Bidding strategy using multivariate distribution and EM algorithm
利用多元分佈及 EM 方法的競標策略
Student thesis: Master's Thesis
Related Research Unit(s)
|Award date||15 Feb 2005|
In many industries, a sizeable proportion of business is obtained through competitive sealed bidding. Construction contracts are a typical example of this. In Hong Kong, construction industry is a major business and there are more than ten thousand registered construction companies. The best way to allocate construction projects to construction companies is by using competitive sealed bidding. Each bidding contractor determines a bidding price based on a cost-plus method. Once the contractor’s cost is estimated, a mark-up as a percentage of the cost is added, and the sum is the tender price. The difficulty facing a competing contractor is in setting the tender price at an appropriate level. A contractor may lose to the competition if his tender is overstated. On the other hand, the project will not be profitable if the tender price is too low. Therefore, the development of a successful bidding strategy is a key factor in the survival of construction companies. Two major criticisms of most bidding strategies are that they use a single distribution with fixed parameters to model the bidding patterns and assume independence of bids among competitors. In this study, a generalized bidding model is proposed to answer these two criticisms by developing realistic multivariate distributions. Instead of assuming a fixed bidding pattern for all contractors, the study estimates bidding patterns and applies them to the bidding model. The EM algorithm will also be used to estimate the covariances of bids by pairs of contractors. Multivariate distribution can be formed once the covariances are determined. Using multivariate distributions, the probabilities of winning and the corresponding mark-ups for projects can be determined by numerical integrations.
- Hong Kong, Multivariate analysis, Expectation-maximization algorithms, Letting of contracts, Construction contracts, China