Classification of typical layout design for Indonesian apartment buildings

. Along with rapid urbanization, housing development increased in major cities worldwide. Public and private apartments have been constructed rapidly in Indonesia to accelerate housing provision, particularly in urbanized areas of major cities. From the early era of apartment construction to the present, many design prototypes have been implemented for various specific purposes such as energy-saving. Nevertheless, no comprehensive study has determined a typical layout of apartment buildings in Indonesia. The typical layout design is useful as the base model for building simulations. This study sought to classify the typical layout design of public and private apartments in Indonesia, using the selected building drawing data ranging from 1986 to 2021. We obtained primary data consisting of 268 public apartments from 22 cities and 268 private apartments from 9 major cities in Indonesia. To determine the typical layouts, both qualitative and quantitative analyses were conducted. As a result of this study, a pair of typical public and private apartment buildings which include the 1-bedroom unit and 2-bedroom unit types have been selected as the typical layout of the Indonesian apartment building. The results of the typical layout classification could be utilized as a standard testing model for designing future apartment building prototypes.


Introduction
Standard design and prototypes have frequently been used by architects as design guidelines. When architecture was regarded more as an industrial product, the standard design had already begun to be used. Neufert published the German building industry's standard design in 1936 in order to promote a faster design process in actual construction [1]. Numerous professional organizations or academic institutions, such as the Architectural Institute of Japan (AIJ), establish standards for residential buildings [2]. In some other countries, government agencies also develop the standard design.
Design prototypes can be developed in accordance with specific objectives. For a better energy-performance building, a prototype design can be proposed to enhance several key parameters such as passive design, natural lighting design, building energy retrofitting, etc [3]. The study conducted by Tereci et al. (2013) confirmed that there is an impact of building typology on building energy performances [4].
Rapid urbanization demands a high growth of the basic infrastructure, especially affordable housing [5]. In the case of Indonesia, housing development in cities kept on increasing in parallel with the high rates of urbanization. Even so, in 2020 the housing backlogs still accounted for about 7.64 million units which mainly (84.8%) to facilitate low-income groups [6]. Globally, towards the target of carbon-neutral by 2050, all new buildings together with approximately 20% of existing building stock would need to be zero-carbon-ready by 2030 [7]. The housing sector is particularly important, because the percentage of residential buildings has usually much higher, compared to other sectors in terms of energy consumption and carbon emissions [8]. Looking at the Indonesian policies on fostering apartment development to reduce housing backlogs, and due to the country's ambitious target to reach net zero emissions, it is important to determine apartment building typical layouts as a base model for future low-carbon design prototypes.
In Indonesia, so far there were only a few studies related to apartment building typology. Some of them classified the Indonesian apartments based on several aspects, including the type of ownership, type of developers, type of construction, type of unit size and type of management with different levels of authority. A specific typology study related to the physical attributes of Indonesian apartment buildings using actual building data has never been conducted. Ever since, low-cost public apartment development has become an important government policy to foster housing provision with a higher density of population in urban areas, especially for low-income groups. For example, since 2007, the Indonesian government has initiated a national project called "One Thousand Towers" to boost the development of apartments, followed by the "One Million Housing Program" to accelerate housing provision in 2015 [6]. From the early era of apartment building construction to the present, many building prototypes of apartment designs have been implemented in Indonesia. However, there are few comprehensive studies on the typical layout of apartment buildings in Indonesia despite its importance.
The objective of this study is to conduct a comprehensive classification of Indonesia's typical apartment building layout. The result will be proposed as the standard testing model for apartment building simulations toward zero-carbon goals.

Data collection
In most developing countries such as Indonesia, the provision of public housing is highly demanded to reduce housing backlogs, especially in urban areas. In Indonesia, public housing especially in relatively densely populated urban areas tends to adopt the type of flats or apartments. The provision of low-cost public apartments for the lowincome groups called Rusunawa (RS) has been started to accumulate in the early 1980s. Specifically for this study, we collected a total of 268 public apartment technical drawings as well as 268 other buildings that represent private apartments from several cities as primary data for the classification. The construction year of the obtained data ranges from 1986 to 2021. 268 public apartment buildings that were collected consist of 162 1-bedroom units (1BR) and 106 2-bedroom units (2BR). While for private apartment buildings, a total of 268 private drawings were investigated, comprising 470 samples of 1BR units and 417 samples of 2BR units, as shown in Table 3.
A total of 22 cities were selected accordingly to acquire data for public apartment buildings, where most of them were classified as metropolitan and large cities (see Fig. 1). As shown in Table 1, 73.8% of the overall data were collected from the megapolitan and metropolitan city, i.e. Jakarta, cities, Surabaya and Bandung, while the rest of data acquired from large and medium cities. The distribution of data samples was reflecting the policy of Indonesian government to promote quality housing settlements for low-income groups in urbanized areas.  On the other hand, since the private apartment has their own specific target groups based on market demands, the sample distribution of private apartments is not evenly spread across Indonesia. Thus, in Indonesia, most private apartments are situated in major cities that are densely populated. We, therefore, conducted the data collection for the following nine major cities, including Jakarta, Bogor, Depok, Tangerang, Bekasi, Bandung, Surabaya, Makassar and Medan (see Fig. 1). All the selected cities are categorized into the megapolitan and metropolitan cities (see Table 2).  Fig. 2 shows one of the samples of building drawings for a public apartment located in Bandung, which is one of the typical mid-rise public apartments. For a public apartment, there is only 1 typical layout plan for each building, whereas private apartments consist of various types of units. The type of architectural drawings that were collected for public apartment buildings was including detailed engineering design plans, as-built drawings and prototype drawings. Meanwhile, for private apartments, the architectural drawings that consist of building key plans, typical unit drawings and 3D visual images were collected from the brochure provided by real estate agencies and from related online sources (see Fig.  3).  In general, private apartments are classified into the following two types with respect to the selling prices: condominium (CD) and Rusunami (RM). According to the Indonesian Government Regulation No. 13/2021, both CD and RM can be categorized as commercial private apartments, but in the housing market, CD mainly targets the upper-middle-class, while RM is for the low-cost owned apartments.

Classification
The classification was conducted through a combination of qualitative and quantitative analysis for public and private apartments, respectively. First, the apartment samples were classified qualitatively through visual observation of the architectural drawings at the building level and unit level. For the public apartments, the building-scale qualitative classification was performed with respect to the building height, type of corridors and type of main upper construction materials. There are 3 building height categories available for public apartments according to the Indonesian government regulation. They are including low-rise buildings (L) which contain 4 or less floors, medium-rise buildings (M) which ranging from 5 to 8 floors and high-rise buildings (H) which has more than 8 floors. Each building has 2 type of building corridors, either single-loaded corridor (SLC) or doubleloaded corridor (DLC). The type of main upper construction was also identified, consisting of the precast system (PC) and conventional reinforced concrete (CRC). On the unit-scale qualitative classification, we manage to classify the type of floor plans which consist of open plan (O) or fixed plan (F), the type of unit sizes which consist of type 21, type 24, type 27, type 30, type 36 and type 45 (all representing their sizes in meter square). On the unit scale observation, we also considered the availability of a balcony. Some units were equipped with a balcony (W), while others were not (N).
Meanwhile, for the private apartments, the qualitative classification on the building scale was performed based on the type of building, either Rusunami (RM) or Condominium (CD), the building heights and the construction years. Building heights classification for private apartments were categorized into 3 types, the 1st type has more than 21 floors (H1), the 2nd type is ranging from 21 to 40 floors (H2) and the 3rd type has more than 40 floors (H3). The category for the years of construction consists of 3 types, below 2011 (Y1), range between 2011-2016 (Y2), and built after 2016 (Y3). The unit scale qualitative classification uses the same classification as the public apartments.
Further classification analyses were made on spatial arrangements on the unit scale, where there are 3 levels of observations. First is due to the availability of an individual living room and the connection with the bedrooms, which is divided into 3 types. Type A of the 1BR represents the type of unit without a dedicated living room, whereas Type B has a dedicated living room adjacent to the bedrooms elongated on one side of the unit. Type C has a dedicated living room adjacent to the bedrooms located near the corridor or the outdoors. For the 2BR units, Type A has 2 bedrooms positioned along the side of the unit, while in Type B, the bedrooms are located towards the outdoors. The second level of observation is to categorize the toilet inside the unit, whether it is positioned near the corridor, near outdoors, or with no toilet. The third level of observation is regarding the type of balconies in the unit. Type a has a balcony inside the unit, Type b has a half-size balcony outside of the unit, Type c has a full-scale balcony on the outside of the unit, whereas Type d has no balcony. The results of the above qualitative classifications were given certain codes that were used as dummy variables in the following quantitative analysis involving Principal Component Analysis (PCA) and cluster analysis. The illustration of this spatial arrangement classification on the unit scale is shown in Fig. 4.

Qualitative analysis
As basic building-scale information, it was found that most public apartments (75%) were constructed of conventional reinforced concrete, compared with those of precast systems (25%). According to the information from several public apartment developers, the precast systems were considered to be insecure due to some leaking that occurs frequently, particularly in utility areas. The majority of public apartment data samples come from low-rise buildings (50.4%), followed by mid-rise and high-rise buildings (35.8% and 13.8%), respectively.
It was also found that the double-loaded corridor types accounted for 59.7% of all the samples, while singleloaded corridor types accounted for 40.3%. As a result of the primary visual observation, it can be concluded that low-rise apartments (L) with conventional reinforced concrete (CRC) and a double-loaded corridor system (D), which is coded as CRC-L-D, has the highest frequency in this study, accounting for 35.4% of all samples. This is followed by the mid-rise apartments (M) with conventional reinforced concrete and a single-loaded corridor system (S), which is coded as CRC-M-S with 20.5%, followed by the high-rise apartments (H) with a precast system (PC) and the double-loaded corridor system (PC-H-D) with 13.1%.
As a result of the above primary qualitative analysis, a total sample of 268 public apartment buildings was classified into 11 major building types as illustrated in Fig. 5. There are various varieties for each type, but this classification is based on similarities in space configuration and arrangements. The simple slab-based building form is the most commonly used with several variations on the position and type of housing units, horizontal circulation (corridor), vertical circulation (stairs or elevators) and the use of vertical voids. According to the data samples, only 34% of buildings were equipped with vertical voids, while the other 66% were not. The use of these vertical voids is decreasing throughout the year whereas, after 2015, they are only 25% of public apartments use this feature. The diagrammatic drawing shown in Fig. 5 is an illustration of how the typical public apartment building layout is according to the data samples categorized by the attributes that are attached to each type. At the unit level, Types 24 and 36 were found to be the most common types with 35.4% and 35.1% respectively. It was found that 60.8% of public apartments in the data sample had balconies, while the rest were not. On the 2BR unit type, the majority of those balconies were positioned inside the units (68%) and not extended toward the outdoors. Oppositely in the 1BR type, most balconies were positioned on the outside of the unit (93.8%). The average size of the balcony on the 2BR type was found bigger than the size of the 1BR type with 2.67m 2 and 2.17m 2 , respectively. This condition made it possible for the user to use the balcony as the extension of a living room, storage room, or even bedroom. This finding is in line with the research conducted by Kisnarini et al (2015), which stresses the space adaptability of balcony space as the extension of the interior space [9]. Most public apartment units were designed of fixed-plan with 71.3% as opposed to the open-plan or studio type with 28.7%. Therefore, from the above primary qualitative analysis, it can be said that the fixed-plan unit type (F) with a unit size of 36 m 2 without a balcony (N), which is coded as 36-N-F is the most prevalent type in the public apartments (22.8%).
Based on the visual observation on the unit scale, we manage to identify 11 types among the 1BR units. Type B1d was found to be the most typical type, accounting for 30.9% of all samples. The B1d is the type of unit where the living room and bedroom are located on the side of the unit, which is adjacent to the other unit, extending from the entrance to the exterior wall. Some of the most typical unit types including B1d likewise Type A1d lack balconies and have interior spaces instead. Those 2 most common unit types were mostly constructed after 2015, whereas the other frequent types that were built between 2007 to 2015, Type B1b, were equipped with balconies. Almost half of the 1BR units (47.5%) can also be referred to as studio-type units, which did not employ permanent dividers to separate the living room from the bedroom.
On the other hand, the 2BR units were found to be varied into 9 types. Type B1a was found to be the most typical type (45.3%), followed by Type A2a and B1d with 20.8% and 15.09, respectively. Most of the 2BR units were built after 2007, after the issue of a 1000-tower-type housing policy by the government to reduce housing backlogs for young families in urban areas. In terms of unit size averages, the 2BR unit was 33% more spacious compared to the 1BR type, and typically has balconies (84.0%). Each of the unit scale types was given the coding to be used as a dummy variable for the quantitative analysis as explained in the method part.

Quantitative analysis
The quantitative analysis was conducted through separated 2 datasets, first for the 1BR unit type (n=162), and second for the 2BR type (n=106). Principal Component Analysis (PCA) was used in quantitative analysis with SPSS. A total of 14 and 11 variables, including the dummy variables obtained from the abovementioned qualitative analysis, were involved in the PCA for the 1BR and 2BR unit types respectively. As a result, four components were identified primarily due to the eigenvalue of >1.
In PCA, KMO test values were all over 0.5 for both 1BR and 2BR datasets, which means that the index of factorial simplicity (IFS) meets the sampling adequacy requirement [10]. Bartlett's test of sphericity values for both 1BR and 2BR datasets was also significant (<0.001), showing that the variables used were suitable for the factor analysis. After meeting the conditions of the KMO and Bartlett's test, we performed the hierarchical cluster analysis using the factors from the PCA analysis. The Ward Linkage method and the Squared Euclidean distancing method were used to form the dendrogram for the cluster analysis. The cluster options were observed to identify the most similarities in characteristics among data samples listed in the clusters.
Based on the result of the hierarchical analysis, we identified several cluster options for the 1BR public apartment unit type. After observing the data samples population of each cluster, we manage to select the 4cluster option as the most appropriate for 1BR public apartment unit type. The 4-cluster option shares a balanced data sample population with 27,8%, 19.1%, 20.4% and 32.7%, respectively (see Table 4). Each cluster has different specific characteristics. Cluster 1.1 dominated with unit Type A1b, while on the other hand, for the 2BR public apartment unit type, we manage to conclude the 2 clusters option, namely cluster 2.1 and cluster 2.2 as the most appropriate compared with the other options. Cluster 2.2 dominated the total data sample population with 71.7% among all 2BR unit types, characterized by unit Type A2a. while Cluster 2.1 were fully characterized by unit Type B1a. The illustration of the 1BR and 2 BR public apartments are shown in Fig. 6. Table 4. Characteristics of (a) 1BR and (b) 2BR RS. We carried out a descriptive frequency analysis on the results of the cluster analysis of typical unit types for 1BR and 2BR public apartment buildings in order to correlate the unit scale typical layout to the building scale types. Fig. 7 and Fig. 8 present the frequencies of combinations of extracted clusters for unit types with the typical building types for 1BR units and 2BR units, respectively. As shown in Fig. 8, 1BR unit types were often used in low-rise (L) and medium-rise (M) public apartments with the use of conventional reinforced concrete (CRC) for the upper structures. The B1d type, the most common kind of 1BR unit, is more correlated with double-loaded corridor systems than are the B1b and A1d unit types which mostly use single-loaded corridors. Type B1a, as mentioned previously as the most common typical unit on a 2BR unit, is strongly correlated to the double-loaded corridor system. On the other hand, Type A2a as the second most frequent unit type in 2BR uses a single-loaded corridor system. The use of 2BR unit type in the data samples is found to be evenly spread among each height category of low-rise, medium-rise and high-rise buildings.   After performing the quantitative analysis, the six different public apartment unit typical layouts were linked to the building scale observation based on the frequency analysis. According to the total number frequency of samples (see Table 4), we conclude that Type A1b and B1d as the most representative types of 1BR unit types, while Type A2a and B1a represent the 2BR unit types. Each unit type representative is then linked to the most correlated types on the building scale. As shown in Fig.  10, Type B1d of 1BR public apartments corresponds the most to the low-rise building type that characteristically uses conventional reinforced concrete for the main upper structure and has the double-loaded corridor system (CRC-L-D). Meanwhile, Type A1b corresponds to type CRC-L-S. For the 2BR types, Type A2a typically corresponds to Type CRC-M-S, while B1a is most typically linked to CRC-L-D building types. The illustration and characteristics of each typical layout for 1BR and 2BR public apartments in relation to their building scale types are presented in Fig. 10, where most frequent data samples are illustrated with thicker lines which gradually decrease for fewer data samples.

Qualitative analysis
After we qualitatively categorized 1BR private apartment units with three levels of observation, type A1 makes up the largest cluster, making up 37% of the 1BR samples (see Table 6). This type is shown as a studio type with an average unit size of 24.1 m 2 , a restroom on the side facing the hallway, and a partial balcony inside the unit. In terms of sizes and balcony style, types A1, A2, A4 differ slightly. Additionally, the balcony was not part of Type A5, while the restroom was outside of Type A3. Type B1 had a separate living room, making it the biggest unit type with 38.7 m 2 . The illustration of the 1BR typical layout is presented in Fig. 11.
Drawings of 2BR private apartments were also visually separated into two primary types: A and B, similar to the 1BR classification. Two bedrooms of type A are organized along the depth of the unit and side by side, whereas two bedrooms of type B are arranged along the width of the apartment. Following that, both categories were divided into more manageable groupings and subgroups using the previously mentioned criteria. Table 6. Characteristics of the typical 1BR private apartment.

Fig. 11
Typical unit layout of 1BR private apartments.

3.2.2
Quantitative analysis We looked at 2BR type A and type B individually to boost the homogeneity among clusters. The final clustering result for 2BR types A and B shows that, with a percentage of 21%, Type A1 was the largest group within all Type A data samples, as indicated in Table 7. This particular apartment type contains a bathroom beside the bedrooms and on the corridor side, as well as a half balcony attached to a separate living area. Additionally, Type A2 and A3 made up 19 and 18 percent of the population, respectively, while Type A4 and A5 made up 13 and 12 percent. Only 9 percent and 8 percent of all types were A6 and A7, respectively. Contrarily, Type B1 was the most prevalent type of 2BR Type B units, with its cluster accounting for 40% of that. In this design, two bedrooms were built horizontally in an effort to have additional windows for ventilation and lighting. The half balcony is reached from the living room and extends outside, while the toilet is situated at the corner between the master bedroom and corridor side. Fig.  13 shows that types B2 and B3 differ slightly from type B1 in terms of bedroom arrangement; their percentages are 30 and 19 respectively (see Table 7). Type B4 makes up 12% of all cases, which is less than other types. Two bedrooms were exhibited in this unit type, facing one another and connected by a tiny balcony in the middle.
For both CD and RM apartments, Type A makes up the majority of unit types with 76 and 88 percent of each share, respectively. Despite the fact that 1BR unit Type C accounted for 16 percent of CD apartments, it only made up 5% of RM apartments. The proportions of type B RM and CD units were equal to each other, or around 8%. It has been observed that between 1983 and 2022, the popularity of CD 1BR units Type A decreased from 91% to 75%, and between 2005 and 2022, the popularity of RM decreased from 100% to 80%. This pattern indicates that rather than a studio apartment, residents in Indonesia now choose to possess a Type B or C apartment, which has a separate bedroom and living area.
The frequency analysis result was used to determine the construction trend of 2BR units, which was also looked into. Form B was the most prevalent type among CD apartments, accounting for 53% of all units compared to 47% for type A. The percentage of CD 2BR unit type B increased significantly between 1988 and 2016 from 39% to 63% before dropping to 38% in the following years. This implies that CD 2BR unit type A also has a reversal layout. In contrast to CD apartments, Type A units were twice as common as type B units in RM apartments, with rates of 70% and 30%, respectively. Between 2005 and 2022, the proportion of Type A increased gradually from 67 to 76 percent, whereas Type B had a sharp fall from 33 to 24 percent.
The correlation between apartment buildings and typical units is covered in this section. We divided CDs and RMs into groups based on heights and building eras. Six RM types and nine CD types in total were used for the analysis. To observe the link between buildings and units independently, we created statistics. The percentages of six typical 1BR unit types in CD and RM buildings are shown in Fig. 13 and Fig. 14. In comparison to other types, Type A1 was the most prevalent 1BR unit type, and it was more prevalent in structures with 21 to 40 stories in height. Additionally, this unit type was popular in RMs from 2011 to 2016, but not in CDs until after 2016. Despite the fact that Type B1 was built on CDs after 2010, none of them appeared in RMs. Similarly, Fig. 15 and Fig.  16 show that the other 2BR unit types in CDs were dominated by types A1 and B1. From 2016 onward, they were primarily built in 21-40 story structures. Before 2011, however, there was only a limited percentage of type B1 in RMs with 21-40 stories.     Based on frequency analysis regarding the percentages of each typical private apartment data sample (see Table 6 and Table 7), we conclude that Type A1 and B1 suit best represent the typical layout of 1BR unit types. The same coding of Type A1 and B1 also concluded as the representative typical layout of the 2BR unit types. The illustration summary of the typical unit for the private apartments is shown in Fig. 17.   Fig. 17 Typical unit and building scale layout of (a.) 1BR and (b) 2BR private apartments.

Conclusion
The objective of this study is to determine the current existing typical layout of Indonesian apartment buildings. Despite having different sources of data samples between public and private apartment buildings, we have applied both qualitative and quantitative methods for apartment building typical layout classification in Indonesia.
As a summary, we proposed 2 typical layouts for each 1BR and 2 BR public and private apartment in Indonesia. A total of 8 typical layouts were selected based on the number of frequencies on the data samples after the qualitative and quantitative analysis was conducted, which include the use of PCA and cluster analysis. Type A1b and B1d were selected for the public apartments to represent the 1BR, while Type A2a and B1a are selected for the 2BR public apartment. The 1BR and 2BR private apartment typical unit layouts were represented by Type A1 and B1.
This study has overcome the barrier of accessing numerous data and information on actual apartment building design from both government documentation and online public data. The use of these data proves to be practical to analyze and leads to straight forwards conclusions.
From the results of the typical layout as the result of the study, it is concluded that the overall objective was achieved. Each cluster showed strong characteristics that differentiate them from the other. Within these varieties of typical layouts in terms of room arrangement, sizes, and other parametric attributes, other later researchers may choose or compare any desired typical layout depending on the purpose of their study.
The identified typical layouts may also be considered the current standard design of apartment buildings in Indonesia. Standard designs play an important role in building simulations and building standardizations, as they can be used as the standard testing base model for further building simulation for various types of study such as energy performance and the development of lowcarbon buildings.