Applications 39 Korea Journal of Construction Engineering and Management (5): 5214–5222. doi:10.1016/j.eswa.2011.11.018. 15 (1): 39–50. doi:10.6106/KJCEM.2014.15.1.039. Kim, D. J., Y. J. Choi, and G. Y. Kim. 2010. “A Study on the Yoon, Y. S., and S. W. Lee. 2012. “Report: System Improvement Agreement Rate Encouraging Restriction Factor or Multi- Plan for Revitalization Remodeling of the Aging Multi-Family family Housing Remodeling.” Korea Journal of Construction Housing.” Construction & Economy Research Institute of Engineering and Management 11 (1): 122–129. Korea(CERIK). doi:10.6106/KJCEM.2010.11.1.122. Yu, X., C. Ye, and L. Xiang. 2016. “Application of Artificial Neural Kim, G.-H. 2004. “Construction Cost Prediction System Based Network in the Diagnostic System of Osteoporosis.” on Artificial Intelligence at the Project Planning Stage.” Neurocomputing 214: 376–381. doi:10.1016/j.neucom.2016. PhD diss., Korea University. 06.023.
Journal
Journal of Asian Architecture and Building Engineering
– Taylor & Francis
Published: May 3, 2020
Keywords: Renovation work; multi-family house; artificial neural network; economic benefit due to renovation
