| Summary: | 碩士 === 中國文化大學 === 資訊管理研究所 === 89 === Data mining is to discover the hidden patterns from large amounts of data, and can be applied to market analysis, business management, decision support, etc. Due to the wide applicability in many areas, data mining has attracted a great deal of attention in recent years.
Data mining methods can be categorized into several classes, such as Cluster analysis, Induction, Neural networks etc. Neural networks excel at large amounts and multi-dimensional data. Through the process of learning, the relationships between input and output data can be extracted. By the use of these relationships, the data prediction model is then constructed. Traditional typhoon rainfall prediction is based on a certain formula, which is incapable of dealing with the highly non-linear factors between typhoons and rainfalls. Also the traditional back-propagation algorithm is poor for such complex non-linear relationships. Therefore, in the research, we first propose a new model to describe the non-linear relationships by adding new functions into the back-propagation algorithm, and then verify it by using thirty-year rainfall data. The results show that the proposed model is more accurate and efficient than traditional formula and back-propagation algorithm in all observatories.
|
|---|
