Supply chain resilience assessment method and system based on artificial intelligence

Abstract

The present invention discloses a supply chain resilience assessment method and system based on artificial intelligence, and the method includes data collection, data preprocessing, time series data conversion, resilience assessment model construction and supply chain resilience assessment. The present invention relates to the field of supply chain resilience assessment technology, and specifically refers to a supply chain resilience assessment method and system based on artificial intelligence. The present invention obtains raw data for resilience assessment through data collection; adopts a data preprocessing method of data cleaning, data labeling, data encoding and data normalization; adopts a deer herd optimization algorithm to convert supply chain time series data into RGB image data for processing, and converts time series patterns into image spatial features, so that nonlinear correlations and dynamic changes that cannot be captured by time series analysis can be displayed; adopts the EfficientHRNet model to perform a method for supply chain resilience assessment, so as to better capture complex patterns, spatial relationships and multi-scale features in time series data.

Description

Technical Field

[0001] The present invention relates to the technical field of supply chain resilience assessment, and specifically to a supply chain resilience assessment method and system based on artificial intelligence. Background Art

[0002] Supply chain resilience assessment measures the ability of a supply chain to maintain normal operations, recover quickly, and adapt to changes when faced with risks and emergencies. Through assessment, companies can identify potential vulnerable links, optimize resource allocation and emergency response strategies, and improve the response speed and recovery capabilities of the supply chain, thereby reducing the impact of external shocks on the business and improving competitiveness.

[0003] However, traditional supply chain resilience assessment methods have technical problems such as difficulty in automatically capturing potential patterns and complex relationships in the data when faced with complex, high-dimensional and nonlinear supply chain time series data, and difficulty in effectively modeling the multi-scale characteristics and local dependencies of supply chain time series data; traditional supply chain resilience assessment methods have technical problems such as high computational complexity when faced with supply chain time series data with large data dimensions, and inability to effectively extract potential global and local features when analyzing multivariate and multi-scale data, resulting in unstable model performance. Summary of the invention

[0004] In view of the above situation, in order to overcome the defects of the prior art, the present invention provides a supply chain resilience assessment method and system based on artificial intelligence. In view of the technical problems that traditional supply chain resilience assessment methods have difficulty in automatically capturing potential patterns and complex relationships in the data when facing complex, high-dimensional and nonlinear supply chain time series data, and the multi-scale characteristics and local dependencies of supply chain time series data are difficult to effectively model, this solution creatively adopts a method based on the deer herd optimization algorithm to convert supply chain time series data into RGB image data for processing. On the basis of retaining feature information to the greatest extent, it can convert complex time series patterns into image space features, so that nonlinear correlations and dynamic changes that cannot be captured by traditional time series analysis methods can be displayed; in view of the technical problems that traditional supply chain resilience assessment methods have high computational complexity when facing supply chain time series data with large data dimensions, and cannot effectively extract potential global and local features in multivariate and multi-scale data analysis, resulting in unstable model performance, this solution creatively adopts the EfficientHRNet model to perform supply chain resilience assessment, which can better capture complex patterns, spatial relationships and multi-scale characteristics in time series data, and at the same time has higher computational efficiency and accuracy.

[0005] The technical solution adopted by the present invention is as follows: The supply chain resilience assessment method based on artificial intelligence provided by the present invention comprises the following steps:

[0006] Step S1: data collection;

[0007] Step S2: data preprocessing;

[0008] Step S3: time series data conversion;

[0009] Step S4: resilience assessment model construction;

[0010] Step S5: Supply chain resilience assessment.

[0011] Furthermore, in step S1, the data collection is used to collect the original data required for evaluating supply chain resilience, specifically, to obtain the original data set for resilience evaluation through data collection;

[0012] The resilience assessment original data set specifically includes supply chain structure data, supply chain operation data, external environment data, enterprise financial data and supply chain risk data. The supply chain structure data specifically includes the number of suppliers, supplier geographical locations, supplier cooperation models and supplier substitutability. The supply chain operation data specifically includes operation order data, production fluctuation data, production line redundancy and demand fluctuation data. The external environment data specifically includes logistics efficiency and market demand fluctuation data. The enterprise financial data specifically includes enterprise financial status data, enterprise inventory turnover rate and enterprise cash flow data. The supply chain risk data specifically includes the number of supply chain stops, supply chain recovery time and supply chain recovery cost.

[0013] Furthermore, in step S2, the data preprocessing is used to preprocess the collected raw data, and specifically includes the following steps:

[0014] Step S21: data cleaning, which is used to clean the original data, specifically to remove missing values ​​and duplicate values ​​in the original data set of the toughness assessment to obtain a roughly processed data set;

[0015] Step S22: data labeling, which is used to label the roughly processed data, specifically labeling the data in the roughly processed data set as inferior, good, and excellent, and using them as data labels to obtain a labeled data set;

[0016] Step S23: data encoding, which is used to encode the labeled data, specifically, using a one-hot encoding method to encode the labeled data set to obtain an encoded data set;

[0017] Step S24: data normalization, which is used to normalize the coded data, specifically using the minimum-maximum method to normalize the coded data set to obtain a preliminary data set for resilience assessment;

[0018] Step S25: performing preprocessing, specifically, preprocessing the original toughness assessment data set through data cleaning, data labeling, data encoding and data normalization to obtain a preliminary toughness assessment data set.

[0019] Further, in step S3, the time series data conversion is used to convert the time series data in the preliminary data into RGB image data, specifically converting the time series data in the preliminary data set of toughness assessment into Gramian horn field image data, and converting the Gramian horn field image data into RGB image data based on the deer population optimization algorithm;

[0020] The time series data conversion specifically includes the following steps:

[0021] Step S31: preliminary conversion, specifically converting the time series data in the preliminary toughness assessment data set into Gramian angle field image data, the steps comprising:

[0022] Step S311: Convert to polar coordinates using the following formula:

[0023] ;

[0024] In the formula, represents the value of the a-th time series data at time point t, Indicates the angle value converted from the value of the a-th time series data at time point t. represents the arccosine function, a represents the index of the time series data, and t represents the index of the time point of the time series data;

[0025] Step S312: construct the Gramian angular field matrix, the formula used is as follows:

[0026] ;

[0027] Where GAF represents the Gramian angular field matrix, whose elements are the similarities between the angles converted from the values ​​of the time series data at different time points. represents the Gramian angular field matrix Line The elements of the column are time series data in The value at the time point and The similarity between the angles converted from the values ​​at the time points, Indicates that the a-th time series data is in The angle value converted from the time point value, Indicates that the a-th time series data is in The angle value converted from the time point value;

[0028] Step S313: constructing Gramian angular field image data, specifically, processing the time series data in the preliminary toughness assessment data set through the conversion into polar coordinate form and the construction of the Gramian angular field matrix to obtain a Gramian angular field image data set;

[0029] Step S32: image data dimensionality reduction, which is used to perform data dimensionality reduction on the Gramian horn field image data, specifically reducing the dimensionality of the multi-channel image data in the Gramian horn field image data set into three-channel image data based on the deer population optimization algorithm, so as to facilitate further conversion into RGB image data;

[0030] The image data dimension reduction step comprises:

[0031] Step S321: Design channel mapping, the formula used is as follows:

[0032] ;

[0033] In the formula, Represents the R channel of RGB image data, represents the G channel of the RGB image data, represents the B channel of the RGB image data, and C represents the number of channels of the Gramian angular field image data. Indicates the mapping weight of the channel mapping of the b-th Gramian angular field image data to the R channel, Indicates the mapping weight of the channel mapping of the b-th Gramian angular field image data to the G channel, Indicates the mapping weight of the channel mapping of the b-th Gramian angular field image data to the B channel, represents the channel of the b-th Gramian angular field image data;

[0034] Step S322: Initialize the deer population optimization algorithm, the steps include:

[0035] Step S3221: Initialize the search space, specifically construct a deer unit set, and initialize the deer unit initial position matrix, the deer unit is used to represent the weight combination of the Gramian angular field image data mapped to the three channels, and its position in different dimensions represents the weight value of the different channels of the Gramian angular field image data mapped to the three channels. The deer unit initial position matrix is ​​expressed as follows:

[0036] ;

[0037] In the formula, represents the initial position matrix of the deer unit, represents the position of the mth deer unit in the nth dimension, represents the position of the Mth deer unit in the Nth dimension, represents the position of the first deer unit, represents the position of the second deer unit, represents the position of the Mth deer unit, M represents the total number of deer units, and N represents the total number of deer unit dimensions;

[0038] Step S3222: Initialize the velocity matrix, which is used to initialize the initial velocity matrix of the deer unit. The initial velocity matrix of the deer unit is expressed as follows:

[0039] ;

[0040] In the formula, represents the initial velocity matrix of the deer unit, represents the speed of the mth deer unit in the nth dimension, represents the speed of the Mth deer unit in the Nth dimension, represents the speed of the first deer unit, represents the speed of the second deer unit, represents the speed of the Mth deer unit;

[0041] Step S3223: determine the fitness function, specifically, use the sum of the variances of the three channels after the Gramian angular field image data is mapped into three channels as the fitness of the deer unit, the fitness function, the formula used is as follows:

[0042] ;

[0043] In the formula, represents the position of the mth deer unit, represents the fitness function, represents the variance calculation function;

[0044] Step S3224: determining an iteration termination condition, wherein the iteration termination condition specifically includes that the fitness of the deer unit is greater than a set threshold and reaches a maximum number of iterations;

[0045] Step S323: global search, the formula used is as follows:

[0046] ;

[0047] In the formula, represents the velocity of the mth deer unit at the dt+1th iteration, represents the inertia weight, represents the velocity of the mth deer unit at the dtth iteration, represents the weight used to control the deer unit to learn its own optimal solution, represents the optimal solution of the mth deer unit itself, represents the position of the mth deer unit at the dtth iteration, represents the weight used to control the deer unit to learn the global optimal solution, represents the global optimal solution at the dtth iteration, represents the position of the mth deer unit at the dt+1th iteration;

[0048] Step S324: local search, the steps include:

[0049] Step S3241: constructing male deer units and female deer units, specifically constructing a male deer unit set and a female deer unit set, sorting the deer units from high to low according to fitness, setting the first, second and third ranked deer units as male deer units, and setting the other deer units as female deer units;

[0050] Step S3242: updating the territory range, specifically dividing the territory range for each stag unit. If the fitness of the stag unit increases after the global search and update position, the territory range is expanded. If the fitness of the stag unit does not increase after the global search and update position, the territory range is reduced, and the position of the stag unit with the largest fitness within the territory range is set as the local optimal position of the territory.

[0051] The formula used to expand the territory is as follows:

[0052] ;

[0053] The formula used to reduce the scope of the territory is as follows:

[0054] ;

[0055] In the formula, Indicates that at the dt+1th iteration The territory of a stag unit, At the dtth iteration The territory of a stag unit, Indicates the step length of territory adjustment;

[0056] Step S3243: The position of the female deer unit within the territory is updated, and the formula used is as follows:

[0057] ;

[0058] In the formula, Indicates the territory at the dt+1th iteration. The location of the doe unit, Indicates the territory at the dtth iteration. The location of the doe unit, represents the weight of the female deer unit within the territory close to the local optimal position of the territory, represents the local optimal position of the territory at the dtth iteration;

[0059] Step S3244: The position of the female deer unit outside the territory is updated, and the formula used is as follows:

[0060] ;

[0061] Where Da represents the dynamic adjustment factor used to control the exploration ability of the female deer unit, and Dw represents the weight of the influence of the dynamic adjustment factor. Indicates that the dt+1th iteration is outside the territory. The position of the female deer unit approaching the first male deer unit, represents the position of the first stag unit, represents a random number in the range [0,1], Indicates that the dtth iteration is outside the territory. The location of the doe unit, Indicates that the dt+1th iteration is outside the territory. The position of the female deer unit after it approaches the second male deer unit. represents the position of the second stag unit, represents a random number in the range [0,1], Indicates that the dt+1th iteration is outside the territory. The position of the female deer unit approaching the third male deer unit, indicates the location of the third stag unit, represents a random number in the range [0,1], Indicates that the dt+1th iteration is outside the territory. The location of the doe unit;

[0062] Step S325: iterative updating, specifically, continuously iteratively updating until the iteration termination condition is reached, and the optimized Gramian angle field image data is mapped into a weighted combination of three channels;

[0063] Step S33: data scale conversion, specifically mapping the data in the Gramian angular field image dataset to a weighted combination of three channels based on the optimized Gramian angular field image data, reducing the dimension to three channels, and then performing scale conversion to linearly map the data to a range of [0, 255] to obtain a three-channel image dataset;

[0064] Step S34: performing data conversion, for converting the image data in the three-channel image data set into RGB image data, specifically, performing data conversion based on a rule that the values ​​of the data in the three-channel image data set correspond to the brightness values ​​of the converted RGB image, to obtain the RGB image data set;

[0065] Step S35: data set segmentation, specifically, performing data segmentation on the RGB image data set to obtain a toughness assessment training set and a toughness assessment test set.

[0066] Further, in step S4, the resilience assessment model is constructed to construct a model required for assessing supply chain resilience, specifically, an EfficientHRNet model is constructed and used as a resilience assessment model;

[0067] The construction of the resilience assessment model specifically includes the following steps:

[0068] Step S41: multi-resolution feature extraction, the formula used is as follows:

[0069] ;

[0070] In the formula, represents the features of the i-th resolution level, represents the convolution operation function, Represents the features of the i+1th resolution level;

[0071] Step S42: Multi-resolution feature fusion, the formula used is as follows:

[0072] ;

[0073] In the formula, represents the fusion feature of the i-th resolution level, Rs represents the set of all resolution levels, Represents the weight of the feature upsampling at the j-th resolution level participating in the fusion feature calculation process at the i-th resolution level. represents the upsampling function, Represents the features of the jth resolution level;

[0074] Step S43: cross-resolution feature fusion, the formula used is as follows:

[0075] ;

[0076] In the formula, represents the cross-fusion features of the i-th resolution level, Represents the weight of the feature downsampling of the j-th resolution level participating in the cross-fusion feature calculation process of the i-th resolution level, represents the downsampling function;

[0077] Step S44: Calculate the final fusion feature, the formula used is as follows:

[0078] ;

[0079] In the formula, represents the final fusion feature;

[0080] Step S45: Calculate the model output using the following formula:

[0081] ;

[0082] In the formula, Represents the evaluation result of the model output, represents the softmax function, represents the output weight of the model, Represents the output bias term of the model;

[0083] Step S46: construct and train the model, specifically, construct the EfficientHRNet model through the multi-resolution feature extraction, the multi-resolution feature fusion, the cross-resolution feature fusion, the calculation of the final fusion feature and the calculation model output, and train the model based on the resilience assessment training set, verify the model performance based on the resilience assessment test set, select the cross entropy loss function as the model loss function, and obtain the EfficientHRNet model as the resilience assessment model.

[0084] Furthermore, in step S5, the supply chain resilience assessment is specifically to assess the supply chain resilience through the resilience assessment model to obtain supply chain resilience level reference data, and based on the supply chain resilience level reference data, comprehensively assess the resilience level of the supply chain.

[0085] The artificial intelligence-based supply chain resilience assessment system provided by the present invention includes a data acquisition module, a data preprocessing module, a time series data conversion module, a resilience assessment model building module and a supply chain resilience assessment module;

[0086] The data acquisition module is used for data acquisition, and obtains a raw data set for toughness assessment through data acquisition, and sends the raw data set for toughness assessment to a data preprocessing module;

[0087] The data preprocessing module is used for data preprocessing, obtaining a preliminary data set for resilience assessment through data preprocessing, and sending the preliminary data set for resilience assessment to the time series data conversion module;

[0088] The time series data conversion module is used for time series data conversion, by converting the time series data into RGB image data and performing data set segmentation to obtain a toughness assessment training set and a toughness assessment test set, and sending the toughness assessment training set and the toughness assessment test set to the toughness assessment model construction module;

[0089] The resilience assessment model building module is used to build a model required for evaluating supply chain resilience, obtain a resilience assessment model by building an EfficientHRNet model, and send the resilience assessment model to the supply chain resilience assessment module;

[0090] The supply chain resilience assessment module is used to assess supply chain resilience. By adopting the resilience assessment model to conduct supply chain resilience assessment, reference data of supply chain resilience level is obtained.

[0091] The beneficial effects achieved by the present invention using the above scheme are as follows:

[0092] (1) In view of the technical problems that traditional supply chain resilience assessment methods have in the face of complex, high-dimensional and nonlinear supply chain time series data, it is difficult to automatically capture the potential patterns and complex relationships in the data, and the multi-scale characteristics and local dependencies of supply chain time series data are difficult to effectively model, this solution creatively adopts a method based on the deer herd optimization algorithm to convert supply chain time series data into RGB image data for processing. On the basis of retaining feature information to the greatest extent, it can convert complex time series patterns into image space features, so that nonlinear correlations and dynamic changes that cannot be captured by traditional time series analysis methods can be displayed.

[0093] (2) Traditional supply chain resilience assessment methods have high computational complexity when faced with supply chain time series data with large data dimensions, and are unable to effectively extract potential global and local features when analyzing multivariate and multi-scale data, resulting in unstable model performance. This solution creatively uses the EfficientHRNet model to perform supply chain resilience assessment, which can better capture the complex patterns, spatial relationships, and multi-scale features in time series data, while having higher computational efficiency and accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0094] Figure 1 A schematic diagram of the flow of the supply chain resilience assessment method based on artificial intelligence provided by the present invention;

[0095] Figure 2 A schematic diagram of a module of an artificial intelligence-based supply chain resilience assessment system provided by the present invention;

[0096] Figure 3 It is a schematic diagram of the process of time series data conversion in step S3;

[0097] Figure 4 This is a schematic diagram of the process of reducing the dimension of image data in step S32;

[0098] Figure 5 Schematic diagram of the process for building the toughness assessment model in step S4.

[0099] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention and do not constitute a limitation of the present invention. DETAILED DESCRIPTION

[0100] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments; based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0101] In the description of the present invention, it should be understood that terms such as “upper”, “lower”, “front”, “back”, “left”, “right”, “top”, “bottom”, “inside” and “outside” indicating directions or positional relationships are based on the directions or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific direction, be constructed and operated in a specific direction, and therefore should not be understood as limiting the present invention.

[0102] Example 1, see Figure 1 The technical solution adopted by the present invention is as follows: The supply chain resilience assessment method based on artificial intelligence provided by the present invention comprises the following steps:

[0103] Step S1: data collection;

[0104] Step S2: data preprocessing;

[0105] Step S3: time series data conversion;

[0106] Step S4: resilience assessment model construction;

[0107] Step S5: Supply chain resilience assessment.

[0108] Example 2, see Figure 1 and Figure 2 In step S1, the data collection is used to collect the original data required for evaluating supply chain resilience, specifically, to obtain the original data set for resilience evaluation through data collection;

[0109] The resilience assessment original data set specifically includes supply chain structure data, supply chain operation data, external environment data, enterprise financial data and supply chain risk data. The supply chain structure data specifically includes the number of suppliers, supplier geographical locations, supplier cooperation models and supplier substitutability. The supply chain operation data specifically includes operation order data, production fluctuation data, production line redundancy and demand fluctuation data. The external environment data specifically includes logistics efficiency and market demand fluctuation data. The enterprise financial data specifically includes enterprise financial status data, enterprise inventory turnover rate and enterprise cash flow data. The supply chain risk data specifically includes the number of supply chain stops, supply chain recovery time and supply chain recovery cost.

[0110] Example 3, see Figure 1 and Figure 2 This embodiment is based on the above embodiment. In step S2, the data preprocessing is used to preprocess the collected raw data, and specifically includes the following steps:

[0111] Step S21: data cleaning, which is used to clean the original data, specifically to remove missing values ​​and duplicate values ​​in the original data set of the toughness assessment to obtain a roughly processed data set;

[0112] Step S22: data labeling, which is used to label the roughly processed data, specifically labeling the data in the roughly processed data set as inferior, good, and excellent, and using them as data labels to obtain a labeled data set;

[0113] Step S23: data encoding, which is used to encode the labeled data, specifically, using a one-hot encoding method to encode the labeled data set to obtain an encoded data set;

[0114] Step S24: data normalization, which is used to normalize the coded data, specifically using the minimum-maximum method to normalize the coded data set to obtain a preliminary data set for resilience assessment;

[0115] Step S25: performing preprocessing, specifically, preprocessing the original toughness assessment data set through data cleaning, data labeling, data encoding and data normalization to obtain a preliminary toughness assessment data set.

[0116] Example 4, see Figure 1 , Figure 2 , Figure 3 and Figure 4, this embodiment is based on the above embodiment, in step S3, the time series data conversion is used to convert the time series data in the preliminary data into RGB image data, specifically converting the time series data in the preliminary data set of toughness assessment into Gramian horn field image data, and converting the Gramian horn field image data into RGB image data based on the deer herd optimization algorithm;

[0117] The time series data conversion specifically includes the following steps:

[0118] Step S31: preliminary conversion, specifically converting the time series data in the preliminary toughness assessment data set into Gramian angle field image data, the steps comprising:

[0119] Step S311: Convert to polar coordinates using the following formula:

[0120] ;

[0121] In the formula, represents the value of the a-th time series data at time point t, Indicates the angle value converted from the value of the a-th time series data at time point t. represents the arccosine function, a represents the index of the time series data, and t represents the index of the time point of the time series data;

[0122] Step S312: construct the Gramian angular field matrix, the formula used is as follows:

[0123] ;

[0124] Where GAF represents the Gramian angular field matrix, whose elements are the similarities between the angles converted from the values ​​of the time series data at different time points. represents the Gramian angular field matrix Line The elements of the column are time series data in The value at the time point and The similarity between the angles converted from the values ​​at the time points, Indicates that the a-th time series data is in The angle value converted from the time point value, Indicates that the a-th time series data is in The angle value converted from the time point value;

[0125] Step S313: constructing Gramian angular field image data, specifically, processing the time series data in the preliminary toughness assessment data set through the conversion into polar coordinate form and the construction of the Gramian angular field matrix to obtain a Gramian angular field image data set;

[0126] Step S32: image data dimensionality reduction, which is used to perform data dimensionality reduction on the Gramian horn field image data, specifically reducing the dimensionality of the multi-channel image data in the Gramian horn field image data set into three-channel image data based on the deer population optimization algorithm, so as to facilitate further conversion into RGB image data;

[0127] The image data dimension reduction step comprises:

[0128] Step S321: Design channel mapping, the formula used is as follows:

[0129] ;

[0130] In the formula, Represents the R channel of RGB image data, represents the G channel of the RGB image data, represents the B channel of the RGB image data, and C represents the number of channels of the Gramian angular field image data. Indicates the mapping weight of the channel mapping of the b-th Gramian angular field image data to the R channel, Indicates the mapping weight of the channel mapping of the b-th Gramian angular field image data to the G channel, Indicates the mapping weight of the channel mapping of the b-th Gramian angular field image data to the B channel, represents the channel of the b-th Gramian angular field image data;

[0131] Step S322: Initialize the deer population optimization algorithm, the steps include:

[0132] Step S3221: Initialize the search space, specifically construct a deer unit set, and initialize the deer unit initial position matrix, the deer unit is used to represent the weight combination of the Gramian angular field image data mapped to the three channels, and its position in different dimensions represents the weight value of the different channels of the Gramian angular field image data mapped to the three channels. The deer unit initial position matrix is ​​expressed as follows:

[0133] ;

[0134] In the formula, represents the initial position matrix of the deer unit, represents the position of the mth deer unit in the nth dimension, represents the position of the Mth deer unit in the Nth dimension, represents the position of the first deer unit, represents the position of the second deer unit, represents the position of the Mth deer unit, M represents the total number of deer units, and N represents the total number of deer unit dimensions;

[0135] Step S3222: Initialize the velocity matrix, which is used to initialize the initial velocity matrix of the deer unit. The initial velocity matrix of the deer unit is expressed as follows:

[0136] ;

[0137] In the formula, represents the initial velocity matrix of the deer unit, represents the speed of the mth deer unit in the nth dimension, represents the speed of the Mth deer unit in the Nth dimension, represents the speed of the first deer unit, represents the speed of the second deer unit, represents the speed of the Mth deer unit;

[0138] Step S3223: determine the fitness function, specifically, use the sum of the variances of the three channels after the Gramian angular field image data is mapped into three channels as the fitness of the deer unit, the fitness function, the formula used is as follows:

[0139] ;

[0140] In the formula, represents the position of the mth deer unit, represents the fitness function, represents the variance calculation function;

[0141] Step S3224: determining an iteration termination condition, wherein the iteration termination condition specifically includes that the fitness of the deer unit is greater than a set threshold and reaches a maximum number of iterations;

[0142] Step S323: global search, the formula used is as follows:

[0143] ;

[0144] In the formula, represents the velocity of the mth deer unit at the dt+1th iteration, represents the inertia weight, represents the velocity of the mth deer unit at the dtth iteration, represents the weight used to control the deer unit to learn its own optimal solution, represents the optimal solution of the mth deer unit itself, represents the position of the mth deer unit at the dtth iteration, represents the weight used to control the deer unit to learn the global optimal solution, represents the global optimal solution at the dtth iteration, represents the position of the mth deer unit at the dt+1th iteration;

[0145] Step S324: local search, the steps include:

[0146] Step S3241: constructing male deer units and female deer units, specifically constructing a male deer unit set and a female deer unit set, sorting the deer units from high to low according to fitness, setting the first, second and third ranked deer units as male deer units, and setting the other deer units as female deer units;

[0147] Step S3242: updating the territory range, specifically dividing the territory range for each stag unit. If the fitness of the stag unit increases after the global search and update position, the territory range is expanded. If the fitness of the stag unit does not increase after the global search and update position, the territory range is reduced, and the position of the stag unit with the largest fitness within the territory range is set as the local optimal position of the territory.

[0148] The formula used to expand the territory is as follows:

[0149] ;

[0150] The formula used to reduce the scope of the territory is as follows:

[0151] ;

[0152] In the formula, Indicates that at the dt+1th iteration The territory of a stag unit, At the dtth iteration The territory of a stag unit, Indicates the step length of territory adjustment;

[0153] Step S3243: The position of the female deer unit within the territory is updated, and the formula used is as follows:

[0154] ;

[0155] In the formula, Indicates the territory at the dt+1th iteration. The location of the doe unit, Indicates the territory at the dtth iteration. The location of the doe unit, represents the weight of the female deer unit within the territory close to the local optimal position of the territory, represents the local optimal position of the territory at the dtth iteration;

[0156] Step S3244: The position of the female deer unit outside the territory is updated, and the formula used is as follows:

[0157] ;

[0158] Where Da represents the dynamic adjustment factor used to control the exploration ability of the female deer unit, and Dw represents the weight of the influence of the dynamic adjustment factor. Indicates that the dt+1th iteration is outside the territory. The position of the female deer unit approaching the first male deer unit, represents the position of the first stag unit, represents a random number in the range [0,1], Indicates that the dtth iteration is outside the territory. The location of the doe unit, Indicates that the dt+1th iteration is outside the territory. The position of the female deer unit after it approaches the second male deer unit. represents the position of the second stag unit, represents a random number in the range [0,1], Indicates that the dt+1th iteration is outside the territory. The position of the female deer unit approaching the third male deer unit, indicates the location of the third stag unit, represents a random number in the range [0,1], Indicates that the dt+1th iteration is outside the territory. The location of the doe unit;

[0159] Step S325: iterative updating, specifically, continuously iteratively updating until the iteration termination condition is reached, and the optimized Gramian angle field image data is mapped into a weighted combination of three channels;

[0160] Step S33: data scale conversion, specifically mapping the data in the Gramian angular field image dataset to a weighted combination of three channels based on the optimized Gramian angular field image data, reducing the dimension to three channels, and then performing scale conversion to linearly map the data to a range of [0, 255] to obtain a three-channel image dataset;

[0161] Step S34: performing data conversion, for converting the image data in the three-channel image data set into RGB image data, specifically, performing data conversion based on a rule that the values ​​of the data in the three-channel image data set correspond to the brightness values ​​of the converted RGB image, to obtain the RGB image data set;

[0162] Step S35: data set segmentation, specifically, performing data segmentation on the RGB image data set to obtain a toughness assessment training set and a toughness assessment test set.

[0163] By performing the above operations, in view of the technical problems that traditional supply chain resilience assessment methods have difficulty in automatically capturing potential patterns and complex relationships in the data when faced with complex, high-dimensional and nonlinear supply chain time series data, and the multi-scale characteristics and local dependencies of supply chain time series data are difficult to effectively model, this solution creatively adopts a method based on the deer herd optimization algorithm to convert supply chain time series data into RGB image data for processing. On the basis of retaining feature information to the greatest extent, it can convert complex time series patterns into image space features, so that nonlinear correlations and dynamic changes that cannot be captured by traditional time series analysis methods can be displayed.

[0164] Example 5, see Figure 1 , Figure 2 and Figure 5 , this embodiment is based on the above embodiment. In step S4, the resilience assessment model is constructed to construct a model required for assessing supply chain resilience, specifically, an EfficientHRNet model is constructed and used as a resilience assessment model;

[0165] The construction of the resilience assessment model specifically includes the following steps:

[0166] Step S41: multi-resolution feature extraction, the formula used is as follows:

[0167] ;

[0168] In the formula, represents the features of the i-th resolution level, represents the convolution operation function, Represents the features of the i+1th resolution level;

[0169] Step S42: Multi-resolution feature fusion, the formula used is as follows:

[0170] ;

[0171] In the formula, represents the fusion feature of the i-th resolution level, Rs represents the set of all resolution levels, Represents the weight of the feature upsampling at the j-th resolution level participating in the fusion feature calculation process at the i-th resolution level. represents the upsampling function, Represents the features of the jth resolution level;

[0172] Step S43: cross-resolution feature fusion, the formula used is as follows:

[0173] ;

[0174] In the formula, represents the cross-fusion features of the i-th resolution level, Represents the weight of the feature downsampling of the j-th resolution level participating in the cross-fusion feature calculation process of the i-th resolution level, represents the downsampling function;

[0175] Step S44: Calculate the final fusion feature, the formula used is as follows:

[0176] ;

[0177] In the formula, represents the final fusion feature;

[0178] Step S45: Calculate the model output using the following formula:

[0179] ;

[0180] In the formula, Represents the evaluation result of the model output, represents the softmax function, represents the output weight of the model, Represents the output bias term of the model;

[0181] Step S46: construct and train the model, specifically, construct the EfficientHRNet model through the multi-resolution feature extraction, the multi-resolution feature fusion, the cross-resolution feature fusion, the calculation of the final fusion feature and the calculation model output, and train the model based on the resilience assessment training set, verify the model performance based on the resilience assessment test set, select the cross entropy loss function as the model loss function, and obtain the EfficientHRNet model as the resilience assessment model.

[0182] By performing the above operations, this solution creatively uses the EfficientHRNet model to perform supply chain resilience assessment, which can better capture the complex patterns, spatial relationships and multi-scale features in time series data, while having higher computational efficiency and accuracy. This method solves the technical problems that traditional supply chain resilience assessment methods have high computational complexity when facing supply chain time series data with large data dimensions, and cannot effectively extract potential global and local features when analyzing multivariate and multi-scale data, resulting in unstable model performance.

[0183] Example 6, see Figure 1 and Figure 2This embodiment is based on the above embodiment. In step S5, the supply chain resilience assessment is specifically to assess the supply chain resilience through the resilience assessment model to obtain supply chain resilience level reference data, and based on the supply chain resilience level reference data, comprehensively assess the resilience level of the supply chain.

[0184] Embodiment 7, see Figure 1 and Figure 2 , this embodiment is based on the above embodiment, and the supply chain resilience assessment system based on artificial intelligence provided by the present invention includes a data acquisition module, a data preprocessing module, a time series data conversion module, a resilience assessment model construction module and a supply chain resilience assessment module;

[0185] The data acquisition module is used for data acquisition, and obtains a raw data set for toughness assessment through data acquisition, and sends the raw data set for toughness assessment to a data preprocessing module;

[0186] The data preprocessing module is used for data preprocessing, obtaining a preliminary data set for resilience assessment through data preprocessing, and sending the preliminary data set for resilience assessment to the time series data conversion module;

[0187] The time series data conversion module is used for time series data conversion, by converting the time series data into RGB image data and performing data set segmentation to obtain a toughness assessment training set and a toughness assessment test set, and sending the toughness assessment training set and the toughness assessment test set to the toughness assessment model construction module;

[0188] The resilience assessment model building module is used to build a model required for evaluating supply chain resilience, obtain a resilience assessment model by building an EfficientHRNet model, and send the resilience assessment model to the supply chain resilience assessment module;

[0189] The supply chain resilience assessment module is used to assess supply chain resilience. By adopting the resilience assessment model to conduct supply chain resilience assessment, reference data of supply chain resilience level is obtained.

[0190] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0191] While the embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that many changes, modifications, substitutions and variations can be made to the embodiments without departing from the principles and spirit of the invention.

[0192] The present invention and its embodiments are described above, and such description is not restrictive. The drawings show only one embodiment of the present invention, and the actual structure is not limited thereto. In short, if ordinary technicians in the field are inspired by it, without departing from the purpose of the invention, they can design a structure and embodiment similar to the technical solution without creativity, which should belong to the protection scope of the present invention.

Claims

1. Supply chain resilience assessment method based on artificial intelligence, characterized by: The method comprises the following steps: Step S1: Data collection: Through data collection, an original data set for resilience assessment is obtained. The original data set for resilience assessment specifically includes supply chain structure data, supply chain operation data, external environment data, enterprise financial data and supply chain risk data. The supply chain structure data specifically includes the number of suppliers, supplier geographical locations, supplier cooperation models and supplier substitutability. The supply chain operation data specifically includes operation order data, output fluctuation data, production line redundancy and demand fluctuation data. The external environment data specifically includes logistics efficiency and market demand fluctuation data. The enterprise financial data specifically includes enterprise financial status data, enterprise inventory turnover rate and enterprise capital flow data. The supply chain risk data specifically includes the number of supply chain stops, supply chain recovery time and supply chain recovery cost. Step S2: data preprocessing, preprocessing the collected raw data for toughness assessment to obtain a preliminary data set for toughness assessment; Step S3: time series data conversion, used to convert the time series data in the preliminary data into RGB image data, specifically converting the time series data in the preliminary data set of toughness assessment into Gramian horn field image data, and converting the Gramian horn field image data into RGB image data based on the deer population optimization algorithm; Step S4: constructing a resilience assessment model, specifically constructing an EfficientHRNet model as a resilience assessment model; Step S5: supply chain resilience assessment, specifically assessing the supply chain resilience through the resilience assessment model to obtain supply chain resilience level reference data; The construction of the resilience assessment model specifically includes the following steps: Step S41: multi-resolution feature extraction; Step S42: multi-resolution feature fusion; Step S43: cross-resolution feature fusion; Step S44: Calculate the final fusion feature; Step S45: Calculate model output; Step S46: construct and train the model, specifically, construct the EfficientHRNet model through the multi-resolution feature extraction, the multi-resolution feature fusion, the cross-resolution feature fusion, the calculation of the final fusion feature and the calculation model output, and train the model based on the resilience assessment training set, verify the model performance based on the resilience assessment test set, select the cross entropy loss function as the model loss function, and obtain the EfficientHRNet model as the resilience assessment model.

2. The supply chain resilience assessment method based on artificial intelligence according to claim 1 is characterized by: The time series data conversion specifically includes the following steps: Step S31: preliminary conversion, specifically converting the time series data in the preliminary toughness assessment data set into Gramian angle field image data, the steps comprising: Step S311: converting to polar coordinate form; Step S312: constructing a Gramian angular field matrix; Step S313: constructing Gramian angular field image data, specifically, processing the time series data in the preliminary toughness assessment data set through the conversion into polar coordinate form and the construction of the Gramian angular field matrix to obtain a Gramian angular field image data set; Step S32: image data dimensionality reduction, which is used to perform data dimensionality reduction on the Gramian horn field image data, specifically reducing the dimensionality of the multi-channel image data in the Gramian horn field image data set into three-channel image data based on the deer population optimization algorithm, so as to facilitate further conversion into RGB image data; Step S33: data scale conversion, specifically mapping the data in the Gramian angular field image dataset to a weighted combination of three channels based on the optimized Gramian angular field image data, reducing the dimension to three channels, and then performing scale conversion to linearly map the data to a range of [0, 255] to obtain a three-channel image dataset; Step S34: performing data conversion, for converting the image data in the three-channel image data set into RGB image data, specifically, performing data conversion based on a rule that the values ​​of the data in the three-channel image data set correspond to the brightness values ​​of the converted RGB image, to obtain the RGB image data set; Step S35: data set segmentation, specifically, performing data segmentation on the RGB image data set to obtain a toughness assessment training set and a toughness assessment test set.

3. The supply chain resilience assessment method based on artificial intelligence according to claim 2 is characterized by: The image data dimension reduction step comprises: Step S321: design channel mapping; Step S322: Initialize the deer population optimization algorithm, the steps include: Step S3221: Initializing the search space, specifically constructing a deer unit set and initializing a deer unit initial position matrix, wherein the deer unit is used to represent the weight combination of the Gramian angular field image data mapped to three channels, and its position in different dimensions represents the weight value of different channels of the Gramian angular field image data mapped to the three channels; Step S3222: Initialize the velocity matrix; Step S3223: determining a fitness function, specifically using the sum of the variances of the three channels after the Gramian angular field image data is mapped into three channels as the fitness of the deer unit; Step S3224: determining an iteration termination condition, wherein the iteration termination condition specifically includes that the fitness of the deer unit is greater than a set threshold and reaches a maximum number of iterations; Step S323: global search; Step S324: local search, the steps include: Step S3241: constructing male deer units and female deer units, specifically constructing a male deer unit set and a female deer unit set, sorting the deer units from high to low according to fitness, setting the first, second and third ranked deer units as male deer units, and setting the other deer units as female deer units; Step S3242: updating the territory range, specifically dividing the territory range for each stag unit. If the fitness of the stag unit increases after the global search and update position, the territory range is expanded. If the fitness of the stag unit does not increase after the global search and update position, the territory range is reduced, and the position of the stag unit with the largest fitness within the territory range is set as the local optimal position of the territory. Step S3243: updating the position of the female deer unit within the territory; Step S3244: updating the position of the female deer unit outside the territory; Step S325: iterative updating, specifically, continuously iteratively updating until the iteration termination condition is reached, and the optimized Gramian angular field image data is mapped into a weighted combination of three channels.

4. The supply chain resilience assessment method based on artificial intelligence according to claim 1 is characterized by: In step S2, the data preprocessing is used to preprocess the collected raw data, and specifically includes the following steps: Step S21: data cleaning, which is used to clean the original data, specifically to remove missing values ​​and duplicate values ​​in the original data set of the toughness assessment to obtain a roughly processed data set; Step S22: data labeling, which is used to label the roughly processed data, specifically labeling the data in the roughly processed data set as inferior, good, and excellent, and using them as data labels to obtain a labeled data set; Step S23: data encoding, which is used to encode the labeled data, specifically, using a one-hot encoding method to encode the labeled data set to obtain an encoded data set; Step S24: data normalization, which is used to normalize the coded data, specifically using the minimum-maximum method to normalize the coded data set to obtain a preliminary data set for resilience assessment; Step S25: performing preprocessing, specifically preprocessing the original toughness assessment data set through data cleaning, data labeling, data encoding and data normalization to obtain a preliminary toughness assessment data set.

5. The supply chain resilience assessment method based on artificial intelligence according to claim 1 is characterized by: In step S5, the supply chain resilience assessment is specifically to assess the supply chain resilience through the resilience assessment model to obtain supply chain resilience level reference data, and based on the supply chain resilience level reference data, comprehensively assess the resilience level of the supply chain.

6. An artificial intelligence-based supply chain resilience assessment system, used to implement the artificial intelligence-based supply chain resilience assessment method as described in any one of claims 1 to 5, characterized in that: It includes data acquisition module, data preprocessing module, time series data conversion module, resilience assessment model building module and supply chain resilience assessment module.

7. The supply chain resilience assessment system based on artificial intelligence according to claim 6, characterized in that: The data acquisition module is used for data acquisition, and obtains a raw data set for toughness assessment through data acquisition, and sends the raw data set for toughness assessment to a data preprocessing module; The data preprocessing module is used for data preprocessing, obtaining a preliminary data set for resilience assessment through data preprocessing, and sending the preliminary data set for resilience assessment to the time series data conversion module; The time series data conversion module is used for time series data conversion, by converting the time series data into RGB image data and performing data set segmentation to obtain a toughness assessment training set and a toughness assessment test set, and sending the toughness assessment training set and the toughness assessment test set to the toughness assessment model construction module; The resilience assessment model building module is used to build a model required for evaluating supply chain resilience, obtain a resilience assessment model by building an EfficientHRNet model, and send the resilience assessment model to the supply chain resilience assessment module; The supply chain resilience assessment module is used to assess supply chain resilience. By adopting the resilience assessment model to conduct supply chain resilience assessment, reference data of supply chain resilience level is obtained.

Images