加油站成品油销量预测算法与样本数据设计  

Gasoline sales prediction and sample data design

作　　者：杨庆[1] 黄德先[1] 

机构地区：[1]清华大学自动化系,北京100084

出　　处：《计算机工程与应用》2007年第13期210-213,219,共5页Computer Engineering and Applications

基　　金：国家自然科学基金(the National Natural Science Foundation of China under Grant No.60574072)。

摘　　要：为了改善目前部分加油站的成品油销量短期预测状况,提出了一个基于线性判定的组合预测算法框架。为了解决建模中样本数据选择问题,在分析加油站每日成品油销售记录和相关辅助信息的基础上,针对BP网络模型提出了几种样本选择方式和网络输入输出设计模式。研究内容可供实际加油站短期预测参考。In order to improve the current shot-term predictive technical status in gas stations,a combined forecasting algorithm skeleton based identification of linear property is proposed.For solving sample selection during the modeling,several methods of selecting samples and designing patterns of network in-out on aiming at BP model are presented on the basis of analyzing daily sale records and specified interrelated assistant information in a gas station.This research could supply references to practical short term prediction in a gas station.

关 键 词：加油站 预测算法 样本选择 输入输出 BP网络 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]