Model Training and Trip Scheduling Methods, Electronic Devices, and Computer Storage Media

Abstract

The embodiments of the present application provide a model training method, a travel scheduling method, an electronic device, and a computer storage medium. Among them, the model training method includes: obtaining travel order sample data, travel service object sample data, and travel and service demand relationship sample data; using a polygon area grid to generate corresponding order polygon grid data, service object polygon grid data, and demand relationship polygon grid data for the travel order sample data, travel service object sample data, and travel and service demand relationship sample data respectively; constructing a training sample based on the order polygon grid data, service object polygon grid data, and demand relationship polygon grid data, and training a machine learning model to obtain a machine learning model for outputting matching information between travel orders and travel service objects.

Description

Technical Field

[0001] The embodiments of the present application relate to the field of computer technologies, and in particular, to a model training method, a travel scheduling method, an electronic device, and a corresponding computer storage medium. Background Art

[0002] With the rapid development of online travel services, more and more platforms not only provide the function of online car-hailing, but also pay more attention to how to reasonably allocate travel orders to provide better services for travel objects (passengers) and travel service objects (drivers and passengers). For this reason, many online travel service platforms adopt the form of machine learning models in order to achieve relatively reasonable travel scheduling.

[0003] Currently, this form of using machine learning models mostly uses simulated data and, on the premise of a completely random travel scheduling strategy, obtains a travel scheduling scheme through large-scale online iterative experiments. However, this method of using large-scale online iterative experiments results in inaccurate model training results because the considered factors do not conform well to the actual situation. Summary of the Invention

[0004] In view of this, the embodiments of the present application provide a model training and travel scheduling solution to at least partially solve the above problems.

[0005] According to a first aspect of the embodiments of the present application, there is provided a model training method, including: obtaining travel order sample data, travel service object sample data, and travel and service demand relationship sample data; using a polygon area grid to generate corresponding order polygon grid data, service object polygon grid data, and demand relationship polygon grid data for the travel order sample data, travel service object sample data, and travel and service demand relationship sample data respectively; constructing a training sample based on the order polygon grid data, service object polygon grid data, and demand relationship polygon grid data, and training a machine learning model to obtain a machine learning model for outputting matching information between travel orders and travel service objects.

[0006] According to a second aspect of the embodiments of the present application, a travel scheduling method is provided, including: receiving a travel order and obtaining travel data in the travel order, where the travel data includes a travel start position and a travel end position; determining data of a plurality of candidate travel service providers for providing travel services for the travel order, and travel and service demand relationship data of a travel area corresponding to the travel order, where the data of the plurality of candidate travel service providers and the service demand relationship data are both dynamically updated data; and selecting a target travel service provider for providing travel services for the travel order from the plurality of candidate travel service providers according to the travel data, the data of the plurality of candidate travel service providers, and the travel and service demand relationship data.

[0007] According to a third aspect of the embodiments of the present application, another travel scheduling method is provided, including: receiving travel information input by a travel object and generating a travel order according to the travel information, where the travel order includes a travel start position and a travel end position; sending the travel order to a travel platform so that the travel platform determines data of a plurality of candidate travel service providers for providing travel services according to the travel order, and travel and service demand relationship data of a travel area corresponding to the travel order; and selecting a target travel service provider for providing travel services for the travel order from the plurality of candidate travel service providers according to the travel data, the data of the plurality of candidate travel service providers, and the travel and service demand relationship data; and receiving information of the target travel service provider fed back by the travel platform.

[0008] According to a fourth aspect of the embodiments of the present application, yet another travel scheduling method is provided, including: receiving a travel order sent by a travel platform, where the travel order includes a travel start position and a travel end position; generating a response message according to an acceptance operation of the travel order by a travel service provider, where the response message carries information of the travel service provider; and feeding back the response message to the travel platform so that the travel platform selects a target travel service provider for providing travel services for the travel order from a plurality of candidate travel service providers according to the travel order, the information of the travel service provider, and travel and service demand relationship data of a travel area corresponding to the travel order.

[0009] According to a fifth aspect of the embodiments of the present application, an electronic device is provided, including: a processor, a memory, a communication interface, and a communication bus, where the processor, the memory, and the communication interface complete communication with each other through the communication bus; the memory is used for storing at least one executable instruction, and the executable instruction enables the processor to perform operations corresponding to the method described in the first aspect or the second aspect or the third aspect or the fourth aspect.

[0010] According to a sixth aspect of the embodiments of the present application, a computer storage medium is provided, on which a computer program is stored, and when the program is executed by a processor, the method described in the first aspect, the second aspect, the third aspect, or the fourth aspect is implemented.

[0011] According to a seventh aspect of the embodiments of the present application, a computer program product is provided, including computer instructions, and the computer instructions direct a computing device to perform operations corresponding to the method described in the first aspect, the second aspect, the third aspect, or the fourth aspect.

[0012] According to the model training solution provided by the embodiments of the present application, when training a machine learning model for outputting matching information between a travel order and a travel service object, on the one hand, relevant data is processed based on a polygon area grid to generate polygon grid data, which not only unifies the data representation form, but also takes the polygon grid corresponding to the travel order as a unit to associate the three parts of data corresponding to the grid (travel order data, travel and service demand relationship data, and travel service object data), realizing the effective integration of multi-dimensional data, and each polygon grid can effectively utilize the data of surrounding grids, thereby making the training of the model more effective and accurate; on the other hand, the data of the travel order, the data of the travel and service demand relationship, and the data of the travel service object are considered in the constructed training samples, so as to incorporate multi-dimensional data related to travel into the model training at the same time, making the information contained in the training samples richer and more comprehensive, and avoiding defects in model training caused by single-dimensional consideration or non-consideration of travel object factors, resulting in the problem that the finally trained machine model cannot consider the travel situation globally and the model training result is inaccurate.

[0013] The machine learning model obtained based on the above model training solution can obtain more accurate matching information between the departure order and the travel service object during actual travel scheduling, so as to be able to allocate travel orders more reasonably. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments recorded in the embodiments of the present application, and those of ordinary skill in the art can also obtain other drawings based on these drawings.

[0015] Figure 1A It is a step flowchart of a model training method according to Embodiment 1 of the present application;

[0016] Figure 1B is Figure 1ASchematic diagram of an exemplary hexagonal grid data generation in the illustrated embodiment;

[0017] Figure 1C For Figure 1A Schematic diagram of an exemplary layer integration process in the illustrated embodiment;

[0018] Figure 2A Flowchart of steps of a model training method according to Embodiment 2 of the present application;

[0019] Figure 2B For Figure 2A Schematic diagram of a model training process in the illustrated embodiment;

[0020] Figure 3 Flowchart of steps of a travel scheduling method according to Embodiment 3 of the present application;

[0021] Figure 4 Flowchart of steps of a travel scheduling method according to Embodiment 4 of the present application;

[0022] Figure 5 Flowchart of steps of a travel scheduling method according to Embodiment 5 of the present application;

[0023] Figure 6 Schematic diagram of the structure of an electronic device according to Embodiment 6 of the present application. Detailed implementation manners

[0024] In order to enable those skilled in the art to better understand the technical solutions in the embodiments of the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments in the present application shall fall within the protection scope of the embodiments of the present application.

[0025] The following further illustrates the specific implementation of the embodiments of the present application with reference to the accompanying drawings of the embodiments of the present application.

[0026] Embodiment 1

[0027] Referring to FIG. 1, it shows a flowchart of steps of a model training method according to Embodiment 1 of the present application.

[0028] The model training method of this embodiment includes the following steps:

[0029] Step S102: Obtain travel order sample data, travel service object sample data, and travel and service demand relationship sample data.

[0030] Among them, the travel order sample data is used to reflect the information of travel orders, including at least the starting location and the ending location of the travel. Optionally, it may also include information such as travel time, order sending time, order sending account, etc. The travel service object sample data is used to reflect the relevant information of the object providing the travel service (such as the driver), including at least the service probability of the service object from the service starting location to the service ending location. Optionally, it may also include evaluation information of the travel service object, order cancellation rate information, etc. The travel and service demand relationship sample data is used to reflect the supply-demand relationship between travel demands and available travel services within a unit time (such as the supply-demand relationship between orders and drivers who can receive orders), including at least travel demand data and service demand data. Optionally, it may also include time information corresponding to the supply-demand relationship, etc. In practical applications, the above data can be obtained based on real historical data with permission; or it can be obtained only based on a small part of real historical data through data simulation or emulation methods.

[0031] In the embodiments of the present application, the inventor believes that the travel order sample data, the travel service object sample data, and the travel and service demand relationship sample data are all closely related to travel and can comprehensively reflect the travel situation from multiple dimensions. Compared with some solutions that only consider travel orders and supply-demand relationships but do not consider travel service objects, the embodiments of the present application also incorporate the information of travel service objects into the consideration scope. In addition to making the information more comprehensive, it also emphasizes the important role of travel service objects in travel services, which helps to conduct a more reasonable and comprehensive evaluation of travel service objects and improve the platform's usage experience and adhesion to travel service objects.

[0032] Step S104: Use a polygon area grid to generate corresponding order polygon grid data, service object polygon grid data, and demand relationship polygon grid data for the travel order sample data, the travel service object sample data, and the travel and service demand relationship sample data respectively.

[0033] A grid system is a system that divides the Earth's space into recognizable grid cells for analyzing massive spatial datasets. The polygon shapes adopted by this grid system include, but are not limited to: triangles, quadrilaterals, and hexagons. Because in a grid-based spatial index, the more sides a polygon used has, the more a grid approximates a circle, which is more convenient for subsequent operations such as buffer queries. Also, because the grid index requires that the space can be filled with grids without gaps. Based on the polygon interior angle sum formula, that is, θ=(x - 2)*180°, the angle of each angle of a regular polygon is where x represents the number of sides of the polygon, and θ is the interior angle sum of the polygon. And if it is required that the polygon can fill the space, at the intersection of the polygon vertices, assuming that there are y polygons intersecting at each vertex, it is necessary to satisfy: The solutions for x obtained therefrom are respectively: x = 3, y = 6; x = 4, y = 4; x = 6, y = 3. It can be seen therefrom that the polygon grid in the embodiments of the present application can be a triangle or a quadrilateral or a hexagon. Also, since the hexagon has the most sides and is closest to a circle, it has become the implementation method for most map grids. In multiple embodiments of the present application, the hexagonal area grid is taken as an example for illustration. However, those skilled in the art should understand that in practical applications, those skilled in the art can also choose a triangular grid or a quadrilateral grid. In addition, a more suitable polygon grid can be selected according to the shape of the actual geographical area, and all are within the protection scope of the present application.

[0034] The polygon area grid divides the earth space into multiple polygon grid cells, and each of the grids is a regular polygon such as a regular hexagon, representing the corresponding geographical area range, and can be used for the visualization and mining of map spatial data. When it is necessary to analyze map spatial data with a fine granularity, data at different geographical locations can be stored in the polygon grid for subsequent processing and analysis. In practical applications, for the hexagonal area grid of a certain geographical area or some geographical areas, existing grids can be used, or those skilled in the art can construct it in a manner such as the ISEA3H (Icosahedral Snyder Equal Area Aperture 3Hexagonal Grid) method or the Uber H3 algorithm method according to actual needs. The specific implementation of constructing the hexagonal area grid using these algorithms can refer to the description in related technologies, and the embodiments of the present application do not limit this.

[0035] In the embodiments of the present application, based on the travel order sample data, the travel service object sample data, and the travel and service demand relationship sample data respectively, order hexagonal grid data corresponding to the travel order sample data, service object hexagonal grid data corresponding to the travel service object sample data, and demand relationship hexagonal grid data corresponding to the travel and service demand relationship sample data are generated. It should be noted that the generation of the order hexagonal grid data, the service object hexagonal grid data, and the demand relationship hexagonal grid data can be in any order or can be executed in parallel in specific implementations.

[0036] Still taking the hexagonal area grid as an example, since different types of grid data are used to reflect different information, when specifically generating the corresponding grid data, those skilled in the art can adopt different generation methods. For example, the difference between the order demand quantity and the number of service-providing objects in a certain geographical area at a certain time period in the sample data of the travel and service demand relationship can be used as the grid data of the corresponding hexagonal grid in this geographical area. Another example is that the starting position and ending position of a trip can be determined based on the trip order sample data, and different reference values or weight values can be set based on the hexagonal grids to which these two positions belong to reflect the corresponding trip order situation. Still another example is that based on the regular travel service positions of the travel service objects in the travel service object sample data, the probability that a travel service object may provide travel services within the geographical area corresponding to a certain hexagonal grid can be determined, and the grid data of this hexagonal grid can be set based on this, and so on. These hexagonal grid data can provide an effective and rich data basis and basis for subsequent data processing and analysis.

[0037] Next, a specific example is used to illustrate the generation of the above-mentioned hexagonal grid data.

[0038] Suppose that in a certain geographical area (such as a certain city), as Figure 1B shown, there are four hexagonal area grids (those skilled in the art should understand that in actual applications, the number of area grids in a city will be much more than this, but for the convenience of example, only four are used for illustration), namely grids 1, 2, 3, and 4.

[0039] For a certain travel order sample data X, suppose its corresponding starting position of the trip is A, and according to its geographical location, it is determined that it belongs to grid 1 and will be associated with grid 1; its corresponding ending position of the trip is B, and according to its geographical location, it is determined that it belongs to grid 2 and will be associated with grid 2. Exemplarily, it can be simply shown in Table 1 below:

[0040] Departure location A Destination location B Hexagonal area grid 1 1 0 Hexagonal area grid 2 0 1 Hexagonal area grid 3 0 0 Hexagonal area grid 4 0 0

[0041] Corresponding to this travel order sample data X, according to its starting position A of the trip, it is determined that the screening of travel service objects needs to be carried out in grid 1. For example, three travel service objects within a certain distance (such as within 2 kilometers) from the starting position A of the trip are screened out in grid 1. The sample data of these three travel service objects are Y1, Y2, and Y3 respectively. Among them, the probability that Y1 provides travel services from A to B is C1, the probability that Y2 provides travel services from A to B is C2, and the probability that Y3 provides travel services from A to B is C3. Next, taking one travel service object as an example, an example method for obtaining the travel service probability of this travel service object is described. As shown in Table 2 below:

[0042]

[0043] In the above table, the first column represents the departure grid, and the first row represents the arrival grid. As can be seen from the table, the service probability provided by the travel service object from grid 1 to grid 1 is 90% (i.e., providing service within this grid), the service probability from grid 1 to grid 2 is 80%, the service probability from grid 1 to grid 3 is also 80%, and the service probability from grid 1 to grid 4 is 60%. Each travel service object has similar information. When it is determined that the travel order X is from grid 1 to grid 2, the travel service probability provided by the travel service object shown in Table 2 above is 80%. And so on, each travel service object has similar tabular data, and based on this, the travel service probabilities of other travel service objects for the travel order X can be obtained. It should be noted that the above service probabilities can also be expressed in the form of fractions or decimals.

[0044] Furthermore, within the preset time period when the travel time of the travel order sample data X is located, assume that the sample data of the travel and service demand relationship is Z. Exemplarily, the sample data of the travel and service demand relationship can be implemented as the difference between the number of travel orders in a certain grid and the number of travel service objects that can provide travel services within the preset time period. Exemplarily, as shown in Table 3 below:

[0045] Number of travel service objects Number of travel orders Demand relationship Hexagonal area grid 1 40 25 +15 Hexagonal area grid 2 30 27 +3 Hexagonal area grid 3 30 50 -20 Hexagonal area grid 4 45 47 -2

[0046] As can be seen, in grid 1, supply exceeds demand, and there are more travel service objects that can provide travel services than travel orders; in grid 2, the supply-demand relationship is basically balanced, and there are slightly more travel service objects that can provide travel services than travel orders; in grid 3, demand exceeds supply, and there is a significant shortage of travel service objects that can provide travel services; in grid 4, the supply-demand relationship is also basically balanced, and there are slightly fewer travel service objects that can provide travel services than travel orders.

[0047] Based on the above settings, as Figure 1BAs shown in , the grid data of grid 1 and grid 2 corresponding to the travel order X will be mainly generated. Exemplarily, the order hexagonal grid data generated for the travel order X includes at least the information of the travel starting point position A in grid 1 and the information of the travel end point position B in grid 2. Taking the travel starting point position A as an example, the longitude and latitude information corresponding to A will be determined first, and the corresponding hexagonal grid will be determined as grid 1 based on the longitude and latitude information. Assuming that the index of grid 1 is simply indicated as ID-1, an association relationship of ID-1-information of travel order X-longitude and latitude information corresponding to A will be formed. Because A is the starting point, there will also be corresponding direction information pointing to B, which is simply indicated as S1. Then a hexagonal grid data similar to ID-1-information of travel order X-longitude and latitude information corresponding to A-S1 is formed. Similarly, the travel end point position B will form a hexagonal grid data similar to ID-1-information of travel order X-longitude and latitude information corresponding to B-E1. If the above relationship is indicated by a triangle, the order hexagonal grid data generated by the travel order X is shown as the triangle in the hexagonal area grids 1 and 2.

[0048] For the three travel service objects, each travel service object generates corresponding service object hexagonal grid data, and each service object hexagonal grid data includes at least the service probability from grid 1 to grid 2 provided by the service object, and the service probability from grid 2 to grid 1. Taking one of the travel service objects 1 as an example, assuming that its service probability in grid 1 is 90% and its service probability in grid 2 is 80%, it will form grid data similar to ID-1-information of travel order X-information of travel service object 1-grid data of grid 1 with a service probability of 90%, and ID-1-information of travel order X-information of travel service object 1-grid data of grid 2 with a service probability of 80%. If the above relationship is illustrated by a circle, the service object hexagonal grid data generated by travel order X is shown as the circle in hexagonal area grids 1 and 2.

[0049] As for the demand relationship hexagonal grid data, it will at least include the travel and service demand relationship in grid 1 and the travel and service demand relationship in grid 2. Taking the travel and service demand relationship of grid 1 as +15 and the travel and service demand relationship of grid 2 as +3 as an example, the grid data of grid 1 with ID-1-information of travel order X-travel and service demand relationship of +15 and the grid data of grid 2 with ID-1-information of travel order X-travel and service demand relationship of +3 will be formed. If the above relationship is indicated by a rectangle, the demand relationship hexagonal grid data generated by travel order X is shown as the rectangle in the hexagonal area grids 1 and 2.

[0050] Step S106: Construct training samples based on the order polygon grid data, the service object polygon grid data, and the demand relationship polygon grid data, and train a machine learning model to obtain a machine learning model for outputting the matching information between the travel order and the travel service object.

[0051] After obtaining the three types of polygon grid data, training samples can be constructed with each type as a unit. That is, the training sample includes three types of data sets, and each data set corresponds to one type of polygon grid data. Based on the constructed training samples, the machine learning model is trained. In practical applications, the machine learning model can be in an appropriate form of machine learning model adopted by those skilled in the art according to actual needs. As long as it can output the matching information between the travel order and the travel service object after training. For example, a convolutional neural network model or a graph convolutional neural network model can be selected.

[0052] In a feasible manner, the data set can also be implemented in the form of a layer. That is, according to the order polygon grid data, the service object polygon grid data, and the demand relationship polygon grid data, the corresponding order layer, service object layer, and supply and demand layer are constructed respectively.

[0053] For an electronic map, it is usually constructed by multiple layers and finally combined to form an electronic map.

[0054] Based on this principle, in the embodiments of the present application, different layers are constructed respectively based on the polygon grids representing different information, that is, the order layer corresponding to the order polygon grid data, the service object layer corresponding to the service object polygon grid data, and the supply and demand layer corresponding to the demand relationship hexagon grid data. The data of these three parts of layers can be integrated to form rich image data reflecting the travel situation. Since the specifications of the hexagonal area grids of each layer are the same, based on the identifier of one grid, the data of this grid in the three different layers can be associated, thereby realizing the integration of the three different parts of data of the same grid. Then, subsequently, the processing of these data can be transformed into the processing of images.

[0055] Furthermore, training samples can be constructed based on the order layer, the service object layer, and the supply and demand layer, and the machine learning model is trained to obtain a machine learning model for outputting the matching information between the travel order and the travel service object.

[0056] As mentioned above, the order layer, the service object layer, and the supply and demand layer can be integrated with data to form image data including multiple layers as training samples to train the machine learning model.

[0057] It should be noted that in the embodiments of the present application, unless otherwise specified, the quantities related to "multiple", "a variety of", etc. in the embodiments of the present application all mean two or more.

[0058] Still using the example of the travel order X mentioned above, after obtaining the order hexagonal grid data, service object hexagonal grid data, and demand relationship hexagonal grid data generated for the travel order X, three corresponding layers can be generated according to these three parts of hexagonal grid data, namely the order layer, the service object layer, and the supply and demand layer.

[0059] An exemplary layer is as Figure 1C shown. Among them, the order hexagonal grid data is represented as a triangle, the service object hexagonal grid data is represented as a circle, and the demand relationship hexagonal grid data is represented as a rectangle. It can be seen that the above-mentioned grid data are all present in Grid 1 and Grid 2. It should be noted that only the service object hexagonal grid data corresponding to one service object is schematically shown in the figure, corresponding to the circle. The service object hexagonal grid data of other service objects can be processed with reference to this method.

[0060] After the order layer, the service object layer, and the supply and demand layer are completed, they can be integrated according to their corresponding grids. Since there are three groups of service object hexagonal grid data, three training samples will be generated in total, Figure 1C which are respectively shown as Training Samples 1, 2, and 3.

[0061] Furthermore, Training Samples 1, 2, and 3 can be used as sample data in the training sample set to train a machine learning model, so as to obtain a machine learning model for outputting the matching information between travel orders and travel service objects. In a feasible manner, the sample data in the training sample set can also be clustered first, such as clustering according to travel orders, or clustering according to travel service objects, etc., and then the sample data is divided into multiple types. When training the machine learning model, it can be trained specifically for different types of sample data, or multiple machine learning models can be used to train different types of sample data respectively to obtain a more targeted machine learning model.

[0062] It should be noted that in the foregoing example, the travel service object grid data and the demand relationship grid data are only exemplarily described by the grids corresponding to the travel starting position and the travel ending position. However, in actual applications, the travel service object grid data for a certain travel service object may include all the polygon grid data corresponding to its travel service; and the demand relationship grid data may include the demand relationship grid data of all the polygon grids within a certain time period.

[0063] It can be seen that, through this embodiment, when training a machine learning model for outputting matching information between travel orders and travel service objects, on the one hand, relevant data is processed based on a polygon area grid to generate polygon grid data. This not only unifies the data representation form, but also takes the polygon grid corresponding to the travel order as a unit to associate the three parts of data corresponding to this grid (travel order data, travel and service demand relationship data, and travel service object data), realizing the effective integration of multi-dimensional data. Moreover, each polygon grid can effectively utilize the data of surrounding grids, making the model training more effective and accurate. On the other hand, the travel order data, travel and service demand relationship data, and travel service object data are considered in the constructed training samples, so as to incorporate multi-dimensional data related to travel into the model training at the same time, making the information contained in the training samples richer and more comprehensive, and avoiding defects in model training caused by single-dimensional consideration or non-consideration of travel object factors, resulting in the problem that the finally trained machine model cannot consider the travel situation globally and the model training result is inaccurate.

[0064] Based on the order polygon grid data, service object polygon grid data, and demand relationship polygon grid data, constructing the corresponding layer form can transform the processing of multi-dimensional data into the processing of layers, further improving the data processing efficiency.

[0065] The model training method of this embodiment can be executed by any suitable electronic device with data processing capabilities, including but not limited to: servers, PCs, and even mobile terminals with high software and hardware performance (such as mobile phones, PADs, etc.).

[0066] Embodiment 2

[0067] Refer to Figure 2A , which shows a step flow chart of a model training method according to Embodiment 2 of the present application.

[0068] In this embodiment, taking a hexagonal area grid as an example, the model training method provided by the embodiments of the present application is described. However, those skilled in the art should understand that other polygon area grids can also refer to this embodiment to implement the corresponding model training.

[0069] The model training method of this embodiment includes the following steps:

[0070] Step S202: Obtain travel order sample data, travel service object sample data, and travel and service demand relationship sample data.

[0071] Among them, the travel order sample data is used to reflect the information of travel orders, including at least information such as the starting location and ending location of the trip; the travel service object sample data is used to reflect the relevant information of the object providing the travel service, including at least information such as the service probability of the service object from the service starting location to the service ending location; the travel and service demand relationship sample data is used to reflect the supply-demand relationship between the travel demand and the available travel services within a unit time, including at least information such as travel demand data and service demand data.

[0072] It should be noted that if, under the condition of obtaining the use permission, the travel order sample data, travel service object sample data, and travel and service demand relationship sample data from actual historical data are used, the subsequent constructed training samples can better reflect the real situation, and the trained machine learning model is also more in line with the actual situation.

[0073] Step S204: Use the hexagonal area grid to generate corresponding order hexagonal grid data, service object hexagonal grid data, and demand relationship hexagonal grid data for the travel order sample data, travel service object sample data, and travel and service demand relationship sample data respectively.

[0074] In this embodiment, first determine the target geographical area, such as a certain city, or a certain area within a certain city, or several adjacent cities, etc. Divide the target geographical area according to the hexagonal grid, such as the H3 grid, and each grid corresponds to a geographical area range.

[0075] After dividing the hexagonal grid, generate corresponding order hexagonal grid data, service object hexagonal grid data, and demand relationship hexagonal grid data according to the travel order sample data, travel service object sample data, and travel and service demand relationship sample data respectively.

[0076] In a feasible solution, the hexagonal area grid can be used to generate order hexagonal grid data according to the starting location and ending location of the trip in the travel order sample data, generate service object hexagonal grid data according to the service probability of the service starting location and service ending location in the travel service object sample data, and generate demand relationship hexagonal grid data according to the travel demand data and service demand data in the travel and service demand relationship sample data. Thus, travel data in multiple dimensions is constructed to comprehensively reflect the corresponding travel information.

[0077] Among them, when generating order hexagonal grid data based on the starting location and ending location of a travel order sample data, for each travel order sample data, the starting location and ending location of the travel can be obtained therefrom; determine the hexagonal grid where the starting location of the travel is located and set a first weight for this hexagonal grid, and, determine the hexagonal grid where the ending location of the travel is located and set a second weight for this hexagonal grid, where the first weight is less than the second weight; generate order hexagonal grid data according to the determined hexagonal grids and the weights corresponding to each hexagonal grid.

[0078] For example, for a certain travel order sample data, assign 0.1 (the first weight) to the H3 grid where its corresponding starting location of the travel is located, and assign 0.9 (the second weight) to the H3 grid where the ending location of the travel is located. In this way, in addition to effectively reflecting the travel location information of the travel order, the higher value assigned to the H3 grid at the ending location of the travel also takes into account that the travel service object may continue to provide services in the area corresponding to this H3 grid in the future. However, it should be noted that the above assignment setting is only for illustrative purposes. In actual applications, those skilled in the art can set it appropriately according to actual needs as long as the first weight is less than the second weight.

[0079] When generating service object hexagonal grid data based on the service probability between the service starting location and the service ending location in the travel service object sample data, the service probability of the travel service object from the service starting location to the service ending location can be obtained according to the historical travel service data of the travel service object corresponding to the travel service object sample data; generate service object hexagonal grid data according to the service probability. The historical travel service data of the travel service object can effectively reflect the likelihood of the travel service object providing services in certain areas. For example, a certain travel service object only provides services in certain fixed areas and is not willing to go to other areas; the service areas of some travel service objects are not fixed, and so on.

[0080] For example, taking driver X (travel service object) as an example, obtain his historical travel service data. Assume that according to statistical analysis, it is determined that driver X provided travel services 50 times in the previous month. Among them, 25 times from place A to place B, 15 times from place A to place C, and 10 times from a specific location in place A to another specific location in place A. Then, the H3 grid data corresponding to place A for driver X is 10 / 50 = 0.2, the H3 grid data corresponding to place B is 25 / 50 = 0.5, and the H3 grid data corresponding to place C is 15 / 50 = 0.3. It can be seen that the likelihood of driver X from the H3 grid with a certain location as the starting location of the travel to other H3 grids can be learned from the historical travel service data of driver X.

[0081] When generating demand relationship hexagonal grid data from travel demand data and service demand data in the sample data of the relationship between travel and service demand, for each hexagonal grid within a preset area, the sample data of the relationship between travel and service demand corresponding to this hexagonal grid can be obtained; based on the sample data of the relationship between travel and service demand corresponding to each hexagonal grid, the number of travel service objects in this hexagonal grid can be predicted; based on the difference between the number of travel service objects and the number of travel orders in this hexagonal grid obtained from the sample data of the relationship between travel and service demand, the demand relationship hexagonal grid data of this hexagonal grid can be generated. The difference between the number of travel service objects and the number of travel orders in a certain hexagonal grid can effectively reflect the supply-demand difference. In addition, for a hexagonal grid, it can also effectively utilize the data information of surrounding grids. Therefore, if there is a subsequent demand for supply-demand allocation, it is also convenient to perform transport capacity allocation according to the demand relationship data of relevant hexagonal grids.

[0082] Among them, when predicting the number of travel service objects in a certain hexagonal grid, those skilled in the art can adopt appropriate methods for prediction according to actual needs, including but not limited to the neural network model method, the prediction algorithm method, etc. In a feasible method, the historical demand data and the position data of the corresponding travel service objects within the geographical area corresponding to this hexagonal grid within a certain time period can be obtained first, and then the number of travel service objects can be predicted based on this.

[0083] In addition, because the demand relationship has timeliness. For example, the demand relationships from 5:00 pm to 5:30 pm and from 5:30 pm to 6:00 pm may be very different. In order to more timely reflect the actual demand relationship, in a feasible method, before obtaining the sample data of the relationship between travel and service demand corresponding to each hexagonal grid within the preset area, it further includes: obtaining multiple groups of sample data sets of the relationship between travel and service demand corresponding to the preset area according to a preset number of time periods; where each group of sample data sets of the relationship between travel and service demand includes multiple sample data of the relationship between travel and service demand. Subsequently, different supply-demand layers can be constructed according to the demand relationship hexagonal grid data corresponding to each group of sample data sets of the relationship between travel and service demand. That is, if there are N groups of sample data sets of the relationship between travel and service demand, there will be N supply-demand layers, which are called supply-demand sub-layers in the embodiments of the present application, and multiple supply-demand sub-layers together are called the supply-demand layer. Based on this, obtaining the sample data of the relationship between travel and service demand corresponding to each hexagonal grid within the preset area can include: for each group of sample data sets of the relationship between travel and service demand, respectively perform the operation of obtaining the sample data of the relationship between travel and service demand corresponding to each hexagonal grid within the preset area according to the time period.

[0084] For example, taking a certain moment as a reference, the demand relationships for the next 5 minutes, 10 minutes, 15 minutes, ..., 60 minutes at this moment are predicted to obtain 11 sub-layers. The value of each H3 grid in each sub-layer is the difference between the predicted number of travel service objects and the number of travel orders. It should be noted that the above 5 minutes, 10 minutes, and 15 minutes are for illustrative purposes only. Those skilled in the art can set the time interval according to actual needs, and the time intervals can be equal or unequal. For example, the next 5 minutes, 15 minutes, 30 minutes, 60 minutes at this moment (correspondingly, 4 sub-layers will be generated), etc. The embodiments of the present application do not limit this.

[0085] Step S206: According to the order hexagonal grid data, service object hexagonal grid data, and demand relationship hexagonal grid data, respectively construct corresponding order layers, service object layers, and supply-demand layers.

[0086] As mentioned above, whether it is an image or a layer, it is presented in the form of data. Based on this, the hexagonal grid data can be processed into a logical layer form, and an order layer corresponding to the order hexagonal grid data, a service object layer corresponding to the service object hexagonal grid data, and a supply-demand layer corresponding to the demand relationship hexagonal grid data are constructed.

[0087] A schematic diagram of a constructed layer is as Figure 2B shown in. Among them, the order layer is shown as channel1 in Figure 2B shown in, the supply-demand layer is shown as channel 2 in Figure 2B shown in, and the service object layer is shown as channel 3 in Figure 2B shown in.

[0088] In addition, as mentioned above, for multiple sets of travel and service demand relationship sample data sets, when constructing a corresponding supply-demand layer according to the demand relationship hexagonal grid data, it includes: taking each set of travel and service demand relationship sample data sets as a unit, respectively constructing corresponding multiple supply-demand sub-layers for the demand relationship hexagonal grid data generated from multiple sets of travel and service demand relationship sample data sets ( Figure 2B not shown in). In Figure 2B shown, only the supply-demand layer channel 2 formed from these multiple supply-demand sub-layers is shown.

[0089] Step S208: Construct a training sample based on the order layer, service object layer, and supply-demand layer.

[0090] In this step, the data of the order layer, service object layer, and supply-demand layer can be integrated to form a training sample. In a feasible manner, the order layer, service object layer, and supply-demand layer can be integrated to form an image, and this image is used as the training sample.

[0091] Step S210: Collect historical trip order data, and analyze the historical trip order data to obtain the model supervision data corresponding to the training samples.

[0092] This model supervision data is the label data of the training samples, which can reflect the matching degree between the real trip order corresponding to the training samples and the trip service object. Through the historical trip order data, information about historical trip orders, information about the service objects providing trip services for historical trip orders, historical trip information, information about the demand relationship at that time corresponding to historical trip orders, etc. can be obtained. In a feasible way, statistical analysis can be performed based on this information to obtain the model supervision data. Optionally, a neural network model for predicting the order cancellation rate can also be used to predict the order cancellation rate, and this order cancellation rate factor can be used as a factor affecting the matching degree between the trip order and the trip service object, so as to determine the final model supervision data.

[0093] It should be noted that this step can be executed during the execution of steps S204 - S208, or before or after the execution of any one of the steps, or executed in parallel.

[0094] Step S212: Train the machine learning model according to the training samples and the model supervision data.

[0095] As mentioned above, the machine learning model in the embodiments of the present application can be a suitable model, including but not limited to a convolutional neural network model, a graph convolutional neural network model, etc. The embodiments of the present application do not limit the specific model used and the specific structure of the model. Figure 2B In, the machine learning model is simply shown as the part in the dashed box. After training, this machine learning model will output the matching information between the trip order and the trip service object, such as Figure 2B the "matchingscore" shown in.

[0096] From Figure 2B and the above process, it can be seen that the supply - demand layer can provide effective context information for the order layer and the service object layer to more accurately reflect the relationship between the trip order and the trip service object and obtain more accurate training results.

[0097] In this embodiment, when training a machine learning model for outputting the matching information between travel orders and travel service objects, on the one hand, relevant data is processed based on hexagonal area grids to generate hexagonal grid data, and corresponding layers are constructed based on this. Thus, the processing of data in multiple dimensions is transformed into the processing of layers, which not only unifies the data representation form, realizes the effective integration of data in multiple dimensions, but also each hexagonal grid can effectively utilize the data of surrounding grids, making the training of the model more effective and accurate. On the other hand, the travel order data, the data of the relationship between travel and service needs, and the travel service object data are considered in the constructed training samples, so as to incorporate data in multiple dimensions related to travel into the model training at the same time, making the information contained in the training samples richer and more comprehensive, and avoiding defects in model training caused by single-dimensional consideration or non-consideration of travel object factors, resulting in the problem that the finally trained machine model cannot consider the travel situation globally and the model training result is inaccurate.

[0098] The model training method of this embodiment can be executed by any suitable electronic device with data processing capabilities, including but not limited to: servers, PCs, and even mobile terminals with high software and hardware performance (such as mobile phones, PADs, etc.).

[0099] Embodiment III

[0100] Refer to Figure 3 , which shows a step flowchart of a travel scheduling method according to Embodiment III of the present application.

[0101] In this embodiment, based on the machine learning model trained in the foregoing Embodiment I or II, travel orders are allocated. The travel scheduling method of this embodiment includes the following steps:

[0102] Step S302: Receive a travel order and obtain the travel data in the travel order.

[0103] Among them, the travel data includes at least the travel starting point location and the travel ending point location.

[0104] Step S304: According to the travel data, determine the data of multiple candidate travel service objects providing travel services for the travel order, and the data of the relationship between travel and service needs in the travel area corresponding to the travel order.

[0105] Among them, the data of multiple candidate travel service objects and the data of service needs relationship are both real-time data or dynamically updated data.

[0106] After obtaining the travel data, the corresponding travel area can be determined. Based on this, the corresponding service objects in the travel area can be obtained. Furthermore, based on the starting position of the travel in the travel data, multiple candidate travel service objects can be determined for the travel order, and the data of the corresponding multiple candidate travel service objects can be obtained. In addition, the travel and service demand relationship data corresponding to the travel area can be obtained.

[0107] In a feasible manner, the travel area corresponding to the travel order can be expressed by using the aforementioned polygon area grid. That is to say, in the specific implementation of this step, the polygon area grid corresponding to the travel order can be determined first. Furthermore, appropriate multiple candidate travel service objects can be determined for the travel order in this grid, and the travel and service demand relationship data corresponding to this grid can be obtained. It should be noted that the specific determination process can refer to the description of the corresponding part in the foregoing embodiments. However, different from the foregoing embodiments, in this embodiment, all the data in the polygon area grid will obtain the use permission in advance and will be updated in real time or dynamically to timely reflect the data changes in the grid. Therefore, the data of the candidate travel service objects determined for the travel order and the travel and service demand relationship data are the most timely, so that a quick and accurate response can be made for the travel order.

[0108] Step S306: Select a target travel service object that provides travel services for the travel order according to the travel data, the data of multiple candidate travel service objects, and the travel and service demand relationship data.

[0109] After obtaining the travel data, the data of multiple candidate travel service objects, and the travel and service demand relationship data, for each candidate travel service object, these data can be input into a pre-trained machine learning model, and the matching information between the travel order and the candidate travel service object output by the machine learning model can be obtained. Through this matching information, the matching degree between the travel order and the travel service object can be determined. Furthermore, based on the high or low matching degree, the target travel service object that provides travel services can be selected from them.

[0110] In a feasible manner, when a polygon area grid is used to represent the travel area corresponding to a travel order, the polygon area grid can be used to generate corresponding order polygon grid data, service object polygon grid data, and demand relationship polygon grid data for travel data, data of multiple candidate travel service objects, and travel and service demand relationship data respectively; based on the order polygon grid data, service object polygon grid data, and demand relationship polygon grid data, a target travel service object that provides travel services for the travel order is selected from multiple candidate travel service objects. Thereby, not only the data representation form is unified, but also the three parts of data corresponding to the polygon grid of the travel order (data of the travel order, data of the travel and service demand relationship, and data of the travel service object) can be associated with each other with the polygon grid corresponding to the travel order as the unit, realizing the effective integration of multi-dimensional data.

[0111] Further optionally, corresponding order layers, service object layers, and supply-demand layers can also be constructed based on the order polygon grid data, service object polygon grid data, and demand relationship polygon grid data respectively; based on the order layers, service object layers, and supply-demand layers, the matching information between the travel order and multiple candidate travel service objects is obtained through a trained machine learning model, and a target travel service object that provides travel services is selected from multiple candidate travel service objects according to the matching information. By constructing the corresponding layers, it not only facilitates the unified processing of data but also better meets the requirements of the actual travel scenario.

[0112] In practical applications, the data corresponding to the order layer, service object layer, and supply-demand layer can be integrated to generate a corresponding order image, which is then input into a pre-trained machine learning model. Among them, the machine learning model is a machine learning model trained by the method described in the foregoing Embodiment 1 or 2.

[0113] Through this embodiment, multiple dimensions of information related to travel are comprehensively considered, realizing the consideration of travel scheduling from a global perspective. Also, since the data of the travel service object and the service demand relationship data are both real-time data or dynamically updated data, the subsequent data is faster and more accurate. Moreover, the machine learning model obtained based on the foregoing model training scheme can output relatively accurate results. Therefore, when performing actual travel scheduling, more accurate matching information between the departure order and the travel service object can be obtained, so that the travel order can be more reasonably allocated.

[0114] The model training method of this embodiment can be executed by any suitable electronic device with data processing capabilities, including but not limited to: servers, PCs, and even mobile terminals with high software and hardware performance (such as mobile phones, PADs, etc.).

[0115] Embodiment 4

[0116] Refer to Figure 4 , which shows a step flowchart of a travel scheduling method according to Embodiment 4 of the present application.

[0117] In this embodiment, the travel scheduling method of the present application is described from the side of the travel terminal of the travel object. The travel scheduling method includes the following steps:

[0118] Step S402: Receive the travel information input by the travel object and generate a travel order according to the travel information.

[0119] Among them, the travel order includes the travel start position and the travel end position.

[0120] Based on the received travel information, the travel terminal can generate a travel order. In addition to carrying the above travel information, the travel order also carries the identifier of the travel terminal.

[0121] Step S404: Send the travel order to the travel platform so that the travel platform can determine the data of multiple candidate travel service objects providing travel services according to the travel order, and the travel and service demand relationship data of the travel area corresponding to the travel order; and select a target travel service object providing travel services for the travel order from multiple candidate travel service objects according to the travel data, the data of multiple candidate travel service objects, and the travel and service demand relationship data.

[0122] The specific implementation of this step on the travel platform can refer to the description of the relevant parts in the foregoing multiple embodiments and will not be elaborated here.

[0123] Step S406: Receive the information of the target travel service object feedback by the travel platform.

[0124] After the travel platform assigns a target travel service object to the travel order, it will feedback the information of the target travel service object to the travel terminal of the travel object. Furthermore, the travel terminal can display the information of the target travel service object to the travel object.

[0125] It can be seen that through this embodiment, the travel object can obtain a better travel service object through the travel platform, improving the travel experience of the travel object (such as a passenger).

[0126] Embodiment 5

[0127] Refer to Figure 5 , which shows a step flowchart of a travel scheduling method according to Embodiment 5 of the present application.

[0128] This embodiment describes the travel scheduling method provided by the embodiments of the present application from the perspective of the travel terminal of the travel service object. The travel scheduling method of this embodiment includes the following steps:

[0129] Step S502: Receive a travel order sent by the travel platform.

[0130] Among them, the travel order contains the starting position and the ending position of the trip.

[0131] The travel platform provides corresponding interfaces, such as corresponding application programs, to the travel terminal of the travel service object. Therefore, the travel terminal can receive the travel order through the application program and display the information of the travel order through the interface of the application program.

[0132] Step S504: Generate a response message according to the acceptance operation of the travel service object for the travel order.

[0133] Among them, the response message carries the information of the travel service object.

[0134] As mentioned above, the information of the corresponding travel order will be displayed in the interface of the application program. If the travel service object is willing to provide travel services for the travel order, the acceptance operation can be performed through corresponding options in the interface, such as the "Confirm" option or the "Grab Order" option, etc. After receiving the acceptance operation, the travel terminal in this embodiment will generate a corresponding response message, and the response message carries the information of the travel service object.

[0135] Step S506: Feedback the response message to the travel platform so that the travel platform can select a target travel service object that provides travel services for the travel order from multiple candidate travel service objects according to the travel order, the information of the travel service object, and the travel and service demand relationship data of the travel area corresponding to the travel order.

[0136] After receiving the response message, the travel platform will obtain the data of the travel service object and the travel and service demand relationship data of the travel area corresponding to the travel order, and then combine these parts of data to determine the matching degree between the travel service object and the travel order. Also, because there will be other travel service objects responding to the travel order, based on the matching degree between each travel service object and the travel order, a target travel service object that provides travel services for the travel order can be selected from multiple candidate travel service objects.

[0137] Among them, for the specific implementation of selecting a target travel service object that provides travel services for the travel order from multiple candidate travel service objects by combining these parts of data, reference can be made to the corresponding descriptions in the foregoing Embodiments 1 to 3, which will not be elaborated here.

[0138] Through this embodiment, the travel service object can obtain more travel order resources through the first travel platform. Also, since the travel platform screens travel service objects based on the service quality data of the travel service objects, it can effectively promote the improvement of the service quality of travel service objects, thereby achieving the improvement of the overall travel service quality and also enhancing the travel experience of travel objects.

[0139] Embodiment Six

[0140] Referring to Figure 6 , a schematic structural diagram of an electronic device according to Embodiment Six of the present application is shown. The specific implementation of the electronic device is not limited in the specific embodiments of the present application.

[0141] As Figure 6 shown, the electronic device may include: a processor 602, a communication interface 604, a memory 606, and a communication bus 608.

[0142] Among them:

[0143] The processor 602, the communication interface 604, and the memory 606 communicate with each other through the communication bus 608.

[0144] The communication interface 604 is used to communicate with other electronic devices or servers.

[0145] The processor 602 is used to execute the program 610, and specifically can execute the relevant steps in the above-mentioned multiple model training method embodiments, or execute the relevant steps in the above-mentioned multiple travel scheduling method embodiments.

[0146] Specifically, the program 610 may include program code, and the program code includes computer operation instructions.

[0147] The processor 602 may be a CPU, or a specific integrated circuit ASIC (Application Specific Integrated Circuit), or one or more integrated circuits configured to implement the embodiments of the present application. One or more processors included in the intelligent device may be of the same type of processor, such as one or more CPUs; or may be of different types of processors, such as one or more CPUs and one or more ASICs.

[0148] The memory 606 is used to store the program 610. The memory 606 may include a high-speed RAM memory, and may also include non-volatile memory, such as at least one disk memory.

[0149] The program 610 can specifically be used to cause the processor 602 to perform the operations described in the foregoing multiple embodiments.

[0150] For the specific implementation of each step in the program 610, reference may be made to the corresponding descriptions in the corresponding steps and units in the foregoing method embodiments of model training or travel scheduling methods, and there are corresponding beneficial effects, which will not be elaborated here. Those skilled in the art can clearly understand that for the convenience and conciseness of description, the specific working processes of the devices and modules described above can refer to the corresponding process descriptions in the foregoing method embodiments, and will not be elaborated here.

[0151] In addition, an embodiment of the present application also provides a computer program product, including computer instructions, where the computer instructions direct a computing device to perform the operations corresponding to the methods described in any one of the foregoing multiple method embodiments of model training; or perform the operations corresponding to the travel scheduling methods described in any one of the foregoing multiple method embodiments of travel scheduling.

[0152] It should be noted that according to the needs of implementation, each component / step described in the embodiments of the present application can be split into more components / steps, or two or more components / steps or partial operations of components / steps can be combined into new components / steps to achieve the purpose of the embodiments of the present application.

[0153] The method according to the embodiments of the present application can be implemented in hardware, firmware, or be implemented as software or computer code that can be stored in a recording medium (such as a CD ROM, RAM, floppy disk, hard disk, or magneto-optical disk), or be implemented as computer code that is originally stored in a remote recording medium or a non-transitory machine-readable medium and downloaded through a network and will be stored in a local recording medium, so that the method described herein can be stored in such software processing on a recording medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware (such as an ASIC or FPGA). It can be understood that a computer, a processor, a microprocessor controller, or programmable hardware includes a storage component (such as a RAM, ROM, flash memory, etc.) that can store or receive software or computer code, and when the software or computer code is accessed and executed by the computer, the processor, or the hardware, the model training method or travel scheduling method described herein is implemented. In addition, when a general-purpose computer accesses the code for implementing the model training method or travel scheduling method shown herein, the execution of the code converts the general-purpose computer into a dedicated computer for executing the model training method or travel scheduling method shown herein.

[0154] Those of ordinary skill in the art can realize that the units and method steps of each example described in combination with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution. Professional technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of the embodiments of this application.

[0155] The above embodiments are only used to illustrate the embodiments of this application, rather than limiting the embodiments of this application. Those of ordinary skill in the relevant technical field can also make various changes and modifications without departing from the spirit and scope of the embodiments of this application. Therefore, all equivalent technical solutions also belong to the scope of the embodiments of this application. The patent protection scope of the embodiments of this application should be defined by the claims.

Claims

1. A model training method, including: obtaining travel order sample data, travel service object sample data, and travel and service demand relationship sample data, wherein the travel service object sample data at least includes the service probability of the service object from the service start position to the service end position, and the service probability is generated based on historical travel service data; using a polygon area grid, respectively generating order polygon grid data according to the travel start position and travel end position in the travel order sample data, generating service object polygon grid data according to the service probability of the service start position and service end position in the travel service object sample data, and generating demand relationship polygon grid data according to the travel demand data and service demand data in the travel and service demand relationship sample data; constructing a training sample according to the order polygon grid data, the service object polygon grid data, and the demand relationship polygon grid data, and training a machine learning model to obtain a machine learning model for outputting matching information between travel orders and travel service objects.

2. The method according to claim 1, wherein, the constructing a training sample according to the order polygon grid data, the service object polygon grid data, and the demand relationship polygon grid data includes: respectively constructing corresponding order layers, service object layers, and supply and demand layers according to the order polygon grid data, the service object polygon grid data, and the demand relationship polygon grid data; constructing a training sample based on the order layer, the service object layer, and the supply and demand layer.

3. The method according to claim 1, wherein, the generating order polygon grid data according to the travel start position and travel end position in the travel order sample data includes: for each travel order sample data, obtaining the travel start position and travel end position therefrom; determining the polygon grid where the travel start position is located and setting a first weight for the polygon grid, and determining the polygon grid where the travel end position is located and setting a second weight for the polygon grid, wherein the first weight is less than the second weight; generating order polygon grid data according to the determined polygon grids and the weights corresponding to each polygon grid.

4. The method according to claim 1, wherein, the generating service object polygon grid data according to the service probability of the service start position and service end position in the travel service object sample data includes: obtaining the service probability of the travel service object from the service start position to the service end position according to the historical travel service data of the travel service object corresponding to the travel service object sample data; generating service object polygon grid data according to the service probability.

5. The method according to claim 1, wherein, the generating demand relationship polygon grid data according to the travel demand data and service demand data in the travel and service demand relationship sample data includes: For each polygon grid within a preset area, obtain sample data on the relationship between travel and service demand corresponding to the polygon grid; Based on the sample data on the relationship between travel and service demand corresponding to each polygon grid, predict the number of travel service objects in the polygon grid; Generate demand relationship polygon grid data for the polygon grid based on the difference between the number of travel service objects and the number of travel orders in the polygon grid obtained from the sample data on the relationship between travel and service demand.

6. The method according to claim 5, wherein, Before obtaining the sample data on the relationship between travel and service demand corresponding to each polygon grid within the preset area, the method further includes: obtaining multiple sets of sample data sets on the relationship between travel and service demand corresponding to the preset area according to a preset plurality of time periods; Obtaining the sample data on the relationship between travel and service demand corresponding to each polygon grid within the preset area includes: for each set of sample data sets on the relationship between travel and service demand, performing, according to the time period, an operation of obtaining the sample data on the relationship between travel and service demand corresponding to each polygon grid within the preset area.

7. The method according to claim 6, wherein, Constructing a corresponding supply-demand layer based on the demand relationship polygon grid data includes: Taking each set of sample data sets on the relationship between travel and service demand as a unit, respectively constructing a corresponding plurality of supply-demand sub-layers for the demand relationship polygon grid data generated from the multiple sets of sample data sets on the relationship between travel and service demand.

8. The method according to claim 1, wherein, Before training the machine learning model, the method further includes: collecting historical travel order data and analyzing the historical travel order data to obtain model supervision data corresponding to the training samples; Training the machine learning model includes: training the machine learning model according to the training samples and the model supervision data.

9. A travel scheduling method, including: Receiving a travel order and obtaining travel data in the travel order, where the travel data includes a travel start position and a travel end position; Based on the travel data, determining data of multiple candidate travel service objects providing travel services for the travel order, and travel and service demand relationship data of the travel area corresponding to the travel order, where the data of the multiple candidate travel service objects and the service demand relationship data are both real-time data or dynamically updated data; Based on the travel data, the data of the multiple candidate travel service objects, and the travel and service demand relationship data, obtaining matching information between the travel order and the multiple candidate travel service objects through a model trained by the method according to any one of claims 1-8, and selecting a target travel service object providing travel services for the travel order from the multiple candidate travel service objects according to the matching information.

10. The method according to claim 9, wherein, Selecting, from the multiple candidate travel service objects, a target travel service object that provides travel services for the travel order according to the travel data, the data of the multiple candidate travel service objects, and the travel and service demand relationship data, includes: Using a polygon area grid to generate corresponding order polygon grid data, service object polygon grid data, and demand relationship polygon grid data for the travel data, the data of the multiple candidate travel service objects, and the travel and service demand relationship data respectively; Selecting, from the multiple candidate travel service objects, a target travel service object that provides travel services for the travel order according to the order polygon grid data, the service object polygon grid data, and the demand relationship polygon grid data.

11. The method according to claim 10, wherein, selecting, from the multiple candidate travel service objects, a target travel service object that provides travel services for the travel order according to the order polygon grid data, the service object polygon grid data, and the demand relationship polygon grid data, includes: Constructing corresponding order layers, service object layers, and supply and demand layers according to the order polygon grid data, the service object polygon grid data, and the demand relationship polygon grid data respectively; Obtaining matching information between the travel order and the multiple candidate travel service objects through a machine learning model that has completed training according to the order layer, the service object layer, and the supply and demand layer, and selecting a target travel service object that provides travel services from the multiple candidate travel service objects according to the matching information.

12. A travel scheduling method, including: Receiving travel information input by a travel object and generating a travel order according to the travel information, where the travel order includes a travel start position and a travel end position; Sending the travel order to a travel platform so that the travel platform determines data of multiple candidate travel service objects that provide travel services according to the travel order, and travel and service demand relationship data of a travel area corresponding to the travel order; And obtaining matching information between the travel order and multiple candidate travel service objects through a model trained by the method according to any one of claims 1-8 according to the travel data in the travel order, the data of the multiple candidate travel service objects, and the travel and service demand relationship data, and selecting a target travel service object that provides travel services for the travel order from the multiple candidate travel service objects according to the matching information; Receiving information about the target travel service object fed back by the travel platform.

13. A travel scheduling method, including: Receiving a travel order sent by a travel platform, where the travel order includes a travel start position and a travel end position; Generating a response message according to an acceptance operation of the travel service object for the travel order, where the response message carries information about the travel service object; Feedback the response message to the travel platform, so that the travel platform can obtain the matching information between the travel order and multiple candidate travel service objects through the model trained by the method according to any one of claims 1-8, based on the travel order, the information of the travel service object, and the travel and service demand relationship data of the travel area corresponding to the travel order, and select a target travel service object that provides travel services for the travel order from multiple candidate travel service objects according to the matching information.

14. An electronic device, comprising: a processor, a memory, a communication interface and a communication bus, and the processor, the memory and the communication interface complete communication with each other through the communication bus; the memory is used to store at least one executable instruction, and the executable instruction causes the processor to perform the operations corresponding to the model training method according to any one of claims 1-8, or to perform the operations corresponding to the travel scheduling method according to any one of claims 9-13.

15. A computer storage medium, on which a computer program is stored, and when the program is executed by a processor, it implements the model training method according to any one of claims 1-8, or when the program is executed by a processor, it implements the travel scheduling method according to any one of claims 9-13.

Images