A method and apparatus for the delivery of pre-prepared beef dishes

Abstract

The application discloses a delivery method and equipment for prefabricated beef dishes, and the method comprises the following steps: obtaining quality control characteristic data of prefabricated beef dishes corresponding to each customer order, delivery time limit requirements, and global delivery scene characteristic data of all customer orders; for each customer order, the weight of the quality control characteristic vector of the prefabricated beef dishes is calculated by the analytic hierarchy process, and an initial weight coefficient set corresponding to the quality control characteristic vector of the customer order is obtained; the initial weight coefficient set of the quality control characteristic vector of the customer order is dynamically corrected by the kernel density probability algorithm, and a dynamic weight coefficient set of the quality control characteristic vector of the customer order is obtained; based on the dynamic weight coefficient set, the quality control characteristic weighted value of each customer order is calculated; and based on the delivery time limit requirements and the quality control characteristic weighted value of the prefabricated beef dishes of each customer order, a delivery route is planned to complete the delivery of each prefabricated beef dish. The method improves the overall delivery quality and quality control safety of the prefabricated beef dishes.

Description

Technical Field

[0001] This application relates to the field of route planning and optimization technology, and in particular to a delivery method and equipment for pre-prepared beef dishes. Background Technology

[0002] With the rapid development of the prepared food industry, the market demand for prepared beef dishes continues to grow. Improving the delivery service level of prepared beef dishes is an important prerequisite for promoting the sustainable development of the prepared beef dish industry.

[0003] Currently, the delivery scheduling of pre-cooked beef dishes mostly adopts traditional planning methods. The general address navigation model only plans the delivery route based on the distance of the delivery address, and in some scenarios, the delivery priority is divided according to customer requirements.

[0004] The above planning methods all have obvious limitations. Neither of them takes into account the core characteristics of pre-prepared food combination delivery, such as temperature control priority, food shelf life, and food mix types, to optimize the delivery route. This makes it easy for non-optimal detours to occur, which greatly increases the exposure time of pre-prepared beef dishes to room temperature during delivery, further aggravates quality loss, and ultimately seriously reduces the overall delivery quality of pre-prepared beef dishes. Summary of the Invention

[0005] This application provides a delivery method and equipment for pre-prepared beef dishes, which addresses the problem that the delivery route for pre-prepared beef dishes is not optimized in a coordinated manner, taking into account core characteristics such as temperature control priority, shelf life, and mixed dish types, which seriously reduces the overall delivery quality of pre-prepared beef dishes.

[0006] The embodiments of this application adopt the following technical solutions: On one hand, embodiments of this application provide a method for delivering pre-cooked beef dishes, the method comprising: The system acquires quality control characteristic data and delivery time limit requirements for each customer's pre-prepared beef dish, as well as global delivery scenario characteristic data for all customer orders. The quality control data includes at least temperature control requirements, shelf life data, and mixed dish types for each pre-prepared beef dish. The global delivery scenario characteristic data includes: ambient temperature of the delivery area, operating status of transportation equipment, and delivery time period type. Standardized vector transformation processing is performed on the quality control feature data of pre-prepared beef dishes corresponding to each customer order and the global delivery scenario feature data to obtain the quality control feature vector and scenario feature vector of pre-prepared beef dishes corresponding to each customer order. For each customer order, the weights of the quality control feature vector of the pre-prepared beef dishes are calculated using the analytic hierarchy process (AHP) to obtain the initial set of weight coefficients corresponding to the quality control feature vector of that customer order. The real-time contribution of each scene feature in the scene feature vector to each quality control feature of the customer order is calculated by the kernel density probability algorithm. The initial weight coefficient set of the quality control feature vector of the customer order is dynamically corrected by the real-time contribution as a correction factor to obtain the dynamic weight coefficient set of the quality control feature vector of the customer order. Based on the dynamic weight coefficient set, the quality control feature vectors of each customer order are weighted and summed dimension by dimension to obtain the quality control feature weighted value of each customer order. The quality control feature weighted value is used as the comprehensive priority value of the pre-made beef dishes in each customer order. Delivery routes are planned based on the aforementioned delivery time requirements and the overall priority of each customer's pre-prepared beef dish order, in order to complete the delivery of each pre-prepared beef dish.

[0007] In one example, the standardization vector transformation processing of the quality control feature data of pre-prepared beef dishes corresponding to each customer order and the feature data of the global delivery scenario specifically includes: The quality control feature data and the global delivery scenario feature data are preprocessed respectively, and the preprocessing includes outlier removal and missing value completion. The preprocessed quality control feature data and global delivery scenario feature data are normalized to obtain standardized quality control feature data and standardized global delivery scenario feature data. According to the preset dimension order, the standardized quality control feature data of each customer order are organized into a quality control feature vector, and the standardized global delivery scenario feature data are organized into a scenario feature vector.

[0008] In one example, for each customer order, the weights of the quality control feature vector of the pre-prepared beef dish are calculated using the analytic hierarchy process (AHP) to obtain the initial set of weight coefficients corresponding to the quality control feature vector of that customer order. Specifically, this includes: The importance relationship between various quality control features is determined based on the quality control feature vector of pre-cooked beef dishes; Based on the importance relationship among various quality control features, a pairwise comparison judgment matrix with quality control features as rows and columns is constructed using the 1-9 scaling method; The eigenvalues ​​of the judgment matrix are calculated using either the summation method or the square root method to obtain the eigenvectors corresponding to the judgment matrix. Perform a consistency check on the eigenvectors corresponding to the judgment matrix; The feature vector that passes the consistency test will be used as the initial set of weight coefficients corresponding to the quality control feature vector of the pre-prepared beef dish in the customer's order.

[0009] In one example, the consistency check of the eigenvectors corresponding to the judgment matrix specifically includes: Calculate the product between the judgment matrix and the corresponding eigenvector of the judgment matrix to obtain the product vector; Substituting the product vector and the number of quality control features into the formula for calculating the maximum eigenvalue yields the maximum eigenvalue. Substituting the maximum eigenvalue and the number of quality control features into the consistency index calculation formula, the consistency index of the judgment matrix is ​​obtained. The consistency ratio of the judgment matrix is ​​calculated based on the consistency index of the judgment matrix and the preset average random consistency index. If the consistency ratio is greater than or equal to the preset average random consistency index, the scale value of the judgment matrix is ​​adjusted, and the calculation of the eigenvector, product vector, maximum eigenvalue, consistency index, and consistency ratio of the judgment matrix is ​​repeated. If the consistency ratio is less than the preset average random consistency index, the consistency test is considered passed.

[0010] In one example, the step of calculating the real-time contribution of each scene feature in the scene feature vector to each quality control feature of the customer order using a kernel density probability algorithm, and then dynamically adjusting the initial weight coefficient set of the customer order quality control feature vector using the real-time contribution as a correction factor, to obtain the dynamic weight coefficient set of the customer order quality control feature vector, specifically includes: For each customer order, the quality control feature vector of the pre-prepared beef dish is associated and concatenated with the scene feature vector to obtain a joint dataset; Based on the aforementioned joint dataset, the kernel density probability algorithm is used to calculate the real-time contribution of each scene feature in the scene feature vector to each quality control feature. The total contribution of a single quality control feature is obtained by summing the real-time contributions of all scene features to the quality control feature. Calculate the proportion of the total contribution of each quality control feature to the sum of the total contributions of all quality control features, and use this proportion as the weight correction factor for the corresponding quality control feature; Based on the initial basic weight coefficients, the weight correction factor corresponding to each quality control feature is multiplied one-to-one with the initial weight coefficient set of the quality control feature to obtain the corrected weights corresponding to each quality control feature. The corrected weights corresponding to each quality control feature are normalized to obtain the dynamic weight coefficient set of the quality control feature vector of the customer order.

[0011] In one example, the calculation of the real-time contribution of each scene feature in the scene feature vector to each quality control feature based on the joint dataset using the kernel density probability algorithm specifically includes: Extract each quality control feature from the quality control feature vector and each scene feature from the scene feature vector from the joint dataset; Calculate the mean and standard deviation of each quality control feature and each scenario feature separately; Based on the Silverman window width criterion, the standard deviation of each scenario feature and the number of customer orders are substituted into the window width calculation formula to obtain the kernel density window width corresponding to each scenario feature; Using the mean of each quality control feature as the kernel density estimation target point, for each quality control feature, the kernel density estimation target point corresponding to the quality control feature, each scene feature and the kernel density window width corresponding to the scene feature are substituted into the Gaussian kernel function to obtain the kernel function value of each quality control feature corresponding to each scene feature. The kernel function values ​​corresponding to the same quality control feature for each scene feature are summed according to the quality control feature dimension. The summation of the kernel function values ​​corresponding to each quality control feature for each scenario feature is combined with the number of customer orders and the kernel density window width corresponding to the scenario feature and substituted into the kernel density estimation formula to obtain the kernel density estimate value for each quality control feature corresponding to it. The kernel density estimates of all scene features under the same quality control feature dimension are normalized to the maximum value to obtain the real-time contribution of each scene feature to each quality control feature.

[0012] In one example, the delivery route planning based on the delivery time limit requirement and the comprehensive priority value of each customer's pre-prepared beef dishes specifically includes: Generate an initial population of a preset size, wherein each individual in the initial population represents a delivery scheme; With the optimization objectives of prioritizing comprehensive priority value, meeting delivery time targets, and minimizing total delivery distance, a fitness function is constructed based on linear weighting combined with a penalty term. Based on the fitness function, and taking into account the comprehensive priority value of each customer order, delivery time limit requirements, and geographical distance between the delivery address and the delivery station, the fitness value of each individual in the initial population is calculated one by one. The fitness values ​​of all individuals are hierarchically divided using a preset sorting method. Individuals whose time difference between the delivery completion time and the latest delivery time of the customer order is not less than the preset buffer time, whose weighted score of comprehensive priority value ranks in the top preset proportion of the initial population, and whose total delivery distance is not greater than the preset distance threshold are selected to form the current planned route set. Continue to perform multi-objective optimization on the current planned route set to reach the maximum number of iterations; The delivery route corresponding to the individual with the highest fitness value after the iteration ends will be used as the delivery route for each customer's pre-prepared beef dish.

[0013] In one example, the continued multi-objective optimization of the current planned route set specifically includes: Based on the fitness function values ​​of individuals in the current planned route set, individual selection, individual crossover, and individual mutation are performed sequentially on the current planned route set to obtain the updated planned route set. Using the fitness function, calculate the fitness function value for each individual in the updated planned route set; Based on the fitness function value of each individual in the updated planned route set, the updated planned route set is hierarchically divided, and individuals whose time difference between the delivery completion time of all customer orders and the latest delivery time of the customer orders is not less than a preset buffer time, whose comprehensive priority weighted score ranks in the top preset proportion of the initial population, and whose total delivery distance is not greater than a preset distance threshold are selected to form a new planned route set.

[0014] In one example, the method further includes: If the overall priority values ​​are the same, the order will be further sorted based on the geographical distance between the customer's order delivery address and the delivery station, from closest to furthest.

[0015] On the other hand, embodiments of this application provide a delivery device for pre-prepared beef dishes, including: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform a delivery method for pre-prepared beef dishes as described above.

[0016] The above-described technical solutions adopted in the embodiments of this application can achieve the following beneficial effects: The method provided in this application employs the Analytic Hierarchy Process (AHP) to construct a judgment matrix based on the quality control characteristics of pre-prepared beef dishes, calculate the initial weight coefficient set of each quality control feature, and use a kernel density probability algorithm to calculate the real-time contribution of each scenario feature to each quality control feature. This contribution is then used to dynamically adjust the initial weight coefficient set. Based on the dynamic weight coefficient set, the quality control feature vectors of each customer order are weighted and summed dimension-by-dimensional to obtain a comprehensive priority value. This deeply binds the core quality control characteristics of pre-prepared beef dishes, the impact of real-time scenarios, and delivery priority. This solves the problem in existing technologies that fail to consider the core characteristics of pre-prepared dish combination delivery, such as temperature control priority, food shelf life, and dish mixing types, to optimize delivery routes. This often leads to suboptimal detours, significantly increasing the exposure time of pre-prepared beef dishes to ambient temperature during delivery, further exacerbating quality loss, and ultimately severely reducing the overall delivery quality of pre-prepared beef dishes. This method adapts the weights of quality control features to real-time delivery scenarios, improving the scenario adaptability and accuracy of priority judgment, and enhancing the overall delivery quality and quality control safety of pre-prepared beef dishes. Attached Figure Description

[0017] To more clearly illustrate the technical solution of this application, some embodiments of this application will be described in detail below with reference to the accompanying drawings, in which: Figure 1 A schematic flowchart illustrating a delivery method for pre-prepared beef dishes provided in this application embodiment; Figure 2 This is a schematic diagram of a delivery device for pre-cooked beef dishes provided in an embodiment of this application. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below in conjunction with specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0019] Some embodiments of this application will now be described in detail with reference to the accompanying drawings.

[0020] Figure 1 This is a flowchart illustrating a method for delivering pre-prepared beef dishes, provided as an embodiment of this application. This process can be executed by a computing device in the relevant field, and certain input parameters or intermediate results in the process can be manually adjusted to help improve accuracy.

[0021] The analysis method involved in the embodiments of this application can be implemented by a terminal device or a server, and this application does not impose any special limitations on it. For ease of understanding and description, the following embodiments are all described in detail using a server as an example.

[0022] It should be noted that the server can be a single device or a system composed of multiple devices, i.e., a distributed server. This application does not make any specific limitations on this.

[0023] Figure 1 The process includes the following steps: S101. Obtain the quality control characteristics and delivery time requirements for each customer's pre-prepared beef dish, as well as the overall delivery scenario characteristics for all customer orders. The quality control data should include at least the temperature control requirements, shelf life, and mixed dish types for each pre-prepared beef dish. The overall delivery scenario characteristics include: ambient temperature of the delivery area, operating status of the transportation equipment, and delivery time period type.

[0024] Among them, quality control characteristic data refers to quantitative data used to characterize the quality control requirements of pre-prepared beef dishes. It is the core basis for judging the quality control risks of dish delivery and includes at least temperature control requirements data, dish shelf life data, and dish mix type data.

[0025] Temperature control requirements data characterize the required refrigeration / freezing temperatures for pre-prepared beef dishes. For example, a higher value indicates a higher requirement for temperature control precision, and the risk of temperature fluctuations during delivery needs close monitoring. Shelf life data characterizes the urgency of the remaining shelf life of pre-prepared beef dishes. For example, a higher value indicates a tighter shelf life, and a higher risk of spoilage due to late delivery. Mixed dish type data characterizes the complexity of the mixed packaging of pre-prepared beef dishes. For example, a higher value indicates more mixed types, and a higher risk of packaging damage and cross-contamination of flavors during delivery.

[0026] Delivery time requirements refer to the latest delivery time specified by the customer when submitting the order.

[0027] Global delivery scenario feature data represents the current overall delivery environment, equipment status, and time period characteristics. It does not change with individual orders and is applicable to weight correction and route planning for all orders, including the ambient temperature of the delivery area, the operating status of transportation equipment, and the type of delivery time period.

[0028] For a concrete example, there are currently three customer orders. Customer order A requires refrigeration at 0-4℃, with 1 day remaining shelf life, and consists of a mix of whole-cut beef and braised beef. Customer order B requires freezing at -18℃, with 3 days remaining shelf life, and consists of a single frozen beef block. Customer order C requires refrigeration at 0-5℃, with 2 days remaining shelf life, and consists of a mix of seasoned beef packets and beef meatballs.

[0029] The ambient temperature of the delivery area refers to the real-time average ambient temperature within the delivery area, directly affecting the load on the refrigerated truck temperature control system. Higher temperatures increase the difficulty of temperature control. For example, an ambient temperature of 32℃ in the delivery area. The transportation equipment operating status refers to the overall operational status of the available refrigerated trucks in the distribution center. This includes, for example, the stability of the temperature control system; a higher value indicates stronger equipment reliability. For example, a transportation equipment operating status indicating high temperature control stability and good operational status for the refrigerated trucks. The delivery time period type refers to the characteristics of the current delivery time period, such as peak or off-peak. A higher value indicates a greater probability of congestion and increased delivery pressure during that time period.

[0030] S102. Standardize the quality control feature data and global delivery scenario feature data of the pre-prepared beef dishes corresponding to each customer order and perform vector transformation to obtain the quality control feature vector and scenario feature vector of the pre-prepared beef dishes corresponding to each customer order.

[0031] Standardized vector transformation refers to converting raw data with different dimensions and numerical ranges into a unified dimension, such as the [0,1] interval, which can be directly used in subsequent weighted calculations and matrix operations in vector form.

[0032] For example, the quality control feature data and the global delivery scenario feature data are preprocessed separately, including outlier removal and missing value completion. The preprocessed quality control feature data and global delivery scenario feature data are then normalized to obtain standardized quality control feature data and standardized global delivery scenario feature data. Following a preset dimensional order, the standardized quality control feature data for each customer order are organized into a quality control feature vector, and the standardized global delivery scenario feature data is organized into a scenario feature vector.

[0033] Preprocessing refers to the purification of raw data to ensure data quality. Outlier removal refers to removing data that is clearly inconsistent with the actual scenario or exceeds a reasonable range, such as temperature control requirements exceeding a preset range. Missing value completion addresses missing raw data by using the average of similar data to fill in the gaps. For example, if quality control feature data for a certain order is missing, it is filled in using the average of similar quality control data from other orders.

[0034] For example, the min-max normalization method can be used, which can be implemented by the following formula (1). (1) in, The raw data after preprocessing. This is the minimum value of this type of data. To find the maximum value for this type of data, map the data to the [0,1] interval to eliminate dimensional differences and make different types of data comparable.

[0035] It should be noted that when performing normalization, the calculation method and mapping logic of normalization can be set according to the actual business rules and data characteristics to make it conform to the quantitative requirements of specific business scenarios. This application does not impose any restrictions on this.

[0036] For example, in this embodiment of the application, when normalizing the temperature control requirement dimension data of pre-cooked beef dishes, the business rule is followed that the lower the temperature, the higher the temperature control requirement of the dish. Furthermore, the actual temperature of the pre-cooked beef dishes is negatively correlated with the pre-processed temperature control requirement dimension value; that is, the lower the temperature, the larger the corresponding temperature control requirement dimension value. To make the normalization result positively correlated with the stringency of the temperature control requirements, i.e., the higher the temperature control requirements, the larger the normalized value, the above formula (1) can be adjusted in reverse. The adjusted normalization formula can be, for example, shown in the following formula (2). (2) This reverse normalization method can achieve a mapping effect where the lower the temperature of the pre-cooked beef dish, the larger the temperature control requirement dimension value, and the larger the normalization value. This allows the normalization result to accurately quantify the stringency of the temperature control requirements of the dish.

[0037] The preset dimension order is the pre-defined arrangement order of feature data, used to unify the vector format and ensure the consistency of subsequent matrix operations and weighted calculations. In this embodiment, the preset order is fixed and can be adjusted according to actual needs.

[0038] Furthermore, to facilitate quantification, before standardizing and vectorizing the quality control feature data of pre-prepared beef dishes corresponding to each customer order and the feature data of the global delivery scenario, various feature data can be discretized and quantized.

[0039] For example, temperature control requirement data is converted into temperature control requirement dimension data. In this embodiment, 0-4℃ corresponds to temperature control requirement dimension data 3, -18℃ corresponds to temperature control requirement dimension data 4, and 0-5℃ corresponds to temperature control requirement dimension data 2. The shelf life data of the dishes is quantified as the remaining days of urgency. In this embodiment, 1 day remaining is 0.9, 3 days remaining is 0.4, and 2 days remaining is 0.6. The mixed packaging type of dishes is quantified as complexity. In this embodiment, the mixed packaging of whole-cut beef and braised beef is 0.8, a single frozen beef chunk is 0.2, and the mixed packaging of seasoned beef packets and beef meatballs is 0.7. The delivery time limits are before 12:00 for customer order A, before 16:00 for customer order B, and before 14:00 for customer order C. The global delivery scenario characteristics are: the ambient temperature of the delivery area is 32℃, the transportation equipment is operating stably and well with temperature control, quantified as 0.9, and the delivery time period is the midday peak, quantified as 0.9.

[0040] In a specific example, after preprocessing, there were no outliers or missing values ​​in the three sets of orders. The quality control feature vectors of customer order A were calculated using formulas (1) and (2) and arranged in the preset order of "temperature control, shelf life, and mixing". The feature vectors of customer order A were [0.5, 1.0, 1.0], customer order B was [1.0, 0.0, 0.0], and customer order C was [0.0, 0.33, 0.8]. The global scene feature vector was [0.73, 0.9, 0.9].

[0041] S103. For each customer order, the weights of the quality control feature vector of the pre-prepared beef dishes are calculated using the analytic hierarchy process (AHP) to obtain the initial set of weight coefficients corresponding to the quality control feature vector of that customer order.

[0042] The initial weight coefficient set does not consider the impact of the global delivery scenario. The weight set of quality control features is consistent with the dimension of the quality control feature vector. Each coefficient corresponds to the initial importance of a quality control feature.

[0043] In some embodiments of this application, the importance relationship between quality control features is determined based on the quality control feature vector of the pre-prepared beef dish. Based on the importance relationship between the quality control features, a pairwise comparison judgment matrix is ​​constructed using a 1-9 scale, with the quality control features as rows and columns. The eigenvalues ​​of the judgment matrix are calculated using either the sum method or the square root method to obtain the eigenvectors corresponding to the judgment matrix. A consistency check is then performed on the eigenvectors corresponding to the judgment matrix.

[0044] Among them, the pairwise comparison judgment matrix, that is, the judgment matrix, is an n-order square matrix constructed with quality control features as rows and columns, where n is the number of quality control features.

[0045] For example, in this embodiment of the application, n=3, the element in the i-th row and j-th column of the matrix This represents the importance scale value of the i-th quality control feature relative to the j-th quality control feature, satisfying... =1, meaning that when compared to itself, it has the same importance. That is, reciprocity, such that if the importance of i is 3 compared to j, then the importance of j compared to i is 1 / 3.

[0046] The 1-9 scaling method is a standard method for setting the scale values ​​of the judgment matrix. The scale values ​​1-9 correspond to the importance levels of equal importance, slightly important, obviously important, and extremely important, respectively. In this embodiment of the application, the scale values ​​need to be set in combination with the quality control characteristics of the pre-prepared beef for each customer order.

[0047] Eigenvalue calculation refers to the process of operating on the judgment matrix to obtain eigenvectors that reflect the relative importance of each quality control feature. These eigenvectors, as initial vectors reflecting the relative importance of each quality control feature, must undergo a consistency check to verify their logical rationality before being used as the initial set of weight coefficients. The eigenvectors that pass the consistency check are used as the initial set of weight coefficients corresponding to the quality control feature vectors of the pre-prepared beef dishes in this customer order.

[0048] For example, the product of the judgment matrix and its corresponding eigenvector is calculated to obtain the product vector. The product vector and the number of quality control features are substituted into the formula for calculating the maximum eigenvalue to obtain the maximum eigenvalue. The maximum eigenvalue and the number of quality control features are then substituted into the formula for calculating the consistency index to obtain the consistency index of the judgment matrix. Based on the consistency index of the judgment matrix and a preset average random consistency index, the consistency ratio of the judgment matrix is ​​calculated. If the consistency ratio is greater than or equal to the preset average random consistency index, the scale value of the judgment matrix is ​​adjusted, and the calculation of the eigenvector, product vector, maximum eigenvalue, consistency index, and consistency ratio of the judgment matrix is ​​repeated. If the consistency ratio is less than the preset average random consistency index, the consistency check is considered passed.

[0049] The consistency check is used to verify the logical rationality of the judgment matrix construction and avoid contradictions in pairwise comparisons. For example, i is more important than j, j is more important than k, but k is more important than i. The check is performed by calculating the maximum eigenvalue, the consistency index (CI), and the consistency ratio (CR), and comparing them with the preset average random consistency index to determine whether the check passes. The maximum eigenvalue is the core feature value of the judgment matrix, reflecting the degree of consistency of the judgment matrix. Ideally, the maximum eigenvalue is equal to the matrix order; the smaller the deviation from the matrix order, the better the consistency.

[0050] The consistency index (CI) reflects the degree of deviation between the largest eigenvalue and the ideal value. For example, it can be expressed by the following formula (3). (3) The consistency ratio (CR) is the ratio of CI to the preset average random consistency index (RI), used to eliminate the influence of matrix order on consistency judgment. For example, it can be expressed by the following formula (4). (4) For example, if CR < 0.1, the consistency check passes; otherwise, the scale value of the judgment matrix needs to be adjusted and recalculated.

[0051] A specific example, taking customer order A, shows that the importance of quality control is temperature control > shelf life > mixing. A third-order judgment matrix formula (5) is constructed as follows: (5) The eigenvectors were calculated using the summation method, resulting in [0.633, 0.260, 0.107]. The largest eigenvalue... CI = (3.041-3) / (3-1) = 0.0205, RI is 0.58, CR = 0.0205 / 0.58 ≈ 0.035 < 0.1, the consistency test is passed, and this feature vector is the initial weight coefficient set of customer order A. Similarly, the initial weights of customer order B [0.800, 0.100, 0.100] and customer order C [0.100, 0.300, 0.600] are obtained.

[0052] S104. Calculate the real-time contribution of each scene feature in the scene feature vector to each quality control feature of the customer order using the kernel density probability algorithm. Use the real-time contribution as a correction factor to dynamically correct the initial weight coefficient set of the quality control feature vector of the customer order, and obtain the dynamic weight coefficient set of the quality control feature vector of the customer order.

[0053] Among them, the kernel density probability algorithm is a non-parametric probability density estimation method used to quantify the real-time impact of each scenario feature on each quality control feature in the global delivery scenario, that is, the real-time contribution.

[0054] The real-time contribution is the degree of influence of a single scene feature on a single quality control feature, calculated by the kernel density probability algorithm. For example, the larger the value, the more significant the influence of the scene feature on the quality control feature.

[0055] In some embodiments of this application, for each customer order, the quality control feature vector of the pre-prepared beef dish is concatenated with the scene feature vector to obtain a joint dataset. Based on the joint dataset, a kernel density probability algorithm is used to calculate the real-time contribution of each scene feature to each quality control feature in the scene feature vector. The real-time contributions of all scene features to the quality control feature are summed column-wise to obtain the total contribution of a single quality control feature. The proportion of the total contribution of each quality control feature to the sum of the total contributions of all quality control features is calculated, and this proportion is used as the weight correction factor for the corresponding quality control feature. Based on the initial base weight coefficients, the weight correction factor corresponding to each quality control feature is multiplied one-to-one with the initial weight coefficient set of the quality control feature to obtain the corrected weights corresponding to each quality control feature. The corrected weights corresponding to each quality control feature are normalized to obtain the dynamic weight coefficient set of the quality control feature vector for the customer order.

[0056] The joint dataset is a comprehensive dataset containing both quality control features and scenario features, formed by concatenating the quality control feature vector of each customer order with the global delivery scenario feature vector. The weight correction factor is a proportional factor used to correct the initial weight coefficients. Its dimension is consistent with the initial weight coefficient set, and each factor corresponds to a quality control feature, determined by the proportion of that quality control feature's total contribution.

[0057] Total contribution refers to the sum of the real-time contributions of all scenario features corresponding to a single quality control feature, reflecting the overall impact of the global scenario on that quality control feature. The dynamic weight coefficient set is a set of quality control feature weights that takes into account the real-time impact of the global delivery scenario, adapting to dynamic changes in the scenario.

[0058] For example, quality control features and scenario features are extracted from the joint dataset. The mean and standard deviation of each quality control feature and scenario feature are calculated separately. Based on the Silverman window criterion, the standard deviation of each scenario feature and the number of customer orders are substituted into the window width calculation formula to obtain the kernel density window width corresponding to each scenario feature. Using the mean of each quality control feature as the kernel density estimation target point, for each quality control feature, the kernel density estimation target point, each scenario feature, and the kernel density window width corresponding to that scenario feature are substituted into the Gaussian kernel function to obtain the kernel function value for each scenario feature corresponding to each quality control feature. The kernel function values ​​corresponding to the same quality control feature for each scenario feature are summed per scenario feature along the quality control feature dimension.

[0059] The kernel function values ​​corresponding to each quality control feature for each scenario feature are summed, and then substituted into the kernel density estimation formula along with the number of customer orders and the kernel density window width corresponding to that scenario feature to obtain the kernel density estimate for each quality control feature. The kernel density estimates of all scenario features under the same quality control feature dimension are normalized to the maximum value to obtain the real-time contribution of each scenario feature to each quality control feature.

[0060] Among them, the Silverman window width criterion is a standard method for calculating the window width (h) in the kernel density probability algorithm. The window width determines the smoothness of the kernel density estimate, and can be expressed, for example, by the following formula (6). (6) in, The standard deviation of scene features. For customer order quantity, The classical statistical index used to calculate the optimal window width in the preset kernel density estimation can be set according to the actual situation, and this application embodiment does not limit it.

[0061] The Gaussian kernel function is used to calculate the kernel density estimate of a single scene feature on a single quality control feature, which can be expressed, for example, by the following formula (7). (7) in, This is the difference between the kernel density estimate target point corresponding to the current quality control feature and the current scene feature. For example, it can be expressed by the following formula (8): (8) in, Let the mean of the j-th quality control characteristic be the target point for kernel density estimation. Let be the feature value of the i-th scene.

[0062] The kernel density estimate is used to statistically standardize and correct the summation of kernel function values ​​by combining the number of customer orders and the kernel density window width, thereby achieving accurate quantification of the real-time impact of scenario features on quality control features. For example, it can be expressed by the following formula (9): (9) in, Let be the kernel density estimate of the i-th scene feature on the j-th quality control feature, and n be the number of customer orders. Let be the kernel density window width corresponding to the i-th scene feature. This is the result of summing the kernel function value corresponding to the i-th scene feature under the j-th quality control feature dimension, after scene-by-scene feature summation. The kernel function value is calculated by combining the i-th scene feature and the j-th quality control feature.

[0063] The real-time contribution obtained by normalizing the kernel density estimate can be directly used as the basis for correcting the weight of the quality control feature, realizing the mapping of the kernel density estimate to the [0,1] interval, eliminating the dimensional differences of the kernel density estimate under different quality control feature dimensions, and making the influence of each scenario feature on the same quality control feature comparable. For example, it can be expressed by the following formula (10). (10) in, The value is the real-time contribution of the i-th scene feature to the j-th quality control feature, and its range is [0,1]. The larger the value, the higher the real-time influence of the scene feature on the quality control feature. Let be the kernel density estimate of the i-th scene feature for the j-th quality control feature. For the j-th quality control feature dimension, the maximum value among all scenario features' kernel density estimates of this quality control feature is denoted as .

[0064] It should be noted that the above formulas (6)-(10) are the core calculation formulas of the kernel density probability algorithm in this application embodiment. The input and output data of each formula are connected in sequence. Through the layer-by-layer calculation of the above series of formulas, the quantitative and standardized solution of the real-time influence of global delivery scenario features on the quality control features of pre-made beef dishes is realized, so that the corrected dynamic weight coefficient set can accurately adapt to the current delivery scenario and improve the accuracy of quality control feature weighting and delivery priority judgment.

[0065] A concrete example is a customer order quantity n=3. Scene feature standard deviation Calculate the window width Taking customer order A as an example, with the mean of the quality control features as the target point, the kernel function value is calculated by substituting it into the Gaussian kernel function. After summation, kernel density estimation, and maximum value normalization, the real-time contribution of the scene feature vector to customer order A is obtained. Specifically, the real-time contribution of the ambient temperature feature to the temperature control feature is 0.49, to the shelf life feature is 1.0, and to the mixed-formation feature is 0.33. The real-time contribution of the equipment status feature to the temperature control feature is 1.0, to the shelf life feature is 0.49, and to the mixed-formation feature is 0.67. The real-time contribution of the time period feature to the temperature control feature is 0.49, to the shelf life feature is 0.49, and to the mixed-formation feature is 1.0.

[0066] The total contribution values ​​are summed to obtain 1.98 for temperature control feature, 1.98 for shelf life feature, and 2.00 for blending type feature. The calculated weight correction factor is [0.332, 0.332, 0.336]. Multiplying this by the initial weight [0.633, 0.260, 0.107] and normalizing, we obtain the dynamic weight coefficient set for customer order A [0.632, 0.259, 0.108]. Similarly, we obtain the dynamic weight coefficient set for customer order B [0.785, 0.107, 0.108] and the dynamic weight coefficient set for customer order C [0.099, 0.298, 0.603].

[0067] S105. Based on the dynamic weight coefficient set, the quality control feature vectors of each customer order are weighted and summed dimension by dimension to obtain the weighted value of the quality control features of each customer order. The weighted value of the quality control features is used as the comprehensive priority value of the pre-made beef dishes of each customer order.

[0068] Among them, the weighted value of quality control features reflects the priority of quality control risk for a single order of pre-made beef dishes. For example, the larger the weighted value of quality control features, the higher the quality control risk, and the more priority should be given to delivery.

[0069] Dimensional weighted summation refers to multiplying the value of each dimension of the quality control feature vector by each coefficient corresponding to the dynamic weight coefficient set one-to-one, and then summing all the products to obtain the weighted value of the quality control feature.

[0070] For a specific example, customer order A has a quality control vector of [0.5, 1.0, 1.0] and dynamic weights of [0.632, 0.259, 0.108]. The weighted sum is approximately 0.5 × 0.632 + 1.0 × 0.259 + 1.0 × 0.108 ≈ 0.683. Customer order B has a quality control vector of [1.0, 0.0, 0.0] and a weighted sum of [1.0 × 0.785] = 0.785. Customer order C has a quality control vector of [0.0, 0.33, 0.8] and a weighted sum of approximately 0.580. The overall priority values ​​are: Customer order A 0.683, Customer order B 0.785, and Customer order C 0.580.

[0071] S106. Based on the delivery time limit requirements and the comprehensive priority value of each customer's pre-prepared beef dishes, plan the delivery route to complete the delivery of each pre-prepared beef dish.

[0072] In some embodiments of this application, the pre-prepared beef dishes in each customer order are sorted according to a preset sorting rule based on their comprehensive priority value to obtain a preliminary delivery order for each customer order. Based on the preliminary delivery order, and considering the geographical distance between each customer order's delivery address and the delivery station, as well as the urgency of the delivery time requirements, a path optimization algorithm based on a genetic algorithm is used to adjust the delivery order, resulting in an optimized delivery order.

[0073] The preset sorting rule can be, for example, sorting in descending order by overall priority value, meaning that the higher the overall priority value, the higher the order delivery priority and the earlier it will be arranged for delivery. This can be specifically set according to the actual order situation, and this embodiment does not impose any limitations on it. The initial delivery order is obtained by sorting all orders according to the preset sorting rule, without considering geographical distance or time urgency.

[0074] For example, an initial population of a preset size is generated, where each individual represents a delivery plan. With the optimization objectives of prioritizing comprehensive priority, meeting delivery time requirements, and minimizing total delivery distance, a fitness function is constructed based on linear weighting combined with a penalty term. Based on the fitness function, and considering the comprehensive priority of each customer order, delivery time requirements, and the geographical distance between the delivery address and the delivery station, the fitness value of each individual in the initial population is calculated. A preset sorting method is used to hierarchically categorize the fitness values ​​of all individuals. Individuals whose delivery completion time is not less than a preset buffer time, whose comprehensive priority weighted score ranks within a preset proportion in the initial population, and whose total delivery distance is not greater than a preset distance threshold are selected to form the current planned route set.

[0075] Continue multi-objective optimization of the current planned route set until the maximum number of iterations is reached. The delivery route corresponding to the individual with the highest fitness value after the iteration terminates will be used as the delivery route for each customer's pre-prepared beef dish order.

[0076] In this application example, multi-objective optimization is further performed on the current planned route set, specifically including: Based on the fitness function values ​​of individuals in the current planned route set, individual selection, crossover, and mutation are performed sequentially to obtain an updated planned route set. The fitness function value of each individual in the updated planned route set is calculated. Based on the fitness function values ​​of each individual in the updated planned route set, the updated planned route set is hierarchically partitioned. Individuals whose delivery completion time and the latest delivery time of all customer orders differ from the latest delivery time in the customer orders by no less than a preset buffer time, whose weighted score based on comprehensive priority values ​​ranks within a preset proportion of the initial population, and whose total delivery distance does not exceed a preset distance threshold are selected to form a new planned route set.

[0077] The fitness function value can be expressed, for example, by the following formula (11). (11) in, The fitness value of the delivery solution. Assuming a comprehensive priority score, This is a penalty item for delivery time limits. Total delivery distance The preset weighting coefficients satisfy the following conditions: .

[0078] Overall priority score The degree to which the delivery plan fits the quality control priority of pre-prepared beef dishes is used to quantify the degree of fit between the delivery plan and the quality control priority. The calculation method is the sum of the comprehensive priority values ​​of all customer orders in the delivery plan, which can be expressed by the following formula (12). (12) Where n is the total number of customer orders, The overall priority value is the pre-prepared beef dish for the i-th customer order. Orders with higher overall priority values ​​are placed earlier in the delivery plan. The larger the value, the better it is suited to delivery needs where quality control is a priority.

[0079] Delivery time limit penalty This system is used to constrain delivery plans to meet the delivery time requirements of each customer's order, preventing quality loss of prepared beef dishes due to late delivery. The calculation method is as follows: if the i-th customer's order is delivered according to the delivery plan within its specified latest delivery time, the penalty is 0. If the delivery time is exceeded, a preset large penalty value is assigned. It needs to be explained. The value is much greater than The maximum possible value is used to ensure that the fitness value of the timeout scheme is significantly reduced, for example, it can be expressed by the following formula (13). (13) In the embodiments of this application, it is preset To adapt to scenarios where the comprehensive priority value range is [0,1] and the number of orders is 3.

[0080] Total delivery distance The efficiency of a delivery plan is quantified by calculating the total geographical distance from the delivery station, passing through each customer's order delivery address in the order of the orders in the delivery plan, and finally returning to the delivery station. This can be represented by, for example, the following formula (14). (14) The distance from the delivery station to the address of the first order. This is the distance between the first order address and the second order address. This is the distance to the delivery station for the nth order address. The distance is calculated using geographic coordinates or obtained from actual road conditions, and the unit is km.

[0081] In this embodiment of the application, the weighting coefficient may be set to, for example, a value of 100%. Prioritizing quality control with limited safeguards To strictly control delivery time limits and prevent food spoilage, To balance delivery efficiency and shorten the total distance, the weighting coefficient can be set based on the core needs of pre-prepared beef dish delivery and can be flexibly adjusted according to the actual delivery scenario. This application embodiment does not impose any restrictions on this, but it always satisfies the following requirements: This is to ensure that quality control and timeliness take precedence over efficiency.

[0082] Furthermore, when the overall priority values ​​are the same, the order is further sorted based on the geographical distance between each customer's order delivery address and the delivery station, from closest to furthest.

[0083] It should be noted that in the embodiments of this application, each individual in the initial population is used to represent a delivery scheme, and each delivery scheme corresponds to a complete delivery route; the delivery route set consists of multiple delivery routes that meet the constraints, that is, it consists of multiple individuals that meet the requirements.

[0084] For example, suppose this pre-prepared beef dish delivery task includes three customer orders: Customer Order A, Customer Order B, and Customer Order C. Customer Order A has a priority value of 0.75, a latest delivery deadline of 12:00, and a distance of 2.5km from the delivery station. Customer Order B has a priority value of 0.82, a latest delivery deadline of 16:00, and a distance of 4.0km from the delivery station. Customer Order C has a priority value of 0.60, a latest delivery deadline of 14:00, and a distance of 1.8km from the delivery station. The delivery distance between Customer Order A and Customer Order B is 3.0km, the delivery distance between Customer Order A and Customer Order C is 1.2km, and the delivery distance between Customer Order B and Customer Order C is 4.5km.

[0085] The preset sorting rule is descending order of comprehensive priority value. The initial population size is 5 individuals. The time difference between the completion time of all orders and the latest delivery time in a customer's order is no less than 15 minutes. Individuals with a comprehensive priority value weighted score ranking in the top 80% of the initial population and a total delivery distance no greater than 1.2 times the shortest baseline distance are selected. The maximum number of iterations is 50. Fitness function weight coefficients. , , That is, satisfying Timeout penalty value This is significantly greater than the maximum possible sum of the comprehensive priority values, which is 2.17. The delivery station departs at 10:00 AM, and the delivery speed is 60 km / h, used to calculate delivery time and determine if it is overdue.

[0086] First, based on this initial delivery sequence, an initial population of a preset size is generated, for example, 5 individuals. Each individual represents a valid delivery plan for a customer order without duplicates. The delivery plans corresponding to the 5 individuals are as follows: Individual 1 (Customer Order B, Customer Order A, Customer Order C), Individual 2 (Customer Order B, Customer Order C, Customer Order A), Individual 3 (Customer Order A, Customer Order B, Customer Order C), Individual 4 (Customer Order A, Customer Order C, Customer Order B), and Individual 5 (Customer Order C, Customer Order A, Customer Order B).

[0087] Secondly, with the optimization objectives of prioritizing comprehensive priority value, meeting delivery time requirements, and minimizing total delivery distance, the comprehensive priority value of all individuals is calculated based on formula (12) to be 2.17, which is independent of the delivery order and only related to the composition of customer orders.

[0088] Calculated based on delivery station, first order, second order, third order, and delivery station, for Individual 1 and Individual 5. For 10 kilometers (km), individual 3 The distance is 11.8km, with individuals 2 and 4. The distance is 12.2 km. The preset distance threshold is 12.0 km, therefore individuals 2 and 4... Exceeding the threshold.

[0089] Based on delivery speed and total delivery distance, the delivery time for each individual was calculated. All individual deliveries took less than 15 minutes, all customer orders were delivered within the latest delivery time limit, and the time difference between the delivery completion time and the latest delivery time was no less than 15 minutes. Therefore, the delivery time for all individuals... =0.

[0090] Based on the above parameters, the fitness value of each individual in the initial population was calculated one by one. The fitness values ​​of individuals 1 and 5 were both -1.132. The fitness value of individual 3 was -1.492. The fitness values ​​of individuals 2 and 4 were both -1.572.

[0091] Based on the above screening criteria, individuals 2 and 4 were eliminated because their total delivery distance exceeded the threshold. Ultimately, individuals 1, 3, and 5 were selected to form the current planned route set.

[0092] Subsequently, multi-objective optimization is performed on the current planned route set, that is, hierarchical partitioning and selection, intersection, and mutation operations are repeatedly performed on the individuals in the route set until the preset maximum number of iterations of 50 is reached. For example, and optionally, high-quality individuals are selected from the current planned route set, i.e., individuals 1, 3, and 5, while low-fitness, low-quality individuals are eliminated to form the parent population. Individuals 1 and 5, with the highest fitness values, are preferentially retained and directly enter the parent population. Alternatively, individuals 1, 3, and 5 can also be retained and directly enter the parent population.

[0093] After the selection operation is completed, the crossover operation is performed only on the selected parent population. Individuals are randomly paired in the parent population, and new offspring distribution schemes are generated through ordered crossover. Two core pairings are randomly selected: individual 1 × individual 5 and individual 3 × individual 1.

[0094] The first parent generation consists of Individual 1 (Customer Order B, Customer Order A, Customer Order C) × Individual 5 (Customer Order C, Customer Order A, Customer Order B). Randomly select the intersection segment as the 1st and 2nd orders. Retain the segment from Individual 1 (Customer Order B, Customer Order A), and fill in the missing Customer Order C from Individual 5 to generate Child Generation 1 (Customer Order B, Customer Order A, Customer Order C). Retain the segment from Individual 5 (Customer Order C, Customer Order A), and fill in the missing Customer Order B from Individual 1 to generate Child Generation 2 (Customer Order C, Customer Order A, Customer Order B).

[0095] The second parent generation consists of individual 3 (customer order A, customer order B, customer order C) × individual 1 (customer order B, customer order A, customer order C). A random intersection segment is selected as the second order. The segment from individual 3 (customer order B) is retained, and customer orders A and C from individual 1 that are not present in this segment are filled in (in the order of the segments not present in individual 1), generating child generation 3 (customer order A, customer order B, customer order C). The segment from individual 1 (customer order A) is retained, and customer orders B and C from individual 3 that are not present in this segment are filled in (in the order of the segments not present in individual 3), generating child generation 4 (customer order B, customer order A, customer order C).

[0096] After crossover, the offspring populations are offspring 1, offspring 2, offspring 3, and offspring 4.

[0097] After the crossover operation is completed, a mutation operation is performed to fine-tune the offspring locally, preventing premature convergence of the algorithm. Two order positions in the offspring are randomly swapped with a preset low probability (0.1) to achieve a local perturbation of the delivery plan. For example, offspring 4 (customer order B, customer order A, customer order C) is randomly selected with a probability of 0.1, and its 2nd and 3rd orders are swapped, resulting in (customer order B, customer order C, customer order A). Other offspring do not mutate, ultimately generating a new generation population. This is continuously iterated and optimized until the preset maximum number of iterations (50) is reached.

[0098] Finally, after the iteration terminates, the high-quality individual with the highest fitness value in the population is individual 1. The delivery plan corresponding to individual 1, which is consistent with the initial delivery order, is selected as the final delivery route for the pre-prepared beef dishes for each customer order.

[0099] This route ensures priority delivery for customer orders with higher overall priority values, while also achieving the optimization goals of meeting delivery time targets and minimizing total delivery distance, effectively preventing quality loss of pre-prepared beef dishes due to improper delivery.

[0100] Furthermore, when the overall priority values ​​are the same, the order is further sorted based on the geographical distance between each customer's order delivery address and the delivery station, from closest to furthest.

[0101] The method provided in this application acquires quality control feature data of pre-prepared beef dishes corresponding to each customer order, as well as global delivery scenario feature data such as ambient temperature in the delivery area and the operating status of transportation equipment, providing comprehensive and accurate basic data support for subsequent optimization of delivery priorities. By performing standardized vector transformation processing on the acquired quality control feature data and global delivery scenario feature data, an operable data foundation is provided for quantifying quality control feature priorities.

[0102] Using the analytic hierarchy process (AHP), a judgment matrix was constructed based on the quality control characteristics of pre-prepared beef dishes. The initial weight coefficient set of each quality control feature was calculated, clarifying the initial importance of each quality control feature. This provides a weight basis for subsequent comprehensive priority value calculation and avoids the neglect of core quality control features.

[0103] The kernel density probability algorithm is used to calculate the real-time contribution of each scenario feature to each quality control feature, and the initial weight coefficient set is dynamically corrected using this contribution, thereby improving the scenario adaptability and accuracy of priority judgment.

[0104] Based on a dynamic weighted coefficient set, the quality control feature vectors of each customer order are weighted and summed dimension by dimension to obtain a comprehensive priority value. This deeply binds the core quality control features of pre-prepared beef dishes, the impact of real-time scenarios, and delivery priorities. Based on delivery time requirements and the comprehensive priority value of each customer order's pre-prepared beef dishes, delivery routes are planned to complete the delivery of each pre-prepared beef dish. This reduces detours and room temperature exposure time, lowers the quality loss of dishes, and improves the overall delivery quality and quality control safety of pre-prepared beef dishes.

[0105] It should be noted that, although the embodiments in this application are based on... Figure 1 Steps S101 to S106 will be described sequentially, but this does not mean that steps S101 to S106 must be performed in a strict order. The reason this embodiment follows this order is... Figure 1The order in which steps S101 to S106 are described is provided to facilitate understanding of the technical solutions of the embodiments of this application by those skilled in the art. In other words, in the embodiments of this application, the order of steps S101 to S106 can be appropriately adjusted according to actual needs.

[0106] based on Figure 1 In addition to the method described herein, this specification also provides some specific implementation schemes and extension schemes of this method, which will be further explained below.

[0107] Based on the same idea, some embodiments of this application also provide devices and non-volatile computer storage media corresponding to the above methods.

[0108] Figure 2 A schematic diagram of a delivery device for pre-cooked beef dishes provided in this application embodiment includes: At least one processor; and, A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform a delivery method for pre-prepared beef dishes as described above.

[0109] The various embodiments in this application are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the device embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0110] The devices and methods provided in this application are one-to-one correspondences. Therefore, the devices also have similar beneficial technical effects as their corresponding methods. Since the beneficial technical effects of the methods have been described in detail above, the beneficial technical effects of the devices and media will not be repeated here.

[0111] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0112] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0113] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0114] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0115] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0116] Memory may include non-persistent storage in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0117] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0118] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0119] The above description is merely an embodiment of this application and is not intended to limit this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the technical principles of this application should fall within the protection scope of this application.

Claims

1. A delivery method for pre-cooked beef dishes, characterized in that, The method includes: The system acquires quality control characteristic data and delivery time limit requirements for each customer's pre-prepared beef dish, as well as global delivery scenario characteristic data for all customer orders. The quality control data includes at least temperature control requirements, shelf life data, and dish mixing type for each pre-prepared beef dish. The global delivery scenario characteristic data includes: ambient temperature of the delivery area, operating status of transportation equipment, and delivery time period type. Standardized vector transformation processing is performed on the quality control feature data of pre-prepared beef dishes corresponding to each customer order and the global delivery scenario feature data to obtain the quality control feature vector and scenario feature vector of pre-prepared beef dishes corresponding to each customer order. For each customer order, the weights of the quality control feature vector of the pre-prepared beef dishes are calculated using the analytic hierarchy process (AHP) to obtain the initial set of weight coefficients corresponding to the quality control feature vector of that customer order. The real-time contribution of each scene feature in the scene feature vector to each quality control feature of the customer order is calculated by the kernel density probability algorithm. The initial weight coefficient set of the quality control feature vector of the customer order is dynamically corrected by the real-time contribution as a correction factor to obtain the dynamic weight coefficient set of the quality control feature vector of the customer order. Based on the dynamic weight coefficient set, the quality control feature vectors of each customer order are weighted and summed dimension by dimension to obtain the quality control feature weighted value of each customer order. The quality control feature weighted value is used as the comprehensive priority value of the pre-made beef dishes in each customer order. Delivery routes are planned based on the aforementioned delivery time requirements and the overall priority of each customer's pre-prepared beef dish order, in order to complete the delivery of each pre-prepared beef dish.

2. The method according to claim 1, characterized in that, The standardization and vector transformation processing of the quality control feature data of pre-prepared beef dishes corresponding to each customer order and the feature data of the global delivery scenario specifically includes: The quality control feature data and the global delivery scenario feature data are preprocessed respectively, and the preprocessing includes outlier removal and missing value completion. The preprocessed quality control feature data and global delivery scenario feature data are normalized to obtain standardized quality control feature data and standardized global delivery scenario feature data. According to the preset dimension order, the standardized quality control feature data of each customer order are organized into a quality control feature vector, and the standardized global delivery scenario feature data are organized into a scenario feature vector.

3. The method according to claim 1, characterized in that, For each customer order, the quality control feature vector of the pre-prepared beef dish is weighted using the analytic hierarchy process (AHP) to obtain the initial set of weight coefficients corresponding to the quality control feature vector of that customer order. Specifically, this includes: The importance relationship between various quality control features is determined based on the quality control feature vector of pre-cooked beef dishes; Based on the importance relationship among various quality control features, a pairwise comparison judgment matrix with quality control features as rows and columns is constructed using the 1-9 scaling method; The eigenvalues ​​of the judgment matrix are calculated using either the summation method or the square root method to obtain the eigenvectors corresponding to the judgment matrix. Perform a consistency check on the eigenvectors corresponding to the judgment matrix; The feature vector that passes the consistency test will be used as the initial set of weight coefficients corresponding to the quality control feature vector of the pre-prepared beef dish in the customer's order.

4. The method according to claim 3, characterized in that, The consistency check of the eigenvectors corresponding to the judgment matrix specifically includes: Calculate the product between the judgment matrix and the corresponding eigenvector of the judgment matrix to obtain the product vector; Substituting the product vector and the number of quality control features into the formula for calculating the maximum eigenvalue yields the maximum eigenvalue. Substituting the maximum eigenvalue and the number of quality control features into the consistency index calculation formula, the consistency index of the judgment matrix is ​​obtained. The consistency ratio of the judgment matrix is ​​calculated based on the consistency index of the judgment matrix and the preset average random consistency index. If the consistency ratio is greater than or equal to the preset average random consistency index, the scale value of the judgment matrix is ​​adjusted, and the calculation of the eigenvector, product vector, maximum eigenvalue, consistency index, and consistency ratio of the judgment matrix is ​​repeated. If the consistency ratio is less than the preset average random consistency index, the consistency test is considered passed.

5. The method according to claim 1, characterized in that, The step involves calculating the real-time contribution of each scene feature in the scene feature vector to each quality control feature of the customer order using a kernel density probability algorithm, and then dynamically adjusting the initial weight coefficient set of the customer order quality control feature vector using the real-time contribution as a correction factor to obtain the dynamic weight coefficient set of the customer order quality control feature vector. Specifically, this includes: For each customer order, the quality control feature vector of the pre-prepared beef dish is associated and concatenated with the scene feature vector to obtain a joint dataset; Based on the aforementioned joint dataset, the kernel density probability algorithm is used to calculate the real-time contribution of each scene feature in the scene feature vector to each quality control feature. The total contribution of a single quality control feature is obtained by summing the real-time contributions of all scene features to the quality control feature. Calculate the proportion of the total contribution of each quality control feature to the sum of the total contributions of all quality control features, and use this proportion as the weight correction factor for the corresponding quality control feature; Based on the initial basic weight coefficients, the weight correction factor corresponding to each quality control feature is multiplied one-to-one with the initial weight coefficient set of the quality control feature to obtain the corrected weights corresponding to each quality control feature. The corrected weights corresponding to each quality control feature are normalized to obtain the dynamic weight coefficient set of the quality control feature vector of the customer order.

6. The method according to claim 5, characterized in that, The step of calculating the real-time contribution of each scene feature to each quality control feature in the scene feature vector based on the joint dataset using the kernel density probability algorithm specifically includes: Extract each quality control feature from the quality control feature vector and each scene feature from the scene feature vector from the joint dataset; Calculate the mean and standard deviation of each quality control feature and each scenario feature separately; Based on the Silverman window width criterion, the standard deviation of each scenario feature and the number of customer orders are substituted into the window width calculation formula to obtain the kernel density window width corresponding to each scenario feature; Using the mean of each quality control feature as the kernel density estimation target point, for each quality control feature, the kernel density estimation target point corresponding to the quality control feature, each scene feature and the kernel density window width corresponding to the scene feature are substituted into the Gaussian kernel function to obtain the kernel function value of each quality control feature corresponding to each scene feature. The kernel function values ​​corresponding to the same quality control feature for each scene feature are summed according to the quality control feature dimension. The summation of the kernel function values ​​corresponding to each quality control feature for each scenario feature is combined with the number of customer orders and the kernel density window width corresponding to the scenario feature and substituted into the kernel density estimation formula to obtain the kernel density estimate value for each quality control feature corresponding to it. The kernel density estimates of all scene features under the same quality control feature dimension are normalized to the maximum value to obtain the real-time contribution of each scene feature to each quality control feature.

7. The method according to claim 1, characterized in that, The delivery route planning based on the delivery time limit requirements and the comprehensive priority value of each customer's pre-prepared beef dishes specifically includes: Generate an initial population of a preset size, wherein each individual in the initial population represents a delivery scheme; With the optimization objectives of prioritizing comprehensive priority value, meeting delivery time targets, and minimizing total delivery distance, a fitness function is constructed based on linear weighting combined with a penalty term. Based on the fitness function, and taking into account the comprehensive priority value of each customer order, delivery time limit requirements, and geographical distance between the delivery address and the delivery station, the fitness value of each individual in the initial population is calculated one by one. The fitness values ​​of all individuals are hierarchically divided using a preset sorting method. Individuals whose time difference between the delivery completion time and the latest delivery time of the customer order is not less than the preset buffer time, whose weighted score of comprehensive priority value ranks in the top preset proportion of the initial population, and whose total delivery distance is not greater than the preset distance threshold are selected to form the current planned route set. Continue to perform multi-objective optimization on the current planned route set to reach the maximum number of iterations; The delivery route corresponding to the individual with the highest fitness value after the iteration ends will be used as the delivery route for each customer's pre-prepared beef dish.

8. The method according to claim 7, characterized in that, The continued multi-objective optimization of the current planned route set specifically includes: Based on the fitness function values ​​of individuals in the current planned route set, individual selection, individual crossover, and individual mutation are performed sequentially on the current planned route set to obtain the updated planned route set. Using the fitness function, calculate the fitness function value for each individual in the updated planned route set; Based on the fitness function value of each individual in the updated planned route set, the updated planned route set is hierarchically divided, and individuals whose time difference between the delivery completion time of all customer orders and the latest delivery time of the customer orders is not less than a preset buffer time, whose comprehensive priority weighted score ranks in the top preset proportion of the initial population, and whose total delivery distance is not greater than a preset distance threshold are selected to form a new planned route set.

9. The method according to claim 7, characterized in that, The method further includes: If the overall priority values ​​are the same, the order will be further sorted based on the geographical distance between the customer's order delivery address and the delivery station, from closest to furthest.

10. A delivery device for pre-cooked beef dishes, characterized in that, include: At least one processor; as well as, A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform a delivery method for pre-prepared beef dishes as described in any one of claims 1-9.

Images