Dissertation > Excellent graduate degree dissertation topics show

Abstract

Data mining has been widely used in all walks of life of modern society, but mostly for historical data analysis and processing, the pursuit of the law no longer just found hidden in the historical data to solve practical problems, but would like to in the competitive society instant access to useful information for static historical data mining static data mining is not well to achieve this demand; design a dynamic data analysis and processing of an information processing technology has great practical significance. Data to predict a major research direction in data mining and prediction of the multi-dimensional aspects of the key issues to be solved, to some extent, also became a multi-dimensional forecast forecasting a bottleneck; combined with dynamic data mining to study the cube dynamic forecasting problems in real-world applications have broad practical significance. Dynamic data mining is not limited to data projections, explore its application area also has great practical significance. Depth analysis of the previous data processing technology development status is given after the dynamic data during the operation of a dynamic data sources combined with historical data, current data and the upcoming data analysis and processing of data mining techniques: the use of a sliding window dynamic access to data through dynamic data window processing data, the use of data test of the dynamic performance of the data mining. The artificial neural network is a network model simulate the human brain work, has a powerful parallel processing power and memory function, has been widely used in various industries; article is mainly combined with neural network technology to study and solve the actual multi-dimensional dynamic data to predict problems, and achieved major breakthrough: the design, single-input single-input single-step autoregressive neural network-based multi-step and multi-input multi-step prediction model of dynamic data; model each hidden layer node using the delay step feedback, so that the network has a good memory; each attribute hidden nodes connected by a feedback delay the step to simulate the relationship between attributes, so that the model has good stability and practicality. Improved BP algorithm fully consider the error function, activation function and learning rate on the performance of the algorithm, combined with the characteristics of the prediction model Improved Integration of BP algorithm suitable for the prediction model. By Matlab simulation results show the performance of the prediction model, experiments show that the three types of multi-dimensional dynamic forecasting model predicts better: works especially well in a multi-dimensional curve prediction, linear prediction error is zero; stability of the model, the predicted results almost independent of the initial weights. The article also discussed theoretically dynamic data mining applications automatically updated in the knowledge base of the expert system, the concept of a theoretical framework to achieve real-time updates of the knowledge in expert systems combined with dynamic data mining. This design dynamic prediction model, consider a cube of dimension attributes that may exist between internal links, and thus more relevant to the prediction of the actual implementation of the multi-dimensional dynamic data, also to some extent for the further study of the issue of multi-dimensional dynamic data ideas .

Related Dissertations

CLC: > Industrial Technology > Automation technology,computer technology > Computing technology,computer technology > Computer software > Program design,software engineering > Programming > Database theory and systems
© 2012 www.DissertationTopic.Net Mobile

The Research of Dynamic Data Mining Based on Neural Network

Abstract

Related Dissertations