Gray attention network logistics workload prediction method and device based on firefly algorithm optimization

Abstract

The invention discloses a gray attention network logistics workload prediction method and device based on firefly algorithm optimization. The method specifically comprises the following steps: an index system determination step, a prediction index analysis step, a gray attention network step and a parameter optimization step. According to the method, prediction can be carried out for a specific operation scene, a gray algorithm and an improved attention network fusion model are utilized, parameter optimization is carried out on the network through a firefly algorithm, and the optimal weight and threshold value of the network are obtained. According to the method, the workload prediction precision and universality are improved, and meanwhile, the robustness and training ability of the model are enhanced. The method can be suitable for various types of logistics operation scenes, achieves the efficient and accurate prediction of the logistics operation amount, remarkably improves the prediction stability and accuracy, and has a considerable application value in the prediction of the logistics operation amount.

Description

Technical Field

[0001] The present invention relates to the technical field of workload prediction in logistics systems, and specifically to a gray attention network logistics workload prediction method and device based on firefly algorithm optimization, which is suitable for workload prediction in various logistics scenarios. Background Art

[0002] Logistics plays an important role in all aspects of modern social production and is an important support for the flow of economic factors. All countries also regard the development of the logistics industry as an important part of economic development. As an important part of this, logistics workload forecasting can help governments and enterprises plan logistics infrastructure, promote the coordinated operation of the entire supply chain, and identify potential operational risks in advance. It is of great significance to improve logistics efficiency, reduce costs, enhance corporate competitiveness, and promote regional economic development.

[0003] The existing forecasting methods for logistics workload can be divided into time series analysis, causal forecasting models and machine learning algorithms. Time series analysis predicts future logistics workload by studying trends, seasonality and cyclical changes in historical data. Common methods include moving average, exponential smoothing and autoregressive integrated moving average (ARIMA) model. Causal forecasting models are based on the relationship between logistics workload and other influencing factors (such as market demand, price changes, economic indicators, etc.), and establish regression models or other statistical models for forecasting. Machine learning algorithms can automatically learn complex patterns in historical data and predict logistics workload. Mainstream methods include support vector machines, gray forecasting and neural networks.

[0004] The above methods have been widely used in a variety of scenarios to predict logistics workload, but there are still many problems. Time series analysis has a weak ability to respond to sudden and random events (natural disasters, market mutations); long-term prediction accuracy is not high, especially when the external environment changes greatly; it requires a large amount of historical data support and is not suitable for situations where the data is incomplete or volatile. The causal prediction model is relatively complex and requires strong statistical analysis capabilities; it is more dependent on the correct selection of influencing factors, and incorrect variable selection will lead to prediction bias; it is difficult to handle nonlinear or highly uncertain relationships. Traditional machine learning algorithms have high requirements for data quality, and noise and outliers may affect the prediction effect.

[0005] These problems limit the use scenarios and conditions of logistics workload prediction, resulting in poor prediction results, high data requirements, and low universality in existing workload prediction methods in more complex scenarios. In order to overcome the above problems, logistics workload prediction based on fusion models has gradually become a research hotspot in recent years. By combining multiple prediction methods, the advantages of each model are obtained to form a more robust prediction result. Based on the gray model and coupled with deep learning technology, the logistics workload prediction model can achieve high accuracy and robustness, and reduce dependence on scenarios. These methods not only improve the accuracy of prediction, but also significantly reduce prediction costs and human errors, providing strong support for the efficient operation of logistics management. Summary of the invention

[0006] The purpose of the present invention is to address the deficiencies in the prior art and propose a method and device for predicting logistics workload based on a grey attention network optimized with a firefly algorithm. The grey model and neural network theory are integrated to construct a grey attention network model for predicting logistics workload. The parameters are optimized through the firefly algorithm to achieve accurate prediction of logistics workload in various situations. This provides a useful reference for governments and enterprises to formulate the selection and construction scale of logistics parks, which is conducive to the efficient operation of logistics management and the maximization of resource utilization.

[0007] To achieve the above objectives, the present invention adopts the following technical solutions: In the first aspect, the present invention provides a method for predicting logistics workload in a grey attention network based on firefly algorithm optimization, the method comprising the following steps: (1) Steps for determining the indicator system: Based on the specific actual logistics scenarios and the characteristics of industrial associations, analyze the factors that affect the workload, establish an indicator system that affects the logistics workload, and pre-process the indicator data; (2) Prediction index analysis steps: Based on the historical workload of logistics scenarios, use grey relational analysis (GRA) to obtain the correlation indicators that affect the logistics workload; (3) Grey attention network step: Integrate the attention mechanism between the hidden layer and the output layer of the grey system model to construct a grey attention network for predicting logistics workload; (4) Parameter optimization step: Use the firefly algorithm to find the optimal weights and thresholds of the attention network, and use them as the initial parameters of the grey attention network model for training. The initial weights of the network are continuously updated through an iterative process, and the associated indicators are used as input to predict the logistics workload based on the optimized grey attention network.

[0008] Furthermore, the establishment of the logistics operation volume impact indicator system is specifically to conduct qualitative description and quantitative analysis of industry correlation characteristics, including obtaining specific correlation indicators from four dimensions: economy, society, nature, and infrastructure.

[0009] Furthermore, the preprocessing of the indicator data is to clean the original data through outlier detection and missing value processing methods.

[0010] Furthermore, the grey relational analysis method is as follows: The historical workload of the logistics scenario is taken as the parent sequence, and the collected indicators affecting the logistics workload are taken as the sub-sequence. All potential influencing indicators are mined, and the correlation coefficient between the parent sequence and the sub-sequence is calculated. The correlation between the parent sequence and the sub-sequence is obtained by taking the average value. Based on the set threshold, several indicators with the highest correlation coefficient with the logistics workload are finally output as input variables of the subsequent prediction model.

[0011] Furthermore, the attention mechanism is integrated between the hidden layer and the output layer of the grey system model to construct a grey attention network for predicting logistics workload, and the specific contents are as follows: Based on the grey system model GM, the future workload is predicted, and the prediction results are used as the input of the attention network AT for further training and learning. The firefly algorithm is used to optimize the input prediction data by gradient descent. Through iterative training and back propagation, the parameters of the attention network model are modified to obtain more accurate prediction results. The attention mechanism is integrated between the hidden layer and the output layer of the original neural network, and the input features are weighted, and finally the predicted amount of logistics operations in the corresponding scenario is output. By constructing the GM-AT combined model, the data is transmitted through GM to obtain preliminary prediction results, which are then passed into the attention network AT for nonlinear calculation, thus achieving accurate prediction of logistics workload.

[0012] Furthermore, the grey system model GM is a dynamic model that can describe multivariable, first-order and linear relationships. The input variables of the model are first placed in GM, and the original data related to the logistics workload is grey modeled to explore the regularity between the associated indicators and the logistics workload, so as to realize the prediction of future workload.

[0013] Furthermore, the firefly algorithm optimizes model parameters, and the specific contents are as follows: The firefly algorithm is used to optimize the weights and thresholds of the attention network AT. The initial weights of the network are continuously updated through an iterative process to find the global optimal solution.

[0014] In the second aspect, the present invention also provides a gray attention network logistics workload prediction device based on firefly algorithm optimization, including a memory and one or more processors, wherein the memory stores executable code, and when the processor executes the executable code, it implements the gray attention network logistics workload prediction method based on firefly algorithm optimization.

[0015] In a third aspect, the present invention also provides a computer-readable storage medium having a program stored thereon, which, when executed by a processor, implements the gray attention network logistics workload prediction method based on firefly algorithm optimization.

[0016] In a fourth aspect, the present invention also provides a computer program product, including a computer program, which, when executed by a processor, implements the gray attention network logistics workload prediction method based on firefly algorithm optimization.

[0017] Beneficial effects of the present invention: The present invention can carry out predictions for specific logistics workload scenarios, has very strong universality, and effectively improves the prediction accuracy of logistics workload. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0019] Figure 1 The present invention provides a gray attention network logistics workload prediction flow chart based on firefly algorithm optimization.

[0020] Figure 2 It is the GM-AT prediction flow chart of the present invention.

[0021] Figure 3 It is the GM-AT topological structure diagram of the present invention.

[0022] Figure 4 It is the attention mechanism diagram of the present invention.

[0023] Figure 5 It is the flow chart of the firefly algorithm optimization GM-AT of the present invention.

[0024] Figure 6 It is a structural diagram of a gray attention network logistics workload prediction device based on firefly algorithm optimization provided by the present invention. DETAILED DESCRIPTION

[0025] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described below in conjunction with the accompanying drawings and implementation examples. It should be understood that the specific implementation examples described herein are only used to explain the present invention and are not used to limit the present invention.

[0026] Logistics workload forecasting is a key link in the entire process of logistics management. This invention improves the accuracy, reliability and universality of forecasting, so as to cope with the fast-changing market, predict development in advance, formulate corresponding logistics plans, and promote the transformation of modern logistics of enterprises. This invention adopts the traditional gray model and deep learning fusion technology to more accurately predict the future logistics workload, avoiding the problems of low accuracy, poor reliability and strong dependence on scenarios in traditional methods.

[0027] This embodiment provides a method for predicting the logistics workload of a gray attention network based on the optimization of the firefly algorithm. Figure 1 As shown, the specific steps include: 1) Steps to determine the indicator system: When forecasting logistics workload, collecting influencing indicators is a complex and critical step. Focusing on specific logistics operation scenarios, considering relevant industry-related characteristics, analyze the factors that affect the workload. Staff can use surveys, historical data queries and other means to cover economic or non-economic factors, determine potential influencing indicators, and build an influencing factor indicator system for the forecasting model. In the construction process, it is necessary to start from a multi-factor perspective, adhere to the principles of science, practicality and independence, and ensure the rationality and effectiveness of indicator selection. After the indicator system is initially determined, the collected data is processed to form an effective data set that can be directly used for subsequent work.

[0028] 2) Prediction indicator analysis steps: While considering the potential influencing indicators of logistics operation volume, it is more necessary to mine from multi-dimensional indicators and screen out highly correlated indicators as model input variables. This embodiment uses grey correlation analysis to complete this task. The specific steps are as follows: Construct the initial decision matrix (X), assuming that there are n data sequences and m candidate indicators, and the jth value under the i-th candidate indicator is represented by x ij In this embodiment, the candidate indicators are summarized influencing indicators, such as the development and distribution of logistics enterprises, the scale of related industries, the demand and consumption capacity of surrounding areas, etc. The data sequence is the operation location. The initial matrix can be obtained :

[0029] Reduce the difference in absolute values ​​between different candidate indicator data, and adjust the initial matrix Perform normalization:

[0030] Generate a normalized matrix:

[0031] Reference sequence:

[0032] in is the reference value relative to the jth data, obtained by processing the reference column (logistics operation volume in this case). Then calculate the difference between each sample value and the reference value, and construct the difference matrix:

[0033]

[0034] Finally, calculate the grey correlation coefficient:

[0035] In the formula is the discrimination coefficient, The smaller the value, the higher the resolution. Taking 0.5 has a moderate recognition effect and good stability.

[0036] 3) Grey attention network steps: After obtaining the input variables, use Figure 2 The grey attention network model shown in the figure is used to predict the workload. This model combines the grey system model and the attention network, so that the network model can efficiently predict the logistics workload in specific scenarios. The grey prediction model and the attention network prediction model are mature and widely used in their respective fields. In order to expand the scope of application, especially for complex scenario prediction problems such as logistics workload, it is difficult to obtain a large number of accurate information samples. By combining the two models and giving full play to their respective advantages, the prediction of small data uncertain systems in real scenarios can be realized. The specific steps are as follows: Grey model processing: It is a dynamic model that can describe multivariate, first-order and linear relationships. Assume that there are variables , And the sequence corresponding to each variable:

[0037] For each , the model describes the relationship between multiple variables by cumulatively generating sequences and differential equations. The cumulatively generated sequence is:

[0038] The differential equation is:

[0039] Here the parameters It can be obtained by the least squares method:

[0040] when and hour, The response formula is:

[0041] use Indicates a cumulative result , Represents the prediction results , the final prediction value obtained by the grey model for:

[0042] In the formula, is the development coefficient, is the gray action. Mapping it to the extended attention network (AT), we get the GM-AT model, and its prediction process is shown in Figure 2 First, a suitable grey attention network is constructed, and the model is trained and optimized to obtain a mature model. The results obtained by the grey algorithm are fed into the network through the input layer, and the gradient descent solution is performed between layers to obtain more accurate prediction results.

[0043] The attention network AT is a further optimization of the neural network. By introducing the attention mechanism, its structure is as follows Figure 3 The attention mechanism helps the model to adaptively focus on important features in the input sequence. Its topological structure is shown in Figure 4 First, initialize the network parameters and preset them according to the sample data values. The initial value of , calculate the network weight And output threshold, learning rate and expected error of the input model.

[0044]

[0045]

[0046]

[0047] in:

[0048] L4 layer:

[0049] For each input sequence of the prediction model Perform calculations, For the setting of network weights, the output of each layer of the network structure is: L1 layer:

[0050] L2 layer:

[0051] Before the L3 layer, an attention layer is introduced to enhance the influence of important information by calculating and distributing the attention coefficients. This is done through a learnable weight matrix and a softmax function, which ensures that the sum of all attention weights is 1. For each input , calculate its corresponding attention weight .

[0052]

[0053] in, is through a learnable weight matrix and bias The calculation results are:

[0054] in, is the weight vector The transpose of represents the hyperbolic tangent function, and the calculated attention weight is applied to the output of the L3 layer, using Represents the attention weight:

[0055]

[0056]

[0057] in, is the attention weight, the output of the L4 layer for:

[0058] In the back propagation, the influence of attention weight on the error is considered, and the error calculation of L3 and L2 layers is updated to ensure that the error can be back propagated through the attention weight. L4 layer error:

[0059] L3 layer error:

[0060] L2 layer error:

[0061] According to the weight and threshold of the error correction network, adjust the weights from L2 to L3 and from L1 to L2:

[0062]

[0063] In the formula For the learning rate, finally, adjust the threshold:

[0064] Determine whether the error requirement is met. If not, return to the step and iterate the model training.

[0065] 6) Parameter optimization steps: Define a method to use the firefly algorithm to optimize the weights and thresholds of the attention network model, improve the robustness of the prediction model network, and solve the problem that its performance depends largely on the configuration of the initial weights, and it is easy to fall into the local optimum and difficult to reach the global optimal solution. Figure 5 As shown, the network parameters of GM-AT are optimized.

[0066] First, initialize the parameters of the firefly algorithm, including the population size , maximum number of iterations , Luminous intensity (brightness) , attraction coefficient , light intensity attenuation coefficient and step size parameter , randomly deploy fireflies in the solution space.

[0067] Calculate the relative brightness of each firefly and attractiveness ,according to Determine the direction of movement and update the position of the firefly.

[0068]

[0069]

[0070] in, It means that the brightness of the firefly in the distance meter is 0. is the Euclidean distance between fireflies. is the attraction of the firefly when the distance is 0, For the least attractive. Firefly and fireflies The attractive force between them is calculated as follows:

[0071] in, and Firefly and fireflies location. is the step size parameter, is a random number in the range of [0,1], used to introduce randomness. After the update, the brightness of the fireflies is recalculated. If the maximum number of iterations is reached or the convergence condition is met, the algorithm stops, otherwise it continues to iterate. The final optimal solution is the position of the brightest firefly among all the fireflies, which means that the optimal solution of the objective function value is achieved.

[0072] Corresponding to the aforementioned embodiment of a method for predicting logistics workload in a grey attention network based on optimization of the firefly algorithm, the present invention also provides an embodiment of a device for predicting logistics workload in a grey attention network based on optimization of the firefly algorithm.

[0073] See also Figure 6 An embodiment of the present invention provides a gray attention network logistics workload prediction device based on firefly algorithm optimization, including a memory and one or more processors, wherein the memory stores executable code, and when the processor executes the executable code, it is used to implement a gray attention network logistics workload prediction method based on firefly algorithm optimization in the above embodiment.

[0074] The embodiment of the gray attention network logistics workload prediction device based on firefly algorithm optimization provided by the present invention can be applied to any device with data processing capability, and the any device with data processing capability can be a device or apparatus such as a computer. The device embodiment can be implemented through software, or through hardware or a combination of software and hardware. Taking software implementation as an example, as a device in a logical sense, it is formed by the processor of any device with data processing capability in which it is located reading the corresponding computer program instructions in the non-volatile memory into the internal memory for execution. From the hardware level, if Figure 6 As shown, it is a hardware structure diagram of a gray attention network logistics workload prediction device based on firefly algorithm optimization provided by the present invention, in which any device with data processing capability is located, except Figure 6 In addition to the processor, memory, network interface, and non-volatile memory shown, any device with data processing capabilities in which the apparatus in the embodiments is located may also include other hardware, generally based on the actual functions of the device with data processing capabilities, which will not be described in detail.

[0075] The implementation process of the functions and effects of each unit in the above-mentioned device is specifically described in the implementation process of the corresponding steps in the above-mentioned method, and will not be repeated here.

[0076] For the device embodiment, since it basically corresponds to the method embodiment, the relevant parts can refer to the partial description of the method embodiment. The device embodiment described above is only schematic, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of the present invention. Ordinary technicians in this field can understand and implement it without paying creative work.

[0077] An embodiment of the present invention also provides a computer-readable storage medium having a program stored thereon. When the program is executed by a processor, a gray attention network logistics workload prediction method based on firefly algorithm optimization in the above embodiment is implemented.

[0078] The computer-readable storage medium may be an internal storage unit of any device with data processing capability described in any of the aforementioned embodiments, such as a hard disk or a memory. The computer-readable storage medium may also be an external storage device of any device with data processing capability, such as a plug-in hard disk, a smart media card (SMC), an SD card, a flash card, etc. equipped on the device. Furthermore, the computer-readable storage medium may also include both an internal storage unit and an external storage device of any device with data processing capability. The computer-readable storage medium is used to store the computer program and other programs and data required by any device with data processing capability, and may also be used to temporarily store data that has been output or is to be output.

[0079] The present invention also provides a computer program product, including a computer program, which, when executed by a processor, implements the gray attention network logistics workload prediction method based on firefly algorithm optimization.

[0080] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Ordinary technicians in the relevant field can still modify or make equivalent substitutions to the specific implementation methods of the present invention with reference to the above embodiments. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention are within the scope of protection of the claims of the present invention to be approved.

Claims

1. A grey attention network logistics workload prediction method based on firefly algorithm optimization, characterized in that: The method comprises the following steps: (1) Steps for determining the indicator system: Based on the specific actual logistics scenarios and the characteristics of industrial associations, analyze the factors that affect the workload, establish an indicator system that affects the logistics workload, and pre-process the indicator data; (2) Prediction index analysis steps: Based on the historical workload of logistics scenarios, use grey relational analysis (GRA) to obtain the correlation indicators that affect the logistics workload; (3) Grey attention network step: Integrate the attention mechanism between the hidden layer and the output layer of the grey system model to construct a grey attention network for predicting logistics workload; (4) Parameter optimization step: Use the firefly algorithm to find the optimal weights and thresholds of the attention network, and use them as the initial parameters of the grey attention network model for training. The initial weights of the network are continuously updated through an iterative process, and the associated indicators are used as input to predict the logistics workload based on the optimized grey attention network.

2. According to claim 1, a method for predicting logistics workload in a grey attention network based on firefly algorithm optimization is characterized in that: The establishment of the logistics workload impact indicator system specifically involves qualitative description and quantitative analysis of industry correlation characteristics, including obtaining specific correlation indicators from four dimensions: economy, society, nature, and infrastructure.

3. According to claim 1, a method for predicting logistics workload in a grey attention network based on firefly algorithm optimization is characterized in that: The preprocessing of the indicator data is to clean the original data through outlier detection and missing value processing methods.

4. According to claim 1, a method for predicting logistics workload in a grey attention network based on firefly algorithm optimization is characterized in that: The grey relational analysis method is as follows: The historical workload of the logistics scenario is taken as the parent sequence, and the collected indicators affecting the logistics workload are taken as the sub-sequence. All potential influencing indicators are mined, and the correlation coefficient between the parent sequence and the sub-sequence is calculated. The correlation between the parent sequence and the sub-sequence is obtained by taking the average value. Based on the set threshold, several indicators with the highest correlation coefficient with the logistics workload are finally output as input variables of the subsequent prediction model.

5. According to claim 1, a method for predicting logistics workload in a grey attention network based on firefly algorithm optimization is characterized in that: The gray attention network for predicting logistics workload is constructed by integrating the attention mechanism between the hidden layer and the output layer of the gray system model. The specific contents are as follows: Based on the grey system model GM, the future workload is predicted, and the prediction results are used as the input of the attention network AT for further training and learning. The firefly algorithm is used to optimize the input prediction data by gradient descent. Through iterative training and back propagation, the parameters of the attention network model are modified to obtain more accurate prediction results. The attention mechanism is integrated between the hidden layer and the output layer of the original neural network, and the input features are weighted, and finally the predicted amount of logistics operations in the corresponding scenario is output. By constructing the GM-AT combined model, the data is transmitted through GM to obtain preliminary prediction results, which are then passed into the attention network AT for nonlinear calculation, thus achieving accurate prediction of logistics workload.

6. According to claim 5, a method for predicting logistics workload in a grey attention network based on firefly algorithm optimization is characterized in that: The grey system model GM is a dynamic model that describes multivariable, first-order and linear relationships. The input variables of the model are first placed in GM, and the original data related to the logistics workload is grey modeled to explore the regularity between the associated indicators and the logistics workload, so as to realize the prediction of future workload.

7. According to claim 1, a method for predicting logistics workload in a grey attention network based on firefly algorithm optimization is characterized in that: The firefly algorithm optimizes the model parameters, and the specific contents are as follows: The firefly algorithm is used to optimize the weights and thresholds of the attention network AT. The initial weights of the network are continuously updated through an iterative process to find the global optimal solution.

8. A gray attention network logistics workload prediction device based on firefly algorithm optimization, comprising a memory and one or more processors, wherein the memory stores executable code, characterized in that: When the processor executes the executable code, a gray attention network logistics workload prediction method based on firefly algorithm optimization as described in any one of claims 1-7 is implemented.

9. A computer-readable storage medium having a program stored thereon, characterized in that: When the program is executed by the processor, a method for predicting logistics workload in a grey attention network based on firefly algorithm optimization as described in any one of claims 1 to 7 is implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by the processor, it implements the gray attention network logistics workload prediction method based on firefly algorithm optimization as described in any one of claims 1 to 7.

Images