CN114626635A - A method and system for predicting steel logistics cost based on hybrid neural network - Google Patents

A method and system for predicting steel logistics cost based on hybrid neural network Download PDF

Info

Publication number
CN114626635A
CN114626635A CN202210349735.XA CN202210349735A CN114626635A CN 114626635 A CN114626635 A CN 114626635A CN 202210349735 A CN202210349735 A CN 202210349735A CN 114626635 A CN114626635 A CN 114626635A
Authority
CN
China
Prior art keywords
neural network
logistics cost
hybrid neural
data
steel
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
CN202210349735.XA
Other languages
Chinese (zh)
Other versions
CN114626635B (en
Inventor
薄胜
李媛
程渤
叶弘毅
赵静怡
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Beijing Lezhi Technology Co ltd
Beijing University of Posts and Telecommunications
Original Assignee
Beijing Lezhi Technology Co ltd
Beijing University of Posts and Telecommunications
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Beijing Lezhi Technology Co ltd, Beijing University of Posts and Telecommunications filed Critical Beijing Lezhi Technology Co ltd
Priority to CN202210349735.XA priority Critical patent/CN114626635B/en
Publication of CN114626635A publication Critical patent/CN114626635A/en
Application granted granted Critical
Publication of CN114626635B publication Critical patent/CN114626635B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/08—Logistics, e.g. warehousing, loading or distribution; Inventory or stock management

Landscapes

  • Engineering & Computer Science (AREA)
  • Business, Economics & Management (AREA)
  • Physics & Mathematics (AREA)
  • Economics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Strategic Management (AREA)
  • Human Resources & Organizations (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Operations Research (AREA)
  • General Business, Economics & Management (AREA)
  • Tourism & Hospitality (AREA)
  • Quality & Reliability (AREA)
  • Marketing (AREA)
  • Development Economics (AREA)
  • General Engineering & Computer Science (AREA)
  • Evolutionary Computation (AREA)
  • Computational Linguistics (AREA)
  • Biophysics (AREA)
  • Data Mining & Analysis (AREA)
  • Biomedical Technology (AREA)
  • Software Systems (AREA)
  • Artificial Intelligence (AREA)
  • Mathematical Physics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Computing Systems (AREA)
  • Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • General Health & Medical Sciences (AREA)
  • Game Theory and Decision Science (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

本发明涉及一种基于混合神经网络的钢铁物流成本预测方法及系统，方法包括获取待测的钢铁物流成本数据；将所述待测的钢铁物流成本数据输入至训练好的预测模型中，得到预测完成的钢铁物流成本；其中预测模型为训练并优化好的混合神经网络。本发明针对传统物流成本预测方法中存在的随机性进行了改进，利用混合神经网络提高了预测结果的精确程度，并同时优化了传统混合神经网络中存在的局部最优及迭代次数过多的问题，有很高的预测精确度以及较高的可实施性。

Figure 202210349735

The invention relates to a method and system for predicting steel logistics cost based on a hybrid neural network. The method includes acquiring steel logistics cost data to be measured; inputting the steel logistics cost data to be measured into a trained prediction model to obtain a prediction The completed steel logistics cost; where the prediction model is a trained and optimized hybrid neural network. The invention improves the randomness existing in the traditional logistics cost prediction method, improves the accuracy of the prediction result by using the hybrid neural network, and at the same time optimizes the problems of local optimum and too many iterations in the traditional hybrid neural network , with high prediction accuracy and high practicability.

Figure 202210349735

Description

一种基于混合神经网络的钢铁物流成本预测方法及系统A method and system for predicting steel logistics cost based on hybrid neural network

技术领域technical field

本发明涉及物流信息技术领域，特别是涉及一种基于混合神经网络的钢铁物流成本预测方法及系统。The invention relates to the technical field of logistics information, in particular to a method and system for predicting the cost of iron and steel logistics based on a hybrid neural network.

背景技术Background technique

随着经济发展，物流产业在各行各业的重要性日益上升，因而对于物流成本的预测至关重要。With the development of economy, the importance of logistics industry in all walks of life is increasing day by day, so the forecast of logistics cost is very important.

在钢铁产业领域，物流同样有着重要作用。然而，由于传统的钢铁物流成本预测方式往往是基于经验进行预测，对于钢铁物流成本预测这类影响因素繁杂的预测往往极度不准确，存在着诸多缺陷，这导致了钢铁企业物流成本预估、核算困难，不利于钢铁企业的经济发展及钢铁物流管理水平的进步。In the steel industry, logistics also plays an important role. However, since the traditional steel logistics cost forecasting method is often based on experience, forecasts with complex influencing factors such as steel logistics cost forecasting are often extremely inaccurate and have many defects, which lead to the estimation and accounting of the logistics cost of iron and steel enterprises. Difficulties are not conducive to the economic development of iron and steel enterprises and the progress of iron and steel logistics management.

发明内容SUMMARY OF THE INVENTION

为了克服现有技术的不足，本发明的目的是提供一种基于混合神经网络的钢铁物流成本预测方法及系统。In order to overcome the deficiencies of the prior art, the purpose of the present invention is to provide a method and system for predicting the cost of steel logistics based on a hybrid neural network.

为实现上述目的，本发明提供了如下方案：For achieving the above object, the present invention provides the following scheme:

一种基于混合神经网络的钢铁物流成本预测方法，包括：A method for predicting steel logistics cost based on hybrid neural network, including:

获取待测的钢铁物流成本数据；Obtain the steel logistics cost data to be tested;

将所述待测的钢铁物流成本数据输入至训练好的预测模型中，得到预测完成的钢铁物流成本；Inputting the steel logistics cost data to be measured into the trained prediction model to obtain the predicted completed steel logistics cost;

所述预测模型的确定方法为：The method for determining the prediction model is:

获取训练的钢铁物流成本数据；Obtain training steel logistics cost data;

对所述训练的钢铁物流成本数据进行预处理，得到训练集和测试集；Preprocessing the trained steel logistics cost data to obtain a training set and a test set;

对预设的混合神经网络进行初始化，并根据所述训练集对初始化后的混合神经网络进行训练，得到训练后的混合神经网络；Initializing the preset hybrid neural network, and training the initialized hybrid neural network according to the training set to obtain a trained hybrid neural network;

基于全局最优的优化方法对所述训练后的混合神经网络进行参数迭代，得到所述训练好的预测模型。Parameter iteration is performed on the trained hybrid neural network based on the global optimal optimization method to obtain the trained prediction model.

优选地，还包括：Preferably, it also includes:

根据所述测试集对所述训练好的预测模型进行模型测试。Model testing is performed on the trained prediction model according to the test set.

优选地，所述训练的钢铁物流成本数据的表达式为S＝{x1，x2，...，xn，y}，其中，S为所述训练的钢铁物流成本数据的历史数据集合，xn为第n个影响因素，y为钢铁物流成本值。Preferably, the expression of the trained steel logistics cost data is S={x 1 , x 2 , . . . , x n , y}, where S is the historical data set of the trained steel logistics cost data , x n is the nth influencing factor, and y is the cost value of steel logistics.

优选地，所述影响因素包括车辆出发起始地点、车辆运输目的地、车辆类型、运输日天气、路况、货物种类和货物重量。Preferably, the influencing factors include the departure and origin of the vehicle, the transportation destination of the vehicle, the type of the vehicle, the weather on the day of transportation, road conditions, the type of goods and the weight of the goods.

优选地，所述对所述训练的钢铁物流成本数据进行预处理，得到训练集和测试集，包括：Preferably, the steel logistics cost data of the training is preprocessed to obtain a training set and a test set, including:

对所述历史数据集合进行矩阵化处理，得到第一数据矩阵和第二数据矩阵；所述第一数据矩阵的表达式为X＝[xij]m×n，所述第二数据矩阵的表达式为Y＝[yi]m×1；其中，X为所述第一数据矩阵，xij为第i条所述历史数据集合中的第j个影响因素；yi为第i条所述历史数据集合中的物流成本值；n为所述影响因素的数量；m为所述钢铁物流成本数据的总计条数；Perform matrix processing on the historical data set to obtain a first data matrix and a second data matrix; the expression of the first data matrix is X=[x ij ] m×n , the expression of the second data matrix is The formula is Y=[y i ] m×1 ; wherein, X is the first data matrix, x ij is the j-th influencing factor in the i-th historical data set; y i is the i-th article The logistics cost value in the historical data set; n is the number of the influencing factors; m is the total number of the steel logistics cost data;

分别对所述第一数据矩阵和所述第二数据矩阵进行归一化处理，得到第一归一化数据和第二归一化数据；所述第一归一化数据的表达式为x*＝2×

Figure BDA0003579214510000021
所述第二归一化数据的表达式为
Figure BDA0003579214510000022
其中，xmax为所述影响因素的最大值，xmin为所述影响因素的最小值；ymax为所述物流成本值的最大值；ymin为所述物流成本值的最小值；The first data matrix and the second data matrix are respectively normalized to obtain the first normalized data and the second normalized data; the expression of the first normalized data is x * =2×
Figure BDA0003579214510000021
The expression of the second normalized data is
Figure BDA0003579214510000022
Wherein, x max is the maximum value of the influence factor, x min is the minimum value of the influence factor; y max is the maximum value of the logistics cost value; y min is the minimum value of the logistics cost value;

分别对所述第一归一化数据和所述第二归一化数据进行矩阵化表示，得到第一处理矩阵和第二处理矩阵；所述第一处理矩阵的表达式为

Figure BDA0003579214510000023
所述第二处理矩阵的表达式为
Figure BDA0003579214510000024
其中，X*为所述第一处理矩阵，
Figure BDA0003579214510000025
为第i条所述历史数据集合中的第j个影响因素的归一化后的值，Y*为所述第二处理矩阵，
Figure BDA0003579214510000026
为第i条所述历史数据集合中的物流成本值的归一化后的值；Perform matrix representation on the first normalized data and the second normalized data respectively to obtain a first processing matrix and a second processing matrix; the expression of the first processing matrix is
Figure BDA0003579214510000023
The expression of the second processing matrix is
Figure BDA0003579214510000024
where X * is the first processing matrix,
Figure BDA0003579214510000025
is the normalized value of the jth influencing factor in the i-th historical data set, Y * is the second processing matrix,
Figure BDA0003579214510000026
is the normalized value of the logistics cost value in the historical data set described in item i;

根据所述第一处理矩阵和所述第二处理矩阵确定所述训练集和所述测试集。The training set and the test set are determined according to the first processing matrix and the second processing matrix.

优选地，所述对预设的混合神经网络进行初始化，并根据所述训练集对初始化后的混合神经网络进行训练，得到训练后的混合神经网络，包括：Preferably, the preset hybrid neural network is initialized, and the initialized hybrid neural network is trained according to the training set to obtain a trained hybrid neural network, including:

获取所述预设的混合神经网络；所述混合神经网络包括依次连接的输入层、隐藏层和输出层；所述输入层的输入节点的数量等于所述影响因素的数量；Obtain the preset hybrid neural network; the hybrid neural network includes an input layer, a hidden layer and an output layer connected in sequence; the number of input nodes of the input layer is equal to the number of the influencing factors;

根据所述影响因素的数量确定所述隐藏层的节点数量；所述隐藏层的节点数量的公式为：

Figure BDA0003579214510000031
其中，n*为所述隐藏层的节点数量，a为预设的经验值；The number of nodes of the hidden layer is determined according to the number of the influencing factors; the formula of the number of nodes of the hidden layer is:
Figure BDA0003579214510000031
Wherein, n * is the number of nodes in the hidden layer, and a is a preset empirical value;

根据所述第二处理矩阵确定所述输出层的节点数量；determining the number of nodes of the output layer according to the second processing matrix;

初始化所述混合神经网络的混合神经网络参数；所述混合神经网络参数包括所述混合神经网络的学习率、训练目标最小误差、最大迭代次数、节点权重、节点偏置、激活函数和输出函数；Initializing the hybrid neural network parameters of the hybrid neural network; the hybrid neural network parameters include the learning rate of the hybrid neural network, the minimum error of the training target, the maximum number of iterations, the node weight, the node bias, the activation function and the output function;

根据所述训练集对初始化混合神经网络参数后的所述混合神经网络进行训练，得到训练后的混合神经网络。The hybrid neural network after initializing the parameters of the hybrid neural network is trained according to the training set to obtain a trained hybrid neural network.

优选地，所述基于全局最优的优化方法对所述训练后的混合神经网络进行参数迭代，得到所述训练好的预测模型，包括：Preferably, the optimization method based on the global optimum performs parameter iteration on the trained hybrid neural network to obtain the trained prediction model, including:

确定钢铁物流成本预测值与准确值之间的损失值；所述损失值的计算公式为C＝(y*-y)2；其中C为所述损失值，y*为所述钢铁物流成本预测值；Determine the loss value between the steel logistics cost prediction value and the accurate value; the calculation formula of the loss value is C=(y * -y) 2 ; wherein C is the loss value, and y * is the steel logistics cost prediction value;

在训练过程中，当所述混合神经网络到一个新的钢铁物流成本损失Ct后，对Ct进行随机的扰动，得到新的钢铁物流成本损失Ct+1，并以概率P选择是否更新本次扰动，并将预设的优化常量乘以系数进行更新；所述概率的表达式为

Figure BDA0003579214510000032
其中，K为所述优化常量；In the training process, when the hybrid neural network reaches a new steel logistics cost loss C t , random perturbation is performed on C t to obtain a new steel logistics cost loss C t+1 , and the probability P is used to select whether to update or not This disturbance is updated by multiplying the preset optimization constant by the coefficient; the expression of the probability is
Figure BDA0003579214510000032
Wherein, K is the optimization constant;

当更新的所述优化常量达到所述最大迭代次数时，取多个连续的所述钢铁物流成本损失并判断是否均在所述概率下不被接受，若是，则得到所述训练好的预测模型。When the updated optimization constant reaches the maximum number of iterations, take a plurality of consecutive steel logistics cost losses and judge whether they are all unacceptable under the probability, and if so, obtain the trained prediction model .

优选地，所述待测的钢铁物流成本数据的表达式为T＝{x1，x2，...，xn}；其中，T为所述待测的钢铁物流成本数据的集合；xn为第n个影响因素。Preferably, the expression of the steel logistics cost data to be measured is T={x 1 , x 2 , . . . , x n }; wherein, T is the set of the steel logistics cost data to be measured; x n is the nth influencing factor.

一种基于混合神经网络的钢铁物流成本预测系统，包括：A steel logistics cost prediction system based on hybrid neural network, including:

待测数据获取模块，用于获取待测的钢铁物流成本数据；The data acquisition module to be tested is used to acquire the steel logistics cost data to be tested;

预测模块，用于将所述待测的钢铁物流成本数据输入至训练好的预测模型中，得到预测完成的钢铁物流成本；A prediction module, used to input the steel logistics cost data to be measured into the trained prediction model to obtain the predicted completed steel logistics cost;

模型确定模块，具体包括：Model determination module, including:

训练数据获取单元，用于获取训练的钢铁物流成本数据；The training data acquisition unit is used to acquire the steel logistics cost data for training;

预处理单元，用于对所述训练的钢铁物流成本数据进行预处理，得到训练集和测试集；a preprocessing unit for preprocessing the training steel logistics cost data to obtain a training set and a test set;

初始化单元，用于对预设的混合神经网络进行初始化，并根据所述训练集对初始化后的混合神经网络进行训练，得到训练后的混合神经网络；an initialization unit, configured to initialize a preset hybrid neural network, and train the initialized hybrid neural network according to the training set to obtain a trained hybrid neural network;

优化单元，用于基于全局最优的优化方法对所述训练后的混合神经网络进行参数迭代，得到所述训练好的预测模型。An optimization unit, configured to perform parameter iteration on the trained hybrid neural network based on a globally optimal optimization method to obtain the trained prediction model.

优选地，还包括：Preferably, it also includes:

测试模块，用于根据所述测试集对所述训练好的预测模型进行模型测试。A test module, configured to perform model testing on the trained prediction model according to the test set.

根据本发明提供的具体实施例，本发明公开了以下技术效果：According to the specific embodiments provided by the present invention, the present invention discloses the following technical effects:

本发明提供了一种基于混合神经网络的钢铁物流成本预测方法及系统，针对传统物流成本预测方法中存在的随机性进行了改进，利用混合神经网络提高了预测结果的精确程度，并同时优化了传统混合神经网络中存在的局部最优及迭代次数过多的问题，有很高的预测精确度以及较高的可实施性。The invention provides a method and system for predicting the cost of iron and steel logistics based on a hybrid neural network, which improves the randomness existing in the traditional logistics cost prediction method, improves the accuracy of the prediction result by using the hybrid neural network, and optimizes the The problems of local optima and too many iterations in the traditional hybrid neural network have high prediction accuracy and high practicability.

附图说明Description of drawings

为了更清楚地说明本发明实施例或现有技术中的技术方案，下面将对实施例中所需要使用的附图作简单地介绍，显而易见地，下面描述中的附图仅仅是本发明的一些实施例，对于本领域普通技术人员来讲，在不付出创造性劳动性的前提下，还可以根据这些附图获得其他的附图。In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the accompanying drawings required in the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some of the present invention. In the embodiments, for those of ordinary skill in the art, other drawings can also be obtained according to these drawings without creative labor.

图1为本发明提供的实施例中的预测方法的流程图；1 is a flowchart of a prediction method in an embodiment provided by the present invention;

图2为本发明提供的实施例中的混合神经网络的结构示意图；2 is a schematic structural diagram of a hybrid neural network in an embodiment provided by the present invention;

图3为本发明提供的实施例中的模型训练和预测流程图；3 is a flow chart of model training and prediction in the embodiment provided by the present invention;

图4为本发明提供的实施例中的混合神经网络的激活函数的示意图。FIG. 4 is a schematic diagram of an activation function of a hybrid neural network in an embodiment provided by the present invention.

具体实施方式Detailed ways

下面将结合本发明实施例中的附图，对本发明实施例中的技术方案进行清楚、完整地描述，显然，所描述的实施例仅仅是本发明一部分实施例，而不是全部的实施例。基于本发明中的实施例，本领域普通技术人员在没有做出创造性劳动前提下所获得的所有其他实施例，都属于本发明保护的范围。The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

在本文中提及“实施例”意味着，结合实施例描述的特定特征、结构或特性可以包含在本申请的至少一个实施例中。在说明书中的各个位置出现该短语并不一定均是指相同的实施例，也不是与其它实施例互斥的独立的或备选的实施例。本领域技术人员显式地和隐式地理解的是，本文所描述的实施例可以与其它实施例相结合。Reference herein to an "embodiment" means that a particular feature, structure, or characteristic described in connection with the embodiment can be included in at least one embodiment of the present application. The appearances of the phrase in various places in the specification are not necessarily all referring to the same embodiment, nor a separate or alternative embodiment that is mutually exclusive of other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

本申请的说明书和权利要求书及所述附图中的术语“第一”、“第二”、“第三”和“第四”等是用于区别不同对象，而不是用于描述特定顺序。此外，术语“包括”和“具有”以及它们任何变形，意图在于覆盖不排他的包含。例如包含了一系列步骤、过程、方法等没有限定于已列出的步骤，而是可选地还包括没有列出的步骤，或可选地还包括对于这些过程、方法、产品或设备固有的其它步骤元。The terms "first", "second", "third" and "fourth" in the description and claims of the present application and the drawings are used to distinguish different objects, rather than to describe a specific order . Furthermore, the terms "comprising" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, including a series of steps, processes, methods, etc. is not limited to the listed steps, but optionally also includes unlisted steps, or optionally also includes inherent to these processes, methods, products or devices. other steps.

本发明的目的是提供一种基于混合神经网络的钢铁物流成本预测方法及系统，能够提高对钢铁物流成本的预测精度。The purpose of the present invention is to provide a method and system for predicting the cost of iron and steel logistics based on a hybrid neural network, which can improve the prediction accuracy of the cost of iron and steel logistics.

为使本发明的上述目的、特征和优点能够更加明显易懂，下面结合附图和具体实施方式对本发明作进一步详细的说明。In order to make the above objects, features and advantages of the present invention more clearly understood, the present invention will be described in further detail below with reference to the accompanying drawings and specific embodiments.

图1为本发明提供的实施例中的预测方法的流程图，如图1所示，本发明提供了一种基于混合神经网络的钢铁物流成本预测方法，包括：Fig. 1 is a flow chart of a prediction method in an embodiment provided by the present invention. As shown in Fig. 1, the present invention provides a method for predicting steel logistics cost based on a hybrid neural network, including:

步骤100获取待测的钢铁物流成本数据；Step 100 obtains the steel logistics cost data to be tested;

步骤200将所述待测的钢铁物流成本数据输入至训练好的预测模型中，得到预测完成的钢铁物流成本；In step 200, the steel logistics cost data to be measured is input into the trained prediction model, and the predicted completed steel logistics cost is obtained;

所述预测模型的确定方法为：The method for determining the prediction model is:

步骤201：获取训练的钢铁物流成本数据；Step 201: Obtain training steel logistics cost data;

步骤202：对所述训练的钢铁物流成本数据进行预处理，得到训练集和测试集；Step 202: Preprocessing the training steel logistics cost data to obtain a training set and a test set;

步骤203：对预设的混合神经网络进行初始化，并根据所述训练集对初始化后的混合神经网络进行训练，得到训练后的混合神经网络；Step 203: Initialize the preset hybrid neural network, and train the initialized hybrid neural network according to the training set to obtain a trained hybrid neural network;

步骤204：基于全局最优的优化方法对所述训练后的混合神经网络进行参数迭代，得到所述训练好的预测模型。Step 204: Perform parameter iteration on the trained hybrid neural network based on a globally optimal optimization method to obtain the trained prediction model.

优选地，还包括：Preferably, it also includes:

步骤300：根据所述测试集对所述训练好的预测模型进行模型测试。Step 300: Perform model testing on the trained prediction model according to the test set.

优选地，所述训练的钢铁物流成本数据的表达式为S＝{x1，x2，...，xn，y}，其中，S为所述训练的钢铁物流成本数据的历史数据集合，xn为第n个影响因素，y为钢铁物流成本值。Preferably, the expression of the trained steel logistics cost data is S={x 1 , x 2 , . . . , x n , y}, where S is the historical data set of the trained steel logistics cost data , x n is the nth influencing factor, and y is the cost value of steel logistics.

图2为本发明提供的实施例中的模型训练的流程图，如图2所示，混合神经网络有三层，第一层输入层的节点都是输入节点，每个节点表示一个输入的数据，钢铁物流成本数据集合用以下集合S表示：Fig. 2 is a flow chart of model training in the embodiment provided by the present invention. As shown in Fig. 2, the hybrid neural network has three layers, the nodes of the input layer of the first layer are input nodes, and each node represents an input data, The steel logistics cost data set is represented by the following set S:

S＝{x1，x2，...，xn，y}S={x 1 , x 2 , ..., x n , y}

如此，输入层输入节点需要分别输入x1，x2，...，xn。In this way, the input layer input nodes need to input x 1 , x 2 , ..., x n respectively.

其中n表示该集合中总的影响钢铁物流成本的影响因素，xi表示第i个影响因素，满足1≤i≤n，y表示钢铁物流成本值。Among them, n represents the total influencing factors of the steel logistics cost in the set, xi represents the i-th influencing factor, and 1≤i≤n is satisfied, and y represents the steel logistics cost value.

优选地，所述影响因素包括车辆出发起始地点、车辆运输目的地、车辆类型、运输日天气、路况、货物种类和货物重量。Preferably, the influencing factors include the departure and origin of the vehicle, the transportation destination of the vehicle, the type of the vehicle, the weather on the day of transportation, road conditions, the type of goods and the weight of the goods.

进一步地，上述影响钢铁物流成本的影响因素是指车辆出发起始地点、车辆运输目的地、车辆类型、运输日天气(晴，雨，雾，雪等)、路况、货物种类、货物重量以及与成本相关的所有此方法验证为可利用的因素。Further, the above-mentioned influencing factors affecting the cost of iron and steel logistics refer to the starting point of the vehicle, the destination of the vehicle, the type of the vehicle, the weather on the transportation day (sunny, rainy, fog, snow, etc.), road conditions, the type of goods, the weight of the goods, and the All cost-related factors for this method are validated as available.

图3为本发明提供的实施例中的模型训练和预测流程图，正如图3所示，本实施例提供的训练和预测方法包括：FIG. 3 is a flow chart of model training and prediction in the embodiment provided by the present invention. As shown in FIG. 3 , the training and prediction method provided by this embodiment includes:

S1，获取历史相关的钢铁物流成本数据；S1, obtain historically related steel logistics cost data;

S2，预处理所述的钢铁物流成本数据；S2, preprocessing the steel logistics cost data;

S3，初始化混合神经网络，并训练神经网络，最后采用全局最优的优化方式进行参数迭代，优化模型；S3, initialize the hybrid neural network, train the neural network, and finally use the global optimal optimization method to iterate the parameters to optimize the model;

S4，利用优化的混合神经网络对钢铁物流成本预测。S4, using the optimized hybrid neural network to predict the logistics cost of steel.

优选地，所述步骤202包括：Preferably, the step 202 includes:

对所述历史数据集合进行矩阵化处理，得到第一数据矩阵和第二数据矩阵；所述第一数据矩阵的表达式为X＝[xij]m×n，所述第二数据矩阵的表达式为Y＝[yi]m×1；其中，X为所述第一数据矩阵，xij为第i条所述历史数据集合中的第j个影响因素；yi为第i条所述历史数据集合中的物流成本值；n为所述影响因素的数量；m为所述钢铁物流成本数据的总计条数；Perform matrix processing on the historical data set to obtain a first data matrix and a second data matrix; the expression of the first data matrix is X=[x ij ] m×n , the expression of the second data matrix is The formula is Y=[y i ] m×1 ; wherein, X is the first data matrix, x ij is the j-th influencing factor in the i-th historical data set; y i is the i-th article The logistics cost value in the historical data set; n is the number of the influencing factors; m is the total number of the steel logistics cost data;

分别对所述第一数据矩阵和所述第二数据矩阵进行归一化处理，得到第一归一化数据和第二归一化数据；所述第一归一化数据的表达式为

Figure BDA0003579214510000071
Figure BDA0003579214510000072
所述第二归一化数据的表达式为
Figure BDA0003579214510000073
其中，xmax为所述影响因素的最大值，xmin为所述影响因素的最小值；ymax为所述物流成本值的最大值；ymin为所述物流成本值的最小值；The first data matrix and the second data matrix are respectively normalized to obtain the first normalized data and the second normalized data; the expression of the first normalized data is
Figure BDA0003579214510000071
Figure BDA0003579214510000072
The expression of the second normalized data is
Figure BDA0003579214510000073
Wherein, x max is the maximum value of the influence factor, x min is the minimum value of the influence factor; y max is the maximum value of the logistics cost value; y min is the minimum value of the logistics cost value;

分别对所述第一归一化数据和所述第二归一化数据进行矩阵化表示，得到第一处理矩阵和第二处理矩阵；所述第一处理矩阵的表达式为

Figure BDA0003579214510000074
所述第二处理矩阵的表达式为
Figure BDA0003579214510000075
其中，X*为所述第一处理矩阵，
Figure BDA0003579214510000076
为第i条所述历史数据集合中的第j个影响因素的归一化后的值，Y*为所述第二处理矩阵，
Figure BDA0003579214510000077
为第i条所述历史数据集合中的物流成本值的归一化后的值；Perform matrix representation on the first normalized data and the second normalized data respectively to obtain a first processing matrix and a second processing matrix; the expression of the first processing matrix is
Figure BDA0003579214510000074
The expression of the second processing matrix is
Figure BDA0003579214510000075
where X * is the first processing matrix,
Figure BDA0003579214510000076
is the normalized value of the jth influencing factor in the i-th historical data set, Y * is the second processing matrix,
Figure BDA0003579214510000077
is the normalized value of the logistics cost value in the historical data set described in item i;

根据所述第一处理矩阵和所述第二处理矩阵确定所述训练集和所述测试集。The training set and the test set are determined according to the first processing matrix and the second processing matrix.

具体的，本实施例中输入的数据是以矩阵形式进行输入：Specifically, the data input in this embodiment is input in the form of a matrix:

X＝[xij]m×n，Y＝[yi]m×1。X=[x ij ] m×n , Y=[y i ] m×1 .

其中元素xij第i条历史数据中集合S中的第j个影响因素，元素yi代表第i条历史数据中的物流成本值。X和Y矩阵的划分需要由实现者手动对数据进行切割处理，使用矩阵运算可以通过向量运算大大降低运算及实现的复杂性。The element x ij is the j-th influencing factor in the set S in the i-th historical data, and the element y i represents the logistics cost value in the i-th historical data. The division of the X and Y matrices requires the implementer to manually cut the data. Using matrix operations can greatly reduce the complexity of operations and implementations through vector operations.

进一步地，本实施例中输入数据之前，对数据矩阵X，Y进行归一化以及训练集，测试集数据的划分。Further, before the data is input in this embodiment, the data matrices X and Y are normalized and the data of the training set and the test set are divided.

归一化处理，让所有数据的值在[-1，1]之间，采用以下公式：Normalization processing, so that all data values are between [-1, 1], using the following formula:

Figure BDA0003579214510000081
Figure BDA0003579214510000081

其中max，min分别代表xi对应的钢铁物流成本影响因素、钢铁物流成本y中的最大最小值。Among them, max and min represent the maximum and minimum value of the iron and steel logistics cost influencing factors corresponding to xi and the iron and steel logistics cost y, respectively.

处理后的数据采用以下矩阵X*，Y*表示：The processed data is represented by the following matrices X * , Y * :

Figure BDA0003579214510000082
Figure BDA0003579214510000082

其中，矩阵元素

Figure BDA0003579214510000083
分别表示xij，yi归一化后的值Among them, the matrix elements
Figure BDA0003579214510000083
respectively represent the normalized values of x ij and y i

更进一步地，对归一化之后的矩阵数据X*，Y*进行矩阵划分，取其中的前80％作为训练数据集

Figure BDA0003579214510000084
剩下20％作为测试集
Figure BDA0003579214510000085
Further, perform matrix division on the normalized matrix data X * , Y * , and take the first 80% of them as the training data set
Figure BDA0003579214510000084
The remaining 20% serve as the test set
Figure BDA0003579214510000085

优选地，所述步骤203包括：Preferably, the step 203 includes:

获取所述预设的混合神经网络；所述混合神经网络包括依次连接的输入层、隐藏层和输出层；所述输入层的输入节点的数量等于所述影响因素的数量；Obtain the preset hybrid neural network; the hybrid neural network includes an input layer, a hidden layer and an output layer connected in sequence; the number of input nodes of the input layer is equal to the number of the influencing factors;

根据所述影响因素的数量确定所述隐藏层的节点数量；所述隐藏层的节点数量的公式为：

Figure BDA0003579214510000086
其中，n*为所述隐藏层的节点数量，a为预设的经验值；The number of nodes of the hidden layer is determined according to the number of the influencing factors; the formula of the number of nodes of the hidden layer is:
Figure BDA0003579214510000086
Wherein, n * is the number of nodes in the hidden layer, and a is a preset empirical value;

根据所述第二处理矩阵确定所述输出层的节点数量；determining the number of nodes of the output layer according to the second processing matrix;

初始化所述混合神经网络的混合神经网络参数；所述混合神经网络参数包括所述混合神经网络的学习率、训练目标最小误差、最大迭代次数、节点权重、节点偏置、激活函数和输出函数；Initializing the hybrid neural network parameters of the hybrid neural network; the hybrid neural network parameters include the learning rate of the hybrid neural network, the minimum error of the training target, the maximum number of iterations, the node weight, the node bias, the activation function and the output function;

根据所述训练集对初始化混合神经网络参数后的所述混合神经网络进行训练，得到训练后的混合神经网络。The hybrid neural network after initializing the parameters of the hybrid neural network is trained according to the training set to obtain a trained hybrid neural network.

具体的，本实施例中根据数据集中X*初始设置混合神经网络的输入节点数量，输入节点的数量设置为集合S＝{x1，x2，...，xn，y}中x的数量，即为影响钢铁物流成本的影响因素数量；Specifically, in this embodiment, the number of input nodes of the hybrid neural network is initially set according to X * in the data set, and the number of input nodes is set to the number of x in the set S={x 1 , x 2 , . . . , x n , y} Quantity, that is, the number of factors that affect the cost of steel logistics;

进一步地，根据影响钢铁物流成本的影响因素数量设置混合神经网络隐藏层节点数量，依据隐藏层节点设置的经验公式：Further, the number of nodes in the hidden layer of the hybrid neural network is set according to the number of influencing factors affecting the cost of steel logistics, and the empirical formula for setting the nodes in the hidden layer is:

Figure BDA0003579214510000091
Figure BDA0003579214510000091

其中，n是影响钢铁物流成本的影响因素数量，1是因为输出节点仅有1个，a可以取1～10内任意值；Among them, n is the number of factors that affect the cost of steel logistics, 1 is because there is only one output node, and a can take any value from 1 to 10;

更进一步地，根据数据集中Y*初始设置混合神经网络输出节点数量，由钢铁物流成本数据集可以知仅有1个输出，输出节点数同样设置为1。Further, according to the initial setting of the number of output nodes of the hybrid neural network in the data set Y * , it can be known from the steel logistics cost data set that there is only one output, and the number of output nodes is also set to 1.

接着继续配置神经网络参数，完善神经网络：Then continue to configure the neural network parameters and improve the neural network:

先设置学习率，训练目标最小误差，最大迭代次数：First set the learning rate, the minimum error of the training target, and the maximum number of iterations:

手动的设置学习率α＝0.01，这个数值不能设置的过大或者过小，过大会导致调整幅度过大，不易得出最优预测钢铁物流成本模型；过小可能导致调整过程迭代次数过多，效率低下。设置训练目标最小误差β＝0.001。Manually set the learning rate α = 0.01, this value cannot be set too large or too small, too large will lead to the adjustment range is too large, it is difficult to obtain the optimal prediction steel logistics cost model; too small may lead to too many iterations of the adjustment process, low efficiency. Set the training target minimum error β=0.001.

再分别手动设置输入层-＞隐藏层，隐藏层-＞输出层各个节点的权重w，偏置b；Then manually set the weight w and bias b of each node of the input layer -> hidden layer, hidden layer -> output layer;

设置隐藏层，输出层的激活函数f(x)：Set up the hidden layer and the activation function f(x) of the output layer:

Figure BDA0003579214510000092
Figure BDA0003579214510000092

f(x)是激活函数，非线性函数，可以把隐藏层输出以及预测物流成本输出映射到[-1，1]之内，如图4所示。f(x) is an activation function, a nonlinear function, which can map the output of the hidden layer and the output of the predicted logistics cost to [-1, 1], as shown in Figure 4.

可以得到隐藏层，输出层的输出：You can get the output of the hidden layer and the output layer:

Figure BDA0003579214510000093
Figure BDA0003579214510000093

其中，hj是隐藏层对输出层的输入，wij，bij分别是第i个输入节点与第j个隐藏层节点之间的权重与偏置。

Figure BDA0003579214510000094
分别是第i个隐藏节点与输出节点之间的权重与偏置，可以得到预测的物流成本y。Among them, h j is the input of the hidden layer to the output layer, w ij and b ij are the weight and bias between the ith input node and the jth hidden layer node, respectively.
Figure BDA0003579214510000094
are the weight and bias between the ith hidden node and the output node, respectively, and the predicted logistics cost y can be obtained.

再进一步地，混合神经网络初始化完成之后，输入前述钢铁物流成本数据集

Figure BDA0003579214510000101
训练，在训练的同时需要不断的依据钢铁物流成本预测输出值与准确输出值的损失使用全局最优的优化方式调整各层之间的权重和偏置，优化混合神经网络，达到预测钢铁物流成本最精准的目的，既防止了过拟合化、局部最优现象，也提升了迭代效率。Further, after the initialization of the hybrid neural network is completed, the aforementioned steel logistics cost data set is input.
Figure BDA0003579214510000101
During training, it is necessary to continuously predict the loss of output value and accurate output value according to the cost of steel logistics. Use the globally optimal optimization method to adjust the weights and biases between layers, optimize the hybrid neural network, and predict the cost of steel logistics. The most accurate purpose is to prevent overfitting and local optimization, and to improve iterative efficiency.

全局最优优化的调整方式如下：The adjustment method of the global optimal optimization is as follows:

首先定义一个较大的常量K＝2000，First define a larger constant K=2000,

并定义出钢铁物流成本预测值与准确值之间的损失C：And define the loss C between the predicted value and the accurate value of the steel logistics cost:

C＝(y*-y)2。C=(y * -y) 2 .

其中y*表示钢铁物流成本的预测值，y表示与之对应的钢铁物流成本的实际值。目的是使得C尽量小。Where y * represents the predicted value of the steel logistics cost, and y represents the corresponding actual value of the steel logistics cost. The purpose is to make C as small as possible.

进一步的，当混合神经网络得到一个新的钢铁物流成本损失Ct后，对其进行随机的扰动，得到新的钢铁物流成本损失Ct+1，并以概率P选择是否更新这次扰动。Further, when the hybrid neural network obtains a new steel logistics cost loss C t , it performs random disturbance to it to obtain a new steel logistics cost loss C t+1 , and chooses whether to update this disturbance with probability P.

Figure BDA0003579214510000102
Figure BDA0003579214510000102

其中Ct，Ct+1为上述的钢铁物流成本损失，K为初始化的常量，Among them, C t , C t+1 are the above steel logistics cost losses, K is the initialization constant,

同时更新K＝K×0.99。Simultaneously update K=K×0.99.

重复上述迭代过程，直到达到最大迭代次数。接着取若干个连续的钢铁损失C，若都在上述P下不被接受，全局最优优化调整结束。The above iterative process is repeated until the maximum number of iterations is reached. Then take several consecutive steel losses C, if all of them are not accepted under the above P, the global optimal optimization adjustment ends.

混合神经网络模型调整完成后，一种基于混合神经网络的钢铁物流成本预测方法也实现完成。之后只需要输入用于测试的钢铁物流成本影响因素数据集

Figure BDA0003579214510000103
进行模型测试即可，当然最后还需要对输出层输出的的钢铁物流预测成本矩阵进行反归一化得到矩阵
Figure BDA0003579214510000104
作为模型预测成果输出与
Figure BDA0003579214510000105
比对验证模型的准确率。After the adjustment of the hybrid neural network model is completed, a method for predicting the cost of steel logistics based on the hybrid neural network is also completed. Only then need to enter the steel logistics cost influencer dataset for testing
Figure BDA0003579214510000103
The model test is enough. Of course, in the end, it is necessary to denormalize the iron and steel logistics prediction cost matrix output by the output layer to obtain the matrix.
Figure BDA0003579214510000104
As the output of the model prediction results and
Figure BDA0003579214510000105
Compare the accuracy of the validation model.

优选地，所述步骤204包括：Preferably, the step 204 includes:

确定钢铁物流成本预测值与准确值之间的损失值；所述损失值的计算公式为C＝(y*-y)2；其中C为所述损失值，y*为所述钢铁物流成本预测值；Determine the loss value between the steel logistics cost prediction value and the accurate value; the calculation formula of the loss value is C=(y * -y) 2 ; wherein C is the loss value, and y * is the steel logistics cost prediction value;

在训练过程中，当所述混合神经网络到一个新的钢铁物流成本损失Ct后，对Ct进行随机的扰动，得到新的钢铁物流成本损失Ct+1，并以概率P选择是否更新本次扰动，并将预设的优化常量乘以系数进行更新；所述概率的表达式为

Figure BDA0003579214510000111
其中，K为所述优化常量；In the training process, when the hybrid neural network reaches a new steel logistics cost loss C t , random perturbation is performed on C t to obtain a new steel logistics cost loss C t+1 , and the probability P is used to select whether to update or not This disturbance is updated by multiplying the preset optimization constant by the coefficient; the expression of the probability is
Figure BDA0003579214510000111
Wherein, K is the optimization constant;

当更新的所述优化常量达到所述最大迭代次数时，取多个连续的所述钢铁物流成本损失并判断是否均在所述概率下不被接受，若是，则得到所述训练好的预测模型。When the updated optimization constant reaches the maximum number of iterations, take a plurality of consecutive steel logistics cost losses and judge whether they are all unacceptable under the probability, and if so, obtain the trained prediction model .

优选地，所述待测的钢铁物流成本数据的表达式为T＝{x1，x2，...，xn}；其中，T为所述待测的钢铁物流成本数据的集合；xn为第n个影响因素。Preferably, the expression of the steel logistics cost data to be measured is T={x 1 , x 2 , . . . , x n }; wherein, T is the set of the steel logistics cost data to be measured; x n is the nth influencing factor.

本实施例中，今后用户若是需要预测钢铁物流成本，只需要按输入顺序将影响因素输入即可，就可以得到按需求输出的预测的钢铁物流成本y。，输入数据用以下集合T表示：In this embodiment, if the user needs to predict the steel logistics cost in the future, he only needs to input the influencing factors in the input order, and the predicted steel logistics cost y output according to the demand can be obtained. , the input data is represented by the following set T:

T＝{x1，x2，...，xn}；其中元素xi表示第i个影响钢铁物流成本的因素，一共有n个影响因素； T ={x 1 , x 2 , .

本实施例还对应提供了一种基于混合神经网络的钢铁物流成本预测系统，包括：This embodiment also provides a hybrid neural network-based steel logistics cost prediction system, including:

待测数据获取模块，用于获取待测的钢铁物流成本数据；The data acquisition module to be tested is used to acquire the steel logistics cost data to be tested;

预测模块，用于将所述待测的钢铁物流成本数据输入至训练好的预测模型中，得到预测完成的钢铁物流成本；A prediction module, used to input the steel logistics cost data to be measured into the trained prediction model to obtain the predicted completed steel logistics cost;

模型确定模块，具体包括：Model determination module, including:

训练数据获取单元，用于获取训练的钢铁物流成本数据；The training data acquisition unit is used to acquire the steel logistics cost data for training;

预处理单元，用于对所述训练的钢铁物流成本数据进行预处理，得到训练集和测试集；a preprocessing unit for preprocessing the training steel logistics cost data to obtain a training set and a test set;

初始化单元，用于对预设的混合神经网络进行初始化，并根据所述训练集对初始化后的混合神经网络进行训练，得到训练后的混合神经网络；an initialization unit, configured to initialize a preset hybrid neural network, and train the initialized hybrid neural network according to the training set to obtain a trained hybrid neural network;

优化单元，用于基于全局最优的优化方法对所述训练后的混合神经网络进行参数迭代，得到所述训练好的预测模型。An optimization unit, configured to perform parameter iteration on the trained hybrid neural network based on a globally optimal optimization method to obtain the trained prediction model.

本发明的有益效果如下：The beneficial effects of the present invention are as follows:

本发明针对传统物流成本预测方法中存在的随机性进行了改进，同时优化了传统神经网络中存在的局部最优及迭代次数过多的问题，有很高的预测精确度以及较高的可实施性。The invention improves the randomness existing in the traditional logistics cost prediction method, and at the same time optimizes the problems of local optimum and excessive iteration times in the traditional neural network, and has high prediction accuracy and high implementability. sex.

本说明书中各个实施例采用递进的方式描述，每个实施例重点说明的都是与其他实施例的不同之处，各个实施例之间相同相似部分互相参见即可。对于实施例公开的系统而言，由于其与实施例公开的方法相对应，所以描述的比较简单，相关之处参见方法部分说明即可。The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments, and the same and similar parts between the various embodiments can be referred to each other. For the system disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant part can be referred to the description of the method.

本文中应用了具体个例对本发明的原理及实施方式进行了阐述，以上实施例的说明只是用于帮助理解本发明的方法及其核心思想；同时，对于本领域的一般技术人员，依据本发明的思想，在具体实施方式及应用范围上均会有改变之处。综上所述，本说明书内容不应理解为对本发明的限制。The principles and implementations of the present invention are described herein using specific examples, and the descriptions of the above embodiments are only used to help understand the method and the core idea of the present invention; There will be changes in the specific implementation and application scope. In conclusion, the contents of this specification should not be construed as limiting the present invention.

Claims (10)

1. A method for predicting the logistics cost of steel based on a hybrid neural network is characterized by comprising the following steps:
acquiring steel logistics cost data to be detected;
inputting the steel logistics cost data to be measured into a trained prediction model to obtain the predicted steel logistics cost;
the determination method of the prediction model comprises the following steps:
acquiring trained steel logistics cost data;
preprocessing the trained steel logistics cost data to obtain a training set and a testing set;
initializing a preset hybrid neural network, and training the initialized hybrid neural network according to the training set to obtain a trained hybrid neural network;
and performing parameter iteration on the trained hybrid neural network based on a global optimal optimization method to obtain the trained prediction model.
2. The method for predicting the logistics cost of steel based on the hybrid neural network as claimed in claim 1, further comprising:
and performing model test on the trained prediction model according to the test set.
3. The hybrid neural network-based steel logistics cost prediction method of claim 1, wherein the expression of the trained steel logistics cost data is S ═ { x ═ x1，x2，…，xnY, wherein S is a historical data set of the trained steel logistics cost data, xnAnd y is the cost value of the steel logistics as the nth influencing factor.
4. The hybrid neural network-based steel logistics cost prediction method of claim 1, wherein the influence factors include a vehicle departure starting location, a vehicle transportation destination, a vehicle type, a transportation day weather, road conditions, a cargo type and a cargo weight.
5. The hybrid neural network-based steel logistics cost prediction method of claim 3, wherein the preprocessing the trained steel logistics cost data to obtain a training set and a test set comprises:
performing matrixing processing on the historical data set to obtain a first data matrix and a second data matrix; the expression of the first data matrix is X ═ Xij]m×nThe expression of the second data matrix is Y ═ Yi]m×1(ii) a Wherein X is the first data matrix, XijThe jth influence factor in the ith historical data set is obtained; y isiThe logistics cost value in the ith historical data set is used as the logistics cost value; n is the number of the influencing factors; m is the total number of the steel logistics cost data;
respectively carrying out normalization processing on the first data matrix and the second data matrix to obtain first normalized data and second normalized data; the expression of the first normalized data is
Figure FDA0003579214500000021
Figure FDA0003579214500000022
The expression of the second normalized data is
Figure FDA0003579214500000023
Wherein x ismaxIs the maximum value of said influencing factor, xminIs the minimum value of the influencing factor; y ismaxIs the maximum value of said logistic cost value; y isminIs the minimum of said logistic cost values;
performing matrixing expression on the first normalized data and the second normalized data respectively to obtain a first processing matrix and a second processing matrix; the expression of the first processing matrix is
Figure FDA0003579214500000024
The expression of the second processing matrix is
Figure FDA0003579214500000025
Wherein, X*For the purpose of the first processing matrix,
Figure FDA0003579214500000026
normalized value of j influence factor in ith historical data set, Y*For the purpose of said second processing matrix,
Figure FDA0003579214500000027
normalized values for logistics cost values in the ith said historical data set;
determining the training set and the test set according to the first processing matrix and the second processing matrix.
6. The method for predicting the logistics cost of steel and iron based on the hybrid neural network according to claim 5, wherein the initializing a preset hybrid neural network and training the initialized hybrid neural network according to the training set to obtain the trained hybrid neural network comprises:
acquiring the preset hybrid neural network; the hybrid neural network comprises an input layer, a hidden layer and an output layer which are sequentially connected; the number of input nodes of the input layer is equal to the number of influencing factors;
determining the number of nodes of the hidden layer according to the number of the influence factors; the formula of the number of nodes of the hidden layer is as follows:
Figure FDA0003579214500000028
wherein n is*A is a preset empirical value, and is the number of nodes of the hidden layer;
determining the number of nodes of the output layer according to the second processing matrix;
initializing hybrid neural network parameters of the hybrid neural network; the parameters of the hybrid neural network comprise the learning rate, the minimum error of a training target, the maximum iteration times, the node weight, the node bias, an activation function and an output function of the hybrid neural network;
and training the hybrid neural network after initializing the parameters of the hybrid neural network according to the training set to obtain the trained hybrid neural network.
7. The method for predicting the steel logistics cost based on the hybrid neural network as claimed in claim 6, wherein the method for optimizing based on the global optimum performs parameter iteration on the trained hybrid neural network to obtain the trained prediction model, and comprises the following steps:
determining a loss value between a predicted value and an accurate value of the steel logistics cost; the loss value is calculated by the formula of (y ═ C)*-y)2(ii) a Wherein C is the loss value, y*Predicting the cost of the steel logistics;
during the training process, when the mixed neural network reaches a new steel logistics cost loss CtThen, to CtCarrying out random disturbance to obtain new steel logistics cost loss Ct+1Selecting whether to update the current disturbance according to the probability P, and multiplying a preset optimization constant by a coefficient for updating; the expression of the probability is
Figure FDA0003579214500000031
Wherein K is the optimization constant;
and when the updated optimization constant reaches the maximum iteration number, taking a plurality of continuous steel logistics cost losses and judging whether the steel logistics cost losses are all not accepted under the probability, and if so, obtaining the trained prediction model.
8. The method according to claim 1, wherein the expression of the to-be-measured steel logistics cost data is T ═ x1，x2，…，xn}; wherein T is the set of the steel logistics cost data to be detected; x is the number ofnIs the nth influencing factor.
9. A steel logistics cost prediction system based on a hybrid neural network is characterized by comprising the following components:
the to-be-detected data acquisition module is used for acquiring to-be-detected steel logistics cost data;
the prediction module is used for inputting the steel logistics cost data to be measured into a trained prediction model to obtain the predicted steel logistics cost;
the model determining module specifically comprises:
the training data acquisition unit is used for acquiring the trained steel logistics cost data;
the preprocessing unit is used for preprocessing the trained steel logistics cost data to obtain a training set and a testing set;
the initialization unit is used for initializing a preset hybrid neural network and training the initialized hybrid neural network according to the training set to obtain a trained hybrid neural network;
and the optimization unit is used for carrying out parameter iteration on the trained hybrid neural network based on a global optimal optimization method to obtain the trained prediction model.
10. The hybrid neural network-based steel logistics cost prediction system of claim 9, further comprising:
and the testing module is used for carrying out model testing on the trained prediction model according to the testing set.
CN202210349735.XA 2022-04-02 2022-04-02 A steel logistics cost prediction method and system based on hybrid neural network Active CN114626635B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202210349735.XA CN114626635B (en) 2022-04-02 2022-04-02 A steel logistics cost prediction method and system based on hybrid neural network

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202210349735.XA CN114626635B (en) 2022-04-02 2022-04-02 A steel logistics cost prediction method and system based on hybrid neural network

Publications (2)

Publication Number Publication Date
CN114626635A true CN114626635A (en) 2022-06-14
CN114626635B CN114626635B (en) 2024-08-06

Family

ID=81905109

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202210349735.XA Active CN114626635B (en) 2022-04-02 2022-04-02 A steel logistics cost prediction method and system based on hybrid neural network

Country Status (1)

Country Link
CN (1) CN114626635B (en)

Citations (15)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6269351B1 (en) * 1999-03-31 2001-07-31 Dryken Technologies, Inc. Method and system for training an artificial neural network
US20040111169A1 (en) * 2002-12-04 2004-06-10 Hong Se June Method for ensemble predictive modeling by multiplicative adjustment of class probability: APM (adjusted probability model)
CN107169599A (en) * 2017-05-12 2017-09-15 东北大学 A kind of Multiobjective Optimal Operation method based on iron and steel enterprise's energy resource system
CN107844860A (en) * 2017-11-01 2018-03-27 山东浪潮云服务信息科技有限公司 A kind of optimal method and device
CN108449295A (en) * 2018-02-05 2018-08-24 西安电子科技大学昆山创新研究院 Joint Modulation Recognition Method Based on RBM Network and BP Neural Network
CN109671272A (en) * 2018-12-29 2019-04-23 广东工业大学 A kind of freeway traffic flow prediction technique
CN109859469A (en) * 2019-02-15 2019-06-07 重庆邮电大学 A kind of vehicle flowrate prediction technique based on integrated LSTM neural network
CN109858127A (en) * 2018-04-13 2019-06-07 北京工商大学 Cyanobacterial bloom prediction technique based on recurrence sequential depth confidence network
CN110097185A (en) * 2019-03-29 2019-08-06 北京大学 A kind of Optimized model method and application based on generation confrontation network
CN110601777A (en) * 2019-08-29 2019-12-20 浙江大学 Method for estimating satellite-ground downlink co-channel interference under low-orbit mobile satellite constellation
WO2020177863A1 (en) * 2019-03-06 2020-09-10 Nokia Technologies Oy Training of algorithms
CN111695731A (en) * 2020-06-09 2020-09-22 中国电力科学研究院有限公司 Load prediction method, system and equipment based on multi-source data and hybrid neural network
CN112529328A (en) * 2020-12-23 2021-03-19 长春理工大学 Product performance prediction method and system
CN113011567A (en) * 2021-03-31 2021-06-22 深圳精智达技术股份有限公司 Training method and device of convolutional neural network model
US20220004839A1 (en) * 2020-07-02 2022-01-06 Ahp-Tech Inc. Artificial-intelligence decision-making core system with neural network

Patent Citations (15)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6269351B1 (en) * 1999-03-31 2001-07-31 Dryken Technologies, Inc. Method and system for training an artificial neural network
US20040111169A1 (en) * 2002-12-04 2004-06-10 Hong Se June Method for ensemble predictive modeling by multiplicative adjustment of class probability: APM (adjusted probability model)
CN107169599A (en) * 2017-05-12 2017-09-15 东北大学 A kind of Multiobjective Optimal Operation method based on iron and steel enterprise's energy resource system
CN107844860A (en) * 2017-11-01 2018-03-27 山东浪潮云服务信息科技有限公司 A kind of optimal method and device
CN108449295A (en) * 2018-02-05 2018-08-24 西安电子科技大学昆山创新研究院 Joint Modulation Recognition Method Based on RBM Network and BP Neural Network
CN109858127A (en) * 2018-04-13 2019-06-07 北京工商大学 Cyanobacterial bloom prediction technique based on recurrence sequential depth confidence network
CN109671272A (en) * 2018-12-29 2019-04-23 广东工业大学 A kind of freeway traffic flow prediction technique
CN109859469A (en) * 2019-02-15 2019-06-07 重庆邮电大学 A kind of vehicle flowrate prediction technique based on integrated LSTM neural network
WO2020177863A1 (en) * 2019-03-06 2020-09-10 Nokia Technologies Oy Training of algorithms
CN110097185A (en) * 2019-03-29 2019-08-06 北京大学 A kind of Optimized model method and application based on generation confrontation network
CN110601777A (en) * 2019-08-29 2019-12-20 浙江大学 Method for estimating satellite-ground downlink co-channel interference under low-orbit mobile satellite constellation
CN111695731A (en) * 2020-06-09 2020-09-22 中国电力科学研究院有限公司 Load prediction method, system and equipment based on multi-source data and hybrid neural network
US20220004839A1 (en) * 2020-07-02 2022-01-06 Ahp-Tech Inc. Artificial-intelligence decision-making core system with neural network
CN112529328A (en) * 2020-12-23 2021-03-19 长春理工大学 Product performance prediction method and system
CN113011567A (en) * 2021-03-31 2021-06-22 深圳精智达技术股份有限公司 Training method and device of convolutional neural network model

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
YYYA77: "模拟退火算法（SA）原理与python实例", HTTPS://BLOG.CSDN.NET/A999999123/ARTICLE/DETAILS/109500746, 5 November 2020 (2020-11-05), pages 1 - 2 *
大写的FENG: "数据预处理：归一化和标准化", HTTPS://ZHUANLAN.ZHIHU.COM/P/296252799, 16 November 2020 (2020-11-16), pages 1 - 2 *
马威, 张国平, 王正欧: "一种基于参数扰动退火策略的神经网络全局优化新算法", 系统工程理论方法应用, no. 04, 25 December 2002 (2002-12-25), pages 314 - 318 *

Also Published As

Publication number Publication date
CN114626635B (en) 2024-08-06

Similar Documents

Publication Publication Date Title
Li et al. Prediction for tourism flow based on LSTM neural network
CN111079989B (en) DWT-PCA-LSTM-based water supply amount prediction device for water supply company
CN110210993B (en) Urban short-term gas load prediction method based on cyclic neural network model
CN111860989B (en) LSTM neural network short-time traffic flow prediction method based on ant colony optimization
CN108830418A (en) A kind of Short-Term Load Forecasting Method
CN111582559B (en) Arrival time estimation method and device
CN106815782A (en) A kind of real estate estimation method and system based on neutral net statistical models
CN112036598B (en) A charging pile usage information prediction method based on multi-information coupling
CN109242265B (en) Urban water demand combined prediction method based on least square sum of errors
CN110543978A (en) Traffic flow data prediction method and device based on wavelet neural network
CN113657028B (en) Online aerosol optical thickness prediction method based on multi-source information
CN117648631A (en) Power battery health state estimation method for electric automobile group
CN108090563A (en) A kind of electric flux Forecasting Methodology based on BP neural network
CN110889530A (en) Destination prediction method based on recurrent neural network and server
CN114595873A (en) Grey-correlation-based DA-LSTM short-term power load prediction method
CN114386142B (en) Building energy consumption prediction method based on multi-source fusion feature selection and fuzzy difference enhanced Stacking framework
CN115115119A (en) OA-GRU short-term power load prediction method based on grey correlation
CN115542427A (en) A method and system for predicting rainfall based on an improved neural network model
CN116303786B (en) Block chain financial big data management system based on multidimensional data fusion algorithm
CN114548475A (en) A classification evaluation method and visualization system of carbon emission intensity based on big data prediction
CN109787821A (en) A kind of Large-scale Mobile customer traffic consumption intelligent Forecasting
CN119167038A (en) A method, device, medium and equipment for predicting load of air conditioning system
CN109214610A (en) A kind of saturation Methods of electric load forecasting based on shot and long term Memory Neural Networks
CN114049162A (en) Model training method, demand prediction method, apparatus, device, and storage medium
CN119250848A (en) A real estate valuation method based on Bayesian optimization and machine learning

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant