基于BP神经网络的延边地区渤海国遗址预测研究

Prediction of Bohai Kingdom Site in Yanbian Area Based on BP Neural Network

摘要: 本文选择同遗址分布有关的高程、坡度、坡向、距离河流的距离、距离村屯的距离等因素值的数据集为样本，利用BP神经网络建立延边地区渤海国遗址预测模型，并对预测结果分析。结果表明：预测模型的预测准确率达88.7%，高概率区面积占研究区域的17.6%。预测结果高概率区高程集中在海拔270~570 m高程区间，坡度集中在0˚~3˚区间，坡向集中在平地和东北方向，高概率区在河流缓冲区上，主要分布在0~1000 m缓冲区间，在道路缓冲区上主要分布在0~1500 m缓冲区间，在村屯缓冲区上主要分布在500~1500 m缓冲区间。研究结果为今后发掘新的遗址和保护现存的遗址方面提供一定的科学依据。

Abstract: In this study, we selected data sets as the sample which was related to the distribution of the site; the factors included height, gradient, slope aspect, the distance from the river, and the distance from village etc. This is based on a back-propagation (BP) neural net work to establish forecasting model to analyze the predicting results on Bohai kingdom site in Yanbian Korean Autonomous Prefecture. The results suggest that the accuracy of the prediction model gets to 88.7%, the high probability region of the whole study area is 17.6%. The elevation value of model prediction results of the high probability area concentrates in 270 - 570 meters, the slope is concentrated in 0 - 3 degree, the slope direction is concentrated in the flat ground and the Northeast. The high proba-bility region of the river buffer is distributed from 0 to 1000 meters. The site areas of road buffer are distributed within 0 - 1500 meters and the site areas of the village buffer are distributed within 500 - 1500 meters. These results provide the scientific foundations for excavation and protection of the sites.