Method and device for determining order priority, equipment, medium and product
By combining deep neural networks and genetic algorithms, the problem of inflexible priority allocation in the order processing system of catering enterprises is solved, and more accurate and efficient order priority determination is achieved, improving customer satisfaction.
Patent Information
- Application Number
- CN202510108138.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-01-23
- Publication Date
- 2025-05-27
AI Technical Summary
The existing catering companies lack flexible priority allocation in order processing systems, resulting in the waiting time of in-store customers and the orders being completed in advance or lagging, affecting the efficiency of meal preparation and customer satisfaction.
A method combining deep neural network and genetic algorithm is adopted to generate multiple order sequences of order lists through the score prediction model, and the initial population is iteratively calculated through the genetic algorithm to calculate the comprehensive score and finally determine the order priority.
It improves the accuracy and efficiency of determining order priorities, ensures the overall order priority ranking optimally, and improves customer satisfaction.
Smart Images
Figure CN120046913A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of artificial intelligence, and in particular to a method, device, equipment, medium, and product for determining order priority. Background Art
[0002] Most catering companies on the market currently lack flexible priority allocation for order processing systems, resulting in long waiting times for customers, early or delayed completion of orders, and other issues, affecting meal preparation efficiency and customer satisfaction. Existing methods for determining order priorities usually include:
[0003] First Come, First Served (FCFS): Lack of consideration for order priority may lead to delayed or early completion of orders, which can easily cause order backlogs during peak periods and affect customer satisfaction.
[0004] Fixed Priority: The priority rules are fixed and highly subjective, so there is a lack of dynamic adaptive mechanism, and it is difficult to adjust the order priority according to the real-time situation. It is impossible to comprehensively consider multiple factors and cannot fit the actual production.
[0005] Based on a single machine learning algorithm: A single machine learning algorithm is susceptible to data bias, has poor generalization ability, and cannot fully capture data features. Especially when processing multiple types of information, the sorting effect is often greatly reduced. Summary of the invention
[0006] In order to determine order priority more accurately, the present application proposes a method for determining order priority.
[0007] The method comprises: obtaining multiple indicators of each order in an order list and indicator weights corresponding to the indicators, and generating multiple groups of order sequences of the order list based on a score prediction model; wherein each of the order sequences comprises the ranking of each order in the order list; selecting a predetermined number of the order sequences from the multiple groups of the order sequences as an initial population of a genetic algorithm; using the genetic algorithm, iterating the initial population to obtain a final population; calculating the comprehensive scores of all the order sequences in the final population, and selecting the order sequence with the highest comprehensive score to determine the priority of the order in the order list; wherein the comprehensive score is obtained through an objective function based on the order data.
[0008] Optionally, the score prediction model is trained by the following steps:
[0009] Obtaining multiple training orders and corresponding indicators and indicator scores;
[0010] Extracting a feature vector from a pair consisting of the training order and the indicator score;
[0011] Based on the ranking function, generating a vector score for each of the feature vectors, and obtaining a vector score list;
[0012] Repeat the above steps to obtain multiple sets of vector score lists;
[0013] The ranking function is continuously iterated and optimized through the Lambda function, so that the loss function of multiple groups of the training order sequences is minimized to obtain the score prediction model; wherein the Lambda function is constructed based on the Listwise method.
[0014] Optionally, the indicators and / or the training indicators include: order type, order complexity, order label, distance between the customer and the restaurant, restaurant type, and the environment near the restaurant.
[0015] Optionally, the comprehensive score is determined by summing the order scores of each order in the order sequence, and the order score of each order in the order sequence is determined by the following formula:
[0016]
[0017] Where i represents the order number in the order sequence, P i Score the order, T i is the order type, w 1 is the weight of the order type, C i is the order complexity, w 2 is the weight of the order complexity, L i is the order label, w 3 is the weight of the order label, D i is the distance between the customer and the restaurant, w 4 is the weight of the distance between the customer and the restaurant, F i Score for taste, w 5 is the weight of taste score, S i is the waiting time after customers arrive at the store, w 6 The weight of the waiting time after the customer arrives at the store.
[0018] Optionally, the genetic algorithm is a non-dominated sorting genetic algorithm.
[0019] On the other hand, the present application proposes a device for determining order priority, comprising:
[0020] A score prediction unit, used to obtain multiple indicators of each order in the order list and the indicator weights corresponding to the indicators, and generate multiple groups of order sequences of the order list based on the score prediction model; wherein each of the order sequences includes the ranking of each order in the order list;
[0021] A population acquisition unit, used for selecting a predetermined number of the order sequences from a plurality of groups of the order sequences as an initial population of the genetic algorithm;
[0022] A genetic algorithm unit, used for iterating the initial population through the genetic algorithm to obtain a final population;
[0023] A priority determination unit is used to calculate the comprehensive scores of all the order sequences in the final population and select the order sequence with the highest comprehensive score to determine the priority of the orders in the order list; wherein the comprehensive score is obtained based on the order data through an objective function.
[0024] On the other hand, the present application proposes an electronic device, comprising a memory storing computer-executable instructions and a processor; when the instructions are executed by the processor, the device implements any of the methods described above.
[0025] On the other hand, the present application proposes a computer-readable medium, wherein the computer-readable medium stores one or more programs, and the one or more programs can be executed by one or more processors to implement any of the methods described above.
[0026] On the other hand, the present application proposes a computer program product, comprising a computer program, wherein when the computer program is executed by a processor, the method as described in any one of the above is implemented.
[0027] This application combines deep neural networks and genetic algorithms to determine order priorities. By training the score prediction model through the Listwise method in the deep neural network, a better initial population for the genetic algorithm is obtained, which not only improves the accuracy of the determined order priority, but also greatly speeds up the iteration speed of the genetic algorithm, and can obtain the optimal final population faster. And through the non-dominated sorting genetic algorithm, all factors that may affect the order priority are comprehensively considered to ensure that the overall ranking of the determined order priority is optimal, rather than the optimal ranking corresponding to one or several orders, but the ranking effect of other orders is very poor. And finally, through the accurate order priority ranking, the satisfaction of users who place orders is improved. BRIEF DESCRIPTION OF THE DRAWINGS
[0028] Figure 1 It is a flowchart of a method for determining order priority according to an implementation of the present application.
[0029] Figure 2 It is a schematic diagram of the cross-operation of the method for determining order priority according to an implementation of the present application.
[0030] Figure 3 It is a schematic diagram of a variation operation of the method for determining order priority according to an implementation of the present application.
[0031] Figure 4 It is a flowchart of a training score prediction model for a method for determining order priority according to an implementation of the present application.
[0032] Figure 5 It is a schematic diagram of a device for determining order priority according to an embodiment of the present application.
[0033] Figure 6 is a schematic diagram of an electronic device according to an embodiment of the present application. DETAILED DESCRIPTION
[0034] This application proposes a method for determining order priorities based on a deep neural network (DNN) and a genetic algorithm. In this application, an algorithm based on the Listwise method is used to train a deep neural network model, and multiple priority sequences of a group of orders that need to be prioritized are predicted through the trained model, and several priority sequences are selected from the multiple priority sequences as the initial population of the genetic algorithm (each priority sequence is an individual of the genetic algorithm). Finally, an optimal priority sequence is obtained through the genetic algorithm to determine the priority order of the group of orders.
[0035] Figure 1 A flowchart of the method for prioritizing orders.
[0036] In step 101, multiple indicators and indicator weights corresponding to each order in the order list are obtained, and multiple groups of order sequences of the order list are generated based on the score prediction model; wherein each order sequence includes the ranking of each order in the order list.
[0037] Order indicators may include the following: order type, such as takeout orders, dine-in orders, and packaged orders; order complexity, which is related to the type and quantity of food included in the order; order tags, such as large orders, VIP orders, etc.; the distance between the customer and the restaurant. The score of this indicator is negatively correlated with the distance value, that is, the closer the distance, the higher the probability of priority; restaurant type, such as mother store, hot sweet, away from sweet, McCafé, etc.; the environment near the restaurant, such as schools, shopping malls, stations, airports, etc.
[0038] In this step, the score prediction model is used to generate multiple groups of order sequences for the order list of a group of orders to be prioritized. Each order sequence determines an order priority ranking, and multiple groups of order sequences are order priorities in multiple different ways. To obtain multiple groups of order sequences, the score prediction model can be trained to output multiple groups of order sequences, or the score prediction model can be trained to output one group of order sequences and output them multiple times.
[0039] In step 102, a predetermined number of order sequences are selected from multiple groups of order sequences as the initial population of the genetic algorithm. In order to obtain more accurate order priority sorting, it is necessary to further process the multiple groups of order sequences obtained by the score prediction model through a genetic algorithm. In addition, in order to further improve the accuracy, the sorting results in the order sequence are further optimized through a genetic algorithm, and several order sequences with higher priority sorting accuracy are artificially selected as the initial population of the genetic algorithm. For example, the comprehensive score of each order sequence can be calculated, and several groups of order sequences can be selected in descending order according to the comprehensive score as needed. Among them, the comprehensive score is obtained based on the order data through the objective function. For example, the order score of each order in the group of order sequences is first determined by the following objective function formula, and then the order scores of all orders in the group of order sequences are summed to obtain the comprehensive score of the group of order sequences:
[0040]
[0041] Where i represents the order number of each order in the order sequence, T i is the order type, w 1 is the weight of the order type, C i is the order complexity, w 2 is the weight of order complexity, L i is the order label, w 3 is the weight of the order label, D i is the distance between the customer and the restaurant, w 4 is the weight of the distance between the customer and the restaurant, F i is the taste score, that is, the taste score of the food in different time periods after it is produced, which is manually determined according to the best tasting period of different food. 5 is the weight of taste score, S i The waiting time after the customer arrives at the store is determined by the priority order of the orders in the group of order sequences. The customer's waiting time is also closely related to customer satisfaction. 6 is the weight of the waiting time after the customer arrives at the store. It can be understood that this objective function is only used as an example, and different indicators can be considered to calculate the comprehensive score.
[0042] In step 103, the initial population is iterated by a genetic algorithm to obtain a final population. Since the objective function of a certain order may be optimal in a certain order sequence, but the objective function of another order is optimal in another order sequence, there is no solution that makes the objective functions of all orders in a certain order sequence optimal, that is, the solutions cannot be compared in terms of the overall objective function, and it is impossible to improve a certain solution without weakening other solutions. Therefore, the genetic algorithm used in this application is a non-dominated sorting genetic algorithm, such as the NSGA-II genetic algorithm. The iterative goal of the genetic algorithm is to maximize the average objective function value of all orders, and after comprehensively considering all dimensions of each order, the average performance of all orders in each dimension is optimal, that is, each individual in the final population. Different termination conditions can be set to determine the termination of the iterative operation, such as setting the number of iterative operations, the iterative operation time, or the optimization target value.
[0043] For the non-dominated sorting genetic algorithm used in this application, first, after obtaining the initial population, it is determined whether it has generated the first-generation sub-population. If it is determined not to be, the initial population is non-dominated sorted and the crowding degree is calculated, and then the first-generation sub-population is obtained through the selection, crossover and mutation operations of the genetic algorithm.
[0044] Starting from the second generation population, that is, the first generation sub-population, the parent population and the sub-population are merged and then non-dominated sorting is performed, and the crowding degree of the individuals in each non-dominated layer is calculated. After selection, crossover, and mutation operations, a new parent population is obtained. For example, after the initial population and the first generation sub-population are merged, the population composed of individuals obtained through the above operations is used as the parent population of the second generation sub-population, instead of directly using the first generation sub-population as the parent population of the second generation sub-population, and the subsequent parent population and sub-population are deduced in the same way. In this way, it is ensured that individuals with excellent performance will not be lost in the iterative operation, and the diversity of individuals in the population is maintained, so that the results of the genetic algorithm can avoid falling into the local optimum, but obtain the global optimal order priority sorting result.
[0045] Finally, when the iteration termination condition is reached, the last generation of sub-population is taken as the final population, and the individual with the highest comprehensive score is selected from the final population to determine the priority ranking of the group of orders.
[0046] Specifically, the selection operation is: solve the fitness of all individuals in the population according to the fitness function, and then the probability of each individual being selected can be obtained according to the following formula:
[0047]
[0048] PROB t is the probability of individual t (i.e. the corresponding order sequence) being selected, M is the total number of the population, F tis the fitness of individual t, sum(F t ) is the sum of the fitness of all individuals in the population. Through the selection operation, individuals with high fitness are more likely to be selected to enter the next generation, while individuals with low fitness are relatively less likely to be selected to enter the next generation, thereby continuously optimizing the order sequence through iterative operations.
[0049] Figure 2 Schematic diagram of the crossover operation. In this example, each order sequence includes seven orders numbered 1, 2, 3, 4, 5, 6, and 7. The priority ordering of the orders in the order sequence as the first parent is 1, 4, 6, 5, 3, 7, and 2, and the priority ordering of the orders in the order sequence as the second parent is 2, 6, 3, 7, 1, 4, and 5.
[0050] First, two intersection points are randomly generated on the first parent generation and the second parent generation. For example, in the first parent generation, the first intersection point is generated between orders 4 and 6, and the second intersection point is generated between orders 3 and 7. In the second parent generation, the third intersection point is generated between orders 6 and 3, and the fourth intersection point is generated between orders 1 and 4. Thus, orders 6, 5, and 3 in the first parent generation are cut off by the first intersection point and the second intersection point, and orders 3, 7, and 1 in the second parent generation are cut off by the third intersection point and the fourth intersection point.
[0051] Then, the first child inherits the order priority ranking of the first parent, and connects the cut orders 3, 7, and 1 in the second parent between the first intersection and the second intersection in the first parent. Since orders 3, 7, and 1 have been inherited from the second parent, the first child arranges the remaining orders 2, 4, 5, and 6 except for orders 3, 7, and 1 into 4, 6, 5, and 2 according to the sorting method of the first parent, and inserts orders 3, 7, and 1 between orders 6 and 5, and finally obtains the order sequence in the first child as 4, 6, 3, 7, 1, 5, and 2. The second child and other possible children are crossed in this way, which will not be repeated here.
[0052] Figure 3 is a schematic diagram of a mutation operation. In this example, each order sequence includes seven orders numbered 1, 2, 3, 4, 5, 6, and 7. Two gene positions are randomly selected in the order sequence (as shown by the shadows in the figure), and the orders on these two gene positions are exchanged, thereby completing the mutation operation. It is understandable that more gene positions can also be generated to perform more complex mutations.
[0053] Through crossover and mutation operations, new individuals can be generated in the population, avoiding the algorithm from falling into a local optimal solution and causing unreasonable order priority sorting for individual customers, thereby improving overall customer satisfaction.
[0054] In step 104, the comprehensive scores of all order sequences in the final population are calculated, and the order sequence with the highest comprehensive score is selected to determine the priority of the orders in the order list.
[0055] The following table shows an example of five orders in the order sequence with the highest comprehensive score determined using this method.
[0056]
[0057] After obtaining the order sequence with the highest comprehensive score, the orders are sorted according to the order numbers from large to small according to the order scores as 2, 4, 3, 1, and 5. When the production line is idle, the production of the orders is arranged in this order.
[0058] Figure 4 It is the training flow chart of the aforementioned score prediction model.
[0059] In step 401, multiple training orders and corresponding indicators and indicator scores are obtained. The indicators of the training orders are the same as the aforementioned indicators, wherein each training order corresponds to a score list, and the score list contains the indicator score corresponding to each indicator of the order. The indicator score can be given manually explicitly or implicitly.
[0060] For example, for a set of orders Order = {d 1 ,d 2 ,…,d i ,…,d n}, where d i Represents the i-th order, each order corresponds to a score list It represents the score y of the order on the jth indicator, which can be given explicitly or implicitly by humans.
[0061] In step 402, feature vectors are extracted from pairs of training orders and indicator scores. After each training order and the indicator score of each indicator form an order-score list pair, the feature vector extracted therefrom is the training set of the score prediction model.
[0062] For example, for the above set of training orders, the number of pairs consisting of each training order-score list is Extract a feature vector from Where (i = 1, 2, ..., n; j = 1, 2, ..., m). Thus, the feature vector can be used as the training set of the score prediction model
[0063] In step 403, based on the ranking function, a vector score is generated for each feature vector, and a vector score list is obtained. The vector score corresponds to the score of the training order, and the vector score list includes the scores of each training order.
[0064] For example, for the feature vectors used as the training set above, a ranking function f is used to output a vector score for each feature vector This generates a list of vector scores for all eigenvectors
[0065] The vector score list represents the score of each indicator in different training orders. Thus, the scores of all training orders on all indicators can be obtained through the vector score list of all indicators.
[0066] In step 404, the above steps are repeated to obtain multiple sets of vector score lists.
[0067] For example, for the aforementioned vector score list, the vector score lists of all indicators are a group, and the scores of each indicator of each training order can be determined through this group of vector score lists, and then the total score of each order can be calculated based on the scores of each indicator, and thus a sequence of training orders can be determined. By obtaining multiple groups of vector score lists, a training order sequence of multiple different sorting methods of a set of training orders Order can be obtained.
[0068] In step 405, the ranking function is continuously iterated and optimized through the Lambda function, so that the loss function of multiple groups of training order sequences is minimized to obtain a score prediction model; wherein the Lambda function is constructed based on the Listwise method.
[0069] For example, for the aforementioned sorting function, the sequence of multiple training orders obtained based on the sorting function is not the optimal sequence, and the sorting function needs to be continuously iterated and optimized based on the obtained sequence of training orders, so that the accuracy of the training order sequence generated by the trained sorting function is higher. To reflect the training optimization effect of the ranking function, the Lambda function is used to continuously update the model to minimize the loss function, and the score prediction model can be trained. In summary, the ranking function is obtained by optimizing and adjusting parameters to minimize the loss function. It captures the complex relationship between feature vectors and ranking by mapping feature vectors to scores, while the score prediction model adjusts the score generation method of the ranking function through continuous training and learning, so that the final ranking result can minimize the loss function, and can generate multiple sets of sequences with the loss function minimized.
[0070] The Lambda function is determined by:
[0071] First, calculate the actual DCG (Discounted Cumulative Gain) using the following formula:
[0072]
[0073] Among them, k is the first k orders selected manually, rel i represents the score of the i-th order, rel 1 This is the score of the first order. For example, the scores and score list of five orders are 3, 2, 3, 0, 1. After manually selecting the scores of the first three orders (3, 2, 3) and substituting them into the above formula, the DCG obtained is 3 It is 5.76.
[0074] Then, calculate the ideal DCG, that is, sort the orders according to the order scores to get the ideal order, and then calculate it by the following formula:
[0075]
[0076] in, represents the score of the ideally ordered i-th order, This is the score of the first order in the ideal order. For example, the score list of five orders is 3, 2, 3, 0, 1. After sorting the scores from large to small, they are 3, 3, 2, 1, 0. After manually selecting the scores of the first three orders (3, 3, 2) and substituting them into the above formula, the IDCG obtained is 3 It is 5.89.
[0077] Finally, the ratio of DCG to IDCG is calculated as NDCG (Normalized Discounted Cumulative Gain). Since NDCG is a discontinuous function and cannot be used for gradient calculation, we introduce the NDCG value NDCG(old) obtained based on the current ranking before the model is updated; and the NDCG value NDCG(new) obtained based on the new ranking after the model is updated. By calculating their difference, we get the adjusted increment |Δ(NDCG)|=|NDCG(new)–NDCG(old)|, and convert the increment into
[0078] |Δ(NDCG)| is approximately the gradient calculated iteratively during model training, which is used to update the model parameters. The Lambda function finally constructed is:
[0079]
[0080] Among them, λ a,b is the Lambda value of a pair of orders a and b (usually the orders ranked high and low in the sorting process during the training of the score prediction model); a 、s b are the scores of order a and order b respectively; Denote NDCG with respect to the fraction s aThe gradient of , that is, the rate at which NDCG changes when the score of order a is adjusted; σ is a constant for adjusting the step size during model training, which can be used to control the amplitude of updates during model training; is a Sigmoid function, which is used to represent the impact of the scores of order a and order b on the model training update. This function is implemented by a -s b Apply smooth transformations so that ranking updates are gradual rather than abrupt.
[0081] The present application also proposes a device for determining order priority, referring to Figure 5 , which includes the following units:
[0082] The score prediction unit 501 is used to obtain multiple indicators of each order in the order list and the indicator weights corresponding to the indicators, and generate multiple groups of order sequences of the order list based on the score prediction model; wherein each order sequence includes the ranking of each order in the order list;
[0083] A population acquisition unit 502 is used to select a predetermined number of order sequences from multiple groups of order sequences as an initial population of the genetic algorithm; wherein the comprehensive score is obtained based on the order data through an objective function;
[0084] A genetic algorithm unit 503 is used to perform iterative operations on the initial population through a genetic algorithm to obtain a final population;
[0085] The priority determination unit 504 is used to calculate the comprehensive scores of all order sequences in the final population and select the order sequence with the highest comprehensive score to determine the priority of the orders in the order list.
[0086] It can be understood that the device can run on various computers and servers, and can use the method for determining order priorities corresponding to the present application to determine a more reasonable ordering of production priorities for all orders, thereby improving overall customer satisfaction.
[0087] Reference now Figure 6 , which is a block diagram of an electronic device 600 according to an embodiment of the present application. The electronic device 600 may include one or more processors 602, a system control logic 608 connected to at least one of the processors 602, a system memory 604 connected to the system control logic 608, a non-volatile memory (NVM) 606 connected to the system control logic 608, and a network interface 610 connected to the system control logic 608.
[0088] The processor 602 may include one or more single-core or multi-core processors. The processor 602 may include any combination of a general-purpose processor and a dedicated processor (e.g., a graphics processor, an application processor, a baseband processor, etc.). In the embodiments of this document, the processor 602 may be configured to execute one or more embodiments of determining order priorities according to various embodiments proposed in this application.
[0089] In some embodiments, system control logic 608 may include any suitable interface controller to provide any suitable interface to at least one of processors 602 and / or any suitable device or component in communication with system control logic 608 .
[0090] In some embodiments, the system control logic 608 may include one or more memory controllers to provide an interface to the system memory 604. The system memory 604 may be used to load and store data and / or instructions. In some embodiments, the system memory 604 of the electronic device 600 may include any suitable volatile memory, such as a suitable dynamic random access memory (DRAM).
[0091] The non-volatile memory 606 may include one or more tangible, non-transitory computer-readable media for storing data and / or instructions. In some embodiments, the non-volatile memory 606 may include any suitable non-volatile memory such as flash memory and / or any suitable non-volatile storage device, such as at least one of a HDD (Hard Disk Drive), a CD (Compact Disc) drive, and a DVD (Digital Versatile Disc) drive.
[0092] The non-volatile memory 606 may include a portion of storage resources installed on a device of the electronic device 600, or it may be accessible by the device but not necessarily a portion of the device. For example, the non-volatile memory 606 may be accessed over a network via the network interface 610.
[0093] In particular, the system memory 604 and the non-volatile memory 606 may include: a temporary copy and a permanent copy of the instruction 620, respectively. The instruction 620 may include: an instruction that causes the electronic device 600 to implement the method provided by the present application when executed by at least one of the processors 602. In some embodiments, the instruction 620, hardware, firmware and / or software components thereof may be additionally / alternatively placed in the system control logic 608, the network interface 610 and / or the processor 602.
[0094] In some embodiments, the network interface 610 may be integrated with other components of the electronic device 600. For example, the network interface 610 may be integrated with at least one of the processor 602, the system memory 604, the non-volatile memory 606, and a firmware device (not shown) having instructions, and when at least one of the processors 602 executes the instructions, the electronic device 600 implements one or more of the various embodiments described herein. The network interface 610 may further include any suitable hardware and / or firmware to provide a multiple-input multiple-output radio interface.
[0095] In one embodiment, at least one of the processors 602 may be packaged together with logic for one or more controllers of the system control logic 608 to form a system in package (SiP). In one embodiment, at least one of the processors 602 may be integrated on the same die with logic for one or more controllers of the system control logic 608 to form a system on chip (SoC).
[0096] The electronic device 600 may further include an input / output (I / O) device 612. The input / output (I / O) device 612 may include a user interface, so that a user can interact with the device 600; the design of the peripheral component interface enables peripheral components to interact with the electronic device 600.
[0097] In some embodiments, the user interface may include, but is not limited to, a display (e.g., an LCD display, a touch screen display, etc.), a speaker, a microphone, one or more cameras (e.g., a still image camera and / or a video camera), a flashlight (e.g., an LED flash), and a keyboard.
[0098] In some embodiments, the peripheral component interface may include, but is not limited to, a non-volatile memory port, an audio jack, and a power interface.
[0099] It is to be understood that the structure illustrated in the embodiment of the present invention does not constitute a specific limitation on the electronic device 600. In other embodiments of the present application, the electronic device 600 may include more or fewer components than shown in the figure, or combine some components, or split some components, or arrange the components differently. The components shown in the figure may be implemented in hardware, software, or a combination of software and hardware.
[0100] Program code can be applied to input instructions to perform the functions described herein and generate output information. The output information can be applied to one or more output devices in a known manner. For purposes of this application, a processing system includes any system having a processor such as, for example, a digital signal processor (DSP), a microcontroller, an application specific integrated circuit (ASIC), or a microprocessor.
[0101] Program code can be implemented with high-level programming language or object-oriented programming language to communicate with the processing system. When necessary, program code can also be implemented with assembly language or machine language. In fact, the mechanism described herein is not limited to the scope of any specific programming language. In either case, the language can be a compiled language or an interpreted language.
[0102] One or more aspects of at least one embodiment may be implemented by representative instructions stored on a computer-readable storage medium, which represent various logic in a processor, and when the instructions are read by a machine, cause the machine to fabricate logic for performing the techniques described herein. These representations, known as "IP cores," may be stored on a tangible computer-readable storage medium and supplied to multiple customers or manufacturing facilities to load into a manufacturing machine that actually manufactures the logic or processor.
[0103] An embodiment of the present application discloses a computer-readable medium storing one or more programs executable by one or more processors to implement the method of determining order priority of the present application.
[0104] An embodiment of the present application discloses a computer program product, including a computer program, which implements the method for determining order priority of the present application when executed by a processor.
[0105] The above specific embodiments illustrate the implementation of the present application, and those skilled in the art can easily understand other advantages and effects of the present application from the contents disclosed in this specification. Although the description of the present application will be introduced in conjunction with the preferred embodiment, this does not mean that the features of this invention are limited to the implementation. In addition, in order to avoid confusion or blurring the focus of the present application, some specific details will be omitted in the description. It should be noted that the embodiments in the present application and the features in the embodiments can be combined with each other without conflict.
[0106] Furthermore, various operations will be described as multiple discrete operations in a manner that is most helpful for understanding the illustrative embodiments; however, the order of description should not be construed as implying that these operations are necessarily order dependent. In particular, these operations do not need to be performed in the order presented.
[0107] Unless the context dictates otherwise, the terms "comprising," "having," and "including" are synonymous. The phrase "A / B" means "A or B." The phrase "A and / or B" means "(A and B) or (A or B)."
[0108] As used herein, the term "module" or "unit" may refer to, be or include: an application specific integrated circuit (ASIC), an electronic circuit, a (shared, dedicated or group) processor and / or memory that executes one or more software or firmware programs, a combinational logic circuit and / or other suitable components that provide the described functionality.
[0109] In the accompanying drawings, some structural or method features are shown in a specific arrangement and / or order. However, it should be understood that such a specific arrangement and / or order may not be required. In some embodiments, these features may be arranged in a manner and / or order different from that shown in the illustrative drawings. In addition, the inclusion of structural or method features in a specific figure does not mean that such features are required in all embodiments, and in some embodiments, these features may not be included or may be combined with other features.
[0110] It should be understood that although the terms "first", "second", etc. may be used here to describe various units or data, these units or data should not be limited by these terms. These terms are used only to distinguish one feature from another. For example, without departing from the scope of the exemplary embodiment, a first feature may be referred to as a second feature, and similarly, a second feature may be referred to as a first feature.
[0111] It should be noted that in this specification, similar reference numerals and letters denote similar items in the following drawings, and therefore, once an item is defined in one drawing, it does not need to be further defined and explained in the subsequent drawings.
[0112] While the present invention has been shown and described with reference to certain preferred embodiments thereof, it will be understood by those skilled in the art that various changes in form and details may be made therein without departing from the spirit and scope of the invention.
Claims
1. A method for determining order priority, characterized in that: include: Acquire multiple indicators of each order in the order list and the indicator weights corresponding to the indicators, and generate multiple groups of order sequences of the order list based on the score prediction model; wherein each of the order sequences includes the ranking of each order in the order list; Selecting a predetermined number of the order sequences from a plurality of groups of the order sequences as an initial population of the genetic algorithm; By using the genetic algorithm, iterative operation is performed on the initial population to obtain a final population; Calculate the comprehensive scores of all the order sequences in the final population, and select the order sequence with the highest comprehensive score to determine the priority of the orders in the order list; wherein the comprehensive score is obtained based on the order data through an objective function.
2. The method according to claim 1, characterized in that The score prediction model is trained by the following steps: Obtaining multiple training orders and corresponding indicators and indicator scores; Extracting a feature vector from a pair consisting of the training order and the indicator score; Based on the ranking function, generating a vector score for each of the feature vectors, and obtaining a vector score list; Repeat the above steps to obtain multiple sets of vector score lists; The ranking function is continuously iterated and optimized through the Lambda function, so that the loss function of multiple groups of the training order sequences is minimized to obtain the score prediction model; wherein the Lambda function is constructed based on the Listwise method.
3. The method according to claim 1 or 2, characterized in that: The indicators and / or the training indicators include: order type, order complexity, order label, distance between the customer and the restaurant, restaurant type, and the environment near the restaurant.
4. The method according to claim 3, characterized in that: The comprehensive score is determined by summing the order scores of each order in the order sequence, and the order score of each order in the order sequence is determined by the following formula: Where i represents the order number in the order sequence, P i Score the order, T i is the order type, w1 is the weight of the order type, C i is the order complexity, w2 is the weight of the order complexity, L i is the order label, w3 is the weight of the order label, D i is the distance between the customer and the restaurant, w4 is the weight of the distance between the customer and the restaurant, F i is the taste score, w5 is the weight of the taste score, S i is the waiting time after the customer arrives at the store, and w6 is the weight of the waiting time after the customer arrives at the store.
5. The method according to claim 1, characterized in that The genetic algorithm is a non-dominated sorting genetic algorithm.
6. A device for determining order priority, characterized in that: include: A score prediction unit, used to obtain multiple indicators of each order in the order list and the indicator weights corresponding to the indicators, and generate multiple groups of order sequences of the order list based on the score prediction model; wherein each of the order sequences includes the ranking of each order in the order list; A population acquisition unit, used for selecting a predetermined number of the order sequences from a plurality of groups of the order sequences as an initial population of the genetic algorithm; A genetic algorithm unit, used for iterating the initial population through the genetic algorithm to obtain a final population; A priority determination unit is used to calculate the comprehensive scores of all the order sequences in the final population and select the order sequence with the highest comprehensive score to determine the priority of the orders in the order list; wherein the comprehensive score is obtained based on the order data through an objective function.
7. An electronic device, characterized in that: The device comprises a memory storing computer executable instructions and a processor; when the instructions are executed by the processor, the device implements the method according to any one of claims 1 to 5.
8. A computer-readable medium, characterized in that The computer-readable medium stores one or more programs, and the one or more programs can be executed by one or more processors to implement the method according to any one of claims 1 to 5.
9. 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 5 is implemented.