Feng Shui Expert System for Evaluating Residential

2011 International Conference on Electrical Engineering and Informatics 17-19 July 2011, Bandung, Indonesia

Feng Shui Expert System for Evaluating Residential Architectural Design

As’ad, Nabila#1, Leylia K, Masayu*2 #Jalan Ganeca no.10 Bandung, Indonesia

[email protected] *Jalan Ganeca no.10 Bandung, Indonesia

[email protected]

Abstract— Nowadays for some particular clients, Feng Shui application is needed in their building’s construction. When those clients use architect's services, they often use Feng Shui expert's services to evaluate the architecture design. The difference arguments between architects and Feng Shui experts brought in by the client cause propelling development of a building. Alternatively, clients can use Feng Shui master architect but relatively few in number in Indonesia and must be paid handsomely. In this paper, an expert system to evaluate residential architectural design with Feng Shui method will be developed. Feng Shui has some method to evaluate the architectural designs ranging from the surrounding environments, external layout, and internal layout. The evaluation methods used in this paper are Form School, Yin / Yang, and Five Elements Theory. Development of expert system using an expert system development life cycle called the Linear Model. Expert system development involves consultation with a master architect, book titled “Feng Shui for the Home” by Evelyn Lip, and references from previous related research by Michael Y. Mak and S. Y Thomas Ng. Knowledge representations used here are block diagram, dependency diagrams, and some sets of rules. Knowledge mapped into a block diagram and three dependency diagrams, one for each evaluation module. This expert system development involves 33 rules that are included in 22 sets of rules. There are 21 sets of rules to generate the data abstraction and a set of rules that generate solutions. This expert system implemented using WebCLIPS. After a user gives input data, the system will display the results and give an explanation why the results have been obtained. Tests conducted internally and externally to the system functionality and user input variation. The results show that the system can function properly and produce the right answers for a wide variety of cases. Keywords— expert system, Feng Shui, residential architectural design, CLIPS.

I. INTRODUCTION Since the Western Zhou dynasty at 1100 BC - 771 BC,

Chinese society has to have confidence that the energy found in nature can be defined by science and intuition [1]. Feng

Shui focuses on the harmony between the buildings to the environment, and has been used to evaluate on a site and building, and used to define the utilization of space in interior design before the implementation process of the building.

Some architects’ clients want to apply Feng Shui from choosing the location of the building plots to design the building. Quite often, clients bring Feng Shui experts when consulting with an architect. The problem that arises is some architects feel disturbed by the Feng Shui expert who came up with the client because each of them has their own domain expertise. As a result, there are argument differences between architects and Feng Shui expert. The difference of the argument will be one of the inhibiting factors in the design of a building. Some architects are interested in studying the science of Feng Shui because it is closely related to their expertise. However, the number of Feng Shui architects in Indonesia is relatively rare and not so prominent so fees for Feng Shui architects in Indonesia can be quite expensive.

This paper offers solutions in the form of an expert system that will evaluate the designs that have been designed by the architect before the design is implemented. Expert system will be suggestions for architects as consideration in designing a building. This expert system had developed earlier by Michael Y. Mak and S. Y Thomas Ng [2], [3]. In those studies, architectural design evaluation process is divided into 24 basic criteria of a building, which will be evaluated.

Expert system developed in this paper will conduct an evaluation starting from the evaluation of the environment around a building site; the evaluation looked outside of the building, until the evaluation appears in the building. Expert involved in this expert system development is the Ms. Dewi Mariana, ST., MT., AIFSA, IAI.

The development of this expert system is described by the expert system life cycle called Linear Model [4]. In the linear model there are several stages of development starts from the planning until the evaluation of system. The stages in the linear model are performed sequentially and repeated from the beginning until the expert system is complete and ready for use. Linear model is illustrated in Fig. 1.

G2 - 3

978-1-4577-0752-0/11/$26.00 ©2011 IEEE

Fig. 1 Linear Model [4]

II. EXPERT SYSTEM DEVELOPMENT

A. Domain Analysis Expert systems are computer programs that reconstruct

human reasoning of experts to solve problems in a certain field [5] to achieve the competence level with an expert [6]. Characteristics of an expert system is to simulate human reasoning, using a heuristic approach, explain and prove the solution, and using a representation of human knowledge [7].

Expert system for evaluating architecture design with Feng Shui methods have been built earlier by [2], [3]. This domain has two classes of solutions; appropriate or not appropriate according to Feng Shui. Problems in this domain are categorized into the classification because there is a direct mapping from observation to class solutions. This paper is built with reference of the study [2], [3]. Representations of knowledge in this also used in the study, namely block diagram and dependency diagrams.

Feng Shui selection of the method used depends on the expert concerned. [2], [3] use the concept of Feng Shui model, the parallel concept, four-module design concept, and the concept of Feng Shui criteria as a method of evaluation. This paper uses the concept of Form School, the Yin / Yang, and Five Elements Theory. Because of the difference Feng Shui method that used for evaluating architectural design, there are differences between this paper and [2], [3]. The differences are in its evaluation module including criteria for each module, inference method, and expert system shell that used for its implementation. The differences for criteria used in this paper and in [2], [3] are shown in Table 1 to Table 3. Detail for each difference will be explained in the next stages.

B. Planning Planning stage includes several activities, such as defining

the problem domain expert system to be built, analyzing user's target goals, and defining features of expert system functionality. This expert system development involves an architect who also pursues Feng Shui as an expert and also has a role as an examiner of this expert system.

Expert system is built to represent expert reasoning in the evaluation. This is in accordance with the characteristics of an expert system that simulates human reasoning (in this case, an expert), a field in taking a decision, and produce solutions from client input. Analysis from an expert is required in conducting the evaluation of architectural design. The analysis is one of the simulations experts reasoning in making decisions. Expert system built in this paper use knowledge gained from the real practice of an expert. In real practice of

Feng Shui, expert carried out a lot of face-to-face meetings between client and expert. [7]

Based on those reasons, this expert system’s functionalities are that system can accept user input, perform inference on user input, and display the results of the evaluation also display an explanation for the results of evaluation of residential architectural design.

C. Knowledge Definition Knowledge definition stage includes several activities to

generate demand expert system knowledge. This stage consists of two tasks, “source identification” and “selection and acquisition, analysis, and extraction”.

There are three types of knowledge sources involved in the development of this expert system. First, from the selected expert; an architect as well as a Feng Shui practitioner who lives in Bandung. Beside that, reference books, “Feng Shui for the Home” book by Evelyn Lip (Lip, 2005), and earlier research papers by Michael Y. Mak and S. Thomas Ng. [2], [3] are used as reference sources.

Evaluation of buildings with Feng Shui includes four modules, namely surrounding environment, external layout, internal layout, and interior arrangement [8]. Each of the modules requires evaluation criteria. Therefore, in this paper, interior arrangements will not be a separate evaluation module. At this paper interior will only be evaluated in general and belongs to internal layout module. Discussions in interior criteria include dominant color, patterns, and materials of the room.

This paper involves the construction of three evaluation modules. There are differences in evaluation criteria for each module. Criteria used in the construction of this paper are the result of knowledge extraction from reference books and [2], [3]. Comparison of criteria for each module used in previous studies and this paper presented in Table 1 to Table 3.

Another thing that affects the evaluation of residential architectural design with Feng Shui method is ages and birth of dates of the resident owners. But this does not become the evaluation criteria in this paper because the possibility of one's birth date is very broad.

Knowledge representation is a step represents a knowledge base into a model that illustrates the relationship between a data with other data. This step is done to help the knowledge engineer in understanding the domain knowledge. The global knowledge used in this paper mapped into a block diagram, dependency diagrams, and some sets of rules. Block diagram is a diagram that represents the relationship between components in the scope of evaluating architecture design with Feng Shui knowledge.

Block diagram illustrates the problems that would limit the scope of the developed expert system. Criteria for each module, which shown in Table 1 to Table 3, are mapped into block diagram. Fig. 2 shows block diagram used in [3] while Fig. 3 shows block diagram used in this paper. In the overall block diagram illustrated in Fig. 3, there are five levels acquired from the level found in Fig. 2. However, this paper did not use the intermediate level and the Feng Shui model level. Both levels are only used in the methods of Feng Shui

study [2], [3], namely the concept model of Feng Shui. As shown in Fig. 3, the five levels used in this paper are decision target level, overall evaluation level, design criteria level, sub-criteria level, and input action level.

Decision target level produces the final solution. The overall evaluation level indicates the level of evaluation at three evaluation modules. This level was acquired from the level of Overall Evaluation on the block diagram [3] in which at these levels to be evaluated against four evaluation modules are used. Design criteria level indicates the level of criteria in each module that has been presented in Table 1 to Table 3. Sub-criteria level indicates sub-criteria for some evaluation criteria. This level is added in this paper because of some criteria that have sub-criteria evaluation. Input action level consists of several conditions that allow one criterion. These conditions are user answer choices.

TABLE I COMPARISON TABLE FOR SURROUNDING ENVIRONMENT MODULE

No Criteria in Mak’s study [2], [3]

Criteria used in this paper

1 Topography Topography 2 Front of Site

Site 3 Rear of Site 4 Sides of Site 5 Street Location Building Location 6 Water View - 7 Wind Direction - 8 - Fence 9 - Garden

TABLE II COMPARISON TABLE FOR EXTERNAL LAYOUT MODULE

No Criteria in Mak’s study [2], [3]

Criteria used in this paper

1 Shape of Site - 2 Entrance Entrance 3 Shape of Building - 4 Orientation - 5 Trees - 6 Pond - 7 - Body of Building

against the Entrance

8 - Shape of Roof 9 - Building Looks

against Neighbours

TABLE III COMPARISON TABLE FOR INTERNAL LAYOUT MODULE

No Criteria in Mak’s study [2], [3]

Criteria used in this paper

1 Layout Rooms 2 Doors - 3 Windows - 4 Shape of Rooms - 5 Staircase Staircase Position 6 Ceiling Bedroom’s Plafon 7 - Interior

Fig. 2 Block Diagram in Mak's Study [3]

Fig. 3 Block Diagram used in this Paper

D. Knowledge Representation

This stage presents the inference process and inference methods used. Representations of connectedness between user input and rules used in this expert system is illustrated in three dependency diagrams. Dependency diagrams for each module are obtained from mapping the representation of knowledge that has been done in previous stages. Fig. 4 shows a dependency diagram for external layout evaluation module.

Fig. 4 Dependency Diagram for External Layout Evaluation Module

In developing an expert system, [3] uses decision tables as the method of inference. There are five solutions to the evaluation results in [3]; excellent, favourable, fair, unfavourable, and bad. The disadvantage is the number of criteria and the amount of the value of each criterion will result in a combination in large numbers. For this paper, several factors are considered as components of the same rank, so do not depend on the possible combinations of existing criteria. This problem can be solved by heuristic classification as the method of inference because from user input data obtained abstraction data for each criteria and modules.

Fig. 5 Inference Structure of Heuristic Classification [9]

According to [9], heuristic classification is a method of

expert systems that solve unstructured problems that require diagnosis. In simple classification, the existing input data and the data stored in the knowledge base relate directly to the solution. Unlike the case with heuristic classification, characteristic of heuristic classification is that data do not

directly relate to a set of solutions [9]. The process of inference by heuristic classification method is illustrated in Fig. 5. Fig. 5 illustrated that performed data abstraction process on user input data and then do the heuristic match process to produce solution abstraction. The solution abstraction is then processed with refinement process to achieve the exact solution of the problem.

Inference process of heuristic classification in this paper, as shown in Fig. 6, uses data-directed search. The program performs a search and matching data with data abstraction. These data abstractions are processed with heuristic match to achieve solution abstraction. After that, conducted refinement process to achieve the exact solution. Search and matching system are done using forward chaining. Illustration for the global inference structure of data-directed search can be seen in Fig. 5. The processes are as follows.

1) Data Abstraction: The facts obtained from action input data are processed by data abstraction process. Data abstraction process was done to get data abstraction. Data abstraction is obtained at the sub-criteria level, design criteria level, and overall evaluation level. Each criterion whose weight at these levels is illustrated in Fig. 3.

2) Heuristic Match: Heuristic match process is done to obtain a solution that will be displayed to the user. The solution is obtained at target solution level. In this paper, refinement process is not performed because

direct relation is performed from data abstractions to the solution sets.

Fig. 6 Heuristic Classification Structure for this Paper

On external layout evaluation module that shown in Fig. 4, there are four evaluation criteria of architectural design as presented in Table 2. The four criteria, which are become data abstraction on this expert system, are symbolized by a circle. Data abstraction is the process of determining the class of solutions to the raw data through a process of inference [9]. The raw data is obtained from user inputs. Data abstraction indicates whether the evaluation criteria according to Feng Shui or not. There are three types of data abstraction; definitional, qualitative, and generalization abstraction [9]. However, only two data abstractions are being used in this expert system; definitional and qualitative abstraction. Definitional abstraction is based on the essential things in data

SetRule

Set Rule

(Yes, No)

External LayoutEvaluation

Left side is higher than the entrance

(Yes, No)

Right side is higher than the entrance

(Yes, No)

Left and right sides are in the same height

(Yes, No)

(Yes, No)

(Yes, No)

Set Rule

(Yes, No)

Body of Building against the Entrance

Entrance

Shape of Roof

Building Looks against Neighbours

SetRule

In a straight line with ladder (Yes, No)

In a straight line with the back door (Yes, No)

SetRule

Gable (Yes, No)

Shield roof(Yes, No)

Cross gable (Yes, No)

Pyramid shield roof (Yes, No)

Cross shield roof (Yes, No)

Combination roof (Yes, No)

Flat roof (Yes, No)

Sloping roof (Yes, No)

Lower (Yes, No)

Higher(Yes, No)

Equal (Yes, No)

Right neighbour is higher(Yes, No)

Left neighbour is higher(Yes, No)

and qualitative abstraction is based on qualitative data [9]. The example of definitional abstraction used in this paper is if the topography of the building is lower than road building, then the topography doesn’t fit according to Feng Shui. The example of qualitative abstraction is if the distance between the fence and the main entrance is less than 1.9 meters, then the fence doesn’t fit according to Feng Shui.

One rule is used for heuristic match process. This rule adds up the results of the third modules’ weight then checks the amount. If its value is greater than or equal to 80, is considered good residential building, and if less than 50 then considered as bad. Meanwhile, if it is between 50 and 80 then the residential building considered as medium. The number of 50 and 80 was given from the involved expert based on expert’s experience.

Expert system built is a web-based application. Expert system is built using webCLIPS. Architecture of expert system for evaluation residential architectural design with Feng Shui methods is illustrated in. Expert system architecture consists of four components; they are knowledge base, working memory, problem solving component, and control. Working memory contains facts related to the specific case and the fact temporary. Knowledge base is an expert’s knowledge about a specific domain. Inference performed by problem solving component is stored in intermediate results and problem solution.

Users of this expert system are architect and knowledge engineer. Problem solving component, case-specific facts, and domain-specific knowledge are implemented in CLIPS. The case-specific facts include surrounding environment, external layout, and internal layout data. Domain-specific knowledge includes domain facts and action rule.

Fig. 7 Architecture of Expert System for Evaluating Residential Architectural

Design with Feng Shui Methods

E. Code and Checkout This expert system is implemented using WebCLIPS.

CLIPS (C Language Integrated Production System) is one of expert system development shells. Shell provides assistance to users who want to develop an expert system. Through a shell, user only needs to define knowledge base. CLIPS programming language supports multiple programming

paradigms, namely rule-based, object-oriented, and procedural [4].

Users enter some knowledge base data and related rules into CLIPS. CLIPS provides a forward chaining inference. Each rule consists of condition and action. From the condition that there is then traced the action to be performed. One of the implementation of CLIPS is WebCLIPS.

Heuristic classification scheme consists of design-related facts and rules. Facts consist of problem domain facts, solution facts, and user input facts.

III. TESTING The purpose of this testing stage is to verify the knowledge

base, to check how system performs the inferences process, and to check expert system functionality features that have been defined previously. Testing activities performed are as follows.

1) Internal Testing Internal testing is conducted to test the specification of an

expert system’s functionality features. There are several tests for each component of an expert system’s architecture. Knowledge engineer performs the test. Knowledge engineer answers questions that being asked by the system and pushes the process button. This step is carried out to test the interview component and explanation component.

Knowledge engineer analyzes the results and explanations issued by the system results and makes comparisons with the results of the counted-manually inference results. This step is carried out to test the problem solving components and ensure validity of the facts stored in working memory and knowledge base.

Test cases amounted to seven cases according to Feng Shui method; one case with the medium evaluation result, four cases with good evaluation results, and two cases with bad evaluation results. Results obtained are system being able to accept and process inputs from user. System also can display the inference results and present an explanation by displaying each criterion result. The illustration is shown in Fig. 8.

Fig. 8 An Example of the Result Screen

2) External Testing This testing was conducted to test the expert system output.

The involved expert performs the test. Expert gives cases to the system and analyzes the result by comparing its result with expert’s manual calculation. For a variety of cases, the system

can give results as expected and did not deviate from the domain knowledge.

IV. CONCLUSIONS There are three conclusions that can be drawn from this

paper. First conclusion is that in conducting the evaluation of residential architectural design with Feng Shui methods, it needs the evaluation of three major modules, i.e. surrounding environment module, external layout module, and internal layout module. There are a total of thirteen evaluation criteria for the modules. Evaluation of residential architectural design is done in five levels of evaluation, in order of the inference that the input action level, sub-criteria level, design criteria level, overall evaluation level, and target solution level.

Second, domain problem of evaluation of residential architectural design with Feng Shui methods are categorized into classification problems because there is a direct mapping from observation to the solution. Inference method used in the construction of this paper is the heuristic classification with the process of data-directed search where the data abstraction process for each level of evaluation. Refinement stage did not performed because the final stage of this expert system is in the form of solutions abstraction.

Third, representation of architecture evaluation design with Feng Shui methods domain knowledge are modelled into block diagram, three dependency diagrams, and 22 sets of rules with a total of 33 rules to conduct the evaluation.

ACKNOWLEDGMENT Writers wish to acknowledge Ms. Dewi Mariana, ST., MT.,

AIFSA, and IAI. for her contribution in developing and testing this expert system.

REFERENCES [1] Lip, Evelyn. Feng Shui for the Home. Singapore: Times Books

International, 1995. [2] Mak, Michael Y and Ng, S Thomas, “Applying Knowledge-Based

Expert Systems Approach for Feng Shui Design Evaluation,” in Joint International Conference on Computing and Decision Making in Civil and Building Engineerin, 2006.

[3] Mak, Michael Y., “Applications of Knowledge-Based Expert Systems to Feng Shui Knowledge,” in Proc. IMECS ‘10, 2010.

[4] Giarratano, Joseph. Expert Systems Principles and Programming, Third Edition. USA: PWS Publising Company, 1998.

[5] Puppe, Frank. Systematic Introduction to Expert System: Knowledge Representations and Problem-Solving Methods. USA: Springer-Verlag, 1993.

[6] Engelmore, Robert S (1993). Expert Systems and Artificial Intelligence. [Online]. Available: http://www.wtec.org/loyola/kb/c1_s1.htm.

[7] Kerber, Manfred. AI Techniques (Part I). UK: School of Computer Science, The University of Birmingham, 1999.

[8] Jun, Xu, “Managing Computer-Based Enviromental Information,” in ARCC Spring Research Conference at Virginia Tech, 2001.

[9] Clansey, Willam J. Heuristic Classification. Palo Alto: Stanford University, 1985.