A Study of Nested Hyper-Rectangle Learning Model for Streamflow Estimation

• Main Authors: Huey-Fen Lin, 林惠芬
• Other Authors: Fi-John Chang
• Format: Others
• Language: zh-TW
• Published: 1994
• Online Access: http://ndltd.ncl.edu.tw/handle/82176644822510441101

碩士 === 國立臺灣大學 === 農業工程學系 === 82 === The main purpose of this study is to present a theory of learning called Nested Hyper-Rectangle Learning model (NHRL) for streamflow estimation which plays an important role in hydrology .The model''s structure has a simple format and easy to extend and use . The NHRL algorithm simulates the way humane learns by experiences for estimation and classification . The learning is accomplished by storing objects in Euclidean n-space , a points in the beginning .And the exemplars are generalized to hyper-rectangles ,if the examples "match" the exemplars.The hyper-rectangles may be nested inside one another to arbitary depth . This model uses "Similarity Metric" to measure the similarity between the new example and exemplar , while the weight parameters in " Similarity Metric " are adjusted dynamically,that is the model has the ability of learning. Since the model was generally employed for classification, this study tries to modify the algorithm for estimation . The modified NHRL are tested with two theoretical data functions which are generated with Monte Carlo method ,and employs it in estimating the streamflow of Taiwan''s rivers .Both the test and employment results support the excellent ability of the modified NHRL in estimation .