Order arrival time prediction model training method, order scheduling method and device

Abstract

The invention discloses a training method of an order arrival time prediction model, and an order scheduling method and device, and relates to the technical field of computers. A specific embodiment of the training method of the order arrival time prediction model comprises the steps of obtaining delivery time and delivery-related data of each first historical order in a first historical order set; carrying out feature extraction on the distribution related data to obtain distribution related features; constructing a training sample set based on the delivery time and delivery related characteristics of each first historical order in the first historical order set, and training according to the training sample set to obtain an order arrival time prediction model; the accuracy of the prediction result of the order arrival time can be improved. A specific embodiment of the order scheduling method comprises the steps of obtaining a to-be-processed order set, and determining predicted arrival time of each to-be-processed order according to an order arrival time prediction model; and the to-be-processed order set is scheduled based on the predicted arrival time of each to-be-processed order, so that the scheduling cost can be reduced, and the user experience can be improved.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to a training method for an order arrival time prediction model, an order scheduling method and a device. Background Art

[0002] The existing solution only roughly calculates the order arrival time based on the pre-set driving speed and fixed handover time, and schedules the order based on this order arrival time.

[0003] In the process of implementing the present invention, the inventors found that there are at least the following problems in the related art:

[0004] The calculation results of order arrival time are inaccurate, and order scheduling does not meet user needs, resulting in high costs and poor user experience. Summary of the invention

[0005] In view of this, the embodiments of the present invention provide a training method, an order scheduling method and a device for an order arrival time prediction model, which can improve the accuracy of the prediction results of the order arrival time, and the order scheduling is more in line with user needs, reducing scheduling costs and improving user experience.

[0006] To achieve the above object, according to one aspect of an embodiment of the present invention, a training method for an order arrival time prediction model is provided, comprising:

[0007] Obtaining the delivery time and delivery-related data of each first historical order in the first historical order set; wherein the delivery-related data includes at least one of the following: delivery address data, delivery item data, delivery vehicle data, delivery station data, delivery refusal behavior data, delivery person data, and weather data;

[0008] Extract features from delivery-related data to obtain delivery-related features;

[0009] A training sample set is constructed based on the delivery time and delivery-related features of each first historical order in the first historical order set, and an order arrival time prediction model is obtained by training the training sample set.

[0010] Optionally, before obtaining the delivery time and delivery-related data of each first historical order in the first historical order set, the method further includes:

[0011] A first historical order with missing values ​​in the first historical order set is identified, and the first historical order with missing values ​​is deleted or the missing values ​​in the first historical order with missing values ​​are filled.

[0012] Optionally, after obtaining the delivery time and delivery-related data of each first historical order in the first historical order set, the method further includes:

[0013] Determine a preliminary estimated arrival time of the first historical order based on the delivery-related data of the first historical order and the real-time map data;

[0014] Calculate the difference between the preliminary estimated arrival time and the delivery time of the first historical order, and in response to the difference exceeding a preset threshold, treat the first historical order as a time-abnormal order;

[0015] The time-abnormal orders are removed from the first historical order set.

[0016] Optionally, before removing the time-abnormal order from the first historical order set, the method further includes:

[0017] Determine that the delivery address data for orders with abnormal time is incomplete.

[0018] Optionally, the first historical order in the first historical order set is an order within a first time range; and the method further includes:

[0019] Obtaining the delivery time and delivery-related data of each second historical order in the second historical order set within a second time range; wherein the second time range is different from the first time range;

[0020] Performing feature extraction on the delivery-related data of the second historical order to obtain delivery-related features of the second historical order;

[0021] A test sample set is constructed based on the delivery time and delivery-related features of each second historical order in the second historical order set, and the order arrival time prediction model is evaluated using the test sample set.

[0022] According to a second aspect of an embodiment of the present invention, there is provided an order scheduling method, comprising:

[0023] Obtain a set of pending orders, and determine a predicted arrival time of each pending order according to an order arrival time prediction model; wherein the order arrival time prediction model is obtained by any method in the first aspect of the embodiment of the present invention;

[0024] The set of pending orders is scheduled based on the predicted arrival time of each pending order.

[0025] Optionally, the set of pending orders is scheduled based on the predicted arrival time of each pending order, including:

[0026] Based on the predicted arrival time of each pending order and the adaptive large neighborhood search algorithm, a scheduling strategy corresponding to the pending order set is constructed; wherein the scheduling strategy includes vehicle information and path information;

[0027] The order set to be processed is scheduled according to the scheduling strategy.

[0028] Optionally, the set of pending orders is scheduled according to a scheduling strategy, including:

[0029] Modify the path information in the scheduling strategy to obtain a new scheduling strategy;

[0030] Based on the predicted arrival time of each pending order and the new scheduling strategy, it is determined whether the new scheduling strategy meets the preset time window constraint, and the new scheduling strategy that meets the preset time window constraint is used as the target scheduling strategy to schedule the pending order set.

[0031] According to a third aspect of an embodiment of the present invention, there is provided a training device for an order arrival time prediction model, comprising:

[0032] A data acquisition module, used to acquire the delivery time and delivery-related data of each first historical order in the first historical order set; wherein the delivery-related data includes at least one of the following: delivery address data, delivery item data, delivery vehicle data, delivery station data, delivery refusal behavior data, delivery person data, and weather data;

[0033] A feature extraction module is used to extract features from delivery-related data to obtain delivery-related features;

[0034] The model training module is used to construct a training sample set based on the delivery time and delivery-related features of each first historical order in the first historical order set, and obtain an order arrival time prediction model through training based on the training sample set.

[0035] According to a fourth aspect of an embodiment of the present invention, there is provided an order scheduling device, comprising:

[0036] A model prediction module, used to obtain a set of pending orders, and determine a predicted arrival time of each pending order according to an order arrival time prediction model; wherein the order arrival time prediction model is obtained by any method in the first aspect of the embodiment of the present invention;

[0037] The order scheduling module is used to schedule the pending order set based on the predicted arrival time of each pending order.

[0038] According to a fifth aspect of an embodiment of the present invention, there is provided an electronic device, including:

[0039] one or more processors;

[0040] a storage device for storing one or more programs,

[0041] When one or more programs are executed by one or more processors, the one or more processors implement the method of any of the above embodiments.

[0042] According to a sixth aspect of an embodiment of the present invention, there is provided a computer-readable medium on which a computer program is stored, and when the program is executed by a processor, the method of any of the above embodiments is implemented.

[0043] According to a seventh aspect of an embodiment of the present invention, there is provided a computer program product, including a computer program, which implements the method of any of the above embodiments when the computer program is executed by a processor.

[0044] An embodiment of the above invention has the following advantages or beneficial effects: by obtaining the delivery time and delivery-related data of each first historical order in the first historical order set; wherein the delivery-related data includes at least one of the following: delivery address data, delivery item data, delivery vehicle data, delivery station data, refusal to deliver behavior data, delivery personnel data, weather data; feature extraction of the delivery-related data to obtain delivery-related features; constructing a training sample set based on the delivery time and delivery-related features of each first historical order in the first historical order set, and obtaining an order arrival time prediction model based on the training sample set; thereby improving the accuracy of the prediction result of the order arrival time. Obtain a set of pending orders, and determine the predicted arrival time of each pending order according to the order arrival time prediction model; wherein the order arrival time prediction model is obtained by any method in the first aspect of the present invention; scheduling the pending order set based on the predicted arrival time of each pending order, so that the order scheduling can be more in line with user needs, reduce scheduling costs, and improve user experience.

[0045] The further effects of the above-mentioned non-conventional optional manner will be described below in conjunction with the specific implementation manner. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] The accompanying drawings are used to better understand the present invention and do not constitute an improper limitation of the present invention.

[0047] Figure 1 is a schematic diagram of the main process of a training method for an order arrival time prediction model according to an embodiment of the present invention;

[0048] Figure 2 is a schematic diagram of a training method for an order arrival time prediction model according to a preferred embodiment of the present invention;

[0049] Figure 3 is a schematic diagram of an order scheduling method according to an embodiment of the present invention;

[0050] Figure 4 is a schematic diagram of an order scheduling method according to a preferred embodiment of the present invention;

[0051] Figure 5is a schematic diagram of main modules of a training device for an order arrival time prediction model according to an embodiment of the present invention;

[0052] Figure 6 is a schematic diagram of main modules of an order scheduling device according to an embodiment of the present invention;

[0053] Figure 7 is an exemplary system architecture diagram to which embodiments of the present invention may be applied;

[0054] Figure 8 It is a schematic diagram of the structure of a computer system of a terminal device or a server suitable for implementing an embodiment of the present invention. DETAILED DESCRIPTION

[0055] The following is a description of exemplary embodiments of the present invention in conjunction with the accompanying drawings, including various details of the embodiments of the present invention to facilitate understanding, which should be considered as merely exemplary. Therefore, it should be recognized by those of ordinary skill in the art that various changes and modifications may be made to the embodiments described herein without departing from the scope and spirit of the present invention. Similarly, for clarity and conciseness, the description of well-known functions and structures is omitted in the following description.

[0056] It should be noted that the acquisition, storage and application of personal information etc. involved in the embodiments of the present invention are in compliance with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0057] The existing solution is to roughly calculate the order arrival time based on the preset driving speed and fixed handover time, and then schedule the order based on the order arrival time. The calculation result of the order arrival time is inaccurate, and the order scheduling does not meet the user's needs, resulting in high costs and poor user experience.

[0058] In view of this, according to a first aspect of an embodiment of the present invention, a method for training an order arrival time prediction model is provided.

[0059] Figure 1 FIG. 1 is a schematic diagram of the main process of the training method of the order arrival time prediction model according to an embodiment of the present invention. Figure 1 As shown, the training method of the order arrival time prediction model according to an embodiment of the present invention mainly includes the following steps S101 to S103.

[0060] Step S101, obtaining the delivery time and delivery-related data of each first historical order in the first historical order set; wherein the delivery-related data includes at least one of the following: delivery address data, delivery item data, delivery vehicle data, delivery site data, delivery refusal behavior data, delivery person data, and weather data.

[0061] The first historical order set refers to the collection of all completed order data, which have been fully processed in the entire business history and all related operations (such as delivery, payment, etc.) have been completed. The first historical order is a single order in the first historical order set. The delivery time of the first historical order refers to the time period from order confirmation to delivery completion. Delivery-related data includes multiple aspects of order information, covering the details of the entire delivery process. Specifically, delivery address data includes the specific location where the customer receives the goods; delivery item data includes the type and quantity of delivered goods; delivery vehicle data includes detailed information on the delivery vehicle or traffic conditions; delivery site data includes various transfer or storage points involved in the delivery process; refusal behavior data includes the situation where the customer refuses to accept the goods and related information; delivery personnel data includes detailed information on the personnel responsible for delivery; weather data includes meteorological conditions during delivery.

[0062] The above data can be obtained in multiple ways, such as extracting the delivery time and delivery-related data of historical orders from the order management system, cooperating with logistics companies to obtain vehicle and delivery site information, obtaining records of delivery rejections from the customer service system, collecting real-time data during the delivery process through sensors and GPS technology (Global Positioning System), and obtaining relevant weather information through meteorological service providers, etc.

[0063] Step S102, extracting features from the delivery-related data to obtain delivery-related features.

[0064] Delivery-related features are key information extracted from delivery-related data for analyzing the delivery process. The feature extraction process includes identifying and extracting meaningful indicators from delivery-related data. For example, from the delivery address data, the coordinates of the geographic location, the distribution density of the delivery area, etc. can be extracted; from the delivery item data, the category, volume, weight and other features of the item can be extracted; from the delivery vehicle data, the type of vehicle, carrying capacity, average speed and other information can be extracted; from the delivery personnel data, the features that can be extracted include years of experience, delivery success rate, service quality score, etc.; from the weather data, the bad weather can be extracted, etc.

[0065] Specifically, feature extraction can extract numerical features through data aggregation and statistical analysis, such as calculating average delivery time, delivery success rate, etc. Machine learning techniques can also be used to identify and extract the most representative features, such as feature selection algorithms and dimensionality reduction techniques, and reduce high-dimensional data to a few main features through methods such as principal component analysis.

[0066] Step S103, constructing a training sample set based on the delivery time and delivery-related features of each first historical order in the first historical order set, and obtaining an order arrival time prediction model through training based on the training sample set.

[0067] All order data in the first historical order set are organized into a training sample set. The sample includes the delivery time of each first historical order and its delivery-related features, such as delivery address, delivery items, delivery vehicles, delivery sites, refusal to deliver, delivery personnel, and weather data. These features provide multi-dimensional information that affects delivery time. In the process of constructing the training sample set, these delivery-related features are paired with the actual delivery time to form a data set with input and output labels. The input includes a summary of all relevant data or extracted key information, and the output is the delivery time. Through such a training sample set, rich historical data can be provided for the machine learning model, thereby capturing the impact pattern of various factors on delivery time.

[0068] The order arrival time prediction model can be trained using the training sample set. Machine learning algorithms can be applied to this task, such as linear regression, decision tree, random forest, or deep learning models. The goal of the training process is to enable the model to learn the relationship between features and delivery time, so as to accurately predict the delivery time of future orders. By performing multiple rounds of training and verification on the training sample set, the performance of the model can be optimized and the accuracy of the prediction can be improved. The trained order arrival time prediction model will be able to predict the arrival time of new orders based on their features, thereby helping to optimize logistics management and improve customer service.

[0069] The embodiment of the present invention obtains the delivery time and delivery-related data of each first historical order in the first historical order set; wherein the delivery-related data includes at least one of the following: delivery address data, delivery item data, delivery vehicle data, delivery site data, delivery refusal behavior data, delivery personnel data, and weather data; performs feature extraction on the delivery-related data to obtain delivery-related features; constructs a training sample set based on the delivery time and delivery-related features of each first historical order in the first historical order set, and obtains an order arrival time prediction model through training based on the training sample set; thereby, the accuracy of the prediction result of the order arrival time can be improved, and the order scheduling is more in line with user needs, reducing the scheduling cost and improving the user experience.

[0070] Optionally, before obtaining the delivery time and delivery-related data of each first historical order in the first historical order set, it also includes: identifying the first historical orders in the first historical order set with missing values, deleting the first historical orders with missing values, or filling the missing values ​​in the first historical orders with missing values.

[0071] Identifying and processing missing values ​​is a specific way of data preprocessing. First, scan the entire order set to determine which first historical orders have missing values ​​in their data. These missing values ​​may appear in multiple fields, such as delivery addresses, delivery items, delivery vehicles, etc., affecting the integrity of the data and the accuracy of the analysis. Processing missing values ​​can be to delete the first historical orders with missing values. This method is simple, direct and efficient, and is suitable for when there are fewer missing values ​​or when the data set is large enough. It can also be to fill in the missing values ​​in the first historical orders with missing values. The filling method can use statistical methods, such as the mean, median or mode to fill in missing numerical data, or fill in categorical data with the most common category. In addition, more complex methods can also be used, such as prediction filling based on similar orders, or interpolation through machine learning models, which is conducive to preserving the integrity of the data set and providing more comprehensive and accurate analysis data.

[0072] Optionally, after obtaining the delivery time and delivery-related data of each first historical order in the first historical order set, it also includes: determining the preliminary estimated arrival time of the first historical order based on the delivery-related data and real-time map data of the first historical order; calculating the difference between the preliminary estimated arrival time and the delivery time of the first historical order, and in response to the difference exceeding a preset threshold, treating the first historical order as a time-abnormal order; and removing the time-abnormal order from the first historical order set.

[0073] In the process of determining the preliminary estimated arrival time of the order based on the delivery-related data of the first historical order and the real-time map data, various relevant features of the first historical order, such as the delivery address, delivery items, the current location of the delivery vehicle and traffic conditions, are comprehensively considered. This information is updated through the real-time map data to generate a relatively accurate preliminary estimated arrival time. Specifically, the geographic information system (GIS) and navigation algorithm can be used to estimate the time from the current delivery location to the final destination. After the preliminary estimated arrival time is determined, it is compared with the actual recorded delivery time, and the difference between the two is calculated to identify which first historical orders have significant differences in actual delivery time and preliminary estimated arrival time, and a preset threshold is used to determine whether the difference exceeds the normal range. If the difference of a first historical order exceeds this preset threshold, it can be marked as a time-abnormal order, which may indicate problems in the delivery process, such as traffic congestion, delivery delays or other unexpected situations. Finally, these time-abnormal orders are removed from the first historical order set, and orders with serious time deviations are eliminated, so that the data based on which subsequent analysis and model training are based is accurate and in line with expectations. This processing helps to improve data quality, optimize the performance of the delivery prediction model, and reduce potential deviations.

[0074] Optionally, before removing the time-abnormal order from the first historical order set, the method further includes: determining that the delivery address data of the time-abnormal order is incomplete.

[0075] If the delivery address data of time-abnormal orders is incomplete, it may cause deviations in the calculation of the preliminary estimated arrival time, resulting in inaccurate effects on the difference in delivery time. Therefore, it is necessary to conduct a detailed check on these time-abnormal orders to confirm whether the delivery address contains all the necessary details, such as street, house number, zip code and other information. If it is determined that the delivery address data of time-abnormal orders is incomplete, these time-abnormal orders are removed from the first historical order set. This processing can retain the first historical orders with complete data, further improve data quality, and provide a more reliable basis for subsequent analysis and model training.

[0076] Optionally, the first historical order in the first historical order set is an order within a first time range; the method also includes: obtaining the delivery time and delivery-related data of each second historical order in the second historical order set within a second time range; wherein the second time range is different from the first time range; performing feature extraction on the delivery-related data of the second historical order to obtain delivery-related features of the second historical order; constructing a test sample set based on the delivery time and delivery-related features of each second historical order in the second historical order set, and using the test sample set to evaluate the order arrival time prediction model.

[0077] In order to further verify the accuracy of the order arrival time prediction model, the data of the second historical order set can be obtained within the second time range. The second historical order set contains historical orders in a time period different from the first time range. The data of these orders include delivery time and delivery-related data. These data are similar to the order data in the first time range, but the time span is different. After obtaining the order data in the second time range, the delivery-related data of these orders are firstly extracted. This process is similar to that of processing the first historical order set. Representative delivery-related features are extracted from multiple data such as delivery address, delivery items, delivery vehicles, delivery sites, refusal to deliver, delivery personnel and weather. These features are used to describe the delivery situation of the order and provide a basis for the model testing. A test sample set is constructed based on the delivery time of each second historical order in the second historical order set and the extracted delivery-related features. The test sample set includes the features of these second historical orders and their actual delivery time, which is used to evaluate the performance of the trained order arrival time prediction model. By inputting the test sample set into the prediction model, the prediction accuracy of the model for new data can be evaluated, thereby verifying the applicability and stability of the model in different time ranges, identifying its possible deficiencies, and providing a basis for further optimizing the model.

[0078] Figure 2 Schematic diagram of a training method for an order arrival time prediction model according to a preferred embodiment of the present invention. Figure 2 As shown, first, relevant data can be extracted from the database to obtain original training data, where the relevant data include address data such as the province, city, and district of the order, time data such as the scheduled pick-up and delivery time, cargo data such as the industry and category of the goods, vehicle data such as vehicle type, weather data, delivery personnel data, whether the delivery point needs to go up and down the stairs, whether there is a history of rejection, etc., as well as actual pick-up and delivery time data. Then, data preprocessing is performed to clean up data with empty data such as transportation task number and delivery time. Fence correction processing is performed to retain data where the delivery operation location is a certain distance away from the customer location. The navigation distance, navigation time, and preliminary handover time of each order are calculated, and extreme data with a large difference between the actual arrival time and the preliminary estimated arrival time are removed, that is, time anomaly data is removed. The task data with complete arrival points is retained to ensure that the training data is not disturbed by the removal of abnormal data. Other corrections can also be made to exclude factors that affect the delivery time estimation, such as the delivery personnel's lunch break. The training set and test set are divided according to the time dimension and model training optimization is performed. Multiple models are used for training using the training set, and the model results are tested on the test set, and the model parameters are optimized. The effect of the model was evaluated, and finally the LGBM model (Light Gradient Boosting Machine) was selected as the final ETA (Estimated Time of Arrival) prediction model.

[0079] The selection of model feature categories is shown in Table 1 below (arranged in descending order of importance). In this embodiment, when selecting a suitable machine learning method to construct an ETA time estimation model, the training results of multiple models are compared. Three months of data are selected as the training set data, and the next month's data is selected as the test set data. The MAE (mean absolute error, in hours) of several models is shown in Table 2 below.

[0080]

[0081]

[0082] Based on the above, the LGBM model was finally selected to build the ETA time estimation model. The data for the arrival time to be predicted was processed with features and then input into the above-mentioned trained arrival time prediction model. In a short period of time, the model will output the estimated arrival time result. ETA prediction can combine real-time map traffic data, consider dynamic factors such as traffic congestion and weather, and special factors such as the personalized habits of the deliveryman, to accurately predict the arrival time.

[0083] According to a second aspect of an embodiment of the present invention, an order scheduling method is provided.

[0084] Figure 3 Schematic diagram of the main process of the order scheduling method according to an embodiment of the present invention; Figure 3 As shown, the order scheduling method according to the embodiment of the present invention mainly includes the following steps S301 to step 302.

[0085] Step S301, obtain a set of pending orders, and determine the predicted arrival time of each pending order according to an order arrival time prediction model; wherein the order arrival time prediction model is obtained by any method in the first aspect.

[0086] Step S302 : scheduling the set of pending orders based on the predicted arrival time of each pending order.

[0087] After obtaining the set of pending orders, the order arrival time prediction model can be used to determine the predicted arrival time for each order. The order arrival time prediction model is obtained through various methods described in the first aspect, such as training based on historical order data and delivery-related features. The model analyzes the delivery status of each pending order and generates an estimated arrival time, which is the result of prediction based on the rules learned by the model from historical data and the current order characteristics. After obtaining the predicted arrival time of each pending order, the pending order set is scheduled based on these predicted arrival times. The scheduling process includes arranging the delivery sequence, optimizing resource allocation, and adjusting the delivery route to ensure that the order can arrive efficiently within the expected time. By using the predicted arrival time, various delivery tasks can be reasonably arranged, waiting time and delays can be reduced, and overall logistics efficiency and customer satisfaction can be improved.

[0088] Optionally, the set of pending orders is scheduled based on the predicted arrival time of each pending order, including: constructing a scheduling strategy corresponding to the set of pending orders based on the predicted arrival time of each pending order and an adaptive large neighborhood search algorithm; wherein the scheduling strategy includes vehicle information and path information; and scheduling the set of pending orders according to the scheduling strategy.

[0089] The Adaptive Large Neighborhood Search (ALNS) algorithm is a heuristic algorithm for solving complex scheduling and optimization problems. It finds high-quality solutions by searching a large range of solutions in the solution space. This algorithm can dynamically adjust the search strategy based on the predicted arrival time, thereby effectively optimizing the scheduling solution. The scheduling strategy can be constructed based on the predicted arrival time of each pending order and combined with the optimization capability of the adaptive large neighborhood search algorithm. The scheduling strategy includes detailed vehicle information, such as the carrying capacity, availability and characteristics of each vehicle, as well as path information, including the optimal route, estimated travel time and address nodes of each delivery path. After the scheduling strategy is constructed, each pending order is assigned to an appropriate vehicle according to the strategy, and the optimal delivery path is planned. In this way, vehicle resources can be effectively utilized and the delivery path can be optimized, thereby improving overall logistics efficiency and customer satisfaction. The delivery plan can also be adjusted and optimized in real time, making the delivery process of each pending order more efficient.

[0090] Optionally, the set of orders to be processed is scheduled according to the scheduling strategy, including: modifying the path information in the scheduling strategy to obtain a new scheduling strategy; based on the predicted arrival time of each order to be processed and the new scheduling strategy, judging whether the new scheduling strategy meets the preset time window constraint, and using the new scheduling strategy that meets the preset time window constraint as the target scheduling strategy to schedule the set of orders to be processed.

[0091] After the preliminary scheduling strategy is formulated, the path information can be modified to optimize the scheduling effect. A new scheduling strategy is generated by adjusting the path information. The adjustment can delete or insert address nodes, re-plan the delivery route, adjust the vehicle driving order or optimize the path selection, etc. The modified new scheduling strategy can be verified according to the predicted arrival time of each pending order to determine whether the new scheduling strategy can meet the preset time window constraint, that is, whether the delivery time of all orders is completed within the specified time range. If the new scheduling strategy can deliver on time within the expected arrival time of each order and meets the time window constraint, the new scheduling strategy is regarded as the target scheduling strategy. According to the target scheduling strategy, the actual scheduling of the pending order set is carried out, for example, how to assign orders to vehicles, plan delivery routes, and adjust the delivery plan to achieve efficient and timely order processing. In this way, it can ensure that each order is delivered according to the expected time window, thereby improving the accuracy and efficiency of the overall logistics.

[0092] Figure 4 is a schematic diagram of an order scheduling method according to a preferred embodiment of the present invention; Figure 4As shown, construct the initial solution, and construct the initial solution with certain rules or algorithms, such as using a greedy algorithm, etc., through the order arrival time prediction model and other related information, calculate the objective function of the solution and whether it meets the constraints. If it meets the constraints, it is used as the initial feasible solution. Select the operator, and select the ruin (delete) and insert (insert) operators used in this iteration according to the weights of different operators. Ruin and insert operators can use multiple ones according to actual business. For example, the ruin operator includes: random deletion, delete the worst, delete the closest distance, linear (same car) deletion, cluster-based deletion, delete the largest model, etc.; insert operators include: optimal insertion, regret insert, distance-based insertion, etc. Operator selection can be selected by roulette method based on weight. Execute the ruin operation, based on the selected ruin operator, perform the ruin operation on the initial solution or the solution selected in the previous round, delete some nodes in the solution, and form an infeasible solution. Execute the insert operation. Based on the selected insert operator, the deleted nodes are reinserted into the solution in different ways. The order arrival time prediction model and other information are used again to calculate whether the solution meets the constraints. The solution that meets the constraints is retained as a feasible solution. Perform objective function evaluation. Based on the objective function constructed by the optimization goal, score the current feasible solution again. The optimization goals include: shortest distance, shortest time, optimal cost, least vehicles, point balance, time balance, and arrival time window strategies. Accept the strategy and compare the objective function values ​​of the solution before the ruin and insert operations of this iteration to determine whether to use the solution. The solution usage strategy includes simulated annealing strategy, etc. Then determine whether to terminate, that is, whether the iteration reaches a certain number of steps, the iteration reaches a specified number of times, and the number of iterations that have not been improved reaches a pre-selected threshold. If so, terminate the iteration and output the latest feasible solution after the current iteration. If not, perform the next step of weight update and continue to execute the selection operator and other processes. Weight update: update the weights of ruin and insert operators according to the quality of the new solution generated by the operator selected this time, such as "better than the optimal solution, better than the solution in the previous round, worse than the solution in the previous round". The better the quality of the solution, the more the weight increases, making it easier for operators that can produce better solutions to be selected.

[0093] The embodiment of the present invention adds a machine learning-based order arrival time prediction model to the generation and insert operation of the initial solution of the ALNS algorithm, and determines the quality of the solution or whether to use the solution based on the estimated results of the model. The model outputs an accurate predicted arrival time. After solving the path planning problem with a time window, it can better output order allocation and path planning solutions that meet customer arrival time requirements. The results of logistics distribution path planning and path transportation time prediction are more in line with current reality. After optimization, the adoption rate and cost reduction effect of the dispatching plan including order allocation and path planning strategies have been significantly improved.

[0094] According to a third aspect of an embodiment of the present invention, a training device for an order arrival time prediction model is provided.

[0095] Figure 5 is a schematic diagram of the main modules of the training device of the order arrival time prediction model according to an embodiment of the present invention, such as Figure 5 As shown, an order arrival time prediction model 500 includes:

[0096] The data acquisition module 501 is used to acquire the delivery time and delivery-related data of each first historical order in the first historical order set; wherein the delivery-related data includes at least one of the following: delivery address data, delivery item data, delivery vehicle data, delivery station data, delivery rejection behavior data, delivery person data, and weather data;

[0097] A feature extraction module 502 is used to extract features from the delivery-related data to obtain delivery-related features;

[0098] The model training module 503 is used to construct a training sample set based on the delivery time and delivery-related features of each first historical order in the first historical order set, and obtain an order arrival time prediction model through training based on the training sample set.

[0099] Optionally, the training device 500 further includes a preprocessing module, which is used to:

[0100] A first historical order with missing values ​​in the first historical order set is identified, and the first historical order with missing values ​​is deleted or the missing values ​​in the first historical order with missing values ​​are filled.

[0101] Optionally, the training device 500 further includes an anomaly removal module, which is used to:

[0102] Determine a preliminary estimated arrival time of the first historical order based on the delivery-related data of the first historical order and the real-time map data;

[0103] Calculate the difference between the preliminary estimated arrival time and the delivery time of the first historical order, and in response to the difference exceeding a preset threshold, treat the first historical order as a time-abnormal order;

[0104] The time-abnormal orders are removed from the first historical order set.

[0105] Optionally, the anomaly removal module is further used to:

[0106] Determine that the delivery address data for orders with abnormal time is incomplete.

[0107] Optionally, the first historical order in the first historical order set is an order within a first time range; the training device 500 further includes a model testing module, which is used to:

[0108] Obtaining the delivery time and delivery-related data of each second historical order in the second historical order set within a second time range; wherein the second time range is different from the first time range;

[0109] Performing feature extraction on the delivery-related data of the second historical order to obtain delivery-related features of the second historical order;

[0110] A test sample set is constructed based on the delivery time and delivery-related features of each second historical order in the second historical order set, and the order arrival time prediction model is evaluated using the test sample set.

[0111] It should be noted that the specific implementation content of the order arrival time prediction device in the embodiment of the present invention has been described in detail in the above order arrival time prediction method, so the repeated content will not be described again here.

[0112] According to a fourth aspect of an embodiment of the present invention, an order scheduling device is provided.

[0113] Figure 6 is a schematic diagram of the main modules of the order scheduling device according to an embodiment of the present invention, such as Figure 6 As shown, an order scheduling device 600 mainly includes:

[0114] A model prediction module 601 is used to obtain a set of pending orders and determine a predicted arrival time of each pending order according to an order arrival time prediction model; wherein the order arrival time prediction model is obtained by any method in the first aspect of the embodiment of the present invention;

[0115] The order scheduling module 602 is used to schedule the set of pending orders based on the predicted arrival time of each pending order.

[0116] Optionally, the order scheduling module 602 is further used to:

[0117] Based on the predicted arrival time of each pending order and the adaptive large neighborhood search algorithm, a scheduling strategy corresponding to the pending order set is constructed; wherein the scheduling strategy includes vehicle information and path information;

[0118] The order set to be processed is scheduled according to the scheduling strategy.

[0119] Optionally, the order scheduling module 602 is further used to:

[0120] Modify the path information in the scheduling strategy to obtain a new scheduling strategy;

[0121] Based on the predicted arrival time of each pending order and the new scheduling strategy, it is determined whether the new scheduling strategy meets the preset time window constraint, and the new scheduling strategy that meets the preset time window constraint is used as the target scheduling strategy to schedule the pending order set.

[0122] It should be noted that the specific implementation content of the order scheduling device in the embodiment of the present invention has been described in detail in the above order scheduling method, so the repeated content will not be described again here.

[0123] According to a fifth aspect of an embodiment of the present invention, there is provided an electronic device, including:

[0124] one or more processors;

[0125] a storage device for storing one or more programs,

[0126] When one or more programs are executed by one or more processors, the one or more processors implement the method provided by the first aspect of the embodiment of the present invention and / or the second aspect of the embodiment of the present invention.

[0127] According to a sixth aspect of an embodiment of the present invention, a computer-readable medium is provided, on which a computer program is stored. When the program is executed by a processor, the method provided by the first aspect of the embodiment of the present invention and / or the second aspect of the embodiment of the present invention is implemented.

[0128] According to a seventh aspect of an embodiment of the present invention, a computer program product is provided, including a computer program, which, when executed by a processor, implements the method provided by the first aspect of the embodiment of the present invention and / or the second aspect of the embodiment of the present invention.

[0129] Figure 7 An exemplary system architecture 700 is shown to which the training method of the order arrival time prediction model or the training device of the order arrival time prediction model according to the embodiments of the present invention can be applied.

[0130] like Figure 7As shown, system architecture 700 may include terminal devices 701, 702, 703, network 704 and server 705. Network 704 is used to provide a medium for communication links between terminal devices 701, 702, 703 and server 705. Network 704 may include various connection types, such as wired, wireless communication links or optical fiber cables, etc.

[0131] The user can use the terminal devices 701, 702, 703 to interact with the server 705 through the network 704 to receive or send messages, etc. Various communication client applications can be installed on the terminal devices 701, 702, 703, such as shopping applications, web browser applications, search applications, instant messaging tools, email clients, social platform software, etc. (only for example).

[0132] The terminal devices 701 , 702 , and 703 may be various electronic devices having a display screen and supporting web browsing, including but not limited to smart phones, tablet computers, laptop computers, and desktop computers.

[0133] The server 705 may be a server that provides various services, such as a backend management server (only an example) that provides support for shopping websites browsed by users using the terminal devices 701, 702, and 703. The backend management server may analyze and process the received data such as the model training request, and feed back the processing results (such as the order arrival time prediction model - only an example) to the terminal device.

[0134] It should be noted that the training method of the order arrival time prediction model provided in the embodiment of the present invention is generally run by the server 705 , and accordingly, the training device of the order arrival time prediction model is generally set in the server 705 .

[0135] It should be noted that the order scheduling method provided in the embodiment of the present invention is generally run by the server 705 , and accordingly, the order scheduling device is generally set in the server 705 .

[0136] It should be understood that Figure 7 The number of terminal devices, networks and servers in the embodiment is only for illustration. Any number of terminal devices, networks and servers may be provided according to implementation requirements.

[0137] Reference below Figure 8 , which shows a schematic diagram of the structure of a computer system 800 of a terminal device suitable for implementing an embodiment of the present invention. Figure 8 The terminal device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present invention.

[0138] like Figure 8As shown, the computer system 800 includes a central processing unit (CPU) 801, which can execute various appropriate actions and processes according to a program stored in a read-only memory (ROM) 802 or a program loaded from a storage part 808 into a random access memory (RAM) 803. In the RAM 803, various programs and data required for the operation of the system 800 are also stored. The CPU 801, the ROM 802, and the RAM 803 are connected to each other via a bus 804. An input / output (I / O) interface 805 is also connected to the bus 804.

[0139] The following components are connected to the I / O interface 805: an input section 806 including a keyboard, a mouse, etc.; an output section 807 including a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker, etc.; a storage section 808 including a hard disk, etc.; and a communication section 809 including a network interface card such as a LAN card, a modem, etc. The communication section 809 runs a communication process via a network such as the Internet. A drive 810 is also connected to the I / O interface 805 as needed. A removable medium 811, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 810 as needed, so that a computer program read therefrom is installed into the storage section 808 as needed.

[0140] In particular, according to the embodiments disclosed in the present invention, the process described above with reference to the flowchart can be implemented as a computer software program. For example, the embodiments disclosed in the present invention include a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program contains program code for running the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from the network through the communication part 809, and / or installed from the removable medium 811. When the computer program is executed by the central processing unit (CPU) 801, the above-mentioned functions defined in the system of the present invention are executed.

[0141] It should be noted that the computer-readable medium shown in the present invention may be a computer-readable signal medium or a computer-readable storage medium or any combination of the above two. The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present invention, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, device or device. In the present invention, a computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, which carries a computer-readable program code. This propagated data signal may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. The computer-readable signal medium may also be any computer-readable medium other than a computer-readable storage medium, which may send, propagate or transmit a program for use by or in conjunction with an instruction execution system, apparatus or device. The program code contained on the computer-readable medium may be transmitted using any appropriate medium, including but not limited to: wireless, wire, optical cable, RF, etc., or any suitable combination of the above.

[0142] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, functions and operations of the system, method and computer program product according to various embodiments of the present invention. In this regard, each box in the flow chart or block diagram can represent a module, a program segment, or a part of a code, and the above-mentioned module, program segment, or a part of a code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two boxes represented in succession can actually run basically in parallel, and they can sometimes run in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram or flow chart, and the combination of the boxes in the block diagram or flow chart can be implemented with a dedicated hardware-based system that runs the specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0143] The modules involved in the embodiments of the present invention may be implemented in software or hardware. The modules described may also be arranged in a processor, for example, they may be described as: a processor including a data acquisition module, a feature extraction module, and a model training module. The names of these modules do not, in some cases, constitute limitations on the modules themselves. For example, the data acquisition module may also be described as a "module for obtaining the delivery time and delivery-related data of each first historical order in a first historical order set". Alternatively, a processor includes a model prediction module and an order scheduling module, wherein the names of these modules do not, in some cases, constitute limitations on the modules themselves. For example, the order scheduling module may also be described as a "module for scheduling a set of pending orders based on the predicted arrival time of each pending order".

[0144] As another aspect, the present invention also provides a computer-readable medium, which may be included in the device described in the above embodiment; or it may exist independently and not be assembled into the device. The above computer-readable medium carries one or more programs, and when the above one or more programs are run by a device, the device includes: obtaining the delivery time and delivery-related data of each first historical order in the first historical order set; wherein the delivery-related data includes at least one of the following: delivery address data, delivery item data, delivery vehicle data, delivery site data, refusal to deliver behavior data, delivery personnel data, weather data; extracting features from the delivery-related data to obtain delivery-related features; constructing a training sample set based on the delivery time and delivery-related features of each first historical order in the first historical order set, and obtaining an order arrival time prediction model based on the training sample set. Alternatively, the device implements the following method: obtaining a set of pending orders, determining the predicted arrival time of each pending order according to the order arrival time prediction model; wherein the order arrival time prediction model is obtained by any method in the first aspect of the embodiment of the present invention; scheduling the set of pending orders based on the predicted arrival time of each pending order.

[0145] The computer program product provided by an embodiment of the present invention includes a computer program, which, when executed by a processor, implements the training method of the order arrival time prediction model in the embodiment of the first aspect of the present invention and / or the order scheduling method in the embodiment of the second aspect of the present invention.

[0146] The technical solution according to the embodiment of the present invention has the following advantages or beneficial effects: by obtaining the delivery time and delivery-related data of each first historical order in the first historical order set; wherein the delivery-related data includes at least one of the following: delivery address data, delivery item data, delivery vehicle data, delivery station data, refusal to deliver behavior data, delivery personnel data, weather data; feature extraction of the delivery-related data to obtain delivery-related features; constructing a training sample set based on the delivery time and delivery-related features of each first historical order in the first historical order set, and obtaining an order arrival time prediction model based on the training sample set; thereby improving the accuracy of the prediction result of the order arrival time. Obtain a set of pending orders, and determine the predicted arrival time of each pending order according to the order arrival time prediction model; wherein the order arrival time prediction model is obtained by any method in the first aspect of the present invention; scheduling the set of pending orders based on the predicted arrival time of each pending order can make the order scheduling more in line with user needs, reduce scheduling costs, and improve user experience.

[0147] The above specific implementations do not constitute a limitation on the protection scope of the present invention. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions may occur depending on design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

[0148] It should be noted that the acquisition, storage and application of user personal information involved in the technical solution of the present disclosure are in compliance with the provisions of relevant laws and regulations and do not violate public order and good morals.

Claims

1. A training method for an order arrival time prediction model, characterized in that: include: Obtaining the delivery time and delivery-related data of each first historical order in the first historical order set; wherein the delivery-related data includes at least one of the following: delivery address data, delivery item data, delivery vehicle data, delivery station data, delivery refusal behavior data, delivery person data, and weather data; Extracting features from the delivery-related data to obtain delivery-related features; A training sample set is constructed based on the delivery time and delivery-related features of each of the first historical orders in the first historical order set, and an order arrival time prediction model is obtained by training based on the training sample set.

2. The method according to claim 1, characterized in that: Before obtaining the delivery time and delivery-related data of each first historical order in the first historical order set, it also includes: A first historical order with missing values ​​in the first historical order set is identified, and the first historical order with missing values ​​is deleted or the missing values ​​in the first historical order with missing values ​​are filled.

3. The method according to claim 1, characterized in that After obtaining the delivery time and delivery-related data of each first historical order in the first historical order set, it also includes: Determining a preliminary estimated arrival time of the first historical order based on the delivery-related data of the first historical order and the real-time map data; Calculating a difference between a preliminary estimated arrival time and a delivery time of the first historical order, and in response to the difference exceeding a preset threshold, treating the first historical order as a time-abnormal order; The time-abnormal order is removed from the first historical order set.

4. The method according to claim 3, characterized in that Before removing the time-abnormal order from the first historical order set, the method further includes: It is determined that the delivery address data of the time abnormal order is incomplete.

5. The method according to claim 1, characterized in that The first historical order in the first historical order set is an order within a first time range; the method further includes: Obtaining delivery time and delivery-related data of each second historical order in a second historical order set within a second time range; wherein the second time range is different from the first time range; performing feature extraction on the delivery-related data of the second historical order to obtain delivery-related features of the second historical order; A test sample set is constructed based on the delivery time and delivery-related features of each of the second historical orders in the second historical order set, and the order arrival time prediction model is evaluated using the test sample set.

6. An order scheduling method, characterized in that: include: Obtain a set of pending orders, and determine a predicted arrival time of each pending order according to an order arrival time prediction model; wherein the order arrival time prediction model is obtained by any one of the methods described in claims 1 to 5; The set of pending orders is scheduled based on the predicted arrival time of each pending order.

7. The method according to claim 6, characterized in that Scheduling the set of pending orders based on the predicted arrival time of each pending order includes: Based on the predicted arrival time of each pending order and the adaptive large neighborhood search algorithm, construct a scheduling strategy corresponding to the pending order set; wherein the scheduling strategy includes vehicle information and path information; The set of pending orders is scheduled according to the scheduling strategy.

8. The method according to claim 7, characterized in that Scheduling the set of pending orders according to the scheduling strategy includes: Modifying the path information in the scheduling strategy to obtain a new scheduling strategy; Based on the predicted arrival time of each pending order and the new scheduling strategy, it is determined whether the new scheduling strategy meets the preset time window constraint condition, and the new scheduling strategy that meets the preset time window constraint condition is used as the target scheduling strategy to schedule the pending order set.

9. A training device for an order arrival time prediction model, characterized in that: include: A data acquisition module, used to acquire the delivery time and delivery-related data of each first historical order in the first historical order set; wherein the delivery-related data includes at least one of the following: delivery address data, delivery item data, delivery vehicle data, delivery station data, delivery refusal behavior data, delivery person data, and weather data; A feature extraction module, used to extract features from the delivery-related data to obtain delivery-related features; The model training module is used to construct a training sample set based on the delivery time and delivery-related features of each of the first historical orders in the first historical order set, and train the order arrival time prediction model according to the training sample set.

10. An order scheduling device, characterized in that: include: A model prediction module, used to obtain a set of pending orders and determine a predicted arrival time of each pending order according to an order arrival time prediction model; wherein the order arrival time prediction model is obtained by any method described in claims 1 to 5; An order scheduling module is used to schedule the set of pending orders based on the predicted arrival time of each pending order.

11. An electronic device, characterized in that: include: one or more processors; a storage device for storing one or more programs, When the one or more programs are executed by the one or more processors, the one or more processors implement the method according to any one of claims 1 to 8.

12. A computer readable medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the method according to any one of claims 1 to 8 is implemented.

13. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the method according to any one of claims 1 to 8 is implemented.

Images