Gender and Age Influence on Energy Consumption in A Selected Malaysian Office Building

Due to the fact that the majority of large buildings are being built in densely populated urban areas, Malaysian buildings use more energy per square metre of floor space than those in the majority of other nations. The researcher and intervention designer will be able to determine the impact of each social parameter on the building energy performance with the aid of their understanding of how and which social parameters contribute to building energy performance. The energy consumption profiles may include gender and age-related variables. When taking into account the correlation between age and energy consumption, this study sheds light on energy consumption and gender. The purpose of this study is to determine the connection between gender, age, and the amount of energy used in an office building. After that, it will go into how this group may be a springboard for fresh, energy-saving techniques that will ultimately benefit everyone. Using a combination of statistical and neural network methods, 1,116 samples from 13 office building locations across 150-day periods were assessed. Large amounts of data are difficult to evaluate using simply traditional statistical approaches, thus neural networks are employed to model and analyse these data sets. The study's findings imply that gender and age play a role in how efficiently a building uses energy. The findings show that women spend much more energy than men, and that the most significant age groups for energy consumption are those under 30 and those between 41 and 50 years old.


Introduction
Building energy performance in different countries varies due to factors such as building codes, standards, laws and regulations, appliances used, building occupant behaviour and several other factors which vary greatly from country to country (Delzendeh et al., 2017). An organization needs to understand these factors and address them through engineering and energy systems design for the optimum performance of the building energy usage (Blok et al., 2007).
There are various ways to approach the idea of building performance. It will be clear that such factors must play a crucial role in the theoretical underpinnings of architecture.
There have been numerous attempts to create descriptive and prescriptive theories, and these have a tendency to focus on particular aspects of the structure (Baird et al., 2018). Employing physics-based models like Energy Plus is an example of building energy simulation tools have been extensively used to research and assess building energy performance. (Deng, Fannon, and Eckelman 2018) Although obtaining very specific architectural information, such as specific space features, might be challenging, realistic simulation is frequently necessary (Nguyen et al., 2014).
This study will address the result using a realistic simulation that based on Artificial Neural Network (ANN). The modelling of complicated and nonlinear patterns is a notable use of artificial neural network (ANN) approaches. The algorithm's training process is comparable to the structure of human brains, which have a set number of layers and neurons. Input, output, and hidden layers are the components of a typical ANN design. Although the output layer produces the finished product, the input layer collects all input values. Because to interference between input and output neurons, the presence of hidden layer(s) essentially ensures that ANN models have non-linear relationships (Simon Haykin (McMaster University et al., 2005).
The Artificial neural networks (ANNs) will identify the pattern which developed a choice (Juan and Valdecantos, 2022). Using numerical models to solve statistical engineering problems is the most common approach. However, the development of machine learning has shown that ANNs are a very good alternative to this traditional approach and even give better results, particularly when the problem is random and contains non-linear patterns (Fan et al., 2021;Maier and Dandy, 2000). ANNs are better suited for problems with a random distribution of variables than traditional statistical models since they lack strict guiding principles that set them apart from the latter (Maier and Dandy, 2000). Further support is provided by the interpretability of artificial neural networks, which is essential in many fields, including engineering and medicine, due to the robustness of the models (Guo et al., 2019;Li et al., 2021;Yu and Carrillo, 2019).

Literature Review
Building energy performance refers to the full range of factors that affect the performance of a building, including design, planning and construction. Various factors could have a positive or negative impact on energy efficiency. Generally, passive and active technical approaches are adopted for building energy performance. Passive technical components include daylight harvesting, solar shading and glazing properties. Active technical components include electrical appliances and lighting (Tang and Chin, 2013;Noor, 2016).
From a social standpoint, social demographics, psychological variables, and technological aspects of the buildings may all have an impact on how energy-efficient a building is. Age, gender, education level, employment status, social economic status, income level, household features, building characteristics, and geology characteristics are examples of social elements (Hess and Sovacool, 2020).
It's crucial to comprehend how and which societal factors affect how efficiently a structure uses its energy. The researcher and intervention designer will be assisted in determining the influence of each social parameter on the building's energy efficiency by the social parameter's contribution. The majority of studies in this field often focused on occupancy as a variable rather than breaking it down into various distinct social parameter variables and, sporadically, technical parameters. Instead of breaking down occupancy into various distinct social parameter variables and occasionally as a technical parameter, these studies typically focused on occupancy as a variable. in additional, usually the study on the impacts of building occupancy, a stochastic model was used (Jang and Kang, 2016).
However according to recent studies, there is a significant difference between a building's expected and actual energy use. Several studies have agreed that the primary causes of the discrepancy between expected and actual building energy performance are human behaviour and tenant preferences (social parameter) (Martinaitis et al., 2015;Yang et al., 2016;Calì et al., 2016).
Analysis of energy consumption has looked at the causes, projections, and modifications in occupant behaviour. Social factors including age and gender, among others, have been investigated to explain why people's energy consumption and energy saving behaviour vary. The social parameter is best defined by the eight social parameters that are the best energy performance predictors and/or parameters (Olli et al., 2001). Figure 1 shows integrative conceptualization of the various individual (socio-demographic and psychological) and situational (contextual and structural) factors that associate with energy performance. Figure 1: Integrative conceptualization of the various individual (socio-demographic and psychological) and situational (contextual and structural) factors that associate with energy performance (Olli et al., 2001).
To associate social factors with building energy performance, several sociological studies have been conducted in the past. There are several societal factors stated. Building energy performance is significantly influenced by factors including gender, age, education, proximity to other people, urban location, employment, income, political orientation, and ecological attitude (Olli et al., 2001).
While Hines et al. focus on educational level, Income, Economic orientation, Age and Gender (Hines et al., 1987). Marcos used Gender, Age and conservation behaviour as their social parameter (Marcos, 2009). Table 1 describe the social research related to building energy performance, method use and other attributes that are related to the research. The Uniform Building By-law of 1984 (UBBL 1984) is the rule that governs the specifications for all structures in Malaysia. This standard is used to define the space standards for computing occupancy loads. The net floor area or space given to the use shall be divided by the square metre per occupant to yield the maximum office occupancy load authorised. According to the UBBL of 1984, the number of installed fixed chairs shall be used to calculate the occupancy load of the office area.

Research Framework
Based on the common social parameter used on Table 1 which are gender and age. This study is to comprehend how age and gender influence office building energy consumption. Data analysis will combine statistical and neural network techniques, with the former being used to identify links between energy use and gender while taking into consideration its association with age.
In order to determine how varied gender and ages effect energy consumption, artificial neural networks are used in the research's examination of space occupancy load. The artificial neural network is capable of making predictions based on actual data for the occupancy value for segregated gender and age as well as the occupancy value for off-range variables value.

Material and Method
This research studied the effect of social parameters on energy consumption in a large office building. Neural networks are used to model and analyse large quantities of data, which are difficult to be analysed using only classical statistical methods such as correlation and regression analysis. The neural network with back propagation is applied to obtain an accurate prediction of the energy consumption model.

Data Collection Samples
A methodical data collection process that used 484 individuals over the course of 150 days to acquire primary data. The source sample data were collected from 13 locations with sample sizes, purposes, and meteorological characteristics that were similar to each other. Every employee's gender and age background are tabulated.
The samples variable parameter, age is divided into four categories which are below 30, 31 -40, 41-50 and above 50 years old. The other parameter is the gender, male and female. The daily total occupancy is determined using the gender and age background of each floor's attendants. Each access door has a magnetic door sensor station that records the attendant. Data loggers are used to track energy consumption in each electrical riser on each location.
Data Pre-processing Raw data logger and Magnetic Door Station Main data source was processed using Microsoft Excel Data Vlookup, which screened and chose the appropriate data. A suitable format for training an AI model will be created by combining, cleaning, normalizing, and transforming each daily sample location's data.

Data Analysis
Data analysis and model development are based on analysis using SPSS Statistical software. Multiple Regression is the method used for in data Analysis. In addition, data testing and validation are based on Neural Network Fitting MatLab analysis. The setting for Neural Network Fitting is based on 70% or 782 training samples, 15% or 167 validation samples and 15% or 167 testing samples. Figure 2 shows the Neural Network Diagram. Result of the training is based on Bayesian Regularization as Training Algorithm.

Result of The Emperical Research
The result of samples descriptive statistics shows that 1,116 samples has been analysed. The respective mean value and standard deviation are given in Table 2. The samples that age more than 50 is the least compared to other age group and male samples mean is 14.6 which is more than female samples.  (UBBL 1984) which discussed before, the space standards for calculating occupancy loads is tabulated in Table 3. Table 3 summarized the maximum Space Occupancy Load in the sample location.  Using the Pearson Correlation, the relationship between dependent variable and independent variables is summarized in Table 5. The result indicates that significant level of all the variable is less than 0.05 or 5%. This shows that there are significant differences cause by each variable. The age group of 41 to 50 years old have the highest correlation related to the energy consumption pattern. Generally, the age group below 50 years old have more correlation with energy consumption compare to those above 50 years old.

Table 4 Pearson Correlations
Due to the limitation of statistical method when handling more than two degrees of freedom, model Neural Network method is used as a tool for data fitting. This model summary result using Levenberg-Marquardt is shown in Figure 4. The result indicates the satisfying R value for training, validation, testing of 0.8891, 0.8801 and 0.8830 respectively.   Figure 5 shows present of data related to the Age Range Parameter and energy consumption using a Matlab Plot Graph. In contrast to the energy consumption, the results demonstrate a continuous consistency in the total Age Range Parameter. As indicated in Figure 6, additional study was done to investigate the connection between a particular Age Range Parameter and energy consumption. Although it is still not able to definitively pinpoint the causes of this fluctuation, the results imply that the data's homogeneity is inconsistent. Using a Matlab Plot Graph, Figure 7 displays data relating to the overall gender parameter and energy consumption. The findings show that the total gender parameter is inconsistent, in contrast to energy consumption. Further research was conducted, as shown in Figure 8, to look into the relationship between male and female gender and energy use. The findings imply that the data's homogeneity is inconsistent, even though it is still unable to conclusively identify the reasons for this variation. The effect of age and gender on energy consumption was shown through the use of Matlab Bar Plotting. A trustworthy assessment of the impact value was hampered by the separation of these parameters and the existence of noise in the data. Hence, to acquire a more thorough understanding of the impact of age and gender characteristics on energy consumption, a comparison between the collected data and a model would be required. An artificial neural network in Matlab was used to build a model from sample data. This method can be used to forecast and comprehend the values of data related to various age and gender characteristics. The Matlab Neural Network Fitting on Regression Plotting is displayed in Figure 9. The Regression Plot in Figure 9 indicates that few samples are observed tabulated on the outer region. This might be due to few isolated issues which cause irregular working routine that interrupting daily routine. These isolated issues might cause less energy consumption and higher energy consumption. Activities that need occupant to attend function outside office during office hours might cause the decrease in energy consumption value, whereas, activities such as meetings increase energy consumption.
A comparison between the Model Plotting Result and Real Model Plotting Result was carried out using Artificial Neural Networks (ANNs). The obtained information was then examined to find similarities between the models. To forecast energy consumption in office buildings, the model created in this work combines high-performance actual data models and computer simulation models. The findings of this investigation are presented in Figure 10. The 15 matching points that show similarities between the models are calculated based on Figure 10. These values are compared to the actual data in order to determine the values for the age range and gender variables at each of the 15 matching points. Using an acquired model from artificial neural networks, the energy consumption value during maximum Space Occupancy Load as shown in Table 3 and 15 matching points educational level is calculated (ANNs). The obtained energy consumption value is summarised in Table 5. According to the results of the variable effect analysis utilising minimal value occupancy, women in general have a significant impact on the office building's energy consumption. Yet, during periods of high occupancy, women under the age of 30 and women aged 41 to 50 have the greatest differences from men.

Conclusion
Therefore, the research will use a neural network approach. This approach enables data analysis and the development of statistical models. Since various demographic categories were used in each nation and no comparable data from another nation was available, overall results varied. This study finding has comprehend how age and gender influence office building energy consumption.
In order to compare actual data and the anticipated data model, neural networks were used in the research. According to the findings, female gender categories use more energy than male gender categories do. Also, the age groups with the highest energy usage were found to be women under the age of 30 and those between the ages of 41 and 51.
In this paper, we have explored in-depth the factors that impact energy consumption in a Malaysian office building. Based on this study contribution, simple action such as work from home group segmentation and area segmentation can be a way to optimize energy usage. Furthermore, because they shed light on how energy conservation should be done in an office building under varied conditions, these findings could be helpful for future research.
Yet, further research is needed on a few related issues because women in Malaysian culture frequently engage in "pot luck" gatherings and other food-related activities that require additional energy consumption. Also, throughout the course of this investigation, the utilization of "nice to have" gadgets like air conditioners, personal freezers, and air purifiers was noted.