Study On Early-Warning System Of Enterprise Based On Information Fusion Technology

Enterprises' early-warning system is a complex and dynamic system.Affected bymany factors including the internal and external environment, its assessment ofreliability of warning state is a multi-factor evaluation. Research shows that no singleindicator was found so far to assess the safety level of business operation accurately.With the in-depth study on environment of organizations, the means of environmentalmonitoring have been diversified, which makes the number and complexity of thecollected variables become more. These data may be redundant, complementary andcontradict with each other. Futhermore, owing to the complicated environment andthe asymmetry of information, the data can be ambiguous and uncertain, which canbe difficult to make accurate judgments.In this thesis, enterprises' early-warning index system is established based onbalanced scorecard,which combines neural network and evidence theory's superiorityto deal with the data for the enterprise. As basic probabilityis not easy to assign inevidence theory and the outputs are instable in single BP neural network, aninformation fusion algorithm combining neural networks and D-S evidence theory isput forward, that is, the outputs of neural networks are adopted as the assignment ofbasic probability of evidence theory.The simulation result shows that this algorithm ismore effective than the single method. Then, applying the joint-programmingtechnology, the fusion algorithm is realized. Also, the process in which C++Buildersoftware call the neural network toolbox and connect SQL Server 2000 database isanalysed in detail and the direct reading and writing of the database is achieved. Theresults improve the intelligentized level and operability of early-warning system,which can support managers to make scientific decisions.