Marketing site selection method and device of vehicle, electronic equipment and storage medium

Abstract

The present application relates to the technical field of vehicles, in particular to a marketing site selection method and device for a vehicle, an electronic device and a storage medium, wherein the method comprises: acquiring sales data of multiple regions within a preset time; inputting the sales data into a pre-constructed marketing prediction model to obtain a probability of sales growth of each region, wherein the pre-constructed marketing prediction model is obtained by training a multi-input long short-term memory (LSTM) network model with target training samples; and determining a marketing address of the vehicle according to the probability of sales growth of each region. Thus, the method of site selection in the related art needs to purchase third-party data, the prediction result is inaccurate, and the cost is high, and the problem of low vehicle sales is solved.

Description

Technical Field

[0001] This application relates to the field of vehicle technology, specifically to a marketing site selection method, apparatus, electronic device, and storage medium for vehicles. Background Technology

[0002] As the automotive market enters a mature phase, especially in the new energy vehicle sector, leveraging existing market data for in-depth sales analysis becomes crucial. A key element in sales growth is marketing location selection. Strategic location planning allows limited resources to be used in key areas, enabling targeted marketing efforts and ultimately boosting sales.

[0003] In related technologies, a method for site selection of automotive service chain enterprises based on the LightGBM (LightGradient Boosting Machine) model is proposed (application number: CN202110015599.6). H3 encoding is used to expand the sample of each store. For each expanded sample, multi-level and multi-granular feature construction is carried out with regional grid, sample store and city as granularity. The corresponding numerical feature data are collected according to the constructed features. The revenue generated by each expanded sample in a recent period and its corresponding feature data constitute a dataset to train the lightGBM model to obtain the store site selection revenue prediction model. The revenue prediction is performed by inputting the feature data corresponding to the store candidate point into the store site selection revenue prediction model.

[0004] However, this method has high data acquisition costs and insignificant site selection effects, which urgently need to be addressed.

[0005] In related technologies, a machine learning-based intelligent location fusion method is proposed (application number: CN202011244753.9). This method involves data cleaning and integration; data analysis and processing based on feature engineering; data segmentation and training to obtain results; and training the results obtained in step three based on the LR model to predict the final result.

[0006] However, this method has not been applied to marketing site selection, and the predicted site selection results are not significant, which urgently needs to be addressed. Summary of the Invention

[0007] One objective of this invention is to provide a marketing site selection method, device, electronic device, and storage medium for vehicles, solving the problems of the need to purchase third-party data, inaccurate prediction results, and high costs in related technologies. Another objective is to provide a marketing site selection device for vehicles. A third objective is to provide an electronic device. A fourth objective is to provide a computer-readable storage medium.

[0008] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0009] A marketing location selection method for vehicles includes the following steps: acquiring sales data from multiple regions within a preset time period; inputting the sales data into a pre-built marketing prediction model to obtain the probability of sales growth in each region, wherein the pre-built marketing prediction model is obtained by training a multi-input long short-term memory (LSTM) network model using target training samples; and determining the marketing location of the vehicle based on the probability of sales growth in each region.

[0010] Based on the above technical means, the problems of site selection methods in related technologies, such as the need to purchase third-party data, inaccurate prediction results, and high costs, are solved, thereby increasing vehicle sales.

[0011] Furthermore, before inputting the sales data into the pre-built marketing prediction model, the method further includes: acquiring vehicle network data of multiple target vehicles, wherein the vehicle network data includes vehicle identification number (VIN) data, vehicle speed data, rain sensor data, and / or Global Positioning System (GPS) data; determining the vehicle's location for each target vehicle based on the GPS data, wherein the vehicle's location includes home address, work address, and / or shopping address; determining the monthly statistics on the number of times vehicles of the same model appear at home address, work address, and shopping address, and the number of newly added vehicles, based on the vehicle's location and VIN data; determining the number of rainy days in a monthly period based on the rain sensor data; and using the monthly statistics on the number of times vehicles of the same model appear at home address, work address, and shopping address, the number of newly added vehicles, and the number of rainy days in a monthly period as the target training samples to train the multi-input LSTM model, thereby obtaining the initial marketing prediction model.

[0012] Based on the aforementioned technical means, a multi-input LSTM model is constructed to predict the location, and vehicle network data and user behavior data are used as references to rank the probabilities of multiple locations, providing more specific options.

[0013] Furthermore, after obtaining the initial marketing prediction model, the method further includes: determining whether the initial marketing prediction model meets the preset training requirements; if the initial marketing prediction model meets the preset training requirements, then the initial marketing prediction model is used as the pre-built marketing prediction model; otherwise, the initial marketing prediction model continues to be trained until the initial marketing prediction model meets the preset training requirements.

[0014] Based on the above technical means, the prediction accuracy of marketing forecasting models can be improved.

[0015] Furthermore, after acquiring the vehicle network data of the multiple vehicles, the method further includes: performing an average value operation, a target symbol removal operation, and / or a space removal operation on the vehicle network data based on a preset cleaning strategy.

[0016] Based on the aforementioned technical means and a pre-defined cleaning strategy, the prediction accuracy of the marketing forecasting model can be improved.

[0017] Furthermore, determining the vehicle's stopping location for each target vehicle based on the GPS data, wherein the vehicle's stopping location includes a home address, a work address, and / or a shopping address, includes: drawing multiple circles with different radii centered on the GPS location of each target vehicle's parking point, obtaining the number of GPS points within each circle; if the number of GPS points within a consecutive preset number of circles is the same, then grouping all GPS points within the preset number of circles into multiple GPS groups; sorting the number of each GPS group, marking the GPS group with the maximum number of stopping points satisfying a first preset condition as the work address, marking the GPS group with the maximum number of stopping points satisfying a second preset condition as the home address, and marking the GPS group with a stopping time greater than a preset duration satisfying a third preset condition and whose GPS points are not in the work address or the home address as the shopping address.

[0018] Based on the aforementioned technical means, user behavior data is used as a reference to build a marketing forecasting model, thereby predicting sales in each region. This approach is cost-effective and improves the accuracy of marketing forecasting models.

[0019] A vehicle marketing location selection device includes: a first acquisition module for acquiring sales data from multiple regions within a preset time period; an input module for inputting the sales data into a pre-built marketing prediction model to obtain the probability of sales growth in each region, wherein the pre-built marketing prediction model is obtained by training a multi-input long short-term memory network (LSTM) model using target training samples; and a second acquisition module for determining the vehicle's marketing address based on the probability of sales growth in each region.

[0020] Furthermore, before inputting the sales data into the pre-built marketing prediction model, the input module is also used to: acquire vehicle network data of multiple target vehicles, wherein the vehicle network data includes vehicle identification number (VIN) data, vehicle speed data, rain sensor data, and / or Global Positioning System (GPS) data; determine the vehicle's location for each target vehicle based on the GPS data, wherein the vehicle's location includes home address, work address, and / or shopping address; determine the monthly statistics on the number of times vehicles of the same type appear at home address, work address, and shopping address, and the number of newly added vehicles, based on the vehicle's location and VIN data; determine the number of rainy days in a monthly period based on the rain sensor data; and use the monthly statistics on the number of times vehicles of the same type appear at home address, work address, and shopping address, the number of newly added vehicles, and the number of rainy days in a monthly period as the target training samples to train the multi-input LSTM model to obtain the initial marketing prediction model.

[0021] Furthermore, after obtaining the initial marketing prediction model, the input module is also used to: determine whether the initial marketing prediction model meets the preset training requirements; if the initial marketing prediction model meets the preset training requirements, then the initial marketing prediction model is used as the pre-built marketing prediction model; otherwise, the initial marketing prediction model continues to be trained until the initial marketing prediction model meets the preset training requirements.

[0022] Furthermore, after acquiring the vehicle network data of the multiple vehicles, the input module is also used to: perform an average value operation, a target symbol removal operation, and / or a space removal operation on the vehicle network data based on a preset cleaning strategy.

[0023] Furthermore, the step of determining the vehicle's stopping location for each target vehicle based on the GPS data, wherein the vehicle's stopping location includes a home address, a work address, and / or a shopping address, is further configured to: draw multiple circles with different radii centered on the GPS location of each target vehicle's parking point, obtaining the number of GPS points within each circle; if the number of GPS points within a consecutive preset number of circles is the same, then group all GPS points within the preset number of circles into multiple GPS groups; sort the number of each GPS group, mark the GPS group with the maximum number of stopping points satisfying the first preset condition as the work address, mark the GPS group with the maximum number of stopping points satisfying the second preset condition as the home address, and mark the GPS group with a stopping time greater than a preset duration satisfying the third preset condition and whose GPS points are not in the work address or the home address as the shopping address.

[0024] A vehicle includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, the processor executing the program to implement the vehicle marketing site selection method as described in the above embodiments.

[0025] A computer-readable storage medium having a computer program stored thereon, the program being executed by a processor to implement the vehicle marketing site selection method as described in the above embodiments.

[0026] The beneficial effects of this invention are: it solves the problems of the site selection method in the related technology requiring the purchase of third-party data, inaccurate prediction results, and high costs, thereby increasing vehicle sales. Attached Figure Description

[0027] Figure 1 This is a flowchart of a vehicle marketing site selection method according to the present invention;

[0028] Figure 2 This is a schematic diagram illustrating the format of the new sales input model for each region in this invention;

[0029] Figure 3 This is a schematic diagram of the multi-input LSTM model of the present invention;

[0030] Figure 4 This is a block diagram of the vehicle marketing site selection device of the present invention;

[0031] Figure 5 This is a schematic diagram of the electronic device of the present invention.

[0032] Among them, 10-vehicle marketing location selection device; 100-first acquisition module; 200-input module; 300-second acquisition module; 501-memory; 502-processor; 503-communication interface. Detailed Implementation

[0033] The embodiments of the present invention will be described below with reference to the accompanying drawings and preferred embodiments. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be understood that the preferred embodiments are only for illustrating the present invention and not for limiting the scope of protection of the present invention.

[0034] It should be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of the present invention. Therefore, the drawings only show the components related to the present invention and are not drawn according to the actual number, shape and size of the components in the actual implementation. In the actual implementation, the form, quantity and proportion of each component can be arbitrarily changed, and the layout of the components may also be more complex.

[0035] This embodiment proposes a marketing site selection method for vehicles, such as... Figure 1 As shown, the marketing site selection method for this vehicle includes the following steps:

[0036] In step S101, sales data for multiple regions within a preset time period are obtained.

[0037] The preset time period can be a threshold set by the user, a threshold obtained through a limited number of experiments, or a threshold obtained through a limited number of computer simulations. Preferably, the preset time period is set to the past 6 months. Sales data refers to the number of newly added vehicle identification numbers (VINs) in the region.

[0038] Specifically, in this embodiment, Spring is used to read data from the database on the number of newly added vehicle identification numbers (VINs) in various regions over the past six months.

[0039] In step S102, sales data is input into a pre-built marketing prediction model to obtain the probability of sales growth in each region. The pre-built marketing prediction model is obtained by training a multi-input long short-term memory network (LSTM) model using target training samples.

[0040] In this embodiment, a pre-built marketing prediction model is obtained by training a multi-input long short-term memory network (LSTM) model. The input value of the pre-built marketing prediction model is sales data, and the output value is the probability of sales growth in each region.

[0041] Furthermore, in some embodiments, before inputting sales data into a pre-built marketing prediction model, the method further includes: acquiring vehicle network data of multiple target vehicles, wherein the vehicle network data includes vehicle identification number (VIN) data, vehicle speed data, rain sensor data, and / or Global Positioning System (GPS) data; determining the vehicle's location for each target vehicle based on the GPS data, wherein the vehicle's location includes home address, work address, and / or shopping address; determining the monthly statistics on the number of times vehicles of the same model appear at home address, work address, and shopping address, and the number of newly added vehicles, based on the vehicle's location and VIN data of each target vehicle; determining the number of rainy days in a monthly period based on the rain sensor data; and using the monthly statistics on the number of times vehicles of the same model appear at home address, work address, and shopping address, the number of newly added vehicles, and the number of rainy days in a monthly period as target training samples to train a multi-input LSTM model, thereby obtaining an initial marketing prediction model.

[0042] Specifically, the vehicle sends the target vehicle's operating data to the TSP (Telematics Service Provider) platform. After receiving the vehicle operating data, the TSP platform stores the data in Hadoop. In this embodiment, Spark reads the vehicle network data stored in HDFS and uses the dataframe.select function of the Spark framework to extract the vehicle identification number (VIN) data, vehicle speed data, rain sensor data, and GPS data from the vehicle network data.

[0043] Furthermore, in some embodiments, the vehicle's stopping location is determined based on GPS data. This stopping location includes a home address, a work address, and / or a shopping address. This involves: drawing multiple circles with different radii centered on the GPS location of each target vehicle's parking spot; obtaining the number of GPS points within each circle; if the number of GPS points within a consecutive preset number of circles is the same, then all GPS points within the preset number of circles are grouped together to obtain multiple GPS groups; sorting the number of GPS groups; marking the GPS group with the maximum number of stopping points satisfying a first preset condition as the work address; marking the GPS group with the maximum number of stopping points satisfying a second preset condition as the home address; and marking the GPS group with a stopping time greater than a preset duration that satisfies a third preset condition and whose points are not in the work address or home address as the shopping address.

[0044] The first preset condition is to meet the maximum number of stops during the daytime on weekdays; the second preset condition is to meet the maximum number of stops at midnight; and the third preset condition is to meet the stay time at the stops during the daytime on holidays (including Saturdays and Sundays) that is greater than the preset duration and that the GPS group's points are not at the work address or home address.

[0045] The preset duration can be a threshold set by the user, a threshold obtained through a limited number of experiments, or a threshold obtained through a limited number of computer simulations. Preferably, the preset duration is set to 45 minutes.

[0046] Specifically, this embodiment determines the vehicle's location based on GPS data, including home address, work address, shopping address, etc. It analyzes data from the past year, distinguishing between holidays, midnight, 11 AM, and 2 PM. This embodiment uses each target vehicle's parking location as the center of a circle with a radius of N kilometers. It calculates the number of parking GPS points contained within each circle of radius N. The radius N is then continuously increased by 0.5 kilometers each time, and the count of GPS points within each circle is repeated. If the number of parking GPS points is the same in three consecutive circles (i.e., a preset number of circles), all GPS points within the preset number of circles are considered a group. This process is repeated, deduplicating the obtained GPS groups, and calculating the center of each GPS group.

[0047] The longitude of the center of the circle is (maximum longitude + minimum longitude) / 2; the latitude of the center of the circle is (maximum latitude + minimum latitude) / 2; and the radius is the maximum value of the circle to the GPS points within the group.

[0048] Furthermore, the number of points in each GPS group within the time period is sorted. The GPS group with the largest number of stops during the daytime on weekdays is marked as the work address, and the GPS group with the largest number of stops at midnight is marked as the home address. On holidays (including Saturdays and Sundays), the GPS group with a stop time of more than 45 minutes during the daytime and whose GPS group points are not in the work address or home address is marked as the shopping address. There can be multiple shopping addresses, and the GPS points of the shopping address must appear more than three times per quarter. The number of stops at the work address and home address must be more than 100 times per year. The effect analysis of other locations not covered by the above rules is not considered in this case. At the same time, this embodiment counts the number of rainy days in the monthly time period of the rain sensor of the target vehicle.

[0049] Furthermore, this embodiment statistically analyzes the monthly occurrences of vehicles of the same model at home addresses, work addresses, and shopping addresses, as well as the number of newly added vehicles. Addresses with monthly sales exceeding 3 units and a sales increase of more than 30% are marked as positive, while other addresses are marked as negative.

[0050] In this embodiment, sales data from the past year are collected to construct a relationship table of new vehicle arrivals in each region over time T, as shown in Table 1:

[0051] Table 1

[0052] Time 1 0 1 3 2 1 Time 2 0 3 4 5 2 Time 3 1 1 2 3 4 Time N Predict whether it will grow Predict whether it will grow Predict whether it will grow Predict whether it will grow Predict whether it will grow

[0053] Furthermore, this embodiment utilizes Python's database connection methods to write INSERT statements and insert sales data into the database, while simultaneously constructing a time series model to predict whether there will be new sales (positive or negative) in the region next month. Next, a calculation window is created every six months, with the seventh month serving as the target for predicting whether there will be new sales in the region. Figure 2 As shown.

[0054] Furthermore, in this embodiment, the number of times vehicles of the same type and model appear at monthly statistical home addresses, work addresses, and shopping addresses, as well as the number of newly added vehicles, and the number of rainy days in a monthly period are used as target training samples to train the multi-input LSTM model, thereby obtaining the initial marketing prediction model.

[0055] Among them, the multi-input long short-term memory network (LSTM) model is as follows: Figure 3 As shown, the first layer is a multi-input layer with two inputs: one is the sliding window data, and the other is the address region number corresponding to the sliding window and the corresponding number of rainy days. The second layer is a concat layer, merging the two inputs from the first layer. The third layer is an LSTM (Long Short-Term Memory) network structure, the fourth layer is a Linear fully connected layer structure, and the fifth layer is the output layer. The loss function is the BCELoss cross-entropy loss function.

[0056] Furthermore, in some embodiments, after obtaining the initial marketing prediction model, the method further includes: determining whether the initial marketing prediction model meets the preset training requirements; if the initial marketing prediction model meets the preset training requirements, then the initial marketing prediction model is used as a pre-built marketing prediction model; otherwise, the initial marketing prediction model continues to be trained until the initial marketing prediction model meets the preset training requirements.

[0057] Among them, the preset training requirements are met, namely, the initial marketing prediction model converges and the auc (Area Under Curve) result is greater than 0.75.

[0058] It should be understood that after obtaining the initial marketing prediction model, the process involves determining whether the initial marketing prediction model has converged and whether the AUC result is greater than 0.75. If the initial marketing prediction model converges and the AUC result is greater than 0.75, then the initial marketing prediction model is used as the pre-built marketing prediction model, and a Python version of the web service is built using the Flask framework. If the initial marketing prediction model does not meet the convergence requirement and the AUC result is greater than 0.75, then the initial marketing prediction model is further trained using monthly statistics on the frequency of vehicles of the same type and type appearing at home addresses, work addresses, and shopping addresses, as well as the number of new vehicles appearing, and the number of rainy days within the monthly time period as target training samples, until the initial marketing prediction model meets the preset training requirements.

[0059] Finally, this embodiment uses the Spring and Vue frameworks to construct a method for displaying the marketing forecasting model. Spring is responsible for reading sales data for each address in the database for the past six months and using Java's WebClient to make calls to web services, thereby obtaining the probability of sales growth in each region predicted by the marketing forecasting model.

[0060] In some embodiments, after acquiring vehicle network data from multiple vehicles, the method further includes: performing an average value operation, a target symbol removal operation, and / or a space removal operation on the vehicle network data based on a preset cleaning strategy.

[0061] It should be understood that, due to the chaotic format of the vehicle network data from multiple target vehicles, this embodiment is based on a preset cleaning strategy, that is, to perform format verification on each line of the acquired vehicle network data through the dataframe.map operation. For vehicle network data that does not conform to the format, such as abnormal data or missing data, an average operation is performed, or a target symbol removal operation is performed, or a space removal operation is performed, or an average operation, a target symbol removal operation, and a space removal operation are performed on vehicle network data that does not conform to the format.

[0062] In step S103, the marketing address of the vehicle is determined based on the probability of sales growth in each region.

[0063] Understandably, the probability of sales growth in each region is predicted based on the obtained marketing prediction model, and then sorted by probability and displayed on a map. The predicted regions are the possible growth areas in automobile marketing. In this embodiment, the possible growth areas are used to determine the marketing addresses of vehicles.

[0064] This invention provides a vehicle marketing location selection method. It acquires sales data from multiple regions within a preset time period, inputs this data into a pre-built marketing prediction model to obtain the probability of sales growth in each region, and determines the vehicle's marketing location based on this probability. This solves the problems of related technologies where location selection methods require purchasing third-party data, resulting in inaccurate predictions and high costs, thereby increasing vehicle sales.

[0065] This embodiment also proposes a marketing site selection device for vehicles.

[0066] Figure 4 This is a block diagram of the vehicle marketing site selection device in this embodiment.

[0067] like Figure 4 As shown, the marketing location selection device 10 for the vehicle includes: a first acquisition module 100, an input module 200, and a second acquisition module 300.

[0068] The first acquisition module 100 is used to acquire sales data from multiple regions within a preset time period; the input module 200 is used to input the sales data into a pre-built marketing prediction model to obtain the probability of sales growth in each region, wherein the pre-built marketing prediction model is obtained by training a multi-input long short-term memory network (LSTM) model using target training samples; the second acquisition module 300 is used to determine the marketing address of the vehicle based on the probability of sales growth in each region.

[0069] Furthermore, in some embodiments, before inputting sales data into the pre-built marketing prediction model, the input module 200 is also used to: acquire vehicle network data of multiple target vehicles, wherein the vehicle network data includes vehicle identification number (VIN) data, vehicle speed data, rain sensor data, and / or global positioning system (GPS) data; determine the vehicle's location for each target vehicle based on the GPS data, wherein the vehicle's location includes home address, work address, and / or shopping address; determine the monthly statistics on the number of times vehicles of the same type appear at home address, work address, and shopping address, and the number of newly added vehicles, based on the vehicle's location and VIN data of each target vehicle; determine the number of rainy days in a monthly period based on the rain sensor data; and use the monthly statistics on the number of times vehicles of the same type appear at home address, work address, and shopping address, the number of newly added vehicles, and the number of rainy days in a monthly period as target training samples to train a multi-input LSTM model to obtain an initial marketing prediction model.

[0070] Furthermore, in some embodiments, after obtaining the initial marketing prediction model, the input module 200 is also used to: determine whether the initial marketing prediction model meets the preset training requirements; if the initial marketing prediction model meets the preset training requirements, then the initial marketing prediction model is used as a pre-built marketing prediction model; otherwise, the initial marketing prediction model continues to be trained until the initial marketing prediction model meets the preset training requirements.

[0071] Furthermore, in some embodiments, after acquiring vehicle network data from multiple vehicles, the input module 200 is also used to: perform averaging operations, target symbol removal operations, and / or space removal operations on the vehicle network data based on a preset cleaning strategy.

[0072] Furthermore, in some embodiments, the vehicle's stopping location is determined based on GPS data, wherein the stopping location includes home address, work address, and / or shopping address. The input module 200 is further configured to: draw multiple circles with different radii centered on the GPS of each target vehicle's parking point, and obtain the number of GPS points within each circle; if the number of GPS points within a preset number of consecutive circles is the same, then all GPS points within the preset number of circles are grouped together to obtain multiple GPS groups; sort the number of each GPS group, mark the GPS group with the maximum number of stopping points that meets the first preset condition as the work address, mark the GPS group with the maximum number of stopping points that meets the second preset condition as the home address, and mark the GPS group with the stopping time of the stopping point that meets the third preset condition, which is greater than a preset duration, and whose GPS points are not in the work address or home address, as the shopping address.

[0073] It should be noted that the foregoing explanation of the vehicle marketing location selection method embodiment also applies to the vehicle marketing location selection device of this embodiment, and will not be repeated here.

[0074] This invention provides a vehicle marketing location selection device that acquires sales data from multiple regions within a preset time period, inputs the sales data into a pre-built marketing prediction model, obtains the probability of sales growth in each region, and determines the vehicle's marketing location based on the probability of sales growth in each region. This solves the problems of related technologies where location selection methods require the purchase of third-party data, resulting in inaccurate predictions and high costs, thereby increasing vehicle sales.

[0075] This embodiment also proposes an electronic device.

[0076] Figure 5 This is a schematic diagram of the structure of the electronic device provided in this embodiment. The electronic device may include:

[0077] The memory 501, the processor 502, and the computer program stored on the memory 501 and capable of running on the processor 502.

[0078] When processor 502 executes the program, it implements the vehicle marketing site selection method provided in the above embodiments.

[0079] Furthermore, electronic devices also include:

[0080] Communication interface 503 is used for communication between memory 501 and processor 502.

[0081] The memory 501 is used to store computer programs that can run on the processor 502.

[0082] The memory 501 may include high-speed RAM (Random Access Memory) memory, and may also include non-volatile memory, such as at least one disk storage.

[0083] If the memory 501, processor 502, and communication interface 503 are implemented independently, then the communication interface 503, memory 501, and processor 502 can be interconnected via a bus to complete communication between them. The bus can be an ISA (Industry Standard Architecture) bus, a PCI (Peripheral Component Interconnect) bus, or an EISA (Extended Industry Standard Architecture) bus, etc. The bus can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 5 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0084] Optionally, in a specific implementation, if the memory 501, processor 502, and communication interface 503 are integrated on a single chip, then the memory 501, processor 502, and communication interface 503 can communicate with each other through an internal interface.

[0085] Processor 502 may be a CPU (Central Processing Unit), an ASIC (Application Specific Integrated Circuit), or one or more integrated circuits configured to implement this embodiment.

[0086] This embodiment also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-described vehicle marketing site selection method.

[0087] The above embodiments are merely preferred embodiments provided to fully illustrate the present invention, and the scope of protection of the present invention is not limited thereto. Equivalent substitutions or modifications made by those skilled in the art based on the present invention are all within the scope of protection of the present invention.

Claims

1. A marketing site selection method for vehicles, characterized in that, Includes the following steps: Obtain sales data from multiple regions within a preset time period; The sales data is input into a pre-built marketing prediction model to obtain the probability of sales growth in each region. The pre-built marketing prediction model is obtained by training a multi-input long short-term memory (LSTM) network model using target training samples. The marketing locations for vehicles are determined based on the probability of sales growth in each region. Before inputting the sales data into the pre-built marketing forecasting model, the method further includes: Acquire vehicle network data of multiple target vehicles, wherein the vehicle network data includes vehicle identification number (VIN) data, vehicle speed data, rain sensor data and / or global positioning system (GPS) data; The location of each target vehicle is determined based on the GPS data, wherein the location of the vehicle includes the home address, work address and / or shopping address; Based on the vehicle's parking location and VIN data for each target vehicle, determine the monthly statistics on the number of times vehicles of the same model appear at home address, work address, and shopping address, as well as the number of times newly added vehicles appear. The number of rainy days in a monthly period is determined based on the rain sensor data. The multi-input LSTM model is trained using the monthly statistics of the number of times vehicles of the same type appear at home addresses, work addresses, and shopping addresses, the number of times new vehicles appear, and the number of rainy days in the monthly period as the target training samples to obtain the initial marketing prediction model. The step of determining the vehicle's stopping location for each target vehicle based on the GPS data includes the home address, work address, and / or shopping address, including: Using the GPS of each target vehicle parking point as the center, draw multiple circles with different radii to obtain the number of GPS points in each circle. If the number of GPS points in a consecutive preset number of circles is the same, then group all GPS points in the preset number of circles together to obtain multiple GPS groups. The GPS groups are sorted by the number of each group. The GPS group with the maximum number of stops that meet the first preset condition is marked as the work address. The GPS group with the maximum number of stops that meet the second preset condition is marked as the home address. The GPS group with the stop time greater than a preset duration that meets the third preset condition and whose GPS group points are not in the work address and the home address is marked as the shopping address.

2. The method according to claim 1, characterized in that, After obtaining the initial marketing forecasting model, the following is also included: Determine whether the initial marketing prediction model meets the preset training requirements; If the initial marketing prediction model meets the preset training requirements, then the initial marketing prediction model is used as the pre-built marketing prediction model; otherwise, the initial marketing prediction model continues to be trained until it meets the preset training requirements.

3. The method according to claim 1, characterized in that, After acquiring the vehicle-to-everything (V2X) data from the multiple vehicles, the process also includes: Based on a preset cleaning strategy, the vehicle network data is subjected to averaging, removal of target symbols, and / or removal of spaces.

4. A marketing site selection device for vehicles, characterized in that, The marketing location selection device is used to implement the marketing location selection method for vehicles as described in any one of claims 1-3, and the marketing location selection device includes: The first acquisition module is used to acquire sales data from multiple regions within a preset time period; An input module is used to input the sales data into a pre-built marketing prediction model to obtain the probability of sales growth in each region, wherein the pre-built marketing prediction model is obtained by training a multi-input long short-term memory (LSTM) network model using target training samples; and The second acquisition module is used to determine the marketing address of the vehicle based on the probability of sales growth in each region.

5. The apparatus according to claim 4, characterized in that, Before inputting the sales data into the pre-built marketing forecasting model, the input module is further configured to: Acquire vehicle network data of multiple target vehicles, wherein the vehicle network data includes vehicle identification number (VIN) data, vehicle speed data, rain sensor data and / or global positioning system (GPS) data; The location of each target vehicle is determined based on the GPS data, wherein the location of the vehicle includes the home address, work address and / or shopping address; Based on the vehicle's parking location and VIN data for each target vehicle, determine the monthly statistics on the number of times vehicles of the same model appear at home address, work address, and shopping address, as well as the number of times newly added vehicles appear. The number of rainy days in a monthly period is determined based on the rain sensor data. The multi-input LSTM model is trained using the monthly statistics of the number of times vehicles of the same type appear at home addresses, work addresses, and shopping addresses, the number of newly added vehicles, and the number of rainy days in the monthly period as the target training samples to obtain the initial marketing prediction model.

6. The apparatus according to claim 5, characterized in that, After obtaining the initial marketing prediction model, the input module is further configured to: Determine whether the initial marketing prediction model meets the preset training requirements; If the initial marketing prediction model meets the preset training requirements, then the initial marketing prediction model is used as the pre-built marketing prediction model; otherwise, the initial marketing prediction model continues to be trained until it meets the preset training requirements.

7. A vehicle, characterized in that, include: A memory, a processor, and a computer program stored in the memory and executable on the processor, the processor executing the program to implement the marketing site selection method for a vehicle as described in any one of claims 1-3.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, The program is executed by the processor to implement the marketing site selection method for vehicles as described in any one of claims 1-3.

Images