时间序列预测模型在航空客运量预测研究中的应用外文翻译资料

 2023-01-01 19:36:00

本科毕业设计(论文)

外文翻译

时间序列预测模型在航空客运量预测研究中的应用

作者: Ze-xu ZHAO, Hong-xu WANG, Yu-qiao QIU and Xiao-li LU

国籍:中国

出处:2018年第二届建模、仿真和优化技术及应用国际会议(MSOTA 2018)

中文译文:

关键词: 时间序列预测模型,Vu(e,f,g),NFMTS的反分数函数,Ju(e,f,g),NFMTS的预测函数,AETSPM自动优化搜索方法

摘要: 这篇文章提出了一种新的时间序列预测模型。NFMTS是一组时间序列预测模型。在1993-2001年成都市航空客运量预测中,以A点(0.0002,0.8,0.0002)为计算起点,通过计算筛选,应用NFMTS的自动优化搜索方法,筛选出最佳时间序列预测模型Ju(0.00002,0.8,0.00002)。它不仅可以使成都市航空客运量预测值达到MSE=0和AFER=0,而且可以预测成都市2002-2003年的未来航空客运量。本文提出的预测模型计算简单,使用方便。1993-2003年成都市航空客运量的统计分析。与传统的灰色模型、灰色神经网络模型和灰色支持向量机模型相比,它具有许多优点。

引言: 刘、曾、曾[1]应用传统的灰色模型、灰色神经网络模型和灰色支持向量机模型对成都市1993-2003年的航空客运量进行了预测。实验结果表明,灰色支持向量机模型预测精度(平均相对误差2.42%)明显优于传统灰色模型(平均相对误差6.38%)和灰色神经网络模型(平均相对误差4.54%)。冯、王、尹、陆[2]提出的一组预测模型包含了这些预测模型。对阿拉巴马大学1971-1992年的招生人数进行预测时,预测的招生人数平均预测误差率可达AFER=0.000000,MSE=0.000000。本文对文献[2]的预测功能提出了进一步的改进:对时间序列的未来数据进行预测。利用时间序列预测模型(NFMTS)对成都市1993-2003年的航空客运量进行了预测。以A点(0.0002,0.8,0.0002)为计算起点,经过三次计算筛选,得到了NFMTS的最优时间序列预测模型Ju(0.00002,0.8,0.00002)。并应用Ju(0.00002,0.8,0.00002)对成都市2004-2006年的未来航空客运量进行了预测。本文的预测模型具有明显的优势。

时间序列预测新模型的定义与定理

定义1。将时间序列域设为C={C1,C2,hellip;,Cn},时间序列的年差分公式为

(1)

时间序列的年差分域为D={D2,D3,hellip;,Dn}。差分比的计算公式为

(2)

得到的差分比域是B={B2,B3,hellip;,Bn-1}。

引理1.将时间序列的理论域设为C={C1,C2,hellip;,Cn},年差分域为D={D2,D3,hellip;,Dn},差分比的理论域为B={B2,B3,hellip;,Bn-1}。

定义2.将时间序列的理论域设为C={C1,C2,hellip;,Cn},年差分域为D={D2,D3,hellip;,Dn},差分比域为B={B2,B3,hellip;,Bn-1}。它可以由引理1定义:

其中自变量eisin;(0,1],fisin;(0,1],g(0,1]。Vu(e,f,g)是NFMTS的u年的反函数,或时间序列C的u年的反函数。Ju(e,f,g)是NFMTS的u年的预测函数,或Ju(e,f,g)是NFMTS的u年的预测公式。Ju(e,f,g)也是C时间序列u年的预测函数,或Ju(e,f,g)也是C时间序列u年的预测公式。Ju(e,f,g)也是C时间序列u年的预测函数,或Ju(e,f,g) 也是时间序列C的u年的预测公式。Cu和Cu-1分别是时间序列域C中u年和u-1年的元素。

定义3.将时间序列的理论域设为C={C1,C2,hellip;,Cn},年差分域为D={D2,D3,hellip;,Dn},差分比的理论域为B={B2,B3,hellip;,Bn-1}。可以得到时间序列C的预测公式Ju(e,f,g)。将理论场C、D、B中的数据代入,应用C时间序列u年的预测公式,计算出C时间序列中Cu元素的预测值。因此,Ju(e,f,g)也是时间序列c中u年的预测模型,Ju(e,f,g)也可以称为NFMTS的预测公式,Ju(e,f,g)也可以称为NFMTS的预测模型。

定义4. 将时间序列的理论域设为C={C1,C2,hellip;,Cn},年差分域为D={D2,D3,hellip;,Dn},差分比的理论域为B={B2,B3,hellip;,Bn-1}。该模型可用于建立NFMTS u年的Ju(e,f,g)预测模型。当取所有uisin;{1,2,hellip;,n},取自变量e(0,1),f(0,1))和g(0,1)时,它可以有无限个预测模型Ju(e,f,g)。它们的集成被称为一种新的时间序列预测模型(NFMTS)。这些元素Ju(e,f,g)被称为u年时间序列预测模型的时间序列C。Ju(e,f,g)也是C时间序列u年的预测函数,Ju(e,f,g)也被称为C时间序列u年的预测公式,Mc(u,p,q)也被称为NFMTS时间序列预测模型。Mc(u,p,q)也是NFMTS的预测函数,Mc(u,p,q)也是NFMTS的预测公式

定理1.将时间序列的理论域设为C={C1,C2,hellip;,Cn},年差分域为D={D2,D3,hellip;,Dn},差分比的理论域为B={B2,B3,hellip;,Bn-1}。因此:

i)(NFMTS的预测函数Ju(e,f,g)的连续性定理)NFMTS的预测函数Ju(e,f,g)是半开半闭区间(0,1)上的连续函数。

ii)(NFMTS的预测函数Ju(e,f,g)的收敛定理)对于任意给定的e(0,1),f(0,1]),g(0,1]),如果f是固定的,当e为0,g为0时,NFMTS的预测函数Ju(e,f,g)收敛,u年NFMTS的预测函数Ju(e,f,g)收敛到u年时间序列C的历史数据C u。

定义5.将时间序列的理论域设为C={C1,C2,hellip;,Cn},年差分域为D={D2,D3,hellip;,Dn},差分比的理论域为B={B2,B3,hellip;,Bn-1}。当用NFMTS的预测模型Ju(e,f,g)模拟时间序列Y的历史数据时,Ju(e,f,g)的平均预测错误率AFER=0.000000,MSE=0.000000。定义6.将时间序列的理论域设为C={C1,C2,hellip;,Cn},年差分域为D={D2,D3,hellip;,Dn},差分比的理论域为B={B2,B3,hellip;,Bn-1}。自动找到了NFMTS中的最优时间序列预测模型Ju(e,f,g)。这种方法被称为NFMTS的自动优化搜索方法。即:对于任意时间序列C,从任意点A(e,f,g)为计算起点,固定f,使e和g逐步缩小、编程、搜索、计算,hellip;,直至筛选出NFMTS中的最优时间序列预测模型Ju(e,f,g)。

定理2.(NFMTS中标准时间序列预测模型Ju(e,f,g)的存在定理)将时间序列的理论域设为C={C1,C2,hellip;,Cn},年差分域为D={D2,D3,hellip;,Dn},差分比的理论域为B={B2,B3,hellip;,Bn-1}。以任意点A(e,f,g)为计算起点,采用NFMTS的自动优化搜索方法,可以选择NFMTS中的最优时间序列预测模型Ju(e,f,g)。以成都市1993-2003年的空中交通量为例,说明了如何利用NFMTS的自动优化搜索方法,筛选出NFMTS中最优时间序列预测模型Ju(e,f,g)的详细过程。

实例

实例1. 模拟预测成都市1993-2003年航空客运量,见表1。以A(0.002,0.8,0.002)为计算起点,依次选择e=g=0.02,f=0.8;e=g=0.002,f=0.8;e=g=0.0002,f=0.8;hellip;,通过编程、搜索、计算,hellip;,直到选择了NFMTS中的最优时间序列预测模型Ju(e,f,g)(AFER=0,MSE=0)。当采用预测模型Ju(0.02,0.8,0.02)进行计算时,计算结果填入表1。因为AFER=0.001801ne;0,MSE=0.964545ne;0;继续搜索,当应用Ju(0.002,0.9,8.002)计算时,将计算结果填入表1。因为AFER=0.000160ne;0,MSE=0.011818ne;0;继续搜索,当应用Ju(0.0002,0.8,0.0002)计算时,将计算结果填入表1,因为AFER=0.000000,MSE=0.000000,停止计算。因此,Ju(0.0002,0.8,0.0002)是成都市1993-2003年空中客运量模拟与预测的NFMTS最优时间序列预测模型

表1。采用NFMTS的自动寻优方法筛选了NFMTS的标准时间序列预测模型1993-2003年成都市航空客运量的模拟预测。

注:在表1中,平均预测误差率(AFER):

均方误差

例2.运用错位计算法对成都市2001-2003年的航空客运量进行预测,对未知的2004年成都市航空客运量进行预测。由于缺乏差分比B2003,NFMTS中的最优时间序列预测模型J2004(0.0002,0.8,0.0002)不能用于2001年预测值的计算。B′=(B2002 B2001 B2000)/3=(0.411528 2.101408 1.324627)/3=1.279188,设B′2003=B′=1.279188,然后用NFMTS中的最优时间序列预测模型J2004(0.0002,0.8,0.0002)计算2004年449.3的预测值。可以计算出B2003=1.280130。设B′2004=B2003=1.280130,代入预测模型J2005(0.0002,0.8,0.0002),计算出2005年预测值499.6。可以计算出B2004=1.279898。设B′2005=B2004=1.279898,代入预测模型J2005(0.0002,0.8,0.0002),计算出2006年预测值664.0。具体计算过程如下

表2。应用Ju(0.0002,0.8,0.0002)对2004-2006年成都市航空客运量进行了预测。

例3。采用本文方法和文献[1]方法对成都市1993-2003年航空客运量进行预测,数据见表3。显然,本文中的预测模型Ju(0.0002,0.8,0.0002)具有最高的精度。

表3。利用预测模型对成都市1993-2003年航空客运量进行了预测比较。

总结

作为时间序列预测模型的补充,NFMTS中的最优时间序列预测模型Ju(e,f,g)结构简单,计算方便。与传统的灰色模型、灰色神经网络模型和灰色支持向量机模型相比,该模型在处理1993-2003年成都市客运量方面具有一定的优势。例2给出了一种更科学的B′2003求法,对提高时间序列未来数据的预测精度具有重要意义

附:外文原文

Keywords: Forecasting model of time series, Vu (e,f,g), The inverse fractional function of NFMTS, Ju(e,f,g), The prediction function of NFMTS, AETSPM automatic optimization searching method.

Abstract. This paper presents a new forecasting model of time series (NFMTS). NFMTS is a set of time series forecasting models. In forecasting air passenger volume in Chengdu from 1993 to 2001, take point A (0.0002,0.8,0.0002) as the starting point of calculation, by calculating and screening, the automatic optimization search method of NFMTS can be applied to screen the optimal time series prediction model in NFMTS as Ju (0.0000

剩余内容已隐藏,支付完成后下载完整资料


Keywords: Forecasting model of time series, Vu (e,f,g), The inverse fractional function of NFMTS, Ju(e,f,g), The prediction function of NFMTS, AETSPM automatic optimization searching method.

Abstract. This paper presents a new forecasting model of time series (NFMTS). NFMTS is a set of time series forecasting models. In forecasting air passenger volume in Chengdu from 1993 to 2001, take point A (0.0002,0.8,0.0002) as the starting point of calculation, by calculating and screening, the automatic optimization search method of NFMTS can be applied to screen the optimal time series prediction model in NFMTS as Ju (0.00002,0.8,0.00002). It can not only make the predicted value of Chengdus air passenger volume reach MSE=0 and AFER=0, but also can predict the future air passenger volume of Chengdu from 2002 to 2003. The prediction model presented in this paper is simple in calculation and easy to use. In dealing with the air passenger volume of Chengdu from 1993 to 2003. It has advantages over traditional grey model, grey neural network model and grey support vector machine model.

Introduction

Liu, Zeng and Zeng [1] applied the traditional grey model, grey neural network model and grey support vector machine model to predict the air passenger volume of Chengdu from 1993 to 2003. The conclusion of the experiment is that the prediction accuracy (average relative error 2.42%) of the model mode of gray support vector machine is much better than the traditional grey model (average relative error 6.38%) and the grey neural network model (average relative error 4.54%). The set of prediction models proposed by Feng, Wang, Yin and Lu[2] contains these prediction models. When predicting the enrollment Numbers of the university of Alabama from 1971 to 1992, the average prediction error rate of the predicted enrollment Numbers can reach AFER=0.000000 and MSE=0.000000. This paper proposes to improve the predictive function of reference [2] one more step: to predict the future data of time series. A new forecasting model for time series (NFMTS) was used to predict air passenger traffic in Chengdu from 1993 to 2003. The optimal time series prediction model Ju (0.00002,0.8,0.00002) of NFMTS was obtained after three calculations and screening from point A (0.0002,0.8,0.0002) as the starting point of calculation. In addition, Ju(0.00002,0.8,0.00002) is applied to predict the future air passenger volume of Chengdu from 2004 to 2006. The prediction model in this paper has obvious advantages.

Definition and Theorem of New Forecasting Model for Time Series (NFMTS)

Definition 1. set the time series domain as C= {C1, C2, hellip;, Cn}.The annual difference formula of time series is

Du = Cu –Cu-1. (1) The annual difference domain of time series as D= {D2, D3, hellip;, Dn}. The calculation formula of difference ratio is

43

Bu-1= Du/ Du-1 (2) The domain of the difference ratio obtained is B={ B2, B3, hellip;, Bn-1}.

Lemma 1. Set the theoretical domain of time series as C= {C1, C2, hellip;, Cn},The annual difference domain is D= {D2, D3, hellip;, Dn},The theoretical domain of difference ratio is B={ B2, B3, hellip;, Bn-1}. Du = Bu-1 Du-1. (3) Definition 2. set the theoretical domain of time series as C= {C1, C2, hellip;, Cn}, The annual difference domain is D= {D2, D3, hellip;, Dn}, The domain of difference ratio is B={ B2, B3, hellip;, Bn-1}. It can be defined by lemma 1:

   

            

Where the independent variables e(0, 1], f(0, 1], g(0, 1]. Vu(e,f,g) is the inverse function of u years of NFMTS, or the inverse function of u years of time series C. Ju(e,f,g) is the prediction function of u years of NFMTS, or Ju(e,f,g) is the prediction formula of u uears of NFMTS. Ju(e,f,g) is also the predictive function for u years of time series C, or Ju(e,f,g) is also the predictive formula for u years of time series C. Ju(e,f,g) is also the predictive function for u years of time series C, or Ju(e,f,g) is also the predictive formula for u years of time series C. Cu and Cu-1 are the elements of u year and u-1 year respectively in the time series domain C.

Definition 3. set the theoretical domain of time series as C= {C1, C2, hellip;, Cn}, The annual difference domain is D= {D2, D3, hellip;, Dn}.The theoretical domain of difference ratio is B={ B2, B3, hellip;, Bn-1}. The prediction formula Ju(e,f,g) of time series C can be made. By substituting the data in the theoretical fields C, D and B and applying the prediction formula of u year of time series C, the predicted value of element Cu in time series C can be calculated. Therefore, Ju(e,f,g) is also the prediction model of u year in time series c. Ju(e,f,g) can also be called the prediction formula of NFMTS, or Ju(e,f,g) is called the prediction model of NFMTS.

Definition 4. set the theoretical domain of time series as C= {C1, C2, hellip;, Cn}, The annual difference domain is D= {D2, D3, hellip;, Dn},The theoretical domain of difference ratio is B={ B2, B3, hellip;, Bn-1}. It can establish the prediction model Ju(e,f,g) for u year of NFMTS. When take all of u{1,2, hellip;,n}, and take the independent variables e (0, 1], f (0, 1]) and g (0, 1), it can have an infinite number of prediction models Ju(e,f,g). The ensemble of them is called A new forecasting model for time series (NFMTS). Any of these elements Ju(e,f,g) is called the time series C of the time series forecast model of u years. Ju(e,f,g) is also the prediction function of u years of time series C, or Ju(e,f,g) is also called the prediction formula of u years of time series C. Mc(u,p,q) is also called a time series prediction model of NFMTS.

剩余内容已隐藏,支付完成后下载完整资料


资料编号:[268494],资料为PDF文档或Word文档,PDF文档可免费转换为Word

您需要先支付 30元 才能查看全部内容!立即支付

发小红书推广免费获取该资料资格。点击链接进入获取推广文案即可: Ai一键组稿 | 降AI率 | 降重复率 | 论文一键排版