Intelligent supply chain logistics management method

Abstract

The invention discloses an intelligent supply chain logistics management method, and the method comprises the steps: generating a multivariate supply feature vector corresponding to a predicted supply time period based on the logistics information of each to-be-scheduled logistics order, inputting the multivariate supply feature vector into a pre-trained supply time prediction model, and outputting the target supply time of the to-be-scheduled order; obtaining a difference value between the order maximum time limit of the to-be-dispatched logistics order and the target supply time of the to-be-dispatched logistics order, taking the difference value as a supply emergency value, and if the supply emergency value is smaller than a preset time threshold, marking the to-be-dispatched logistics order as an abnormal order; and carrying out descending sorting to obtain an abnormal order sequence, and based on the abnormal order sequence, notifying a logistics center to adjust logistics information of the abnormal order sequence, and carrying out distribution on the abnormal order sequence. Therefore, the multivariate supply feature vectors of the to-be-dispatched logistics orders are introduced, the factors influencing the logistics supply time are considered in multiple dimensions, the accuracy of the prediction model is improved, and the prediction model can dynamically adapt to the actual change of supply chain logistics.

Description

Technical Field

[0001] The present invention relates to the field of logistics and transportation technology, and in particular to a smart supply chain logistics management method. Background Art

[0002] Supply chain refers to a complete supply chain formed by integrating multiple logistics around the core logistics enterprise. In simple terms, it is a supply chain structure that connects suppliers, manufacturers, distributors and end users, starting from matching parts, making intermediate products and final products, and finally delivering products to consumers through sales networks. As an organic logistics chain, it is a chain formed by all logistics activities involved in economic activities from the time when the market demand for products or services begins to meet the demand. However, in the process of supply chain management, due to the large variety of logistics products, high frequency of mobilization, large information differences between logistics products, and scattered distribution of supply areas, conventional supply chain logistics management methods are inefficient and cannot achieve dynamic supervision and optimization for many logistics products. Therefore, it is impossible to timely plan corresponding supply plans for products that may have supply risks to improve the supervision of products with supply risks, so as to ensure that all kinds of products can be supplied efficiently, quickly and in a timely manner.

[0003] The Chinese invention patent with patent application number 202210294506.2 discloses a method for intelligent optimization management of smart supply chain logistics. By establishing a dynamic supply time warning model, the supply time of numerous logistics products in supply chain logistics management can be predicted. Based on the prediction time and time limit requirements, early warning analysis is performed on each logistics product. Corresponding supply plans can be planned for products that may have supply risks to improve the supervision of products with supply risks, and logistics product supply plans can be adjusted and optimized in a timely manner to facilitate intelligent optimization management of smart supply chain logistics.

[0004] However, due to the complexity and dynamism of supply chain logistics in actual scenarios, the supply time of supply chain logistics orders is affected by many factors. The prediction model is only trained based on the sample data relationship between historical locations, routes and supply times, which limits the prediction evaluation indicators and leads to inaccurate model prediction results. At the same time, due to the existence of different supply chain nodes in the order supply route, the accuracy of the model prediction is inevitably questionable and has a large error space if the overall supply time of a large number of logistics orders in history is directly used to train a prediction model for the supply time of goods. Therefore, how to accurately improve the prediction accuracy of the model is extremely critical. Summary of the invention

[0005] This application provides an intelligent supply chain logistics management method, introduces a multi-dimensional supply feature vector of logistics orders to be scheduled, considers the factors affecting logistics supply time in multiple dimensions, improves the accuracy of the prediction model, and enables the prediction model to dynamically adapt to actual changes in supply chain logistics.

[0006] This application provides a smart supply chain logistics management method, including: S101, based on the logistics information of each logistics order to be scheduled, generate a multivariate supply feature vector [a, b, c] corresponding to its expected supply period, where a is route feature information, b is environmental feature information, and c is traffic feature information; the logistics information includes logistics starting point, end point, transfer point, logistics supply route, expected supply period, and maximum order time limit; S102, inputting the multivariate supply feature vector into a pre-trained supply time prediction model, and outputting the target supply time of the order to be scheduled; S103, obtaining the difference between the maximum order time limit of the logistics order to be scheduled and its target supply time as the supply urgency value. If the supply urgency value is less than the preset time threshold, the logistics order to be scheduled is marked as an abnormal order; S104, obtaining the supply urgency values ​​of all current abnormal orders, sorting them in descending order to obtain an abnormal order sequence, and notifying the logistics center to adjust its logistics information based on the abnormal order sequence; S105, delivering the logistics order to be dispatched according to its logistics supply route.

[0007] Preferably, the method of generating a multivariate supply feature vector corresponding to the expected supply period based on the logistics information of each logistics order to be scheduled specifically includes: A1. Generate route curvature values ​​and longitude and latitude coordinate sequences of key point sequences according to the logistics supply route to form route feature information; the longitude and latitude coordinate sequences are composed of the longitude and latitude coordinates of key points that the logistics supply route passes through in sequence; the route curvature values ​​are obtained by cutting the logistics supply route into several sub-route segments according to all key points, calculating the curvature value of each sub-route segment, and taking the average value to obtain the route curvature value; A2. Obtain the sub-supply period corresponding to each sub-route segment, generate the environmental stability value of each sub-route segment in the corresponding sub-supply period, and take the average value as the environmental characteristic information; A3. Based on the traffic flow data pattern of each sub-route segment in a preset historical time period, the traffic flow value of each sub-route segment in the sub-supply period is estimated as traffic characteristic information.

[0008] Preferably, the pre-trained supply time prediction model is obtained by: B1. Collect a large number of multivariate supply feature vectors of logistics orders to be scheduled in history, and label them with the actual supply time of the logistics orders to be scheduled; B2. Divide all the labeled multivariate supply feature vectors into a training sample set and a validation sample set. Use the training sample set to train the pre-set neural network structure, and use the validation sample set to optimize the model parameters to obtain the optimal supply time prediction model.

[0009] Preferably, the output end of the supply time prediction model is also connected to a pre-trained verification prediction model, and the verification prediction model includes two input ends, which are respectively used to input the target supply time output by the supply time prediction model and the verification supply feature vector sequence of the logistics order to be scheduled. The verification supply feature vector sequence is obtained as follows: C1. Obtain the logistics information corresponding to all sub-route segments of the logistics order to be scheduled, and generate a multivariate supply feature vector corresponding to each sub-route segment; C2. Combine the multivariate supply feature vectors of all sub-route segments to obtain a verification supply feature vector sequence of the logistics order to be scheduled. The verification supply feature vector sequence includes at least one verification supply feature vector, and each verification supply feature vector corresponds to a sub-route segment.

[0010] Preferably, the pre-trained verification prediction model is obtained by: D1. Collect the verification supply feature vectors corresponding to all sub-route segments of a large number of logistics orders to be scheduled in history, and label them with the actual supply time of the sub-route segments; D2. Divide all the marked verification supply feature vectors into a training sample set and a verification sample set, use the training sample set to train the pre-set neural network structure, and use the verification sample set to optimize the model parameters to obtain the optimal verification prediction model; The input end of the verification prediction model also includes a splitting module, which is used to split the input verification supply feature vector sequence to obtain a number of verification supply feature vectors. The verification prediction module predicts the several verification supply feature vectors after the splitting, and outputs the verification supply time corresponding to each verification supply feature vector; The output end of the verification prediction model also includes an accumulation module, which is used to: obtain the verification supply time corresponding to each verification supply feature vector and sum them to obtain the target verification supply time, and combine the target verification supply time, the target supply time, and the verification supply time series as the output of the verification prediction model; Among them, the verification supply time series is composed of the verification supply time corresponding to all the verification supply feature vectors of the logistics orders to be scheduled, and the verification supply time corresponds to the sub-route segment one by one.

[0011] Preferably, the S102 further includes: S201, obtaining the target verification supply time, target supply time, and verification supply time series output by the verification prediction model; S202, taking the weighted sum of the target verification supply time and the target supply time as the new target supply time for the logistics order to be scheduled, replacing the original target supply time.

[0012] Preferably, the S105 further includes: S106, based on each logistics order to be scheduled, periodically update the target supply time of the logistics order to be scheduled according to the preset time interval, and repeat steps S103 to S105 until the delivery of the logistics order to be scheduled is completed.

[0013] Preferably, the S106 specifically includes: S301, according to the logistics status of the current logistics order to be scheduled, update the logistics information of the logistics order to be scheduled, and based on the latest logistics information, repeat steps S101 to S102 to determine the target supply duration as the post-supply duration; wherein the logistics status includes dynamic in-transit and static transit. When the logistics status is dynamic in-transit, update the logistics information of the logistics order to be scheduled, including: Obtain the last closest key point of the logistics order to be scheduled, update the logistics starting point in the current logistics information to the key point, filter out the transit points that have been passed, and update the logistics supply route, transit points, and expected supply period according to the new logistics starting point; When the logistics status is static transit, the logistics information of the logistics order to be scheduled is updated, including: updating the logistics starting point in the current logistics information to the current transit point of the logistics order to be scheduled, filtering out the transit points that have been passed, and updating the logistics supply route, transit point, and expected supply period according to the new logistics starting point; S302, obtaining the leading supply time of the logistics order to be scheduled, and updating the sum of the leading supply time and the post-target supply time as the target supply time of the logistics order to be scheduled.

[0014] Preferably, the step of obtaining the lead time of the logistics order to be scheduled includes: If the logistics status of the current logistics order to be scheduled is dynamically in transit, the difference between the time node when the logistics order to be scheduled starts to be delivered and the time node when the logistics order to be scheduled starts from the last closest key point is used as the lead time; If the logistics status of the current logistics order to be scheduled is static transit, the difference between the time node when the logistics order to be scheduled starts delivery and the time node when the logistics order to be scheduled is expected to depart at the current transit point will be used as the lead supply time.

[0015] Preferably, the S106 further includes: periodically updating the weight values ​​of the target verification supply time and the target supply time, specifically including: S401, respectively calculating the difference between the target verification supply time and the target supply time in step S201 in the previous cycle and the target supply time updated in the current cycle, and recording them as the first difference and the second difference; S402, when the first difference is greater than the second difference: E1, obtain all the sub-route segments that the logistics order to be scheduled has passed through, record them as invalid sub-route segments, and obtain the actual supply time of each invalid sub-route segment; E2, obtain the verification supply time series in the previous cycle, match the verification supply time corresponding to each invalid sub-route segment, and calculate the difference between each invalid sub-route segment and its corresponding verification supply time, and take the average value to obtain the verification credibility value; E3, if the verification credibility value is greater than the credibility threshold, it means that the accuracy of the verification prediction model is low, and the weight value of the target verification supply time is reduced according to the preset ratio, and the weight value of the target supply time is adaptively adjusted; otherwise, keep the original weight value unchanged; S403, when the first difference is less than the second difference: execute steps E1 to E2, if the verification credibility value is greater than the credibility threshold, keep the original weight value unchanged; otherwise, increase the weight value of the target verification supply time according to the preset ratio, and adaptively adjust the weight value of the target supply time.

[0016] One or more technical solutions provided in this application have at least the following technical effects or advantages: By comprehensively considering the characteristic performance of the logistics orders to be scheduled on different sub-route segments, a more accurate and sophisticated multi-element supply feature vector (route, environment, and traffic) of the logistics orders to be scheduled is introduced, and the factors affecting the logistics supply time are more comprehensively considered, thereby improving the accuracy of the prediction model and enabling the prediction model to dynamically adapt to the actual changes in supply chain logistics; by calculating the difference between the predicted supply time and the maximum time limit of differentiated orders for different orders, the supply urgency value is determined, and abnormal orders are generated and processed in a timely manner, which can detect and adjust potential logistics delays in advance during distribution and optimize logistics scheduling efficiency; therefore, more accurate predictions and more efficient logistics scheduling can help reduce order delays and improve order customer satisfaction.

[0017] By introducing a verification prediction model at the output end of the supply time prediction model, a more detailed differentiated prediction is made for each sub-route segment of the logistics order to be scheduled, which will realize mutual verification and combination of the verification prediction model and the supply time prediction model, reducing the possible errors of a single model, enhancing the robustness of the entire system, and further improving the accuracy of the prediction model; due to the existence of different supply chain nodes in the order supply route, directly using the global overall supply time training model will ignore possible large errors. The target supply time is obtained by combining local predictions with global predictions. The updated target supply time is more accurate, which helps the logistics center make more optimized scheduling decisions and improve logistics efficiency.

[0018] By periodically updating the logistics information and target supply time of the logistics orders to be scheduled, the forecasting model can more accurately reflect the actual status and progress of the logistics orders, thereby improving the accuracy of the forecast; in the face of the volatility and variability in the logistics scheduling process, it can adjust the strategy in a timely manner, reduce errors caused by information lags or unexpected changes, and enhance the robustness of the system; more accurate and dynamic target supply times help logistics centers make more optimized scheduling decisions, improve logistics efficiency, and reduce delivery delays.

[0019] By dynamically adjusting the evaluation weight values ​​of the global and local prediction models, the advantages of the models can be more effectively combined to improve the accuracy of the overall prediction. In the face of the complexity and variability of the logistics scheduling process, the model weights can be automatically adjusted periodically to enhance adaptability to different situations. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 Schematic diagram of the process of the intelligent supply chain logistics management method according to an embodiment of the present invention. DETAILED DESCRIPTION

[0021] To facilitate the understanding of the present invention, the present application will be described more comprehensively below with reference to the relevant drawings; the drawings show preferred embodiments of the present invention, but the present invention can be implemented in many different forms and is not limited to the embodiments described herein; on the contrary, the purpose of providing these embodiments is to enable a more thorough and comprehensive understanding of the disclosed content of the present invention.

[0022] It should be noted that the terms “vertical”, “horizontal”, “up”, “down”, “left”, “right” and similar expressions used in this document are only for illustrative purposes and do not represent the only implementation method.

[0023] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by technicians in the technical field to which the present invention belongs; the terms used in the specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention; the term "and / or" used herein includes any and all combinations of one or more related listed items.

[0024] Embodiment 1: Figure 1 It is a flow chart of the intelligent supply chain logistics management method according to an embodiment of the present invention.

[0025] like Figure 1 As shown, a smart supply chain logistics management method is applied to a logistics center and includes the following steps: S101, based on the logistics information of each logistics order to be scheduled, generate a multivariate supply feature vector [a, b, c] corresponding to its expected supply period, where a is the route feature information, b is the environment feature information, and c is the traffic feature information.

[0026] Specifically, there are currently several logistics orders to be scheduled in a logistics center. The logistics information of each logistics order to be scheduled includes but is not limited to the logistics starting point (the current location of the logistics center), the end point (the delivery address of the logistics order to be scheduled), the transfer point, the logistics supply route, the estimated supply time period, and the maximum time limit of the order. The estimated supply time period is generated by the estimated departure time and the estimated arrival time. The estimated arrival time is the time when the logistics order to be scheduled is expected to arrive at the end point. It can be pre-estimated based on the GPS navigation software by inputting the starting point and the end point. The present invention does not elaborate or limit this.

[0027] Among them, the maximum time limit of the order can be set according to the arrival time pre-set by the order customer, and can also be set according to the type of order product. For example, if the order product is fresh food, it is set according to its shelf life. The maximum time limit of the order needs to be less than its shelf life to the greatest extent to meet the satisfaction and demand of the order customer. The present invention does not impose any restrictions on this and is set according to the actual situation of the logistics order to be scheduled.

[0028] Furthermore, based on the logistics information of each logistics order to be scheduled, a multivariate supply feature vector corresponding to its expected supply period is generated, specifically including: A1. Based on the logistics supply route displayed in the GPS navigation software, the route curvature value and the longitude and latitude coordinate sequence of the key point sequence are generated to form the route feature information.

[0029] Among them, the longitude and latitude coordinate sequence is composed of the longitude and latitude coordinates of the key points that the logistics supply route passes through in sequence. The key points are set as the starting point, the transfer point, and the end point. The route curvature value is obtained by cutting the logistics supply route into several sub-route segments based on all the key points. The number of sub-route segments is the number of key points minus 1. The curvature value of each sub-route segment is calculated, and the average value is taken to obtain the route curvature value.

[0030] It should be noted that the curvature value of the sub-route segment can be calculated using the following formula:

[0031] in, is the curvature value of the ith sub-route segment, is the area of ​​the triangle formed by the three target points on the sub-route segment, , , are the straight-line distances between each pair of target points. The target points are determined by evenly dividing the sub-route segment into two segments based on its actual mileage, and setting the endpoints of the two segments as target points.

[0032] For example, the route feature information is represented as: a=[k,(x1,y1),...,( , )], where k is the route curvature value, ( , ) is the latitude and longitude coordinates of the nth key point, and n is the total number of key points in the logistics supply route.

[0033] A2. Obtain the estimated sub-supply period corresponding to each sub-route segment, which is generated by the estimated arrival time at the start and end points of the sub-route segment, generate the environmental stability value of each sub-route segment within the corresponding sub-supply period, and take the average value as the environmental characteristic information.

[0034] Among them, the environmental stability value is determined according to the environmental parameters obtained from the weather forecast API port. The environmental parameters include temperature, humidity, and rainfall. According to the pre-set environmental parameter interval-level value rule, the level value of the interval hit by the environmental parameter is determined, and the average is taken to obtain the environmental stability value. The range of the level value is set to [1,5]. Taking rainfall as an example, when setting the environmental parameter interval-level value rule, different level values ​​are divided according to different ranges of rainfall and the degree of its impact on the environment. The greater the impact, the smaller the level value. For example, when the rainfall is between 0-10mm, its impact on the environment is relatively small, so the level value is set to 5; when the rainfall exceeds 100mm, its threat to the environment increases significantly, so the level value is set to 1; the same is true for other environmental parameters, which will not be elaborated in the present invention.

[0035] A3. Based on the traffic flow data pattern of each sub-route segment in a preset historical time period, the traffic flow value of each sub-route segment in the sub-supply period is estimated as traffic characteristic information.

[0036] Specifically, the preset historical time period is set according to the duration of the sub-supply period, and can be set to 5 times the duration of the sub-supply period, so as to fully represent the historical traffic characteristics of the sub-route segment in the sub-supply period. The traffic flow data of the sub-route segment in multiple historical time periods with the same time attributes as the sub-supply period in the preset historical time period is obtained, and the traffic characteristic information is obtained by averaging. Among them, the time attributes include duration, start and end time (in hours, for example, the start and end time is 9:00-13:00, the duration is 4 hours, and the duration is added as a time attribute to avoid the situation where the start and end times are not in the same day).

[0037] S102, input the multivariate supply feature vector into a pre-trained supply time prediction model, and output the target supply time of the order to be scheduled.

[0038] Specifically, the pre-trained supply time prediction model can be obtained by: B1. Collect a large number of multivariate supply feature vectors of logistics orders to be scheduled in history, and label them with the actual supply time of the logistics orders to be scheduled; B2. Divide all the labeled multivariate supply feature vectors into a training sample set and a validation sample set. Use the training sample set to train the pre-set neural network structure, and use the validation sample set to optimize the model parameters to obtain the optimal supply time prediction model.

[0039] S103, obtaining the difference between the maximum order time limit of the logistics order to be scheduled and its target supply time as the supply urgency value. If the supply urgency value is less than the preset time threshold, the logistics order to be scheduled is marked as an abnormal order.

[0040] Among them, the preset time threshold is set according to the actual situation of the logistics order to be scheduled and the experience of experts. It is used to measure whether the logistics efficiency of the logistics order to be scheduled is in a dangerous situation, that is, how close it is to the maximum time limit of the order. It is used to judge whether the logistics order to be scheduled needs to change the scheduling route in time to speed up the efficiency of supply chain logistics, or give abnormal orders priority in selecting scheduling tools.

[0041] Specifically, the smaller the supply urgency value of the logistics order to be scheduled, the greater the urgency of the logistics order to be scheduled, and the logistics supply route needs to be adjusted as soon as possible to avoid exceeding the maximum time limit of the order and ensure the efficiency of the supply chain logistics.

[0042] S104, obtaining the supply urgency values ​​of all current abnormal orders, sorting them in descending order to obtain an abnormal order sequence, and notifying the logistics center to adjust its logistics information based on the abnormal order sequence.

[0043] Specifically, according to the order of the abnormal order sequence, the scheduling adjustment priority of each abnormal order is determined. The scheduling adjustment may include the adjustment of the logistics supply route and its bound scheduling tools. The specific adjustment is set by the management personnel of the logistics center according to the actual situation (ensuring that the adjusted logistics information can improve its supply efficiency). The present invention does not impose any restrictions on this.

[0044] S105, delivering the logistics order to be dispatched according to its logistics supply route.

[0045] The technical solutions in the above embodiments of the present application have at least the following technical effects or advantages: By integrating the characteristics of logistics orders to be scheduled on different sub-route segments, a more accurate and sophisticated multi-supply feature vector (route, environment, and traffic) of logistics orders to be scheduled is introduced, and the factors affecting logistics supply time are considered more comprehensively, thereby improving the accuracy of the prediction model and enabling the prediction model to dynamically adapt to the actual changes in supply chain logistics; By calculating the difference between the predicted supply time and the maximum time limit of differentiated orders for different orders, determining the supply urgency value, and generating and promptly processing abnormal orders, it is possible to discover and adjust potential logistics delays in advance during delivery, thereby optimizing logistics scheduling efficiency. Thus, more accurate predictions and more efficient logistics scheduling can help reduce order delays and improve order customer satisfaction.

[0046] Embodiment 2: In embodiment 1, each order in the actual supply chain logistics scenario may contain multiple sub-route segments, that is, multiple transit distribution logistics are required. The environment, traffic and route characteristics of each sub-route segment may be quite different. A single prediction model may not be able to accurately capture the characteristic differences of each sub-route segment, and it may overemphasize the overall characteristics and ignore the local characteristics, resulting in deviations in the prediction results.

[0047] Therefore, the embodiments of the present application are optimized to a certain extent based on the above embodiments.

[0048] In some embodiments, the output end of the supply time prediction model is also connected to a pre-trained verification prediction model, and the verification prediction model includes two input ends, which are respectively used to input the target supply time output by the supply time prediction model and the verification supply feature vector sequence of the logistics order to be scheduled. The verification supply feature vector sequence is obtained as follows: C1. Obtain the logistics information corresponding to all sub-route segments of the logistics order to be scheduled, and generate a multivariate supply feature vector corresponding to each sub-route segment.

[0049] Specifically, the logistics information of the sub-route segment includes the starting point, the end point, the sub-supply route, and the expected supply sub-time period (which can be estimated based on the GPS navigation software) of the sub-route segment, and generates a multivariate supply feature vector corresponding to each sub-route segment in its expected supply sub-time period. The generation method can refer to step S101. Since the generation method is similar, the present invention will not elaborate on it.

[0050] C2. Combine the multivariate supply feature vectors of all sub-route segments to obtain a verification supply feature vector sequence of the logistics order to be scheduled. The verification supply feature vector sequence includes at least one verification supply feature vector, and each verification supply feature vector corresponds to a sub-route segment.

[0051] In some embodiments, the pre-trained verification prediction model is obtained by: D1. Collect the verification supply feature vectors corresponding to all sub-route segments of a large number of logistics orders to be scheduled in history, and label them with the actual supply time of the sub-route segment (set as the time between the departure time of the order at the starting point of the sub-route segment and the departure time at the end point); D2. Divide all the marked verification supply feature vectors into a training sample set and a verification sample set, use the training sample set to train the pre-set neural network structure, and use the verification sample set to optimize the model parameters to obtain the optimal verification prediction model.

[0052] In some embodiments, the input end of the verification prediction model also includes a splitting module, which is used to split the input verification supply feature vector sequence to obtain a number of verification supply feature vectors. The verification prediction module predicts the several verification supply feature vectors after the splitting, and outputs the verification supply time corresponding to each verification supply feature vector; the output end of the verification prediction model also includes an accumulation module, which is used to: obtain the verification supply time corresponding to each verification supply feature vector and sum it to obtain the target verification supply time, and combine the target verification supply time, target supply time, and verification supply time sequence as the output of the verification prediction model. Among them, the verification supply time sequence is composed of the verification supply time corresponding to all the verification supply feature vectors of the logistics order to be scheduled, and the verification supply time corresponds to the sub-route segment one by one.

[0053] In some embodiments, step S102 further includes: S201, obtaining the target verification supply time, target supply time, and verification supply time series output by the verification prediction model.

[0054] S202, taking the weighted sum of the target verification supply time and the target supply time as the new target supply time for the logistics order to be scheduled, replacing the original target supply time.

[0055] Specifically, the weight values ​​corresponding to the target verification supply time and the target supply time are set according to actual conditions and historical experience, and are used to represent the credibility of the supply time prediction model and the verification prediction model respectively, and the sum of their weight values ​​is kept at 1.

[0056] The technical solutions in the above embodiments of the present application have at least the following technical effects or advantages: By introducing the verification prediction model at the output end of the supply time prediction model, a more detailed and differentiated prediction is made for each sub-route segment of the logistics order to be scheduled, and the verification prediction model and the supply time prediction model are mutually verified and combined, which reduces the possible errors of a single model, enhances the robustness of the entire system, and further improves the accuracy of the prediction model. Since there are different supply chain nodes in the order supply route, directly using the global supply time training model will ignore possible large errors. The target supply time is obtained by combining local predictions with global predictions. The updated target supply time is more accurate, which helps the logistics center make more optimized scheduling decisions and improve logistics efficiency.

[0057] Example 3: There are a lot of dynamic changes and uncertainties in the logistics scheduling process, such as traffic congestion, weather changes, cargo transshipment, etc. These changes may cause the original distribution plan to fail to be completed on time. The dynamic changes in the logistics process make the prediction results based on static information unreliable. Therefore, a method that can dynamically adjust the prediction and scheduling strategies is needed to improve logistics efficiency.

[0058] Therefore, the embodiments of the present application are optimized to a certain extent based on the above embodiments.

[0059] In some embodiments, step S105 further includes: S106, based on each logistics order to be scheduled, periodically update the target supply time of the logistics order to be scheduled according to the preset time interval, and repeat steps S103 to S105 until the delivery of the logistics order to be scheduled is completed.

[0060] In this way, the target supply time of different logistics orders to be scheduled can be updated dynamically and periodically, thereby updating the abnormal order sequence, overcoming the scheduling efficiency risks brought about by the volatility and variability of the logistics scheduling process, timely discovering abnormal orders and reminding the logistics center of possible distribution cycle risks, enabling the logistics center to take corresponding remedial measures in a timely manner, thereby improving distribution efficiency and completing delivery before the maximum time limit of the order.

[0061] Specifically, based on each logistics order to be scheduled, the target supply time of the logistics order to be scheduled is periodically updated according to the preset time interval, specifically including: S301, update the logistics information of the logistics order to be scheduled according to the logistics status of the current logistics order to be scheduled, and repeat steps S101 to S102 based on the latest logistics information, and determine the target supply time as the post-supply time.

[0062] Specifically, the logistics status includes dynamic in-transit and static transit. When the logistics status is dynamic in-transit, the logistics information of the logistics order to be dispatched is updated, including: Get the last closest key point of the logistics order to be scheduled, update the logistics starting point in the current logistics information to the key point, filter out the transit points that have been passed, and update the logistics supply route, transit points, and estimated supply period (starting time from the last key point - estimated arrival time) according to the new logistics starting point; When the logistics status is static transit, update the logistics information of the logistics order to be scheduled, including: Update the logistics starting point in the current logistics information to the current transit point of the logistics order to be scheduled, filter out the transit points that have been passed, and update the logistics supply route, transit point, and estimated supply period (estimated departure time of the current transit point - estimated arrival time) according to the new logistics starting point.

[0063] S302, obtaining the lead time of the logistics order to be scheduled, and updating the sum of the lead time and the post-target supply time as the target supply time of the logistics order to be scheduled (the time from the start of delivery to the end of delivery).

[0064] This ensures that the target supply time can be updated according to the actual progress and possible changes of the logistics order.

[0065] Specifically, obtaining the lead time of the logistics order to be scheduled includes: If the logistics status of the current logistics order to be scheduled is dynamically in transit, the difference between the time node when the logistics order to be scheduled starts to be delivered and the time node when the logistics order to be scheduled starts from the last closest key point is used as the lead time; If the logistics status of the current logistics order to be scheduled is static transit, the difference between the time node when the logistics order to be scheduled starts delivery and the time node when the logistics order to be scheduled is expected to depart at the current transit point will be used as the lead supply time.

[0066] The technical solutions in the above embodiments of the present application have at least the following technical effects or advantages: By periodically updating the logistics information and target supply time of the logistics orders to be scheduled, the forecasting model can more accurately reflect the actual status and progress of the logistics orders, thereby improving the accuracy of the forecast; In the face of the volatility and variability in the logistics scheduling process, strategies can be adjusted in a timely manner to reduce errors caused by information lags or unexpected changes and enhance the robustness of the system; more accurate and dynamic target supply times can help logistics centers make more optimized scheduling decisions, improve logistics efficiency and reduce delivery delays.

[0067] Embodiment 4: In actual logistics scheduling, the verification prediction model and the supply time prediction model each have their own advantages and disadvantages. In Embodiment 2, the predicted evaluation weight value is fixed. The fixed weight value may not be able to adapt to the dynamic changes in the logistics environment, which may cause the prediction results to deviate from the actual situation and affect the order delivery efficiency.

[0068] Therefore, the embodiments of the present application are optimized to a certain extent based on the above embodiments.

[0069] In some embodiments, step S106 further includes: periodically updating the weight values ​​of the target verification supply time and the target supply time, specifically including: S401, respectively calculate the difference between the target verification supply time and the target supply time in step S201 in the previous cycle and the target supply time updated in the current cycle, and record them as the first difference and the second difference. In this way, the prediction accuracy of the two prediction models (the verification prediction model and the supply time prediction model) in the previous cycle is evaluated.

[0070] S402, when the first difference is greater than the second difference (the credibility of the prediction model needs to be further verified): E1. Obtain all sub-route segments that the logistics order to be scheduled has passed through, record them as invalid sub-route segments, and obtain the actual supply time of each invalid sub-route segment; E2. Obtain the verification supply time series in the previous cycle, match the verification supply time corresponding to each failed sub-route segment, and calculate the difference between each failed sub-route segment and its corresponding verification supply time, and take the average value to obtain the verification credibility value; Among them, the larger the verification credibility value, the lower the credibility of the verification prediction model; E3. If the verification credibility value is greater than the credibility threshold, it means that the accuracy of the verification prediction model is low. The weight value of the target verification supply time is reduced according to the preset ratio, and the weight value of the target supply time is adaptively adjusted; otherwise, the original weight value remains unchanged; S403, when the first difference is less than the second difference: execute steps E1 to E2, if the verification credibility value is greater than the credibility threshold, keep the original weight value unchanged; otherwise, increase the weight value of the target verification supply time according to the preset ratio, and adaptively adjust the weight value of the target supply time.

[0071] The preset ratio is set according to actual conditions, for example, it is set to 0.2 to achieve a fine-tuning effect.

[0072] The technical solutions in the above embodiments of the present application have at least the following technical effects or advantages: By dynamically adjusting the evaluation weight values ​​of the global and local prediction models, the advantages of the models can be more effectively combined to improve the accuracy of the overall prediction. In the face of the complexity and variability of the logistics scheduling process, the model weights can be automatically adjusted periodically to enhance adaptability to different situations.

[0073] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A smart supply chain logistics management method, characterized in that: include: S101, based on the logistics information of each logistics order to be scheduled, generate a multivariate supply feature vector [a, b, c] corresponding to its expected supply period, where a is route feature information, b is environmental feature information, and c is traffic feature information; the logistics information includes logistics starting point, end point, transfer point, logistics supply route, expected supply period, and maximum order time limit; S102, inputting the multivariate supply feature vector into a pre-trained supply time prediction model, and outputting the target supply time of the order to be scheduled; S103, obtaining the difference between the maximum order time limit of the logistics order to be scheduled and its target supply time as the supply urgency value. If the supply urgency value is less than the preset time threshold, the logistics order to be scheduled is marked as an abnormal order; S104, obtaining the supply urgency values ​​of all current abnormal orders, sorting them in descending order to obtain an abnormal order sequence, and notifying the logistics center to adjust its logistics information based on the abnormal order sequence; S105, delivering the logistics order to be dispatched according to its logistics supply route.

2. The intelligent supply chain logistics management method according to claim 1, characterized in that: The method of generating a multivariate supply feature vector corresponding to the expected supply period based on the logistics information of each logistics order to be scheduled specifically includes: A1. Generate route curvature values ​​and longitude and latitude coordinate sequences of key point sequences according to the logistics supply route to form route feature information; the longitude and latitude coordinate sequences are composed of the longitude and latitude coordinates of key points that the logistics supply route passes through in sequence; the route curvature values ​​are obtained by cutting the logistics supply route into several sub-route segments according to all key points, calculating the curvature value of each sub-route segment, and taking the average value to obtain the route curvature value; A2. Obtain the sub-supply period corresponding to each sub-route segment, generate the environmental stability value of each sub-route segment in the corresponding sub-supply period, and take the average value as the environmental characteristic information; A3. Based on the traffic flow data pattern of each sub-route segment in a preset historical time period, the traffic flow value of each sub-route segment in the sub-supply period is estimated as traffic characteristic information.

3. The intelligent supply chain logistics management method according to claim 1, characterized in that: The pre-trained supply time prediction model is obtained by: B1. Collect a large number of multivariate supply feature vectors of logistics orders to be scheduled in history, and label them with the actual supply time of the logistics orders to be scheduled; B2. Divide all the labeled multivariate supply feature vectors into a training sample set and a validation sample set. Use the training sample set to train the pre-set neural network structure, and use the validation sample set to optimize the model parameters to obtain the optimal supply time prediction model.

4. The intelligent supply chain logistics management method according to claim 2, characterized in that: The output end of the supply time prediction model is also connected to a pre-trained verification prediction model. The verification prediction model includes two input ends, which are respectively used to input the target supply time output by the supply time prediction model and the verification supply feature vector sequence of the logistics order to be scheduled. The verification supply feature vector sequence is obtained as follows: C1. Obtain the logistics information corresponding to all sub-route segments of the logistics order to be scheduled, and generate a multivariate supply feature vector corresponding to each sub-route segment; C2. Combine the multivariate supply feature vectors of all sub-route segments to obtain a verification supply feature vector sequence of the logistics order to be scheduled. The verification supply feature vector sequence includes at least one verification supply feature vector, and each verification supply feature vector corresponds to a sub-route segment.

5. The intelligent supply chain logistics management method according to claim 4, characterized in that: The pre-trained verification prediction model is obtained by: D1. Collect the verification supply feature vectors corresponding to all sub-route segments of a large number of logistics orders to be scheduled in history, and label them with the actual supply time of the sub-route segments; D2. Divide all the marked verification supply feature vectors into a training sample set and a verification sample set, use the training sample set to train the pre-set neural network structure, and use the verification sample set to optimize the model parameters to obtain the optimal verification prediction model; The input end of the verification prediction model also includes a splitting module, which is used to split the input verification supply feature vector sequence to obtain a number of verification supply feature vectors. The verification prediction module predicts the several verification supply feature vectors after the splitting, and outputs the verification supply time corresponding to each verification supply feature vector; The output end of the verification prediction model also includes an accumulation module, which is used to: obtain the verification supply time corresponding to each verification supply feature vector and sum them to obtain the target verification supply time, and combine the target verification supply time, the target supply time, and the verification supply time series as the output of the verification prediction model; Among them, the verification supply time series is composed of the verification supply time corresponding to all the verification supply feature vectors of the logistics orders to be scheduled, and the verification supply time corresponds to the sub-route segment one by one.

6. The intelligent supply chain logistics management method according to claim 5, characterized in that: The S102 further includes: S201, obtaining the target verification supply time, target supply time, and verification supply time series output by the verification prediction model; S202, taking the weighted sum of the target verification supply time and the target supply time as the new target supply time for the logistics order to be scheduled, replacing the original target supply time.

7. The intelligent supply chain logistics management method according to claim 6, characterized in that: The S105 further includes: S106, based on each logistics order to be scheduled, periodically update the target supply time of the logistics order to be scheduled according to the preset time interval, and repeat steps S103 to S105 until the delivery of the logistics order to be scheduled is completed.

8. The intelligent supply chain logistics management method according to claim 7, characterized in that: The S106 specifically includes: S301, according to the logistics status of the current logistics order to be scheduled, update the logistics information of the logistics order to be scheduled, and based on the latest logistics information, repeat steps S101 to S102 to determine the target supply duration as the post-supply duration; wherein the logistics status includes dynamic in-transit and static transit. When the logistics status is dynamic in-transit, update the logistics information of the logistics order to be scheduled, including: Obtain the last closest key point of the logistics order to be scheduled, update the logistics starting point in the current logistics information to the key point, filter out the transit points that have been passed, and update the logistics supply route, transit points, and expected supply period according to the new logistics starting point; When the logistics status is static transit, the logistics information of the logistics order to be scheduled is updated, including: updating the logistics starting point in the current logistics information to the current transit point of the logistics order to be scheduled, filtering out the transit points that have been passed, and updating the logistics supply route, transit point, and expected supply period according to the new logistics starting point; S302, obtaining the leading supply time of the logistics order to be scheduled, and updating the sum of the leading supply time and the post-target supply time as the target supply time of the logistics order to be scheduled.

9. The intelligent supply chain logistics management method according to claim 8, characterized in that: The lead time for obtaining the logistics order to be scheduled includes: If the logistics status of the current logistics order to be scheduled is dynamically in transit, the difference between the time node when the logistics order to be scheduled starts to be delivered and the time node when the logistics order to be scheduled starts from the last closest key point is used as the lead time; If the logistics status of the current logistics order to be scheduled is static transit, the difference between the time node when the logistics order to be scheduled starts delivery and the time node when the logistics order to be scheduled is expected to depart at the current transit point will be used as the lead supply time.

10. The intelligent supply chain logistics management method according to claim 7, characterized in that: The step S106 further includes: periodically updating the weight value of the target verification supply time and the target supply time, specifically including: S401, respectively calculating the difference between the target verification supply time and the target supply time in step S201 in the previous cycle and the target supply time updated in the current cycle, and recording them as the first difference and the second difference; S402, when the first difference is greater than the second difference: E1, obtain all the sub-route segments that the logistics order to be scheduled has passed through, record them as invalid sub-route segments, and obtain the actual supply time of each invalid sub-route segment; E2, obtain the verification supply time series in the previous cycle, match the verification supply time corresponding to each invalid sub-route segment, and calculate the difference between each invalid sub-route segment and its corresponding verification supply time, and take the average value to obtain the verification credibility value; E3, if the verification credibility value is greater than the credibility threshold, it means that the accuracy of the verification prediction model is low, and the weight value of the target verification supply time is reduced according to the preset ratio, and the weight value of the target supply time is adaptively adjusted; otherwise, keep the original weight value unchanged; S403, when the first difference is less than the second difference: execute steps E1 to E2, if the verification credibility value is greater than the credibility threshold, keep the original weight value unchanged; otherwise, increase the weight value of the target verification supply time according to the preset ratio, and adaptively adjust the weight value of the target supply time.

Images