A multi agent systems based contractor selection model

Selection of the contractor, as a main participant of a construction project, is the most important and challenging decision process for a client. Construction projects consist of various uncertainties and risks. In addition, construction projects consist of commonly unique and complicated tasks that the participants have to be dealt with in order to provide objectives and to achieve to the success. In order to attain the objectives of a construction project all of the resources should be used effectively and efficiently. Therefore, the success of a construction project is directly related to organize the project team and to select the building production process participants accurately who will use these resources. It means a strong and consistent relationship between client and contractor, who are the two main participants of construction projects, is expected for objectives. Contractor selection (CS) problem in construction sector must be handled in two directions as contractor pre-qualification (CP) and CS. Pre-qualification is the most effective way to support client's decision in order to select contractors from the short list of qualified contractors. However, there should be some factors or criteria in order to differentiate contractors. In practices, "bid price" is the dominant and significant factor that affects the result of selection process directly. However, bid price is not sufficient to be a criterion alone in order to take decision. Therefore, in order to support decision on the selection of contractor by clients, several CP criteria were determined and used within the scope of Multi Criteria Decision Making (MCDM) discipline. In addition, many decision support methods are proposed to qualify contractors. In construction industry, according to the project initials, tender process consists of some decision points such as project delivery method, tender type and contract type. In various researches, the direct and indirect influences of this points on the selection of contractor are examined and identified. Therefore, CP and CS processes must be considered as a sub-process of tender process and must be analyzed in the frame of tender management system. However, the sector type, which effects the flexibility of processes and the use of qualification methods, mustn't be ignored as a factor. In general practices of public tenders, it is observed that the restriction of evaluation processes by various laws, regulations and procedures causes price-driven evaluation methods. On the other hand, private sector provides a much flexible environment to the examination of criteria other than price. Despite various proposals and models to evaluate those criteria in literature, it is observed that there will always be missing points in the evaluation of criteria and in the selection of a right contractor without being human interference. The main objective of this research project is to propose a Multi Agent Systems (MAS) based contractor selection model. Moreover this study is aimed (1) to introduce the findings of the literature review about the contractor selection and pre-qualification models and systems, (2) to classify academic studies dealing with CP, CS and selection criteria of contractors using a Meta classification and analysis method, (3) to analyze the contractor selection criteria in Turkish construction sector, which are carried out by survey study, (4) to examine and to compare the level of weightiness and effectiveness of the contractor selection criteria from the perspective of the Turkish construction sector, (5) to propose a Web based Online Tender management SYStem (WOTSYS) to perform CP and CS processes by end users and (6) to propose a Petri Nets (PN) based simulation model in order to validate and to analyze system results. In this study, 133 peer-reviewed academic studies between 1992 – 2013 have been analyzed and classified in the domain of contractor selection, contractor pre-qualification and weighting of criteria. A Meta-classification system is adapted and used in order to present the state of the art of contractor pre-qualification and selection challenge. The most preferred methods in order to solve contractor selection problem are identified and an increment in the studies of contractor selection in the last period is determined. In addition, by the meta-classification analysis all related publications are examined in detail and existing CP and CS criteria are identified. A survey study, which is validated with the experts' view and statistical tests, is carried out with the sample of 106 participants, who are from different professions and cities in the sample of Turkish public and private construction sectors. Quantitative research consists of 5 sections in order to obtain professions perspectives on tendering systems and contractor selection problem. A statistical study is carried out on the perspectives of participants about the importance and effective usage of the CP and CS criteria in real applications. The relationships between the level of weightiness and effectiveness of the contractor selection criteria, which are divided into main and sub-criteria, are observed with comparative analysis. In this study, CP and CS problem is handled in the machine learning discipline and a multi-agent based solution approach is brought in the environment of artificial intelligence. Therefore, knowledge based systems, MAS and PN are introduced in detail with some necessary definitions. MAS and PN are examined in the scope of construction industry and CP and CS problem. Three basic tender types, which are open, selective limited and negotiated tender, were studied and a negotiation based solution is applied for all tender types. Moreover, auction based tender type is examined and conceptually given in the frame of modeling study. In this study, a MAS based contractor selection model is proposed. The proposed model consists of two sub-models such as a pre-qualification model and a selection model. PN is used to construct workflow models of the proposed model. In pre-qualification model, agents are used to evaluate criteria and to estimate criteria scores of contractors, by regression analysis and Bayesian interval estimation methods. In the selection model, agents are first used to estimate opponents' strategies by statistical methods and to perform negotiations by Bayesian Learning Mechanism. In the competitive construction industry, it is very important for clients and contractors to develop a strategy. For instance, in the open type of tender, contractors should be estimated not only the strategies of client, but also the competitors. Agents are also modelled for the purpose of data management such as storing weighting of criteria, tender and contractor information, etc. It is aimed that the agents could be specialized in updating criteria weighting, according to the subject and type of tender and observing recorded contractors in the proposed overall models. This study is focused to examine CP and CS problem within the frame of tender management. Therefore, the given overall models are used to develop of a part of tender management system and to integrate it with web based platform, which is named as WOTSYS. In order to validate and simulate the given system a PN based simulation model is proposed and integrated with the overall system. In this study, for the end users a prototype program, which consists of programmed MAS environment (by JAVA programming language), a WOTSYS platform and a programmed PN simulation, is created and introduced. The three case studies are examined in order to present initial results of proposed system. The results are indicated that the system gives consistent solutions and provides opportunities for the clients and contractors in the optimization of demands and maximization of total utilities. This thesis is structured as follows: the idea, objectives, scope, problem, methodology and hypothesis of the study are clarified in the introduction section. In the second part tender processes in the construction sector are intensively examined and the new tendering processes such as e-tendering are introduced. In the next part contractor pre-qualification and selection problem is examined with meta classification and analysis study and a quantitative survey is carried out in order to identify CP criteria. In the fourth part structures of the modelling methods are given and MAS and PN based modelling approaches and WOTSYS system development are introduced in the fifth part. Part six consists of the case studies within the scope of given thesis scope. And finally thesis concluded with the discussions and conclusion.