Cold chain logistics data management method, device, equipment and storage medium

Abstract

The present application relates to the field of data processing technology, and discloses a cold chain logistics data management method, device, equipment and storage medium. The method includes: performing multidimensional data monitoring and data preprocessing on the target cold chain logistics, obtaining the target multidimensional cold chain logistics data set and generating the first cold chain logistics cost information; performing short-term time correlation feature extraction to obtain multiple short-term time correlation features; performing long-term time correlation feature extraction to obtain multiple long-term time correlation features; performing feature weight analysis and feature fusion to obtain fused time correlation features and optimizing the cold chain logistics cargo storage cost and damage risk cost to obtain the first cold chain logistics cost optimization index; performing cold chain logistics transportation task mode cost optimization to obtain the second cold chain logistics cost optimization index; performing cost control strategy optimization to generate the second cold chain logistics cost information. The present application improves the logistics efficiency of cold chain logistics and realizes accurate cold chain logistics cost control.

Description

Technical Field

[0001] The present application relates to the field of data processing technology, and in particular to a cold chain logistics data management method, device, equipment and storage medium. Background Art

[0002] In the current cold chain logistics industry, the efficiency and cost control of cold chain logistics have become the main challenges faced by enterprises. Traditional cold chain logistics management mainly relies on experience judgment and simple data processing technology. This method is incapable of processing large-scale, high-dimensional and dynamically changing logistics data, and it is difficult to achieve real-time monitoring and precise management of the cold chain logistics process. Especially in the monitoring and analysis of multi-dimensional data such as temperature, humidity, and location, there is a lack of effective data management methods to comprehensively process this information, which in turn affects the optimization of cold chain logistics costs and the minimization of risks.

[0003] In addition, with the development of Internet of Things technology, a large number of multimodal sensors are deployed in cold chain logistics systems, which generate a large amount of real-time data. How to extract valuable information from these huge data sets and make fast and accurate decisions based on this information has become the key to improving the efficiency of cold chain logistics and reducing operating costs. However, existing data processing and analysis methods are difficult to effectively process and analyze these complex multidimensional data, resulting in the failure to fully realize the potential of cold chain logistics cost optimization and risk control. Summary of the invention

[0004] The present application provides a cold chain logistics data management method, device, equipment and storage medium for improving the logistics efficiency of cold chain logistics and achieving accurate cold chain logistics cost control.

[0005] In a first aspect, the present application provides a cold chain logistics data management method, the cold chain logistics data management method comprising:

[0006] Based on the multimodal sensor network, multidimensional data monitoring and data preprocessing are performed on the target cold chain logistics to obtain a target multidimensional cold chain logistics data set, and a spectral clustering algorithm is used to perform logistics cost analysis on the target multidimensional cold chain logistics data set to generate first cold chain logistics cost information;

[0007] Using a spatiotemporal graph convolutional network to extract short-term temporal correlation features from the target multidimensional cold chain logistics data set to obtain multiple short-term temporal correlation features;

[0008] Performing time series data extraction on the target multidimensional cold chain logistics data set by using the SPARSS algorithm to obtain multiple time series cold chain logistics data, and performing long-term time correlation feature extraction on the multiple time series cold chain logistics data by using a long short-term memory neural network to obtain multiple long-term time correlation features;

[0009] Performing feature weight analysis and feature fusion on the multiple short-term time correlation features and the multiple long-term time correlation features to obtain fused time correlation features, and inputting the fused time correlation features into a preset multi-layer perceptron model to optimize the cold chain logistics cargo storage cost and damage risk cost to obtain a first cold chain logistics cost optimization index;

[0010] The cold chain logistics transportation task mode cost is optimized by using the k-NN algorithm on the fused time-related features to obtain a second cold chain logistics cost optimization index;

[0011] Through a non-dominated sorting genetic algorithm, the cost control strategy of the first cold chain logistics cost information is optimized according to the first cold chain logistics cost optimization index and the second cold chain logistics cost information of the target cold chain logistics is generated.

[0012] In a second aspect, the present application provides a cold chain logistics data management device, the cold chain logistics data management device comprising:

[0013] A monitoring module, used for performing multi-dimensional data monitoring and data preprocessing on a target cold chain logistics based on a multimodal sensor network, obtaining a target multi-dimensional cold chain logistics data set, and performing logistics cost analysis on the target multi-dimensional cold chain logistics data set using a spectral clustering algorithm to generate first cold chain logistics cost information;

[0014] A first feature extraction module is used to extract short-term time correlation features from the target multi-dimensional cold chain logistics data set using a spatiotemporal graph convolutional network to obtain multiple short-term time correlation features;

[0015] A second feature extraction module is used to perform time series data extraction on the target multidimensional cold chain logistics data set by using the SPARSS algorithm to obtain a plurality of time series cold chain logistics data, and to perform long-term time correlation feature extraction on the plurality of time series cold chain logistics data by using a long short-term memory neural network to obtain a plurality of long-term time correlation features;

[0016] A first cost optimization module is used to perform feature weight analysis and feature fusion on the multiple short-term time correlation features and the multiple long-term time correlation features to obtain fused time correlation features, and input the fused time correlation features into a preset multi-layer perceptron model to optimize the cold chain logistics cargo storage cost and damage risk cost to obtain a first cold chain logistics cost optimization index;

[0017] A second cost optimization module is used to optimize the cold chain logistics transportation task mode cost of the fused time-related features through a k-NN algorithm to obtain a second cold chain logistics cost optimization index;

[0018] The strategy optimization module is used to optimize the cost control strategy of the first cold chain logistics cost information according to the first cold chain logistics cost optimization index and the second cold chain logistics cost optimization index through a non-dominated sorting genetic algorithm, and generate the second cold chain logistics cost information of the target cold chain logistics.

[0019] The third aspect of the present application provides a cold chain logistics data management device, comprising: a memory and at least one processor, wherein the memory stores instructions; the at least one processor calls the instructions in the memory so that the cold chain logistics data management device executes the above-mentioned cold chain logistics data management method.

[0020] A fourth aspect of the present application provides a computer-readable storage medium, which stores instructions that, when executed on a computer, enable the computer to execute the above-mentioned cold chain logistics data management method.

[0021] In the technical solution provided by the present application, the multi-dimensional data monitoring and preprocessing by multimodal sensor network can fully capture the key parameters of the cold chain logistics environment. This all-round data collection provides a rich and accurate basis for the data analysis of cold chain logistics, ensuring the efficiency and accuracy of data management. Through the spectral clustering algorithm, the data set is analyzed for logistics cost, which can effectively identify the cost-related patterns and trends. The spatiotemporal graph convolution network and the long short-term memory neural network are used to extract short-term and long-term time correlation features respectively, which not only considers the spatiotemporal correlation of the data, but also can capture the deep characteristics of the time series data in the cold chain logistics process. Such a feature fusion strategy greatly improves the accuracy and robustness of the prediction model. By combining the k-NN algorithm and the non-dominated sorting genetic algorithm for cost optimization, the cost can be minimized under the premise of ensuring the efficiency of cold chain logistics and the safety of goods. In particular, the application of the non-dominated sorting genetic algorithm in multi-objective optimization problems enables the cold chain logistics cost control strategy to achieve the optimal solution under the constraints of various costs and service quality, thereby achieving the dual improvement of cost-effectiveness and service quality, thereby improving the logistics efficiency of cold chain logistics and realizing accurate cold chain logistics cost control. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0023] Figure 1 This is a schematic diagram of an embodiment of the cold chain logistics data management method in the embodiment of the present application;

[0024] Figure 2 This is a schematic diagram of an embodiment of a cold chain logistics data management device in an embodiment of the present application. DETAILED DESCRIPTION

[0025] Embodiments of the present application provide a cold chain logistics data management method, device, equipment and storage medium. The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of this application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments described here can be implemented in an order other than that illustrated or described here. In addition, the terms "including" or "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0026] For ease of understanding, the specific process of the embodiment of the present application is described below. Figure 1 , an embodiment of the cold chain logistics data management method in the embodiment of the present application includes:

[0027] Step S101: performing multidimensional data monitoring and data preprocessing on a target cold chain logistics based on a multimodal sensor network to obtain a target multidimensional cold chain logistics data set, and performing logistics cost analysis on the target multidimensional cold chain logistics data set using a spectral clustering algorithm to generate first cold chain logistics cost information;

[0028] It is understandable that the execution subject of the present application may be a cold chain logistics data management device, or a terminal or a server, which is not limited here. The present application embodiment is described by taking a server as the execution subject as an example.

[0029] Specifically, based on the multimodal sensor network, multidimensional data monitoring of the target cold chain logistics is performed to capture various important information in the cold chain process, such as temperature, humidity, location, etc., to form an initial multidimensional cold chain logistics data set. The initial data set is cleaned and standardized. Data cleaning removes errors and duplicate records in the data, while data standardization ensures that data from different sources and in different formats can be processed uniformly to obtain a standardized multidimensional cold chain logistics data set. Outliers are removed by calculating the quartiles and interquartile range of the standard multidimensional cold chain logistics data set. Based on statistical principles, quartiles and interquartile ranges are used to identify and remove outliers in the data, which may be caused by sensor failures, data transmission errors, or atypical logistics activities. By removing outliers, a more accurate and reliable target multidimensional cold chain logistics data set is obtained. The preprocessed target multidimensional cold chain logistics data set is further analyzed by the spectral clustering algorithm. The spectral clustering algorithm calculates the local standard deviation data, calculates the difference between each point in the data set and its neighboring points, and thus evaluates the local volatility of the data. The local standard deviation data reflects the stability and consistency of different regions in the cold chain logistics dataset. Based on these data, a preliminary analysis of the cluster centers is performed to obtain N initial cluster centers. These initial cluster centers represent different possible logistics cost patterns in the dataset. The initial cluster centers are optimized through the similarity matrix in the spectral clustering algorithm and the optimized initial center selection method. The optimization process takes into account the similarity between data points and aims to find the cluster center that best represents the entire dataset. The target multidimensional cold chain logistics dataset is comprehensively optimized in terms of fixed costs and transportation costs through the target cluster center. By analyzing the data points corresponding to different cluster centers, opportunities for cost optimization in cold chain logistics are identified, and finally the first cold chain logistics cost information is generated.

[0030] Step S102: extract short-term time correlation features from the target multi-dimensional cold chain logistics data set using a spatiotemporal graph convolutional network to obtain multiple short-term time correlation features;

[0031] Specifically, a spatial convolution operation is performed on the target multidimensional cold chain logistics dataset through a spatiotemporal graph convolution network, and the spatial component of the spatiotemporal graph convolution network is used to capture the spatial relationship and interaction between logistics nodes. The spatial convolution output features reflect the interdependence between different locations in cold chain logistics, for example, how the spatial relationship between temperature control points and cargo storage areas affects the overall logistics efficiency and product quality maintenance. The temporal convolution operation is performed through the spatiotemporal graph convolution network to analyze the data changes in the time dimension. The temporal convolution operation utilizes the time component of the network to capture the dynamic characteristics of the cold chain logistics process that change over time, such as the temperature change pattern of the cargo from the departure point to the destination. The obtained temporal convolution output features reveal the law of data change in the time dimension. Short-term temporal correlation features are identified for the spatial convolution output features and the temporal convolution output features. By analyzing the interaction of spatial and temporal convolution features, the key short-term temporal correlation features that affect the efficiency and cost of cold chain logistics are identified. For example, a sudden increase in temperature in a specific area within a certain period of time may have an impact on logistics costs. By analyzing these short-term temporal correlation features, logistics strategies can be adjusted in time to minimize costs and risks.

[0032] Step S103, extracting time series data from the target multidimensional cold chain logistics data set by using the SPARSS algorithm to obtain multiple time series cold chain logistics data, and extracting long-term time correlation features from the multiple time series cold chain logistics data by using a long short-term memory neural network to obtain multiple long-term time correlation features;

[0033] Specifically, the SPARSS algorithm is used to identify the time series pattern of the target multidimensional cold chain logistics data set. The time series pattern represents the specific law of change over time in the cold chain logistics process, such as temperature change, humidity adjustment, or periodic fluctuation of the flow speed of goods. The SPARSS algorithm can accurately capture these patterns by analyzing the time series of the data set. The original multidimensional cold chain logistics data set is grouped according to these patterns, and each group represents a specific time series pattern. The time series hidden features of multiple time series cold chain logistics data are extracted through the forward long short-term unit layer in the long short-term memory neural network. The hidden time series features are extracted from the grouped time series cold chain logistics data through the forward long short-term unit layer and the reverse long short-term unit layer of the LSTM network. The forward layer focuses on the time sequence from the past to the future, capturing the features that appear in the time series as time progresses; while the reverse layer analyzes from the future to the past, extracting those time features that may be overlooked in traditional analysis methods. Through forward and reverse analysis, LSTM can extract rich time series hidden features from the data, which comprehensively cover the data from two different time directions. The forward and reverse time series hidden features are fused to obtain more comprehensive long-term time correlation features. The fusion process takes into account the information obtained from both directions of the time series, so that the obtained long-term time correlation features can more comprehensively and accurately reflect the temporal dynamics of cold chain logistics data, thereby effectively identifying long-term time correlation patterns that have an important impact on cold chain logistics management.

[0034] Step S104: performing feature weight analysis and feature fusion on multiple short-term time correlation features and multiple long-term time correlation features to obtain fused time correlation features, and inputting the fused time correlation features into a preset multi-layer perceptron model to optimize the cold chain logistics cargo storage cost and damage risk cost to obtain a first cold chain logistics cost optimization index;

[0035] Specifically, the mean operation is performed on multiple short-term time-related features to capture the average performance of these features in a shorter time frame and obtain the first feature mean. The standard deviation of the first feature is calculated based on the mean, which provides a way to measure the volatility of the short-term time-related features and reflects the degree of variation of cold chain logistics data in the short term. Similarly, by calculating the mean and standard deviation of the long-term time-related features, the central position of the long-term trend is revealed and its fluctuation in a longer time span is also reflected. After obtaining the mean and standard deviation of the short-term and long-term time-related features, the feature weight coefficient is calculated. Considering the relative importance of short-term and long-term features for the entire cold chain logistics cost and risk management, the weight allocation mechanism is used to ensure that the proportion of each feature in the final fused time-related features matches its actual contribution to the cold chain logistics cost optimization. The weight allocation method based on statistical analysis enhances the prediction accuracy and optimization effect of the model. Based on the feature weight coefficient, the short-term and long-term time-related features are effectively fused to generate the fused time-related features. By integrating information from different time scales, the influence of each feature in the fusion process is adjusted through the weight coefficient, so as to ensure that the obtained fused time-related features can fully reflect the key time dynamics in the cold chain logistics process. The fused time-related features are input into the preset multi-layer perceptron (MLP) model to optimize the storage cost and damage risk cost of cold chain logistics goods. The multi-layer perceptron model learns the complex nonlinear relationship between the fused features and cost optimization through training, and finally obtains the first cold chain logistics cost optimization index. This index reflects the cost optimization potential after considering short-term and long-term time dynamics.

[0036] Step S105, optimizing the cold chain logistics transportation task mode cost by integrating the time-related features through the k-NN algorithm to obtain a second cold chain logistics cost optimization index;

[0037] Specifically, the k-NN algorithm is used to measure the feature distance of the fused time-related features to determine the similarity or difference between each time-related feature. The feature distance measurement data reflects the positional relationship between different cold chain logistics transportation task modes in the feature space. Based on the feature distance measurement data, the k-NN algorithm is used to predict the decision rules of the cold chain logistics transportation task mode. By analyzing the distance relationship between the fused time-related features and the known transportation task mode cost, the most likely applicable decision rule is predicted. The prediction process takes into account the relative position of the fused features in the multidimensional feature space, and determines the transportation task mode and corresponding cost optimization strategy that the target feature is most likely to correspond to by identifying the known samples that are the nearest neighbors to the target feature. This similarity-based prediction method can highly automatically identify the optimal or near-optimal transportation task mode. According to the predicted target decision rule, the cold chain logistics transportation task mode cost optimization is performed on the fused time-related features. The predicted optimal or near-optimal transport task mode decision rules are applied to actual cold chain logistics management. By adjusting the transport route, transport mode, cargo loading strategy, etc., in order to achieve the purpose of minimizing costs or maximizing efficiency, the second cold chain logistics cost optimization index is obtained, which reflects the cost optimization effect after adopting the optimal transport task mode.

[0038] Step S106: Optimize the cost control strategy of the first cold chain logistics cost information according to the first cold chain logistics cost optimization index and the second cold chain logistics cost optimization index through a non-dominated sorting genetic algorithm to generate the second cold chain logistics cost information of the target cold chain logistics.

[0039] Specifically, a multi-objective optimization function of the first cold chain logistics cost information is defined according to the first cold chain logistics cost optimization index and the second cold chain logistics cost optimization index, and the multi-dimensional objective of cold chain logistics cost optimization is converted into a form that can be solved by mathematical methods. The cost control strategy group of the first cold chain logistics cost information is initialized by a non-dominated sorting genetic algorithm to obtain multiple primary cost control strategies. The initialization of these strategy groups is based on the diversified needs of the cost optimization index, ensuring the comprehensiveness and diversity of the optimization process and forming an initial solution space. The fitness value of each primary cost control strategy is calculated by the defined multi-objective optimization function. Based on the fitness value, all primary cost control strategies are sorted and analyzed to obtain a cost control strategy sequence sorted according to the cost optimization effect and implementation feasibility. The top 50% of the cost control strategies are selected as the parent generation, and a new generation of cost control strategies, namely the second cost control strategy, is generated through reproduction, mutation and crossover operations. By simulating the natural selection and genetic mechanism in biological evolution, a better cost control strategy is continuously iterated to ensure the diversity of the understanding space and the effectiveness of the optimization process. The fitness value of the second cost control strategy is calculated and optimized using the multi-objective optimization function. In this process, the cost optimization effect and feasibility of implementation of each strategy are evaluated and ranked again, and the optimal cost control strategy is finally selected. After repeated iterations, the cost control strategy is further refined, so that the final target cost control strategy is optimized in multiple dimensions. The target cost control strategy is applied to the first cold chain logistics cost information for cost control optimization. Through the implementation of the strategy, the cost efficiency of cold chain logistics is improved, the sustainability and stability of the cold chain logistics system are enhanced, and finally the optimized target cold chain logistics second cold chain logistics cost information is generated.

[0040] In the embodiment of the present application, the multi-dimensional data monitoring and preprocessing by multimodal sensor network can fully capture the key parameters of the cold chain logistics environment. This all-round data collection provides a rich and accurate basis for the data analysis of cold chain logistics, ensuring the efficiency and accuracy of data management. Through the spectral clustering algorithm, the data set is analyzed for logistics cost, which can effectively identify the cost-related patterns and trends. The spatiotemporal graph convolution network and the long short-term memory neural network are used to extract short-term and long-term time correlation features respectively, which not only considers the spatiotemporal correlation of the data, but also can capture the deep-level characteristics of the time series data in the cold chain logistics process. Such a feature fusion strategy greatly improves the accuracy and robustness of the prediction model. By combining the k-NN algorithm and the non-dominated sorting genetic algorithm for cost optimization, the cost can be minimized under the premise of ensuring the efficiency of cold chain logistics and the safety of goods. In particular, the application of the non-dominated sorting genetic algorithm in multi-objective optimization problems enables the cold chain logistics cost control strategy to achieve the best optimization under the conditions of meeting a variety of cost and service quality constraints, thereby achieving a dual improvement in cost-effectiveness and service quality, thereby improving the logistics efficiency of cold chain logistics and achieving accurate cold chain logistics cost control.

[0041] In a specific embodiment, the process of executing step S101 may specifically include the following steps:

[0042] (1) Based on the multimodal sensor network, multidimensional data monitoring of the target cold chain logistics is performed to obtain the initial multidimensional cold chain logistics data set;

[0043] (2) Perform data cleaning and data standardization on the initial multidimensional cold chain logistics dataset to obtain a standard multidimensional cold chain logistics dataset;

[0044] (3) Calculate the quartiles and interquartile range of the standard multidimensional cold chain logistics data set, and remove outliers from the standard multidimensional cold chain logistics data set based on the quartiles and interquartile range to obtain the target multidimensional cold chain logistics data set;

[0045] (4) The local standard deviation of the target multidimensional cold chain logistics data set is calculated by the spectral clustering algorithm to obtain the local standard deviation data, and the cluster center analysis of the target multidimensional cold chain logistics data set is performed based on the local standard deviation data to obtain N initial cluster centers;

[0046] (5) Through the similarity matrix in the spectral clustering algorithm, the optimal clustering center analysis is performed on the N initial clustering centers by using the method of optimizing the initial center selection to obtain the target clustering center;

[0047] (6) The fixed cost and transportation cost of the target multidimensional cold chain logistics data set are comprehensively optimized through the target clustering center to generate the first cold chain logistics cost information.

[0048] Specifically, multi-dimensional data monitoring of the target cold chain logistics is performed based on a multimodal sensor network, including temperature, humidity, location, and transportation status monitoring equipment. The data collected by these sensors form an initial multi-dimensional cold chain logistics data set. Data cleaning and data standardization are performed on the initial multi-dimensional cold chain logistics data set. Data cleaning mainly removes erroneous data and duplicate records, while data standardization ensures that data from different sources can be compared and analyzed under the same standard, thereby obtaining a standard multi-dimensional cold chain logistics data set. Outliers in the data set are identified and removed by calculating quartiles and interquartile ranges. Quartiles provide information on the central tendency and dispersion of data distribution, while interquartile ranges reflect the variability of data concentration. The data set is analyzed by a spectral clustering algorithm. Local standard deviation calculations are performed to assess the volatility of data in a local range. Areas with high local standard deviations may indicate potential cost or efficiency issues in specific links in cold chain logistics. Using the similarity matrix in the spectral clustering algorithm, cluster center analysis is performed on the data set based on local standard deviation data, and N cluster centers are initially identified. These cluster centers represent the main patterns or trends in the data set. For example, they may represent different temperature control requirements or transportation distances. In order to improve the accuracy of clustering, the N cluster centers are refined by optimizing the selection method of the initial center to obtain the best cluster center. Adjust the parameters of the clustering algorithm, or introduce additional information to assist in judgment. By determining the cluster centers, understand the patterns contained in the cold chain logistics data set, and provide a basis for cost optimization. Use cluster centers to guide the comprehensive optimization of fixed costs and transportation costs of cold chain logistics. For example, more efficient logistics planning can be achieved by identifying the transportation mode corresponding to a specific cluster center, or the loss rate of certain goods can be reduced by adjusting storage conditions, thereby reducing loss costs. Data-driven optimization methods can significantly improve the efficiency and economy of cold chain logistics, and the first cold chain logistics cost information finally generated provides management with a comprehensive and optimized cost control plan.

[0049] In a specific embodiment, the process of executing step S102 may specifically include the following steps:

[0050] (1) Perform spatial convolution operations on the target multi-dimensional cold chain logistics dataset through the spatiotemporal graph convolutional network to obtain spatial convolution output features;

[0051] (2) Perform temporal convolution operations on the target multidimensional cold chain logistics dataset through the spatiotemporal graph convolutional network to obtain the temporal convolution output features;

[0052] (3) Perform short-term temporal correlation feature recognition on the spatial convolution output features and the temporal convolution output features to obtain multiple short-term temporal correlation features.

[0053] Specifically, by performing spatial convolution operations on the target multidimensional cold chain logistics dataset, the spatial relationships and interactions between different nodes in the cold chain network are obtained. The spatial convolution operation utilizes the ability of the graph convolution network to aggregate the sensor node data with similar physical locations or logical relationships to represent spatial interdependencies. For example, in the cold chain logistics network, the data collected by sensors loaded on different warehouses and transport vehicles can be aggregated and analyzed through spatial convolution to obtain spatial convolution output features. These features represent the interactions and dependencies between nodes in space, such as the consistency of temperature control and the stability of goods during transportation. The target multidimensional cold chain logistics dataset is subjected to temporal convolution operations through the spatiotemporal graph convolution network to obtain data patterns that change over time. The temporal convolution operation focuses on the changes in data in the time series, and captures the dynamic characteristics in the time dimension by analyzing the changes in data at continuous time points. For example, for temperature monitoring data in cold chain logistics, temporal convolution can reveal the temperature fluctuation pattern caused by the temperature difference between day and night, or the temperature adaptability changes when goods are transferred from one climate zone to another. Through temporal analysis, we obtain temporal convolution output features, which show the key information that changes over time in the cold chain logistics process. We identify short-term temporal correlation features. By comprehensively considering the analysis results of both spatial and temporal dimensions, we identify short-term temporal correlation features that have a direct impact on cold chain logistics management. For example, if the spatial convolution features of a specific area show an abnormally high temperature over a certain period of time, and the temporal convolution features also indicate that this area has frequently experienced temperature fluctuations in a short period of time, this short-term temporal correlation feature may indicate potential risks in the cold chain process, such as an increased possibility of cargo damage.

[0054] In a specific embodiment, the process of executing step S103 may specifically include the following steps:

[0055] (1) The SPARSS algorithm is used to perform time series pattern recognition on the target multidimensional cold chain logistics data set to obtain multiple time series patterns;

[0056] (2) According to multiple time series patterns, the target multidimensional cold chain logistics data set is grouped into time series data to obtain multiple time series cold chain logistics data;

[0057] (3) The time series hidden features of multiple time series cold chain logistics data are extracted through the positive long short-term unit layer in the long short-term memory neural network to obtain multiple positive time series hidden features;

[0058] (4) The reverse long short-term unit layer in the long short-term memory neural network is used to extract the time series hidden features of multiple time series cold chain logistics data to obtain multiple reverse time series hidden features;

[0059] (5) Perform long-term temporal correlation feature fusion on multiple forward temporal hidden features and multiple reverse temporal hidden features to obtain multiple long-term temporal correlation features.

[0060] Specifically, the target multidimensional cold chain logistics data set is subjected to time series pattern recognition by the SPARSS algorithm. The SPARSS algorithm can identify multiple time series patterns, which may represent various typical states in the cold chain logistics process, such as periodic patterns of temperature changes, mutation patterns of humidity, etc. For example, if a cold chain logistics data set contains temperature and humidity data collected from multiple sensors, the SPARSS algorithm can identify patterns of steady temperature rise within a specific time period, or patterns of sudden changes in humidity during transportation. The target multidimensional cold chain logistics data set is grouped according to the time series pattern. Similar time series patterns in the data set are aggregated together to form multiple time series cold chain logistics data groups. For example, all data points showing a steady increase in temperature are grouped together, while data points showing a sudden change in humidity are grouped together into another group. The time series cold chain logistics data is analyzed by the forward long short-term unit layer and the reverse long short-term unit layer in the long short-term memory neural network. The long short-term memory neural network is suitable for processing time series data and can capture long-term and short-term dependencies in the data. Through the forward long and short time unit layer, the time series hidden features that unfold in time are extracted; while the reverse long and short time unit layer extracts hidden features from the reverse direction of the time series, capturing patterns that may be overlooked by the forward analysis. For example, when analyzing the temperature rise data processed by the forward long and short time unit layer, the temperature change pattern that increases over time is extracted; while the reverse long and short time unit layer may reveal a specific pattern of temperature change before a specific time point. The forward and reverse time series hidden features are fused to obtain more comprehensive long-term time correlation features. The fusion process takes into account the information obtained from both directions of the time series to ensure a comprehensive understanding of the time dynamics in the cold chain logistics process. For example, by fusing the forward and reverse time series hidden features, the long-term risk pattern of possible damage to goods under specific environmental conditions is revealed.

[0061] In a specific embodiment, the process of executing step S104 may specifically include the following steps:

[0062] (1) performing feature mean calculation on multiple short-term time correlation features to obtain a first feature mean, and calculating a first feature standard deviation of the multiple short-term time correlation features based on the first feature mean;

[0063] (2) performing feature mean calculation on multiple long-term time correlation features to obtain a second feature mean, and calculating the second feature standard deviation of the multiple long-term time correlation features based on the second feature mean;

[0064] (3) calculating feature weight coefficients between a plurality of short-term time correlation features and a plurality of long-term time correlation features according to the first feature mean, the first feature standard deviation, the second feature mean, and the second feature standard deviation;

[0065] (4) Fusing multiple short-term time correlation features and multiple long-term time correlation features according to feature weight coefficients to obtain fused time correlation features;

[0066] (5) The fused time-correlation features are input into the preset multi-layer perceptron model to optimize the storage cost and damage risk cost of cold chain logistics goods, and the first cold chain logistics cost optimization index is obtained.

[0067] Specifically, the feature mean operation is performed on multiple short-term time-related features to capture the average performance of data in a shorter time range and understand the general status and trend of cold chain logistics in the short term. The standard deviation of the feature is calculated based on the mean to evaluate the volatility or stability of the data in the short term. The smaller the standard deviation, the more stable the data; the larger the standard deviation, the stronger the volatility of the data. Similarly, the mean and standard deviation of the long-term time-related features are calculated to capture the average performance and stability of the cold chain logistics status in the long term, providing information for long-term planning and risk assessment. Based on the mean and standard deviation, the feature weight coefficient between the short-term and long-term time-related features is further calculated to determine the proportion that each feature should occupy when performing feature fusion. The calculation of the feature weight coefficient takes into account the importance of the feature and the contribution to the target variable, ensuring that more influential features can obtain higher weights during the fusion process. For example, if the long-term temperature stability has a greater impact on the storage cost of goods than the short-term temperature fluctuation, the long-term feature will be given a higher weight when fused. According to the feature weight coefficient, the short-term and long-term time-related features are fused to generate fused time-related features. The fusion process integrates information from different time scales, and ensures that the influence of each feature in the fused feature is reasonably reflected through weight adjustment. The fused time-related features can more comprehensively reflect the complex dynamics of the cold chain logistics process. The fused time-related features are input into the preset multi-layer perceptron model for analysis to optimize the storage cost and damage risk cost of cold chain logistics goods. By learning the nonlinear relationship between fused features and costs, the multi-layer perceptron model can predict the optimal storage and transportation strategies under different conditions, thereby minimizing costs and risks. For example, the model may find that under a specific temperature and humidity combination, the use of specific packaging materials can significantly reduce the risk of damage to goods, thereby reducing related costs.

[0068] In a specific embodiment, the process of executing step S105 may specifically include the following steps:

[0069] (1) Using the k-NN algorithm to measure the feature distance of the fused time-related features, we can obtain feature distance measurement data;

[0070] (2) Predict the decision rules of the cold chain logistics transportation task mode based on the fused time-related features according to the feature distance measurement data to obtain the target decision rules;

[0071] (3) According to the target decision rule, the cold chain logistics transportation task mode cost is optimized by integrating the time-related features to obtain the second cold chain logistics cost optimization index.

[0072] Specifically, the k-NN algorithm is used to measure the feature distance of the fused time-related features. The algorithm identifies the most similar data points by calculating the distance between the feature to be analyzed and other features in the data set. This distance is usually the Euclidean distance, or other types of distance metrics, such as Manhattan distance or cosine similarity. Based on the feature distance metric data, the algorithm further predicts the cold chain logistics transportation task mode decision rule for the fused time-related features. The algorithm assigns each feature point to the most likely transportation task mode category based on the distance metric. The classification is based on the core principle of the k-NN algorithm: a feature point is most likely to belong to the category of its nearest neighbor. In this embodiment, if a specific fused time-related feature is closest to a feature point of a known transportation task mode (for example, low-temperature long-distance transportation), then this feature point is also most likely to represent similar transportation needs. Through the above steps, target decision rules are obtained, which guide how to optimize the cold chain logistics transportation task mode based on the comprehensive analysis of time-related features. For example, a rule derived from historical data analysis may indicate that under specific climatic conditions and cargo types, the use of specific packaging and transportation methods can minimize transportation costs and reduce cargo losses. According to the target decision rule, the cold chain logistics transportation task mode cost optimization is performed on the fused time-related features. The predicted optimal transportation mode is used to adjust the actual logistics operation to maximize cost efficiency. This may involve adjusting the transportation route, choosing a more appropriate transportation method, or changing the loading strategy of the goods. Through optimization, the second cold chain logistics cost optimization index is obtained, which reflects the cost savings that can be achieved by optimizing the transportation task mode after considering all relevant time-related features.

[0073] In a specific embodiment, the process of executing step S106 may specifically include the following steps:

[0074] (1) defining a multi-objective optimization function of the first cold chain logistics cost information according to the first cold chain logistics cost optimization index and the second cold chain logistics cost optimization index;

[0075] (2) Initializing the cost control strategy group of the first cold chain logistics cost information through a non-dominated sorting genetic algorithm to obtain multiple first cost control strategies;

[0076] (3) calculating the first fitness value of each first cost control strategy respectively through the multi-objective optimization function, and performing sorting analysis on the multiple first cost control strategies according to the first fitness value to obtain a cost control strategy sequence;

[0077] (4) The first 50% of the first cost control strategies in the cost control strategy sequence are used as parents for reproduction, mutation and crossover to generate multiple second cost control strategies;

[0078] (5) calculating the second fitness value of each second cost control strategy respectively through a multi-objective optimization function, and optimizing the plurality of second cost control strategies according to the second fitness values ​​to obtain a target cost control strategy;

[0079] (6) Optimize the cost control of the first cold chain logistics cost information through the target cost control strategy to generate the second cold chain logistics cost information of the target cold chain logistics.

[0080] Specifically, a multi-objective optimization function is defined according to the first cold chain logistics cost optimization index and the second cold chain logistics cost optimization index to minimize the total cost of cold chain logistics under the premise of ensuring the quality of goods and reducing transportation losses. For example, the first cost optimization index may focus on the energy consumption and storage cost during the storage of goods, while the second cost optimization index may focus on the fuel consumption and time cost during transportation. The multi-objective optimization function needs to integrate these indicators to find a balance point, that is, to minimize the cost without sacrificing the quality of goods. The cost control strategy group is initialized for the first cold chain logistics cost information through the non-dominated sorting genetic algorithm to generate multiple primary cost control strategies. A set of cost control strategies is randomly generated, each of which is a possible solution to the cold chain logistics operation mode, including transportation route selection, cargo loading method, temperature control setting, etc., providing a diverse search space for the algorithm to ensure that as many potential solutions as possible can be explored. The fitness value of each primary cost control strategy is calculated through the defined multi-objective optimization function. The performance of each strategy is evaluated. The higher the fitness value, the better the performance of the strategy in cost optimization. Based on the fitness value, all strategies are sorted and analyzed to obtain a sequence of cost control strategies arranged by performance. The top 50% of the cost control strategies are selected as parents, and reproduction, mutation and crossover operations are performed to generate a new generation of cost control strategies. This process simulates the process of natural selection and genetic variation, and generates offspring strategies with new characteristics by exchanging and combining the genes (i.e., strategy parameters) of the parent strategies. These operations help the algorithm to jump out of the local optimal solution and explore a broader solution space. After that, the fitness value of the new generation of cost control strategies is calculated and optimized. After repeated iterations, the algorithm can continuously approach the optimal solution and finally determine the target cost control strategy under the current model and data conditions. This strategy is the best solution after weighing all the considerations, aiming to minimize costs while ensuring the efficient and reliable operation of cold chain logistics. The target cost control strategy is applied to the first cold chain logistics cost information, and cost control optimization is performed to generate the final target cold chain logistics second cold chain logistics cost information. This cost information not only contains the optimized cost budget, but also reflects the economic benefits that can be brought about by adopting this strategy. For example, by optimizing transportation routes and adjusting cargo loading methods, it is possible to significantly reduce transportation costs. At the same time, through sophisticated temperature control management, cargo losses can be reduced, thereby reducing overall costs while ensuring cargo quality.

[0081] The above describes the cold chain logistics data management method in the embodiment of the present application. The following describes the cold chain logistics data management device in the embodiment of the present application. Figure 2 In the embodiment of the present application, an embodiment of the cold chain logistics data management device includes:

[0082] The monitoring module 201 is used to perform multi-dimensional data monitoring and data preprocessing on the target cold chain logistics based on the multimodal sensor network to obtain a target multi-dimensional cold chain logistics data set, and use a spectral clustering algorithm to perform logistics cost analysis on the target multi-dimensional cold chain logistics data set to generate first cold chain logistics cost information;

[0083] A first feature extraction module 202 is used to extract short-term time correlation features from a target multi-dimensional cold chain logistics data set using a spatiotemporal graph convolutional network to obtain a plurality of short-term time correlation features;

[0084] The second feature extraction module 203 is used to extract time series data from the target multidimensional cold chain logistics data set by using the SPARSS algorithm to obtain multiple time series cold chain logistics data, and to extract long-term time correlation features from the multiple time series cold chain logistics data by using a long short-term memory neural network to obtain multiple long-term time correlation features;

[0085] The first cost optimization module 204 is used to perform feature weight analysis and feature fusion on multiple short-term time correlation features and multiple long-term time correlation features to obtain fused time correlation features, and input the fused time correlation features into a preset multi-layer perceptron model to optimize the cold chain logistics cargo storage cost and damage risk cost to obtain a first cold chain logistics cost optimization index;

[0086] The second cost optimization module 205 is used to optimize the cold chain logistics transportation task mode cost by integrating the time-related features through the k-NN algorithm to obtain a second cold chain logistics cost optimization index;

[0087] The strategy optimization module 206 is used to optimize the cost control strategy of the first cold chain logistics cost information according to the first cold chain logistics cost optimization index and the second cold chain logistics cost optimization index through a non-dominated sorting genetic algorithm to generate the second cold chain logistics cost information of the target cold chain logistics.

[0088] Through the cooperation of the above components, the multi-dimensional data monitoring and preprocessing by multimodal sensor network can fully capture the key parameters of the cold chain logistics environment. This all-round data collection provides a rich and accurate basis for the data analysis of cold chain logistics, ensuring the efficiency and accuracy of data management. The logistics cost analysis of the data set by spectral clustering algorithm can effectively identify the cost-related patterns and trends. The spatiotemporal graph convolutional network and long short-term memory neural network are used to extract short-term and long-term time correlation features respectively, which not only considers the spatiotemporal correlation of the data, but also captures the deep characteristics of the time series data in the cold chain logistics process. Such a feature fusion strategy greatly improves the accuracy and robustness of the prediction model. By combining the k-NN algorithm and the non-dominated sorting genetic algorithm for cost optimization, the cost can be minimized under the premise of ensuring the efficiency of cold chain logistics and the safety of goods. In particular, the application of the non-dominated sorting genetic algorithm in multi-objective optimization problems enables the cold chain logistics cost control strategy to achieve the optimal solution under the constraints of multiple costs and service quality, thereby achieving the dual improvement of cost-effectiveness and service quality, thereby improving the logistics efficiency of cold chain logistics and realizing accurate cold chain logistics cost control.

[0089] The present application also provides a cold chain logistics data management device, which includes a memory and a processor. The memory stores computer-readable instructions. When the computer-readable instructions are executed by the processor, the processor executes the steps of the cold chain logistics data management method in the above-mentioned embodiments.

[0090] The present application also provides a computer-readable storage medium, which may be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium. Instructions are stored in the computer-readable storage medium. When the instructions are executed on a computer, the computer executes the steps of the cold chain logistics data management method.

[0091] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described systems, systems and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0092] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk and other media that can store program codes.

[0093] As described above, the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A cold chain logistics data management method, It is characterized in that The cold chain logistics data management method comprises: Based on a multimodal sensor network, multidimensional data monitoring and data preprocessing are performed on the target cold chain logistics to obtain a target multidimensional cold chain logistics data set, and a spectral clustering algorithm is used to perform logistics cost analysis on the target multidimensional cold chain logistics data set to generate first cold chain logistics cost information; specifically including: based on a multimodal sensor network, multidimensional data monitoring is performed on the target cold chain logistics to obtain an initial multidimensional cold chain logistics data set; wherein, based on the multimodal sensor network, multidimensional data monitoring is performed on the target cold chain logistics, including temperature, humidity, location and transportation status monitoring equipment, and the data collected by these sensors form an initial multidimensional cold chain logistics data set; data cleaning and data standardization are performed on the initial multidimensional cold chain logistics data set to obtain a standard multidimensional cold chain logistics data set; the standard multidimensional cold chain logistics data is calculated The method comprises the following steps: calculating the quartiles and interquartile ranges of the standard multidimensional cold chain logistics data set, removing outliers from the standard multidimensional cold chain logistics data set according to the quartiles and the interquartile ranges, and obtaining a target multidimensional cold chain logistics data set; calculating the local standard deviation of the target multidimensional cold chain logistics data set by a spectral clustering algorithm, obtaining local standard deviation data, and performing cluster center analysis on the target multidimensional cold chain logistics data set according to the local standard deviation data, and obtaining N initial cluster centers; performing optimal cluster center analysis on the N initial cluster centers by using a similarity matrix in the spectral clustering algorithm and a method for selecting an optimized initial center, and obtaining a target cluster center; comprehensively optimizing the fixed cost and transportation cost of the target multidimensional cold chain logistics data set by using the target cluster center, and generating first cold chain logistics cost information; Using a spatiotemporal graph convolutional network to extract short-term temporal correlation features from the target multidimensional cold chain logistics data set to obtain multiple short-term temporal correlation features; Performing time series data extraction on the target multidimensional cold chain logistics data set by using the SPARSS algorithm to obtain multiple time series cold chain logistics data, and performing long-term time correlation feature extraction on the multiple time series cold chain logistics data by using a long short-term memory neural network to obtain multiple long-term time correlation features; Performing feature weight analysis and feature fusion on the multiple short-term time correlation features and the multiple long-term time correlation features to obtain fused time correlation features, and inputting the fused time correlation features into a preset multi-layer perceptron model to optimize the cold chain logistics cargo storage cost and damage risk cost to obtain a first cold chain logistics cost optimization index; The cold chain logistics transportation task mode cost is optimized by using the k-NN algorithm on the fused time-related features to obtain a second cold chain logistics cost optimization index; Through a non-dominated sorting genetic algorithm, the cost control strategy of the first cold chain logistics cost information is optimized according to the first cold chain logistics cost optimization index and the second cold chain logistics cost information of the target cold chain logistics is generated.

2. The cold chain logistics data management method according to claim 1, It is characterized in that The spatiotemporal graph convolutional network is used to extract short-term time correlation features from the target multi-dimensional cold chain logistics data set to obtain multiple short-term time correlation features, including: Performing spatial convolution operation on the target multi-dimensional cold chain logistics data set through a spatiotemporal graph convolutional network to obtain spatial convolution output features; Performing a temporal convolution operation on the target multidimensional cold chain logistics data set through the spatiotemporal graph convolution network to obtain a temporal convolution output feature; Short-term time correlation features are identified on the spatial convolution output features and the temporal convolution output features to obtain a plurality of short-term time correlation features.

3. The cold chain logistics data management method according to claim 2, It is characterized in that The target multidimensional cold chain logistics data set is subjected to time series data extraction by the SPARSS algorithm to obtain a plurality of time series cold chain logistics data, and the long short-term memory neural network is used to extract long-term time correlation features of the plurality of time series cold chain logistics data to obtain a plurality of long-term time correlation features, including: Performing time series pattern recognition on the target multidimensional cold chain logistics data set by using the SPARSS algorithm to obtain multiple time series patterns; Performing time series data grouping on the target multi-dimensional cold chain logistics data set according to the multiple time series patterns to obtain multiple time series cold chain logistics data; Extracting time series hidden features from the plurality of time series cold chain logistics data through a positive long short-term unit layer in a long short-term memory neural network to obtain a plurality of positive time series hidden features; Extracting time series hidden features from the plurality of time series cold chain logistics data through a reverse long short-term unit layer in a long short-term memory neural network to obtain a plurality of reverse time series hidden features; The multiple forward temporal hidden features and the multiple reverse temporal hidden features are subjected to long-term time correlation feature fusion to obtain multiple long-term time correlation features.

4. The cold chain logistics data management method according to claim 3, It is characterized in that The performing feature weight analysis and feature fusion on the multiple short-term time correlation features and the multiple long-term time correlation features to obtain fused time correlation features, and inputting the fused time correlation features into a preset multi-layer perceptron model to optimize the cold chain logistics cargo storage cost and damage risk cost, to obtain a first cold chain logistics cost optimization index, including: Performing feature mean calculation on the multiple short-term time correlation features to obtain a first feature mean, and calculating a first feature standard deviation of the multiple short-term time correlation features according to the first feature mean; Performing feature mean calculation on the multiple long-term time-related features to obtain a second feature mean, and calculating a second feature standard deviation of the multiple long-term time-related features according to the second feature mean; Calculating feature weight coefficients between the multiple short-term time association features and the multiple long-term time association features according to the first feature mean, the first feature standard deviation, the second feature mean, and the second feature standard deviation; Performing feature fusion on the multiple short-term time correlation features and the multiple long-term time correlation features according to the feature weight coefficient to obtain a fused time correlation feature; The fused time-related features are input into a preset multi-layer perceptron model to optimize the cold chain logistics cargo storage cost and damage risk cost, and obtain a first cold chain logistics cost optimization index.

5. The cold chain logistics data management method according to claim 4, It is characterized in that The cold chain logistics transportation task mode cost optimization is performed on the fused time-related features by using the k-NN algorithm to obtain a second cold chain logistics cost optimization index, including: Performing feature distance measurement on the fused time correlation feature by using a k-NN algorithm to obtain feature distance measurement data; According to the feature distance measurement data, the cold chain logistics transportation task mode decision rule is predicted for the fused time correlation feature to obtain a target decision rule; The cold chain logistics transportation task mode cost is optimized for the fused time-related features according to the target decision rule to obtain a second cold chain logistics cost optimization index.

6. The cold chain logistics data management method according to claim 5, It is characterized in that The method of optimizing the cost control strategy of the first cold chain logistics cost information according to the first cold chain logistics cost optimization index and the second cold chain logistics cost optimization index by using the non-dominated sorting genetic algorithm to generate the second cold chain logistics cost information of the target cold chain logistics includes: Defining a multi-objective optimization function of the first cold chain logistics cost information according to the first cold chain logistics cost optimization index and the second cold chain logistics cost optimization index; Initializing a cost control strategy group on the first cold chain logistics cost information by using a non-dominated sorting genetic algorithm to obtain a plurality of first cost control strategies; Calculating the first fitness value of each first cost control strategy respectively by using the multi-objective optimization function, and performing sorting analysis on the multiple first cost control strategies according to the first fitness value to obtain a cost control strategy sequence; The first cost control strategies of the first 50% in the cost control strategy sequence are used as parents for reproduction, mutation and crossover to generate a plurality of second cost control strategies; Calculating the second fitness value of each second cost control strategy respectively by using the multi-objective optimization function, and optimizing the multiple second cost control strategies according to the second fitness value to obtain a target cost control strategy; The first cold chain logistics cost information is optimized for cost control through the target cost control strategy to generate second cold chain logistics cost information for the target cold chain logistics.

7. A cold chain logistics data management device, It is characterized in that The cold chain logistics data management device comprises: A monitoring module is used to perform multidimensional data monitoring and data preprocessing on the target cold chain logistics based on a multimodal sensor network to obtain a target multidimensional cold chain logistics data set, and use a spectral clustering algorithm to perform logistics cost analysis on the target multidimensional cold chain logistics data set to generate first cold chain logistics cost information; specifically comprising: performing multidimensional data monitoring on the target cold chain logistics based on a multimodal sensor network to obtain an initial multidimensional cold chain logistics data set; wherein, performing multidimensional data monitoring on the target cold chain logistics based on the multimodal sensor network, including temperature, humidity, location and transportation status monitoring equipment, the data collected by these sensors form an initial multidimensional cold chain logistics data set; performing data cleaning and data standardization on the initial multidimensional cold chain logistics data set to obtain a standard multidimensional cold chain logistics data set; calculating the standard multidimensional cold chain The quartiles and interquartile ranges of the logistics data set are obtained, and outliers are removed from the standard multidimensional cold chain logistics data set according to the quartiles and the interquartile range to obtain a target multidimensional cold chain logistics data set; local standard deviation calculations are performed on the target multidimensional cold chain logistics data set through a spectral clustering algorithm to obtain local standard deviation data, and cluster center analysis is performed on the target multidimensional cold chain logistics data set according to the local standard deviation data to obtain N initial cluster centers; the similarity matrix in the spectral clustering algorithm is used to perform optimal cluster center analysis on the N initial cluster centers using a method for selecting an optimized initial center to obtain a target cluster center; the fixed cost and transportation cost of the target multidimensional cold chain logistics data set are comprehensively optimized through the target cluster center to generate the first cold chain logistics cost information; A first feature extraction module is used to extract short-term time correlation features from the target multi-dimensional cold chain logistics data set using a spatiotemporal graph convolutional network to obtain multiple short-term time correlation features; A second feature extraction module is used to perform time series data extraction on the target multidimensional cold chain logistics data set by using the SPARSS algorithm to obtain a plurality of time series cold chain logistics data, and to perform long-term time correlation feature extraction on the plurality of time series cold chain logistics data by using a long short-term memory neural network to obtain a plurality of long-term time correlation features; A first cost optimization module is used to perform feature weight analysis and feature fusion on the multiple short-term time correlation features and the multiple long-term time correlation features to obtain fused time correlation features, and input the fused time correlation features into a preset multi-layer perceptron model to optimize the cold chain logistics cargo storage cost and damage risk cost to obtain a first cold chain logistics cost optimization index; A second cost optimization module is used to optimize the cold chain logistics transportation task mode cost of the fused time-related features through a k-NN algorithm to obtain a second cold chain logistics cost optimization index; The strategy optimization module is used to optimize the cost control strategy of the first cold chain logistics cost information according to the first cold chain logistics cost optimization index and the second cold chain logistics cost optimization index through a non-dominated sorting genetic algorithm, and generate the second cold chain logistics cost information of the target cold chain logistics.

8. A cold chain logistics data management device, It is characterized in that The cold chain logistics data management device comprises: a memory and at least one processor, wherein instructions are stored in the memory; The at least one processor calls the instructions in the memory so that the cold chain logistics data management device executes the cold chain logistics data management method as described in any one of claims 1-6.

9. A computer-readable storage medium having instructions stored thereon, It is characterized in that When the instruction is executed by the processor, the cold chain logistics data management method as described in any one of claims 1-6 is implemented.

Images