Research Of Data-Mining-based Decision Support System And Its Key Techniques

With China's successful entry into WTO, the globalization of world economy has caused Telecom Corporation to face unprecedented challenge. Rapid and scientific decision concerns the fate of Telecom Corporation to a great extent. At present, Telecom operators generally adopt the traditional decision-making way, which is analyzing data from manual reports. A lot of problems result from this way, such as huge workload, slowness and being easy to lose the data etc. Therefore, it has become the common understanding to study and develop the Decision Support System (DSS) adapted to Chinese Telecom Corporation among all the Telecom operators. This paper focuses on the study of the data-mining-based decision support system and its key techniques, which is the core of the project of "Telecom IP Decision Support System (IPDSS)". A telecom decision support system solution integrating data warehouse and data mining techniques is designed, and the key techniques used in the "IPDSS", such as clustering, prediction and the modeling technology, are analyzed.Aiming at the feature and requirement of IP services of telecommunication operators, a decision-making support system (DSS) based on the .Net framework is designed. This system adopts ASP. NET as the model of the Web application, ADO.NET is used as the foundation of data access, and SQL server 2000 is as a solution of database and data warehouse. The technology of data mining and artificial intelligence is embedded effectively. It can provide quick, scientific reference and foundation for decision-making of telecom IP operation.K-Means algorithm is one of common clustering algorithms, but the cluster center initialization is a hard problem. In this paper, a hierarchical-based initialization approach is proposed for K-Means algorithm. The general clustering problem is treated as weighted clustering problem, and the original data is sampled level by level to reduce the data amount. Then clustering is carried out at each level by top-down. The initial center of each level is mapped from the clustering of upper level and repeat this procedure until the original data level is reached. As a result, the initial center for the original data is obtained. Both the experimental results on simulated data and real data show that the proposed method has fast converging speed, high quality of clustering and is insensitive to noise, which is superior to some existing clustering algorithms.EPNET is an evolutionary model for Artificial Neural Network (ANN), which can evolve the connection weight and architectures simultaneously. In this model, the crossover operator of Genetic Algorithms (GAs) is not adopted. However, 5 mutation operators are used instead to get a better result. This paper mainly investigates the forecasting time series based on the EPNET (Evolutionary Programming Net). The time series forecasting algorithm based on the model is proposed. Besides, problems for implementation in detail are introduced. In the same time, this paper introduces the plan rule based reinforcement learning bias analysis algorithm for rapidly calibrating model parameters.Analytic Hierarchy Process (AHP) is a practical decision method, whose core idea is quantifying the decision makers' qualitative experiences so as to offer the basis for decision in quantitative form. In this paper, AHP is adopted as the basic modeling theory and technology to develop investment model of network construction. At the same time, considering the hierarchical model is a kind of graphic model, the method, that is, representing and storing hierarchical model with database tables is proposed, which makes it easy to manage the hierarchical model. Besides, it also brings forward the management scheme to merge model library and method library as a unity so as to manage easily, which is used for implementation and design of IP DSS model parts.