论文部分内容阅读
Major construction projects play significant roles in the strategic development of the society. The decision makers (DMs) participating in major constructions often represent various interest preferences of stakeholders. The decision making process for major construction projects is usually complex with a large number of DMs’ participation, which can be regarded as a Large Scale Decision Making (LSDM) process.
The potential relationships, especially the conflict relationships and the improper rela-tionships, among DMs have big influences on the impartiality and decision making result for major construction projects. Therefore, this research is focused on the relationship investigation and conflict relationship elimination for LSDM problems in social network environment. Social network analysis, sparse representation, stochastic and nonlinear optimization models, and dynamic programming algorithms are utilized.
The contributions of this thesis can be summarized in the following:
1. In the current social network-based LSDM models, trust relationships are widely utilized to support LSDM models. Usually, the trust relationships are provided in an incomplete form. The impact of relationship strength on trust propagation efficiency is ignored in the process of calculating“indirect”trust degrees via a mediator in current researches. In addition, conflict performances of DMs are ignored in the existing LSDM models. There is no mechanism to investigate the conflict relationships in LSDM problems. To tackle these two gaps, a relationship strength-based trust propagation is proposed to obtain the self-recorded complete trust network with multi-path. By using the obtained complete trust network, fusing with the assessment information, a trust relationship-based conflict detection and elimination (TR-CDE) LSDM model is proposed. A stochastic optimization model is utilized in the TR-CDE LSDM model to govern the conflict relationship performance of DMs. Compared with traditional consensus-degree based LSDM models, the TR-CDE model can well eliminate the conflict performances of DMs and improve the efficiency for LSDM events.
2. In most of the relationship-based LSDM models, the provided relationship information is self-recoded, which is lack of objectivity. To further investigate the conflict relationships of DMs in a more objective way, by using sparse representation, a social network-based conflict relationship investigation process (S-CRIP) is proposed. DMs’ conflict relationships can be recognized by S-CRIP by using assessment information only and can be classified into opinion conflict and behavior conflict. Based on S-CRIP, a dynamic conflict degree-based consensus reaching process (S-CRIP and CD-CRP) LSDM model is advanced with a nonlinear optimization model. By using the proposed S-CRIP and CD-CRP LSDM model, the efficiency of the conflict elimination for LSDM events is further improved.
3. In the existing clustering methods for LSDM events, there is not an algorithm pro-posed to investigate intra-relationships and leaders in each clusters for the LSDM problems. The intra-relationships are of great importance to recognize the improper relationships among these multiple stakeholders in an LSDM event. Moreover, most of the existing clustering methods are supervised algorithms. This thesis advances a sparse representation-based intuitionistic fuzzy clustering (SRIFC) approach to detect the trust relationship among DMs with assessment information only. Taking the relationship strength into consideration, the SRIFC approach consists of two algorithms for different relationship precision and scales. The SRIFC approach is unsupervised and can investigate the intra-relationships as well as recognizing the group leaders and key figures in each cluster for LSDM problems.
The holistic knowledge gained through the study and analysis can be employed to fill up these blanks in the relationship investigation and conflict relationship-based LSDM methodology. The application of the proposed LSDM models in a major construction project can well reduce the conflict among DMs when reaching the final selection, which can help further to prevent unhappiness situations, complaints, demonstrations, and so on.
The potential relationships, especially the conflict relationships and the improper rela-tionships, among DMs have big influences on the impartiality and decision making result for major construction projects. Therefore, this research is focused on the relationship investigation and conflict relationship elimination for LSDM problems in social network environment. Social network analysis, sparse representation, stochastic and nonlinear optimization models, and dynamic programming algorithms are utilized.
The contributions of this thesis can be summarized in the following:
1. In the current social network-based LSDM models, trust relationships are widely utilized to support LSDM models. Usually, the trust relationships are provided in an incomplete form. The impact of relationship strength on trust propagation efficiency is ignored in the process of calculating“indirect”trust degrees via a mediator in current researches. In addition, conflict performances of DMs are ignored in the existing LSDM models. There is no mechanism to investigate the conflict relationships in LSDM problems. To tackle these two gaps, a relationship strength-based trust propagation is proposed to obtain the self-recorded complete trust network with multi-path. By using the obtained complete trust network, fusing with the assessment information, a trust relationship-based conflict detection and elimination (TR-CDE) LSDM model is proposed. A stochastic optimization model is utilized in the TR-CDE LSDM model to govern the conflict relationship performance of DMs. Compared with traditional consensus-degree based LSDM models, the TR-CDE model can well eliminate the conflict performances of DMs and improve the efficiency for LSDM events.
2. In most of the relationship-based LSDM models, the provided relationship information is self-recoded, which is lack of objectivity. To further investigate the conflict relationships of DMs in a more objective way, by using sparse representation, a social network-based conflict relationship investigation process (S-CRIP) is proposed. DMs’ conflict relationships can be recognized by S-CRIP by using assessment information only and can be classified into opinion conflict and behavior conflict. Based on S-CRIP, a dynamic conflict degree-based consensus reaching process (S-CRIP and CD-CRP) LSDM model is advanced with a nonlinear optimization model. By using the proposed S-CRIP and CD-CRP LSDM model, the efficiency of the conflict elimination for LSDM events is further improved.
3. In the existing clustering methods for LSDM events, there is not an algorithm pro-posed to investigate intra-relationships and leaders in each clusters for the LSDM problems. The intra-relationships are of great importance to recognize the improper relationships among these multiple stakeholders in an LSDM event. Moreover, most of the existing clustering methods are supervised algorithms. This thesis advances a sparse representation-based intuitionistic fuzzy clustering (SRIFC) approach to detect the trust relationship among DMs with assessment information only. Taking the relationship strength into consideration, the SRIFC approach consists of two algorithms for different relationship precision and scales. The SRIFC approach is unsupervised and can investigate the intra-relationships as well as recognizing the group leaders and key figures in each cluster for LSDM problems.
The holistic knowledge gained through the study and analysis can be employed to fill up these blanks in the relationship investigation and conflict relationship-based LSDM methodology. The application of the proposed LSDM models in a major construction project can well reduce the conflict among DMs when reaching the final selection, which can help further to prevent unhappiness situations, complaints, demonstrations, and so on.