A Study On VPRS Based Approach For Data Mining And Risk Avoidance In Software Project Bidding

Software project bidding is high-risk the world over, and many IT companies often do not achieve their expected profits, even when they win their bids. Risk avoidance is necessary to reduce risk of software project bidding and improve bidders'benefit. The study reviews the research status of bidding theory, analyzes the risk categories and risk factors in software project bidding, then, discuss the applicability of the variable precision rough set (VPRS) model on knowledge discovery of risk avoidance in software project bidding. VPRS is the extension of Pawlak's rough set, and introduces a threshold valueβ(0.5<β≤1), which makes the model permit error classification at a ratio of 1-β. The character can be used to remove the error classification in risk information system of software project bidding, and obtain strong rules to support decision of risk avoidance.The use of VPRS on Web searching and text mining provides a new approach to obtain risk information. The procedure of Web searching and text mining, latent semantic index method, document ranking based on rough set, and document representation and search based on VPRS are introduced. VPRS makes the probability that the lower approximation of query and document is empty reduce by permitting partial inclusion in object sets. Text mining based on VPRS is explored in the study, and lower limit l and upper limit u are used to denote classification precision. According to the relation between condition probability of object set and approximation region under different sets of attribute value, probabilistic rules to find Web document on software project bidding risk information are extracted.The concept of software project and bidding risk avoidance is defined in the study. Character of software project bidding, based on several theories, the risk indices of software project bidding are analyzed and classified systemically. Experts score the risk exposure (RE) of software projects and indices under a group decision pattern. VPRS and judgment matrix method in analytical hierarchy process (AHP) are combined to design two risk measurement algorithms under two conditions: the experts have the same weight and they may have different weight. The algorithms are used to aggregate each expert's score into integrated risk exposure (IRE), and prioritize the significance of risk indices. It is different from traditional statistic sum of risk evaluation that VPRS compute the significance of risk indices based on their classification ability, which enhances the precision of risk measurement scientifically.Risk rule extraction is also studied in the thesis. a series of algorithms based on VPRS and its generalization: dominance based VPRS and variable precision fuzzy rough set (VPFRS) are designed to process the decision data of RE in software project bidding. Then, risk rule of software project bidding is obtained, and it can be used to forecast risk before the bidding. The impact ofβon the rule extraction is analyzed.Dynamic risk avoidance system in software project bidding is established. The stages in the risk-driven life cycle of software projects and main risk factors existing in the each stage are introduced. The study explores how to integrate software project management risk into bidding risk, and makes use of life cycle management theory to study risk avoidance in bidding for software projects. As many types of risk factors exist in the bidding, the possible risk response measures for different risk indices and the measures'corresponding strength are analyzed. The basic methodology for risk avoidance in bidding for software projects based on life cycle management theory is described and an example is illustrated.A group decision support system (GDSS) framework is established based on VPRS, and it provides a platform for experts to implement risk avoidance decision in software project bidding in a high speed and lower cost. The principle of risk avoidance on the basis of GDSS is analyzed. Structure of expert group decision system, knowledge base, and its update are discussed. At last, the prototype of GDSS is established.