Study On Red Seed-using Watermelon Quantitative Character Based On Artificial Neural Networks

In the process of breeding,huge amount of quantitative character data is acquired, which contains complex,various and huge primitive information.How to find and use the useful and potential character in those data has become a key question for breeding experts.With the development of Artificial neural networks technology,Artificial neural networks has been wildly used.Artificial neural networks advantages such as parallel process,self organized,associate memory and error adapt,its can abstract useful information and knowledge from data which is mass,noisy and fuzzy.The reaserch is in the support of natural science foundation of Anhui province Research on germplasm resources characteristics hereditary regularity and dominative combination for Red-using watermelon.Inbred line economical character data set of Red seed-using watermelon of multi-generation was gained by field tests and measures in our research.According to the data set,the BP neural network and self-organizing competitive neural networks were discussed and following results were obtained.The database was designed according to database principles and methods,where the quantitative character data set of Red seed-using watermelon were stored.The establishment of database could improve the ability of decision maker to organize and use the data set,and make information serve better for decision-making.The BP neural network was studied in details to establish data mining based on prediction function of BP.The prediction results of quantitative character of Red seed-using watermelon aimed at choosing the seed with good characters and making good characters become new variety,which could be treated as a pre-work for breeding new variety.Papers focused on BP neural networks of Clementin technically and self-organizing competitive neural networks design and implementation.Cluster function of self-organized competition neural network was applied to hereditary works,and the results of cluster reflected the different characters well.Red seed-using watermelon seed resources were reasonably used considering several aspects of quantitative characters,and excellent inbred line from different breeds was selected to copulate.The results were meaningful for parents selection and minimization of the blindness of crossbreed composition.Research shows that using of artificial neural networks study the red seed-using watermelon is technical feasibility and has a well-knit theoretical foundation.We could find the Inbred line regularity from result of Artificial neural networks,which provided theory basis for the inbred line purge,parent choosing and predominance hybrid combination cultivating in process of breed cultivation.