Advertisement data processing method and system

Abstract

The invention relates to the technical field of vehicle-mounted advertisement putting, in particular to an advertisement data processing method and system. The method comprises the following steps that user position data and vehicle-mounted position data are collected, and user getting-on and getting-off position extraction is carried out to obtain user getting-on and getting-off position data; obtaining user retention data according to the user getting-on and getting-off position data; performing retention area heat map calculation according to the user retention data to obtain retention area heat map data; carrying out retention area attribute extraction according to the retention area heat map data to obtain retention area attribute data; performing four-dimensional feature space construction according to the retention area attribute data to obtain retention four-dimensional space feature data; obtaining intelligent advertisement recommendation data according to the retention four-dimensional space feature data; and performing advertisement exposure optimization calculation according to the intelligent advertisement recommendation data and the user retention data to obtain vehicle-mounted advertisement putting data. According to the invention, intelligent and efficient vehicle-mounted advertisement putting is realized, and the commercial value is enhanced.

Description

Technical Field

[0001] The present invention relates to the technical field of vehicle-mounted advertisement placement, and in particular to an advertisement data processing method and system. Background Art

[0002] The Internet of Vehicles refers to a network system that interconnects vehicles, road infrastructure, pedestrians and cloud platforms through wireless communication technology, sensors, artificial intelligence and cloud computing to achieve vehicle information sharing, intelligent driving, remote control and intelligent services. With the rapid development of intelligent transportation and the Internet of Vehicles (IoV), in-vehicle advertising delivery systems have gradually attracted market attention as an emerging advertising communication method. In-vehicle advertising can combine mobile user data, environmental information and intelligent recommendation technology to dynamically adjust advertising content during driving, improve advertising exposure and user matching, and thus achieve precision marketing. However, traditional in-vehicle advertising is based on preset locations or fixed playback logic, resulting in the problem of invalid exposure. How to accurately deliver in-vehicle advertising has become a problem. Summary of the invention

[0003] In order to solve the above technical problems, the present invention proposes an advertisement data processing method and system to solve at least one of the above technical problems.

[0004] The present application provides an advertisement data processing method for a vehicle-mounted advertisement delivery device, the method comprising: S1. Collecting user location data and vehicle location data through a vehicle camera or a mobile device, and extracting the user's boarding and alighting location based on the user location data and the vehicle location data to obtain the user's boarding and alighting location data; performing retention calculation based on the user's boarding and alighting location data to obtain the user retention data; S2. Calculate the heat map of the retention area according to the user retention data to obtain the heat map data of the retention area; extract the attributes of the retention area according to the heat map data of the retention area to obtain the attribute data of the retention area; construct a four-dimensional feature space according to the attribute data of the retention area to obtain the feature data of the retention four-dimensional space; S3, performing intelligent advertising recommendation based on the retained four-dimensional spatial feature data to obtain intelligent advertising recommendation data; S4. Perform advertising exposure optimization calculation based on the intelligent advertising recommendation data and user retention data to obtain the in-vehicle advertising delivery data.

[0005] The present invention adopts a trajectory clustering algorithm to analyze the user's vehicle route data, accurately divide the user's activity area, reduce the misjudgment caused by GPS errors, and improve the reliability of the identification of the retention area. By classifying the retention area, the user's retention area is divided into commercial areas, residential areas, transportation hubs, scenic spots, etc., which is convenient for targeted advertising, ensuring that the advertising content is highly matched with the user scene and improving the conversion rate. By calculating the user's stay time in different areas, short stays (such as gas stations, convenience stores) and long stays (such as shopping centers, office areas) are distinguished, thereby optimizing the advertising delivery strategy, such as placing flash ads in short-stay areas and placing deep brand ads in long-stay areas. Combined with the user's retention time, retention area classification and historical behavior, high-potential consumer groups can be accurately identified, invalid advertising can be reduced, advertising click-through rate and conversion rate can be improved, and advertising delivery ROI (return on investment) can be optimized.

[0006] Optionally, S1 includes: S11, collecting user location data and vehicle location data through a vehicle camera or a mobile device; S12, extracting the user's boarding and getting-in position data based on the user's position data and the vehicle's position data to obtain the user's boarding and getting-in position data; S13, calculating the user's vehicle route according to the user's boarding and alighting location data, and obtaining the user's vehicle route data; S14, performing regional retention calculation according to the user's vehicle route data to obtain user regional retention data; S15, calculating the retention behavior according to the user's vehicle route data and the user's regional retention data to obtain the user's retention behavior data; S16, calculating the likelihood of attention to the advertisement area according to the user's vehicle route data and the user's retention behavior data, and obtaining the likelihood of attention to the advertisement area; S17: Integrate the user area retention data, the user retention behavior data, and the advertisement area attention possibility data to obtain the user retention data.

[0007] In the present invention, the user's regional retention data is combined with the vehicle route data to divide the time characteristics, identify the user's driving mode in different time periods, such as morning and evening peaks, commuting time, temporary parking, etc., and provide high-precision data for the retention behavior analysis. The vehicle route time characteristic data and the retention area classification data are mapped to the electronic map, and the user's stay location is accurately matched to avoid the wrong classification caused by GPS error, improve the accuracy of the retention area positioning, and optimize the selection of advertising delivery areas. By analyzing the electronic map mapping data, the road type (main road, branch road, highway, commercial district) is associated to accurately judge whether the user's stay is due to traffic congestion, parking demand, or commercial environment attraction, and optimize the matching degree of advertising content, such as adding retail advertising in commercial areas and adding tourism and gas station advertising in high-speed service areas. Combined with the road type and the vehicle route time characteristic data, time attribution analysis is performed to determine whether the user is detained due to factors such as short-term stop at traffic lights, traffic congestion, and purposeful parking, and then adjust the advertising delivery strategy to ensure that the advertisement is exposed in the most appropriate time period and improve user acceptance. By integrating user retention behavior, road type, and time patterns, we can accurately identify high-value advertising points and avoid ineffective advertising during high-speed driving. Instead, we can place ads in parking areas and low-speed driving areas to improve ad visibility and conversion rates.

[0008] Optionally, the regional retention calculation is specifically: Perform regional clustering according to the user's vehicle route data to obtain regional clustering data; Classify the detention areas according to the regional clustering data to obtain the detention area classification data; The user retention time is calculated based on the retention area classification data to obtain the user area retention data.

[0009] The present invention extracts vehicle speed feature data to distinguish whether the user is driving at high speed, driving at low speed or parking, accurately judges whether the advertisement is in the best visible range, ensures that the advertisement placement matches the user's driving status, and reduces invalid exposure. The user's image data is obtained through the vehicle-mounted camera, and the eye tracking technology is combined to analyze whether the user is looking at the advertisement area, avoiding the problem that traditional advertisement exposure cannot evaluate the user's actual attention, and improving the scientific nature of advertising placement decisions. Facial expression analysis is combined with emotion accumulation extreme value calculation to evaluate the user's emotional response to advertisements, such as interest, boredom, indifference, etc., to build an attention scoring model to ensure that the advertisement push is in line with the user's current emotional state and improve the acceptance of advertisements. Through eye tracking data + geometric features of the advertisement area, key data such as the user's gaze duration and frequency on the advertisement area are extracted, the visual weight of the advertisement area is calculated, the advertisement playback content and duration are optimized, and the effective exposure rate of the advertisement is improved. Through user ad attention data + vehicle speed characteristics + retention behavior data, we can accurately identify the difference between users actively paying attention to ads (consciously watching) and passively paying attention to ads (staying for a long time but not paying strong attention), optimize the delivery strategy, ensure that ads are delivered in areas with high traffic but low attention, and use high-impact ads in areas with low traffic but high attention. Through the data on the likelihood of attention in the ad area, we can build an intelligent ad delivery optimization model, automatically adjust the ad delivery frequency, exposure time, and content matching, ensure the accuracy of ad delivery, increase user click-through rate and interaction rate, reduce ad waste, and improve the overall ROI (return on investment).

[0010] Optionally, the retention behavior calculation is specifically: According to the user's regional stay data, the user's vehicle route data is divided into vehicle route time characteristics to obtain the user's vehicle route time characteristic data; Mapping the preset electronic map data according to the user's vehicle route time characteristic data and the detention area classification data to obtain electronic map mapping data; Associating the road types according to the electronic map mapping data to obtain the associated data of the stranded roads; Time attribution processing is performed based on the stranded road association data and the user's vehicle route time feature data to obtain the user's stranded behavior data.

[0011] The present invention uses density estimation to calculate the density of user retention points, accurately identifies high-traffic areas, avoids the misjudgment caused by traditional single-point GPS trajectory analysis, and improves the accuracy of user retention area identification. Combined with the user retention point density data, a spatiotemporal dynamic heat map is constructed to distinguish the user retention characteristics of different time periods such as morning peak, lunch break, and evening peak, optimize the time strategy for advertising delivery, ensure that the advertisement is delivered during the time period with the highest user attention, and improve exposure efficiency. Through POI data fusion, the user's retention area type (such as shopping malls, office buildings, catering areas, transportation hubs, etc.) is automatically matched to ensure that the advertising content is highly matched with the user's scene. For example, retail advertisements are delivered in shopping malls, and corporate service advertisements are delivered in office areas to optimize advertising conversion rates. Through the user's residence time pattern at different locations, permanent locations, short-term locations, and temporary stays are distinguished, and the user's behavioral characteristics in different places are accurately analyzed to improve the user portrait accuracy of the intelligent advertising delivery system and ensure the timeliness and applicability of advertising push. A four-dimensional user behavior feature space is constructed by combining location (spatial dimension) + time (behavior sequence) + scene (POI type) + interactive behavior (user attention). The advertising recommendation strategy is optimized based on deep learning modeling to improve the accuracy of advertising recommendations.

[0012] Optionally, the advertisement area attention probability calculation is specifically as follows: Extract vehicle speed characteristics according to the user's vehicle route data to obtain vehicle speed characteristic data; Obtain user's vehicle image data; Perform eye tracking and facial expression recognition based on the user's in-vehicle image data to obtain eye tracking data and facial expression data; Calculate the extreme value of emotional accumulation according to the facial expression data to obtain the extreme value data of emotional accumulation; Extracting gaze features of the advertisement area according to the eye tracking data to obtain gaze feature data of the advertisement area; Calculate the advertisement attention based on the advertisement area gaze feature data and the emotion accumulation extreme value data to obtain the user advertisement attention data; Based on the user's advertising attention data, vehicle speed feature data and user retention behavior data, direct passive attention behavior association is performed to obtain the advertising area attention possibility data.

[0013] The present invention adopts the extraction of retention time and the calculation of visit frequency to identify the permanent places (such as home and workplace) where users visit frequently and stay for a long time, distinguish the non-resident places visited infrequently, optimize the user portrait, and ensure that the advertising delivery strategy is highly matched with the user's daily behavior. Combined with the calculation of geographical proximity, it distinguishes non-resident near places (such as supermarkets and gyms near home) and non-resident far places (such as business trip hotels and tourist attractions), and improves the spatial adaptability of advertising delivery, such as pushing daily consumption advertisements in near places, and pushing tourism and travel-related advertisements in far places. Through the analysis of the changes in the time attributes of different types of places, the main activity areas of users in specific time periods are identified, such as at home in the morning, at the company during the day, and at the mall at night, to improve the time accuracy of advertising delivery, and to deliver advertisements in the time window most likely to be accepted by users, so as to improve the click-through rate. Through the dynamic annotation of the location type classification data, the classification of users' stay patterns in different places is optimized, so that the advertising recommendation system can adjust the recommendation strategy based on the different activity scenes of users, avoid the advertising content from being out of touch with the actual needs of users, and improve the relevance of advertising and user experience. By building a permanent / non-resident near / far location classification system, we can automatically optimize the timing, content and frequency of advertising exposure, ensure that advertising is delivered in the right place, at the right time and to the right people, and improve the advertising exposure conversion rate and ROI (return on investment).

[0014] Optionally, S2 includes: S21, calculating the user retention point density according to the user retention data to obtain the user retention point density data; S22, performing multi-period dynamic heat calculation based on user retention point density data to obtain retention area heat map data; S23, performing location type classification matching according to the retention area heat map data to obtain location type classification data; S24, performing time attribute change processing according to the location type classification data to obtain retention area attribute data; S25. Construct a four-dimensional feature space according to the retention area attribute data to obtain retention four-dimensional space feature data.

[0015] The present invention adopts low-interaction modeling to analyze the stable behavior patterns of users in permanent locations such as home and office, identify long-term retention but low-interaction characteristics, provide optimized delivery strategies for long-term brand advertising, financial insurance, and subscription services, and improve advertising adaptability. Through high-frequency short-stay analysis, the high-frequency access behavior of users in short-stay locations such as supermarkets, coffee shops, and convenience stores is accurately identified, which is suitable for instant consumption advertising (such as takeout, fast-moving consumer goods, and limited-time discounts), improves the ability to accurately deliver advertisements, and avoids advertising waste. Combined with random stay modeling, the short-term or low-frequency stay patterns of users in non-resident remote locations such as tourist attractions, airports, and highway service areas are analyzed to optimize the advertising recommendation logic, such as pushing hotel, car rental, and scenic spot ticket advertisements to tourist users, and improve the situational adaptability and user conversion rate of advertisements. The four-dimensional feature modeling of location (spatial distribution) + time (stay duration) + scene (location type) + interaction (user behavior) is adopted to finely distinguish user behaviors in different location categories, optimize the content, timing, and frequency of advertising exposure, and improve user experience. Through intelligent classification of the first, second, and third retention four-dimensional spatial feature data, dynamic adaptive optimization of the advertising system is achieved, ensuring that advertisements accurately reach target users in different locations and scenarios, thereby increasing advertising click-through rate, interaction rate, and ROI (return on investment).

[0016] Optionally, the time attribute change processing is specifically as follows: According to the location type classification data, the length of stay is extracted and the frequency of stay is calculated to obtain the length of stay data and the frequency of stay data respectively; Identify the permanent location based on the residence duration data and the residence visit frequency data to obtain the permanent location data; Identify non-resident locations based on the resident location data and location type classification data to obtain non-resident location data; Calculate the location proximity based on the permanent location data and the non-resident location data to obtain the permanent location proximity data; When the resident location proximity data is less than or equal to the preset resident location proximity threshold data, the non-resident location data corresponding to the resident location proximity data is determined as the non-resident near location data; When the resident location proximity data is greater than the preset resident location proximity threshold data, the non-resident location data corresponding to the resident location proximity data is determined as non-resident remote location data; The location type classification data is labeled according to the resident location data, the non-resident near location data and the non-resident far location data to obtain the retention area attribute data; The retained four-dimensional space feature data includes the first retained four-dimensional space feature data, the second retained four-dimensional space feature data and the third retained four-dimensional space feature data. The four-dimensional feature space is constructed as follows: Low-interaction modeling is performed on the retention area attribute data corresponding to the number of permanent locations to obtain the first retention four-dimensional spatial feature data; Perform high-frequency short-dwelling analysis on the retention area attribute data corresponding to the non-resident perigee data to obtain the second retention four-dimensional space feature data; The random residence model is performed on the residence area attribute data corresponding to the non-resident remote location data to obtain the third residence four-dimensional space feature data.

[0017] The present invention adopts the extraction of retention time and the calculation of visit frequency to identify the permanent places (such as home and workplace) where users visit frequently and stay for a long time, distinguish the non-resident places visited infrequently, optimize the user portrait, and ensure that the advertising delivery strategy is highly matched with the user's daily behavior. Combined with the calculation of geographical proximity, it distinguishes non-resident near places (such as supermarkets and gyms near home) and non-resident far places (such as business trip hotels and tourist attractions), and improves the spatial adaptability of advertising delivery, such as pushing daily consumption advertisements in near places, and pushing tourism and travel-related advertisements in far places. Through the analysis of the changes in the time attributes of different types of places, the main activity areas of users in specific time periods are identified, such as at home in the morning, at the company during the day, and at the mall at night, to improve the time accuracy of advertising delivery, and to deliver advertisements in the time window most likely to be accepted by users, so as to improve the click-through rate. Through the dynamic annotation of the location type classification data, the classification of users' stay patterns in different places is optimized, so that the advertising recommendation system can adjust the recommendation strategy based on the different activity scenes of users, avoid the advertising content from being out of touch with the actual needs of users, and improve the relevance of advertising and user experience. By building a permanent / non-resident near / far location classification system, we can automatically optimize the timing, content and frequency of advertising exposure to ensure that advertising is delivered in the right place, at the right time and to the right people.

[0018] Low-interaction modeling is used to analyze the stable behavior patterns of users in permanent locations such as home and office, identify long-term stay but low-interaction characteristics, and provide optimized delivery strategies for long-term brand advertising, financial insurance, and subscription services to improve advertising adaptability. Through high-frequency short-stay analysis, users' high-frequency visit behaviors in short-stay locations such as supermarkets, coffee shops, and convenience stores are accurately identified, which is suitable for instant consumption advertising (such as takeout, fast-moving consumer goods, and limited-time discounts), improving the ability to accurately deliver advertisements and avoiding advertising waste. Combined with random stay modeling, users' short-term or low-frequency stay patterns in non-resident remote locations such as tourist attractions, airports, and highway service areas are analyzed to optimize the advertising recommendation logic, such as pushing hotel, car rental, and scenic spot ticket advertisements to tourist users, improving the contextual adaptability of advertisements and user conversion rates. The four-dimensional feature modeling of location (spatial distribution) + time (stay duration) + scene (location type) + interaction (user behavior) is used to finely distinguish user behaviors in different location categories, optimize the content, timing, and frequency of advertising exposure, and improve user experience.

[0019] Optionally, S3 includes: S31, performing relevance calculation on preset user portrait data according to the retained four-dimensional spatial feature data to obtain advertisement relevance data; S32, generating advertisement delivery content according to the advertisement relevance data and a preset advertisement material database to obtain advertisement delivery content data; S33, calculating the advertisement delivery timing according to the retained four-dimensional space feature data and the advertisement delivery content data to obtain advertisement delivery timing data; S34: Integrate the advertisement content data and the advertisement timing data to obtain intelligent advertisement recommendation data.

[0020] In the present invention, the correlation calculation is performed by retaining the four-dimensional space feature data and the preset user portrait, accurately matching the user's interests, travel habits and consumption preferences, ensuring that the advertising content is highly consistent with the target user, and improving the advertising effect. Combined with the advertising relevance data and the advertising material database, based on the user's stay location, time and behavior pattern, the most suitable advertising content for the user scenario is automatically generated to ensure the adaptability of the advertisement in different travel environments and retention areas, and improve user acceptance. The advertising delivery timing is calculated by retaining the four-dimensional space feature data + advertising delivery content data, analyzing the user's behavior pattern in different time periods, ensuring that the advertisement is exposed in the time period when the user's attention is high and the advertisement is easily accepted, and improving the advertising conversion rate. By integrating the advertising delivery content data + advertising delivery timing data, the advertising recommendation system is optimized, the advertising delivery strategy is dynamically adapted to the changes in user behavior, the intelligence and precision of the advertising system are improved, and the advertising delivery effect is ensured to be the best. Through the intelligent optimization of advertising recommendations, invalid advertising delivery is reduced, user fatigue and disgust with advertising are avoided, and it is ensured that the advertisement is pushed to high-potential consumer users at the right time and place, and the advertising click-through rate and user interaction rate are improved.

[0021] Optionally, S4 includes: S41, calculating the advertisement exposure suitability according to the intelligent advertisement recommendation data and the user retention data, and obtaining advertisement exposure suitability data; S42, calculating the advertisement playing frequency according to the advertisement exposure suitability data to obtain advertisement playing frequency data; S43, generating an advertisement content display mode according to the advertisement play frequency data and the user retention data, and obtaining advertisement content display mode data; S44. Generate advertisement placement for intelligent advertisement recommendation data according to advertisement content display method data and advertisement play frequency data to obtain vehicle-mounted advertisement placement data.

[0022] The present invention combines intelligent advertising recommendation data with user retention data to analyze the user's advertising acceptance in different locations, times, and behavior modes, ensure that the advertising delivery scenario is highly matched with user needs, and improve advertising visibility and user acceptance. Through the advertising exposure adaptation data, the optimal advertising playback frequency is calculated to ensure that high-relevance advertising obtains higher exposure, and low-relevance advertising is played less, so as to avoid user advertising fatigue and improve advertising delivery effect and click-through rate. Combined with the advertising playback frequency data and user retention behavior, the advertising display method (such as short video, static picture, interactive advertising, etc.) is intelligently selected to ensure that the advertising display form adapts to the user's current state (such as static viewing vs. listening while driving) and improve the advertising conversion rate. By integrating the advertising content display method + advertising playback frequency data, the vehicle advertising system is optimized to dynamically adjust the advertising playback strategy in real time, ensure that the advertising is delivered at the best time, best location, and best frequency, and improve advertising revenue. By accurately calculating the advertising exposure adaptation + dynamically optimizing the playback frequency, the exposure of invalid advertising is reduced, the efficient use of advertising resources is ensured, the user's advertising click-through rate and interaction rate are improved, the advertising delivery ROI is improved, and the commercial value of advertising is enhanced.

[0023] Optionally, the present application further provides an advertisement data processing system for executing the advertisement data processing method as described above, the advertisement data processing system comprising: The vehicle behavior collection and retention detection module is used to collect user location data and vehicle location data through a vehicle camera or a mobile device, and extract the user's boarding and alighting location based on the user location data and the vehicle location data to obtain the user's boarding and alighting location data; perform retention calculation based on the user's boarding and alighting location data to obtain the user's retention data; The vehicle-mounted regional retention analysis module is used to calculate the retention area heat map according to the user retention data to obtain the retention area heat map data; extract the retention area attributes according to the retention area heat map data to obtain the retention area attribute data; construct the four-dimensional feature space according to the retention area attribute data to obtain the retention four-dimensional space feature data; An intelligent advertisement recommendation module, used for making intelligent advertisement recommendations based on the retained four-dimensional spatial feature data to obtain intelligent advertisement recommendation data; The advertising delivery optimization module is used to perform advertising exposure optimization calculations based on intelligent advertising recommendation data and user retention data to obtain in-vehicle advertising delivery data.

[0024] The objects of the present invention are: 1. Adopt the fusion perception technology of vehicle-mounted cameras and mobile devices to collect user location data and vehicle-mounted location data in real time, and accurately extract the user's boarding and alighting locations through trajectory calculation, to ensure that advertising is based on real scenes rather than static area matching, thereby improving the fit between advertising and user behavior.

[0025] 2. Through the analysis of user retention data, we build a heat map of retention areas in multiple time periods, accurately locate high-traffic and high-stay areas, and optimize the advertising delivery areas by combining geographic location, time distribution and environmental information, making advertising display more strategic and improving advertising click-through rate and conversion rate.

[0026] 3. Adopt the four-dimensional modeling method of location (space) + time (behavior cycle) + scene (POI type) + interaction (user attention) to deeply analyze users' travel and consumption habits, realize the precision and personalization of advertising, ensure that advertising meets user needs, and improve user acceptance.

[0027] 4. Combine the retention four-dimensional space feature data with user portrait data to calculate the advertising relevance, ensure that advertising is highly matched with user interests, travel habits and consumption scenarios, and improve advertising appeal and user interaction rate. Through advertising exposure adaptability calculation + playback frequency optimization + intelligent display method generation, the timing, playback frequency and display method of advertising are adjusted in real time to ensure that advertising is delivered at the most appropriate time, the best frequency, and the most acceptable way to users, thereby improving advertising ROI (return on investment). BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Other features, objects and advantages of the present application will become more apparent by reading the detailed description of non-limiting implementations made with reference to the following drawings: Figure 1 A flowchart showing a method for processing advertisement data according to an embodiment of the present invention is provided; Figure 2 A flowchart showing a method for collecting vehicle-mounted behavior and detecting retention in accordance with an embodiment of the present invention is provided; Figure 3 A flowchart showing a method for analyzing a vehicle-mounted regional retention according to an embodiment of the present invention is provided; Figure 4 A flowchart showing a method for intelligent advertisement recommendation according to an embodiment of the present invention is shown; Figure 5 A flowchart showing a method for optimizing advertisement delivery according to an embodiment of the present invention is provided; The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0029] The technical method of the present invention is described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by technicians in this field without creative work are within the scope of protection of the present invention.

[0030] In addition, the drawings are only schematic illustrations of the present invention and are not necessarily drawn to scale. The functional entities can be implemented in software form, or in one or more hardware modules or integrated circuits, or in different network and / or processor methods and / or microcontroller methods.

[0031] It should be understood that, although the terms "first", "second", etc. may be used herein to describe various units, these units should not be limited by these terms. These terms are used only to distinguish one unit from another unit. For example, without departing from the scope of the exemplary embodiments, the first unit may be referred to as the second unit, and similarly the second unit may be referred to as the first unit. The term "and / or" used herein includes any and all combinations of one or more of the listed associated items.

[0032] See also Figures 1 to 5 The present application provides an advertisement data processing method for a vehicle-mounted advertisement delivery device, the method comprising: S1. Collecting user location data and vehicle location data through a vehicle camera or a mobile device, and extracting the user's boarding and alighting location based on the user location data and the vehicle location data to obtain the user's boarding and alighting location data; performing retention calculation based on the user's boarding and alighting location data to obtain the user retention data; In one embodiment, the user's location data is collected through a vehicle-mounted camera (such as an in-vehicle face recognition camera, an outside vehicle environment camera) or a user's mobile device (such as a mobile phone GPS, Wi-Fi, or Bluetooth Beacon). The vehicle-mounted device's GPS or Beidou positioning system synchronously collects the vehicle's location data and records it with a timestamp. The user's location data and the vehicle-mounted location data are analyzed to calculate the relative distance between the user and the vehicle. , if the user meets the following conditions within the time period: , , is the relative distance between the user and the vehicle, The distance threshold for boarding is defined (e.g. 1.5 meters). Define the distance threshold for getting off the bus (e.g. 2.5 meters). Extract the user's getting on and off location data through the user's boarding and getting off timestamps. Calculate the user's stay time in a certain area, such as the cumulative calculation of the stay time of the vehicle route in a certain area, or set the retention time threshold (e.g. 5 minutes), calculate the user's location change since getting off the bus. If the user is still within the radius (such as 50m), it is considered that the user is stranded.

[0033] S2. Calculate the heat map of the retention area according to the user retention data to obtain the heat map data of the retention area; extract the attributes of the retention area according to the heat map data of the retention area to obtain the attribute data of the retention area; construct a four-dimensional feature space according to the attribute data of the retention area to obtain the feature data of the retention four-dimensional space; In one embodiment, a two-dimensional gridding method is used to divide the urban area into N M grid cells (e.g. 50m 50m). Count the number of users staying in each grid: , For grid cells The number of retained users in is the number of the grid in the vertical direction, is the grid number in the horizontal direction, For all users To sum, For which is the indicator function, indicating that the user Is it in the grid? For users location information, It is a two-dimensional grid unit with a size of 50m×50m. The retention data is smoothed by Gaussian kernel density estimation (KDE) to generate retention area heat map data. Time series analysis is performed on the heat data to determine the peak flow period. The average retention time of the area and the high-frequency travel time period of users are calculated. Combined with external data (such as POI data, shopping malls, restaurants, etc.), the retention area attribute data is extracted: whether it is a commercial area / office area / residential area; the main consumer groups in the area (such as young people, family users, business people); the distribution of main transportation tools (walking, cycling, public transportation, etc.).

[0034] The four-dimensional spatial feature data of retention is constructed based on the following four dimensions: spatial dimension, such as the geographical location (latitude and longitude) of the user's retention; temporal dimension, such as the time period and duration of the user's retention; behavioral dimension, such as the user's travel mode (commuting, leisure, shopping); and community dimension, such as combining user portrait data (such as gender, age, and consumption preferences) to analyze the main population types in the area and obtain user demographic characteristics (gender, age, and consumption preferences).

[0035] In one embodiment, the heat, time, space, and user characteristics are combined to form retention four-dimensional feature data.

[0036] S3, performing intelligent advertising recommendation based on the retained four-dimensional spatial feature data to obtain intelligent advertising recommendation data; In one embodiment, an advertisement delivery model is established to match the most suitable advertisement content, based on user behavior tags (such as "business person, female, 20-30 years old, coffee lover"); based on the user's historical click behavior, a collaborative filtering algorithm is used to calculate advertisement similarity: , For advertising User interests The matching degree between For advertising User interests The number of elements in the intersection (i.e. the common interests of the two), For advertising User interests The number of elements in the union (i.e., all points of interest in both).

[0037] S4. Perform advertising exposure optimization calculation based on the intelligent advertising recommendation data and user retention data to obtain the in-vehicle advertising delivery data.

[0038] In one embodiment, based on the smart advertising recommendation data and user retention data , calculate the best ad exposure time, if the user stays for (e.g. 30 seconds), then the ad is placed. If the user's historical preference matches the ad , then priority is given. Set the reward function , is the reward value for advertising. is the weight coefficient of CTR (click-through rate), is the click-through rate, the probability that the ad is clicked, is the weight coefficient of ROI (return on investment), For the return on investment, an indicator to measure the effectiveness of advertising, use Q-learning to learn the best delivery strategy , For the current state Take Action The Q value (i.e. expected reward), is the learning rate, For current action In Status Instant rewards for is a discount factor (between 0 and 1) measuring the importance of future rewards, For the future state The best value, optimize the timing and content of advertising, and generate in-vehicle advertising data, including information such as advertising playback time, playback area, target audience matching, and click-through rate prediction.

[0039] Optionally, S1 includes: S11, collecting user location data and vehicle location data through a vehicle camera or a mobile device; In one embodiment, computer vision technology (such as face recognition and pedestrian detection) is used to identify the user's boarding / disembarking location. Alternatively, the user's geographic location is obtained through mobile phone GPS, Wi-Fi, Bluetooth Beacon, etc.

[0040] S12, extracting the user's boarding and getting-in position data based on the user's position data and the vehicle's position data to obtain the user's boarding and getting-in position data; In one embodiment, the distance between the user and the vehicle is calculated , is the distance between the user and the vehicle, is the user's longitude coordinate, is the longitude coordinate of the vehicle, is the user's longitude coordinate, is the latitude coordinate of the vehicle. Furthermore, within the specified time window, analyze the changes in user and vehicle positions. If the distance between the user and the vehicle is less than the set boarding threshold (e.g., 1.5 m) and the vehicle's driving trajectory or position is highly coincident with the user's position, the position that first meets the conditions is identified as the boarding position. Continue to monitor the position changes of the user and the vehicle. If, after the vehicle stops, the user is still within the set threshold range (e.g., the boarding threshold = 2.5 m) and the vehicle is stationary, the position that last met the conditions is identified as the boarding position.

[0041] S13, calculating the user's vehicle route according to the user's boarding and alighting location data, and obtaining the user's vehicle route data; In one embodiment, the user's movement path in the car is recorded. The user's travel distance is calculated: , is the total distance traveled by the user on the vehicle, is the time point parameter, is the quantity data at a time point, For users in The longitude coordinates of a time point, For users in The longitude coordinates of a time point, For users in The latitude coordinates of a time point, For users in The latitude coordinates of the time points. Calculate the driving speed: , For users in The average speed of the trip, For users in The longitude coordinates of a time point, For users in The longitude coordinates of a time point, For users in The latitude coordinates of a time point, For users in The latitude coordinates of a time point, For the The time parameters of the data points, For the The time parameter of the data point.

[0042] S14, performing regional retention calculation according to the user's vehicle route data to obtain user regional retention data; In one embodiment, a retention time threshold is set Based on the user's vehicle route data, monitor the user's entry and exit time in a certain area and calculate the user's stay time in a certain area: .like , the user is considered to be staying in the area. Combining the area coordinate information, stay time and stay judgment result, the user area stay data (area coordinates, stay time, stay status) is generated. The stay status is whether the stay condition is met (yes / no).

[0043] S15, calculating the retention behavior according to the user's vehicle route data and the user's regional retention data to obtain the user's retention behavior data; In one embodiment, the user's stay time and area type are combined to classify the stay behavior. Short-term stay (such as 2-5 minutes) is for waiting at a traffic light or getting off the bus. Medium-term stay (such as 5-20 minutes) is for dining, shopping, and work-related activities. Long-term stay (such as more than 20 minutes) is for in-depth activities such as entertainment and office work. The stay behavior is classified by calculation: , To minimize the objective, is the user retention sample sequence item, is the number of samples retained by the user, is the category order item, is the number of categories of detention behavior, For sample points Belong to The probability of the class, The user's stay data in this area, For the The center of the class. If the user stays for 2≤ <5 minutes, it is classified as a short stay. <20 minutes, it is classified as medium-time stay. ≥20 minutes is classified as a long stay. Combining the stay time, area type and stay behavior category, the user stay behavior data (stay area coordinates, stay time, user area type (commercial area / office area / residential area), user stay behavior type) is formed.

[0044] S16, calculating the likelihood of attention to the advertisement area according to the user's vehicle route data and the user's retention behavior data, and obtaining the likelihood of attention to the advertisement area; In one embodiment, the influence of user dwell time on advertisement attention is calculated: , is the probability of users paying attention to the advertising area, is the time the user stays in the advertising area. is the mean of the global residence time. >1, it means that the user stays in the advertising area longer than the global average and pays more attention. <1, indicating that the user's stay time in the advertising area is lower than the global average, and the user's attention is low. Calculate the correlation between user stay behavior and advertising: , The correlation between user stay behavior and advertisement, is the user behavior category ordinal item, is the weight of each behavior category, For advertising User behavior categories The similarity between them (using cosine similarity). , For the possibility of attention of the advertising area, is the weight of the residence time factor, is the probability of users paying attention to the advertising area, is the weight of the behavior relevance factor, The relevance of user behavior to advertising.

[0045] S17: Integrate the user area retention data, the user retention behavior data, and the advertisement area attention possibility data to obtain the user retention data.

[0046] In one embodiment, the user region retention data, the user retention behavior data and the advertisement region attention probability data are integrated to obtain the user retention data: , The horizontal and vertical coordinates of the user area retention data, The dwell / residence time.

[0047] Optionally, the regional retention calculation is specifically: Perform regional clustering according to the user's vehicle route data to obtain regional clustering data; In one embodiment, regional clustering is performed: , For point The local density of are the indices of all trajectory points, is the indicator function, if The density is then taken into account (if the trajectory point At the track point Neighborhood radius If it is within, the count is 1, otherwise the count is 0). For point and The distance between The neighborhood radius set for density clustering. Based on the core trajectory point set , classify the trajectory points in the neighborhood into the same area to form regional clustering data: , Clustering data for user regions, is the latitude and longitude coordinates of the user's trajectory point, For trajectory points The cluster category to which it belongs. , is the core trajectory point set, is the trajectory point index, For trajectory points The local density of is the density threshold (determines the minimum number of neighborhood points).

[0048] Classify the detention areas according to the regional clustering data to obtain the detention area classification data; In one embodiment, based on the geographical location of the area, the data of points of interest in the area and the categories of points of interest (such as restaurants, shops, office buildings, and residences) are queried. The retention area is classified according to the regional clustering data (such as commercial areas, residential areas, and office areas). The categories of points of interest in each cluster area are queried: , For Region The classification type (such as "commercial area", "office area", "residential area"), To select the category with the highest probability, are possible region types, For Category In clustering If commercial POIs (restaurants, shops) account for the highest proportion, the area is classified as a commercial area. If office buildings account for the highest proportion, the area is classified as an office area. If residential POIs (apartments, houses) account for the highest proportion, the area is classified as a residential area.

[0049] The user retention time is calculated based on the retention area classification data to obtain the user area retention data.

[0050] In one embodiment, the user's stay time in different retention areas is calculated to form user regional retention data. Calculate entry and exit time , For users in the area A stay time within For users leaving the zone time, Enter the area for users time. , For users in the area The total residence time, For users entering this area Visits, For users in this area Secondary residence time. , It is a data set of users’ stay in various regions. For the detention area category, For users in the area The user's regional retention data is formed by combining the retention area category and the cumulative retention time.

[0051] Optionally, the retention behavior calculation is specifically: According to the user's regional stay data, the user's vehicle route data is divided into vehicle route time characteristics to obtain the user's vehicle route time characteristic data; In one embodiment, different time period categories are set: peak period (07:00-09:00 in the morning, 17:00-19:00 in the evening): commuting-related behaviors; non-peak period (09:00-17:00): daily travel; night time (19:00-23:00): entertainment and leisure activities; late night (23:00-07:00): involving night shifts, accommodation or workers in specific industries; the user's driving trajectory in each time period is calculated.

[0052] In one embodiment, the vehicle route time axis is divided into different time periods according to the user's entry time and exit time. The pre-travel period records the time period of the user driving from the previous area to the current stranded area; the stranded period records the length of time the user stays in the stranded area; and the post-travel period records the time period of the user driving from the current stranded area to the next area. Combined with the driving trajectory data, time period division information and the type of the staying area, the user's vehicle route time feature data is generated.

[0053] Mapping the preset electronic map data according to the user's vehicle route time characteristic data and the detention area classification data to obtain electronic map mapping data; In one embodiment, the user's driving route data and the classification data of the staying area are mapped to an electronic map. The electronic map data uses a high-precision GIS map, which contains information such as roads, buildings, commercial areas, and public facilities. Calculate the nearest point between the user's coordinate point and the electronic map element , is the map element closest to the user's trajectory point. To take the minimum value of the map element, To traverse the set of points of interest on the electronic map, is a collection of points of interest on the electronic map. is the longitude of the user's track point, is the longitude of the point of interest P on the electronic map, is the latitude of the user's trajectory point, is the latitude of the point of interest P on the electronic map.

[0054] Associating the road types according to the electronic map mapping data to obtain the associated data of the stranded roads; In one embodiment, the type of road the user is on is determined based on the electronic map mapping data, and the impact of different road types on the user's retention behavior is analyzed. Expressways do not involve retention and are mainly used for long-distance travel. Urban trunk roads involve traffic lights and short stops. Commercial roads have a high retention rate, and there are behaviors related to shopping, dining, and office work. Retention on residential roads is related to living and visiting. , is the shortest distance from the user's current location to the nearest road, To calculate the set of all roads The minimum value in To traverse all the roads in the road, is the longitude of the user's current location, is the longitude of the road centerline, is the latitude of the user's current location, is the latitude of the road centerline.

[0055] Time attribution processing is performed based on the stranded road association data and the user's vehicle route time feature data to obtain the user's stranded behavior data.

[0056] In one embodiment, a time-road-detention behavior mapping table is set: morning rush hour + city main roads → waiting for traffic lights on the way to work; night + commercial roads → detention in restaurants and entertainment venues; late night + residential roads → detention related to going home or accommodation, or other time-road-detention behavior mapping tables extracted based on historical experience data. , is the attribution probability of user retention behavior, where the behavior categories include "commuting", "shopping", "leisure and entertainment", etc. For different behavior categories (such as commuting, shopping, leisure and entertainment), is the weight learned from historical data, Behavior matching for different times, routes, and road types.

[0057] Optionally, the advertisement area attention probability calculation is specifically as follows: Extract vehicle speed characteristics according to the user's vehicle route data to obtain vehicle speed characteristic data; In one embodiment, the speed characteristics of the user during driving are calculated to analyze whether there is a long period of low-speed driving or parking, thereby affecting the attention to the advertisement. , For users in time The instantaneous speed, For users The latitude and longitude coordinates of the time, For users The latitude and longitude coordinates of the time, For users The latitude and longitude coordinates of the time, For users The latitude and longitude coordinates of the time, For users The timestamp of the moment, For users Timestamp of the moment. , is the average speed of the user, is the total number of speed data points, is the order term of the velocity data points, For users in time The instantaneous speed at . , is the variance of vehicle speed, which measures the degree of change in vehicle speed. is the total number of speed data points, is the order term of the velocity data points, For users in time The instantaneous speed at is the average vehicle speed of the user.

[0058] Obtain user's vehicle image data; In one embodiment, image acquisition is performed by a vehicle-mounted camera (cockpit monitoring camera).

[0059] Perform eye tracking and facial expression recognition based on the user's in-vehicle image data to obtain eye tracking data and facial expression data; In one embodiment, the eye image data and facial expression data are captured by a deep learning algorithm, the eye image data is extracted for sight angle (i.e., the sight matrix of the pupil center is calculated) and the numerical value of the facial expression data is mapped into the facial emotion value. Calculate the sight direction: , is the user's sight angle, is the inverse tangent function, is the pixel ordinate of the user's eye, is the camera reference ordinate, is the pixel horizontal coordinate of the user's eye, The horizontal coordinate of the camera reference. Combined with the user's sight angle, gaze time, and number of sight jumps, eye tracking data is generated. . Calculate expression score (based on CNN model, such as OpenFace): , Score the emotion. is the facial feature point sequence item, is the total number of facial landmarks, is the weighting coefficient, These are facial feature points (upward corners of the mouth, raised eyebrows, etc.). .

[0060] Calculate the extreme value of emotional accumulation according to the facial expression data to obtain the extreme value data of emotional accumulation; In one embodiment, the cumulative change of the user's emotions over a period of time is calculated to determine whether the user has a long-term emotional response to the advertisement content. , is the total amount of emotion during the entire time period, For users in time The timestamp of the moment, is the starting time for cumulative emotion calculation, is the end time of the cumulative emotion calculation, For time Moment score for the moment. , It is the maximum accumulated emotion value of the user in this time period.

[0061] Extracting gaze features of the advertisement area according to the eye tracking data to obtain gaze feature data of the advertisement area; In one embodiment, the boundaries of the advertisement screen area are set: , is the horizontal coordinate of the lower left corner of the advertising area, is the lower left corner ordinate of the advertising area, is the horizontal coordinate of the user's eye focus, is the vertical coordinate of the user's eye focus, is the horizontal coordinate of the upper right corner of the advertising area, is the upper right corner ordinate of the advertising area. If the above conditions are met, it is determined that the user is looking at the advertising area. , The cumulative time that users gaze at the ad area. The timestamp of the user's line of sight coordinate data. The time interval for the user to continuously gaze at the advertising area. The advertising area gaze feature data is generated by combining the user's line of sight coordinates, gaze time, and advertising area boundaries.

[0062] Calculate the advertisement attention based on the advertisement area gaze feature data and the emotion accumulation extreme value data to obtain the user advertisement attention data; In one embodiment, , Score the user's attention to the ad. is the fixation time weight coefficient, is the fixation time on the advertising area, is the extreme value weight coefficient of emotion accumulation, Accumulate extreme values ​​for emotions.

[0063] Based on the user's advertising attention data, vehicle speed feature data and user retention behavior data, direct passive attention behavior association is performed to obtain the advertising area attention possibility data.

[0064] In one embodiment, , Note the possibility for advertising areas, is the weight coefficient of advertising attention, Score the user's ad attention. is the weight coefficient of the average vehicle speed, is the average speed of the user, is the weight coefficient of retention behavior, The weights are obtained by training based on historical experience data and represent the user's retention behavior in the area.

[0065] In one embodiment, the user's advertising attention data is associated with the vehicle speed characteristic data and the user's retention behavior data to obtain a corresponding data group.

[0066] Optionally, S2 includes: Calculate the user retention point density according to the user retention data to obtain the user retention point density data; In one embodiment, the retention point density is calculated: , For the location The density of stagnation points at is the number of retained data points, is the smoothing parameter (bandwidth) of the kernel function, is the retention point order term, is a kernel function (such as a Gaussian kernel), is the horizontal coordinate of the target point, For the user's The horizontal coordinate of the retention point, is the ordinate of the target point, For the user's The vertical coordinate of the retention point.

[0067] Perform multi-period dynamic heat calculation based on user retention point density data to obtain retention area heat map data; In one embodiment, the heat of the retention area in different time periods is calculated, and the distribution changes of the crowd at different times are analyzed. , For time Location The retention heat is the number of retained data points, is the smoothing parameter (bandwidth) of the kernel function, is the retention point order term, is a kernel function (such as a Gaussian kernel), is the horizontal coordinate of the target point, For the user's The horizontal coordinate of the retention point, is the ordinate of the target point, For the user's The vertical coordinate of the retention point, is the time decay function. , is the natural exponential term, is the time attenuation coefficient (controls the influence of time), is the time data corresponding to the sequence item of the retention point, The current time data.

[0068] Perform location type classification matching based on the retention area heat map data to obtain location type classification data; In one embodiment, matching the electronic map POI to query the stay point Nearby POI records: , For retention point Type of location, To select the location type with the highest probability, Possible location types for the detention point, For the retention point belongs to the category The probability of (based on historical data statistics), is the retention point type data, is a collection of all possible location types (such as "shopping mall", "office area", "restaurant"). Classify the detention points: , To minimize the objective function, is the order term of the sample points, is the number of sample points, is the cluster center order item, is the cluster quantity data, For sample points Belong to The probability of the class, For user A retention point, For the The cluster center of the class.

[0069] According to the location type classification data, the time attribute change is processed to obtain the retention area attribute data; In one embodiment, the data is classified according to the location type, the user's stay type at different times and locations is obtained, and the time proportion of the location type is calculated. , For time The proportion of a certain location type, For time The number of detention points belonging to this location type, For time The total number of detention points in the place. Combined with the proportion of place types in different time periods, the detention area attribute data is generated.

[0070] A four-dimensional feature space is constructed according to the retention area attribute data to obtain the retention four-dimensional space feature data.

[0071] In one embodiment, the spatial dimension is such as the user's retention location, the temporal dimension is such as the retention time, the behavioral dimension is such as the location type of the retention point, and the community dimension is such as the user group characteristics (age, occupation); the feature vector is constructed: , is the user's retention four-dimensional feature vector, is the longitude of the user's stay point, is the latitude of the user's stay point, The time the user stays at the point, For this place type at time The proportion of the place.

[0072] Optionally, the time attribute change processing is specifically as follows: According to the location type classification data, the length of stay is extracted and the frequency of stay is calculated to obtain the length of stay data and the frequency of stay data respectively; In one embodiment, a time threshold is set (e.g. 10 minutes). Calculate the cumulative stay time at each location , For users at location The total residence time, For user Places visited , For users Places visited The visit frequency calculation is to calculate the number of times a user visits a certain location: , For users at location The frequency of visits, For user Places visited , is the indicator function, if If the value is greater than the threshold, it is counted as a valid access.

[0073] Identify the permanent location based on the residence duration data and the residence visit frequency data to obtain the permanent location data; In one embodiment, a resident determination threshold is set. (If the total duration of stay per day is > 4h), (For example, the number of visits per week is > 3 times). Combining the length of stay and the frequency of visits to determine the resident status: , For the The residence determination of the stay point, whether it is a permanent location (1 for yes, 0 for no), For Ruodi If the residence time of a retention point is greater than the permanent determination threshold, it is 1. For Ruodi If the visit frequency of a retention point is greater than the visit threshold, it is 1.

[0074] Identify non-resident locations based on the resident location data and location type classification data to obtain non-resident location data; In one embodiment, the non-resident determination condition is set as follows: , For non-resident determination, whether it is a non-resident location (1 for yes, 0 for no), For permanent determination, whether it is a permanent location (1 for yes, 0 for no).

[0075] Calculate the location proximity based on the permanent location data and the non-resident location data to obtain the permanent location proximity data; In one embodiment, the similarity between the permanent location and the non-permanent location is calculated for classification into "near location" and "apparent location". , For permanent location With non-resident locations The geographical distance between For permanent location The longitude coordinates of Non-resident location The longitude coordinates of For permanent location The latitude coordinate of Non-resident location The latitude coordinate of .

[0076] When the resident location proximity data is less than or equal to the preset resident location proximity threshold data, the non-resident location data corresponding to the resident location proximity data is determined as the non-resident near location data; When the resident location proximity data is greater than the preset resident location proximity threshold data, the non-resident location data corresponding to the resident location proximity data is determined as non-resident remote location data; In one embodiment, based on the proximity threshold of the permanent location Classify perigee and apogee. , , For non-resident near-point gatherings, are the latitude and longitude coordinates of the permanent perigee, are the latitude and longitude coordinates of the non-resident perigee, is the distance threshold between the permanent location and the non-permanent location, Assemble for non-resident remote locations.

[0077] The location type classification data is labeled according to the resident location data, the non-resident near location data and the non-resident far location data to obtain the retention area attribute data; In one embodiment, resident, non-resident near and far locations are classified and labeled with location type data. , assign values ​​to non-resident perigees: , assign values ​​to non-resident remote locations: .

[0078] The retained four-dimensional space feature data includes the first retained four-dimensional space feature data, the second retained four-dimensional space feature data and the third retained four-dimensional space feature data. The four-dimensional feature space is constructed as follows: Low-interaction modeling is performed on the retention area attribute data corresponding to the number of permanent locations to obtain the first retention four-dimensional spatial feature data; In one embodiment, the average length of stay of a user at a permanent location is set as follows: , For location The average residence time, For users at location The number of detentions that occurred, For users at location No. Visits, For user Places visited Calculate the user's interaction frequency at this location (i.e. the number of times the user switches to other locations): , For users at location The interaction frequency, For users at location The number of detentions that occurred, For users from location The number of times you switch to other locations. Set low interaction rules: , For location Is it a low-interaction location (1 for yes, 0 for no), Whether the average time users stay at the location exceeds the low interaction time threshold, Whether the user's interaction frequency at the location is lower than the low interaction frequency threshold, To define the minimum dwell time for low interaction (e.g. >4h per day), To define the maximum switching frequency of low interaction, the first retention four-dimensional spatial feature data (the coordinates of the permanent location, the average retention time of the location, the user's interaction frequency, whether it is a low interaction location (1 for yes, 0 for no)) is generated by combining the low interaction determination result and the four-dimensional feature vector.

[0079] Perform high-frequency short-dwelling analysis on the retention area attribute data corresponding to the non-resident perigee data to obtain the second retention four-dimensional space feature data; In one embodiment, the mean short dwell time is calculated: , Non-resident nearby locations The average short-term residence time at For users at location The number of short stays that occurred, For users at location No. Short stay, For user Places visited The residence time, To define the maximum stay time for a short stay (e.g. <30min). Calculate the user's visit frequency to the location: , For users in non-resident nearby locations The number of short-stay visits at For users at location No. Visits, For users at location The number of short stays that occurred, is the indicator function, if , it is counted as a short visit. , For location Whether it is a high-frequency short-stay location (1 for yes, 0 for no), Whether the user's short-stay visit frequency at the location exceeds the threshold. The second retention four-dimensional spatial feature data (the coordinates of non-resident nearby locations, the average short-stay time at the location, the user's short-stay visit frequency, whether it is a high-frequency short-stay location (1=yes, 0=no)) is constructed based on the four dimensions of space, time, behavior and community.

[0080] The random residence model is performed on the residence area attribute data corresponding to the non-resident remote location data to obtain the third residence four-dimensional space feature data.

[0081] In one embodiment, the random stay feature is calculated to calculate the average stay time of a single visit: , For non-resident remote locations The average duration of a single visit at For users at location The number of stops that occurred, For users at location No. Visits, For user Places visited of stay time. For users in non-resident remote locations The variance of the time interval between visits at For users at location The number of stops that occurred, For users at location No. Visits, For user Places visited timestamp, For user Places visited The timestamp of the , For location Whether it is a random stop location (1 for yes, 0 for no), Whether the variance of the user's visit interval at this location exceeds the threshold, is the set random stay interval variance threshold. Combined with the random stay judgment result and the four-dimensional feature vector, the third stay four-dimensional spatial feature data (the coordinates of non-resident remote locations, the average single stay time of the location, the user's visit interval variance, whether it belongs to a random stay location (1=yes, 0=no)) is generated.

[0082] Optionally, S3 includes: Calculate the relevance of the preset user portrait data based on the retained four-dimensional spatial feature data to obtain advertising relevance data; In one embodiment, , Retention characteristics With user portrait The relevance of advertisements between is the retained four-dimensional space eigenvector, is the user portrait feature vector, is the Euclidean norm of the vector.

[0083] Generate advertisement delivery content according to the advertisement relevance data and a preset advertisement material database to obtain advertisement delivery content data; In one embodiment, the matching score between the advertisement material and the retention feature is calculated: , Retention characteristics With creatives The matching degree, Retention characteristics With user portrait The relevance of advertisements between Create user portraits With creatives Select the top N ads with the highest matching degree: , Retention characteristics The best advertising content in To obtain Largest creative index , Retention characteristics With creatives degree of matching.

[0084] Calculate the advertisement delivery timing according to the retained four-dimensional space feature data and the advertisement delivery content data to obtain advertisement delivery timing data; In one embodiment, the advertising attention index is set: , For users in the stranded area Advertising attention index at is the weight coefficient of fixation duration, is the duration of looking at the advertising area, is the weight coefficient of the extreme value of emotion, is the extreme value of emotion, and the weight is obtained through historical data training (the other weights are obtained in this way unless the specific method is specified). Calculate the best time for advertising exposure: , For the retention area The best time to run ads, To obtain the probability of ad exposure Maximum time , For time Probability of ad exposure at.

[0085] Integrate the advertising content data and the advertising timing data to obtain intelligent advertising recommendation data.

[0086] In one embodiment, the advertisement content and the delivery time are combined: , Retention characteristics Advertising recommendation program at Retention characteristics Recommended ads at Retention characteristics The timing of advertising playback.

[0087] Optionally, S4 includes: Calculate the advertisement exposure suitability based on the intelligent advertisement recommendation data and the user retention data to obtain the advertisement exposure suitability data; In one embodiment, the smart advertising recommendation data and user retention data are combined to calculate the advertising exposure suitability and measure the degree of fit between the advertising content and the user retention characteristics. The weighted matching model is used to calculate the matching degree between the advertising content and the user retention characteristics: , Ad exposure suitability. is the weight coefficient of the advertising content matching degree, The matching degree between the advertising content and the user characteristics. is the weight coefficient of residence time, The time the user stays in the area.

[0088] Calculate the advertisement playing frequency according to the advertisement exposure suitability data to obtain advertisement playing frequency data; In one embodiment, the number of times an advertisement is played is calculated: , For the retention area The number of times ads are played at is the maximum playback frequency, is the minimum playback frequency, Ad exposure suitability. is the minimum value of exposure adaptation, The maximum value of exposure adaptation.

[0089] Generate an advertisement content display mode according to advertisement play frequency data and user retention data to obtain advertisement content display mode data; In one embodiment, the display method is determined according to the user's stay time. , For users in the stranded area The residence time at the location (in seconds), is the set residence time threshold, For the retention area The number of times ads are played at Set the video ad playback frequency threshold.

[0090] According to the advertisement content display method data and the advertisement playing frequency data, the intelligent advertisement recommendation data is used to generate advertisement delivery, and the in-vehicle advertisement delivery data is obtained.

[0091] In one embodiment, , Retention characteristics Advertising plans at To recommend advertising content, For the timing of advertising, is the frequency of advertisement playback, The advertising display method (static / video / voice).

[0092] Optionally, the present application further provides an advertisement data processing system for executing the advertisement data processing method as described above, the advertisement data processing system comprising: The vehicle behavior collection and retention detection module is used to collect user location data and vehicle location data through a vehicle camera or a mobile device, and extract the user's boarding and alighting location based on the user location data and the vehicle location data to obtain the user's boarding and alighting location data; perform retention calculation based on the user's boarding and alighting location data to obtain the user's retention data; The vehicle-mounted regional retention analysis module is used to calculate the retention area heat map according to the user retention data to obtain the retention area heat map data; extract the retention area attributes according to the retention area heat map data to obtain the retention area attribute data; construct the four-dimensional feature space according to the retention area attribute data to obtain the retention four-dimensional space feature data; An intelligent advertisement recommendation module, used for making intelligent advertisement recommendations based on the retained four-dimensional spatial feature data to obtain intelligent advertisement recommendation data; The advertising delivery optimization module is used to perform advertising exposure optimization calculations based on intelligent advertising recommendation data and user retention data to obtain in-vehicle advertising delivery data.

[0093] Therefore, from any point of view, the embodiments should be regarded as illustrative and non-restrictive, and the scope of the present invention is limited by the attached application documents rather than the above description, and it is intended that all changes falling within the meaning and scope of equivalent elements of the application documents are included in the present invention.

[0094] The above description is only a specific embodiment of the present invention, so that those skilled in the art can understand or implement the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but should conform to the widest scope consistent with the principles and novel features invented herein.

Claims

1. A method for processing advertisement data, characterized in that: For an in-vehicle advertising device, the method includes: S1. Collecting user location data and vehicle location data through a vehicle camera or a mobile device, and extracting the user's boarding and alighting location based on the user location data and the vehicle location data to obtain the user's boarding and alighting location data; performing retention calculation based on the user's boarding and alighting location data to obtain the user retention data; S2. Calculate the heat map of the retention area according to the user retention data to obtain the heat map data of the retention area; extract the attributes of the retention area according to the heat map data of the retention area to obtain the attribute data of the retention area; construct a four-dimensional feature space according to the attribute data of the retention area to obtain the feature data of the retention four-dimensional space; S3, performing intelligent advertising recommendation based on the retained four-dimensional spatial feature data to obtain intelligent advertising recommendation data; S4. Perform advertising exposure optimization calculation based on the intelligent advertising recommendation data and user retention data to obtain the in-vehicle advertising delivery data.

2. The method according to claim 1, characterized in that S1 includes: Collect user location data and vehicle location data through vehicle cameras or mobile devices; Extract the user's boarding and getting-in position data based on the user's location data and the vehicle's location data to obtain the user's boarding and getting-in position data; Calculate the user's vehicle route based on the user's boarding and alighting location data to obtain the user's vehicle route data; Perform regional retention calculation based on the user's vehicle route data to obtain the user's regional retention data; Calculate the retention behavior based on the user's vehicle route data and the user's regional retention data to obtain the user's retention behavior data; Calculate the likelihood of attention to the advertising area based on the user's vehicle route data and the user's retention behavior data to obtain the likelihood of attention to the advertising area; The user area retention data, the user retention behavior data and the advertisement area attention possibility data are integrated to obtain the user retention data.

3. The method according to claim 2, characterized in that The calculation of regional retention is as follows: Perform regional clustering according to the user's vehicle route data to obtain regional clustering data; Classify the detention areas according to the regional clustering data to obtain the detention area classification data; The user retention time is calculated based on the retention area classification data to obtain the user area retention data.

4. The method according to claim 2, characterized in that: The retention behavior calculation is as follows: According to the user's regional stay data, the user's vehicle route data is divided into vehicle route time characteristics to obtain the user's vehicle route time characteristic data; Mapping the preset electronic map data according to the user's vehicle route time characteristic data and the detention area classification data to obtain electronic map mapping data; Associating the road types according to the electronic map mapping data to obtain the associated data of the stranded roads; Time attribution processing is performed based on the stranded road association data and the user's vehicle route time feature data to obtain the user's stranded behavior data.

5. The method according to claim 2, characterized in that: The calculation of the probability of attention in the advertising area is as follows: Extract vehicle speed characteristics according to the user's vehicle route data to obtain vehicle speed characteristic data; Obtain user's vehicle image data; Perform eye tracking and facial expression recognition based on the user's in-vehicle image data to obtain eye tracking data and facial expression data; Calculate the extreme value of emotional accumulation according to the facial expression data to obtain the extreme value data of emotional accumulation; Extracting gaze features of the advertisement area according to the eye tracking data to obtain gaze feature data of the advertisement area; Calculate the advertisement attention based on the advertisement area gaze feature data and the emotion accumulation extreme value data to obtain the user advertisement attention data; Based on the user's advertising attention data, vehicle speed feature data and user retention behavior data, direct passive attention behavior association is performed to obtain the advertising area attention possibility data.

6. The method according to claim 1, characterized in that S2 include: Calculate the user retention point density according to the user retention data to obtain the user retention point density data; Perform multi-period dynamic heat calculation based on user retention point density data to obtain retention area heat map data; Perform location type classification matching based on the retention area heat map data to obtain location type classification data; According to the location type classification data, the time attribute change is processed to obtain the retention area attribute data; A four-dimensional feature space is constructed according to the retention area attribute data to obtain the retention four-dimensional space feature data.

7. The method according to claim 6, characterized in that The specific processing of time attribute changes is as follows: According to the location type classification data, the length of stay is extracted and the frequency of stay is calculated to obtain the length of stay data and the frequency of stay data respectively; Identify the permanent location based on the residence duration data and the residence visit frequency data to obtain the permanent location data; Identify non-resident locations based on the resident location data and location type classification data to obtain non-resident location data; Calculate the location proximity based on the permanent location data and the non-resident location data to obtain the permanent location proximity data; When the resident location proximity data is less than or equal to the preset resident location proximity threshold data, the non-resident location data corresponding to the resident location proximity data is determined as the non-resident near location data; When the resident location proximity data is greater than the preset resident location proximity threshold data, the non-resident location data corresponding to the resident location proximity data is determined as non-resident remote location data; The location type classification data is labeled according to the resident location data, the non-resident near location data and the non-resident far location data to obtain the retention area attribute data; The retained four-dimensional space feature data includes the first retained four-dimensional space feature data, the second retained four-dimensional space feature data and the third retained four-dimensional space feature data. The four-dimensional feature space is constructed as follows: Low-interaction modeling is performed on the retention area attribute data corresponding to the number of permanent locations to obtain the first retention four-dimensional spatial feature data; Perform high-frequency short-dwelling analysis on the retention area attribute data corresponding to the non-resident perigee data to obtain the second retention four-dimensional space feature data; The random residence model is performed on the residence area attribute data corresponding to the non-resident remote location data to obtain the third residence four-dimensional space feature data.

8. The method according to claim 1, characterized in that S3 include: Calculate the relevance of the preset user portrait data based on the retained four-dimensional spatial feature data to obtain advertising relevance data; Generate advertisement delivery content according to the advertisement relevance data and a preset advertisement material database to obtain advertisement delivery content data; Calculate the advertisement delivery timing according to the retained four-dimensional space feature data and the advertisement delivery content data to obtain advertisement delivery timing data; Integrate the advertising content data and the advertising timing data to obtain intelligent advertising recommendation data.

9. The method according to claim 1, characterized in that: S4 includes: Calculate the advertisement exposure suitability based on the intelligent advertisement recommendation data and the user retention data to obtain the advertisement exposure suitability data; Calculate the advertisement playing frequency according to the advertisement exposure suitability data to obtain advertisement playing frequency data; Generate an advertisement content display mode according to advertisement play frequency data and user retention data to obtain advertisement content display mode data; According to the advertisement content display method data and the advertisement playing frequency data, the intelligent advertisement recommendation data is used to generate advertisement delivery, and the in-vehicle advertisement delivery data is obtained.

10. An advertisement data processing system, characterized in that: For executing the advertisement data processing method according to claim 1, the advertisement data processing system comprises: The vehicle behavior collection and retention detection module is used to collect user location data and vehicle location data through a vehicle camera or a mobile device, and extract the user's boarding and alighting location based on the user location data and the vehicle location data to obtain the user's boarding and alighting location data; perform retention calculation based on the user's boarding and alighting location data to obtain the user's retention data; The vehicle-mounted regional retention analysis module is used to calculate the retention area heat map according to the user retention data to obtain the retention area heat map data; extract the retention area attributes according to the retention area heat map data to obtain the retention area attribute data; construct the four-dimensional feature space according to the retention area attribute data to obtain the retention four-dimensional space feature data; An intelligent advertisement recommendation module, used to perform intelligent advertisement recommendation based on the retained four-dimensional spatial feature data to obtain intelligent advertisement recommendation data; The advertising delivery optimization module is used to perform advertising exposure optimization calculations based on intelligent advertising recommendation data and user retention data to obtain in-vehicle advertising delivery data.

Images