A method for accurately pushing oil sales based on user portraits

Abstract

This invention relates to a method for precise oil sales targeting based on user profiles, comprising: acquiring vehicle attribute data, historical refueling behavior data, geographic location trajectory data, and promotional response feedback data of target users; and constructing a multi-dimensional user profile based on the vehicle attribute data, historical refueling behavior data, geographic location trajectory data, and promotional response feedback data, including vehicle type tags, refueling frequency tags, fuel preference tags, geographically active area tags, and promotion sensitivity tags. This method for precise oil sales targeting based on user profiles constructs a multi-dimensional user profile system specifically for oil sales scenarios, deeply integrating four core dimensions: vehicle attributes, refueling behavior, geographic location, and promotional response. It overcomes the shortcomings of general advertising push models that rely on online behavioral data and are disconnected from offline refueling scenarios.

Description

Technical Field

[0001] This invention relates to the field of computer technology, specifically to a method for precise oil sales recommendations based on user profiles. Background Technology

[0002] With the optimization of energy consumption structure and the development of digital marketing technologies, the petroleum sales industry is gradually introducing big data and artificial intelligence to improve customer service levels and marketing accuracy. Precision push technology based on user profiles has been widely used in e-commerce, advertising, and finance. Its core lies in constructing a user tagging system through multi-source data fusion and distributing personalized information based on behavioral preferences. However, the petroleum retail scenario has significant industry-specific characteristics: low user purchase frequency, high transaction value, and behavioral patterns highly dependent on geographical location. Furthermore, it is affected by multiple operational factors such as oil price fluctuations, promotional cycles, and gas station inventory, making it difficult for general-purpose profiling models to accurately depict the real needs and response mechanisms of fuel consumers.

[0003] Among these, the precise push notifications for petroleum sales focus on combining dimensions such as vehicle attributes, refueling habits, location movement patterns, and price sensitivity to construct a dedicated user profile system tailored to the characteristics of high-frequency offline location interactions and low-frequency, high-value transactions. This technological direction aims to break through the limitations of traditional advertising or push notification logic in the power sector, deeply integrating users' offline refueling behavior with online marketing strategies, thereby improving promotional conversion efficiency and customer lifetime value while ensuring user experience.

[0004] In existing technologies, some solutions attempt to migrate general user profiling methods to the energy consumption field, but they suffer from significant adaptation deficiencies. For example, while some electricity-carbon collaborative push methods consider electricity price elasticity and user classification, they fail to incorporate key features such as the geographical distribution of gas stations, fuel type preferences, and the correlation between vehicle displacement and fuel consumption. Other advertising push systems, while possessing strong data integration capabilities, rely on online behavioral data such as web clickstreams and social interactions, which have weak correlation with actual refueling decisions by oil users. More importantly, existing methods generally neglect the timeliness constraints of promotional activities in oil sales, the dynamic inventory of regional gas stations, and the differentiated response patterns of users to composite incentive mechanisms such as coupons, points, and membership levels. This leads to a disconnect between the pushed content and the user's true intentions, resulting in low conversion rates. These problems are particularly pronounced in scenarios of drastic oil price fluctuations or intensified regional competition, severely restricting the improvement of refined operational capabilities of oil retail enterprises. A precise push method that deeply integrates industry characteristics is urgently needed to solve this technical challenge. Summary of the Invention

[0005] The purpose of this invention is to provide a method for precise oil sales targeting based on user profiles, in order to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a method for precise oil sales targeting based on user profiles, comprising:

[0007] Acquire vehicle attribute data, historical refueling behavior data, geographic location trajectory data, and promotional response feedback data of target users;

[0008] Based on the vehicle attribute data, historical refueling behavior data, geographic location trajectory data, and promotion response feedback data, a multi-dimensional user profile is constructed, including vehicle type tags, refueling frequency tags, fuel preference tags, geographic active area tags, and promotion sensitivity tags.

[0009] Real-time data collection includes current oil price fluctuations, gas station inventory status, promotional activity configurations, and user location data.

[0010] The multi-dimensional user profile is dynamically matched and analyzed with the current oil price fluctuation data, the inventory status data of each gas station, the promotional activity configuration data, and the user's current location data to generate a personalized push strategy for the target user.

[0011] According to the personalized push strategy, marketing information containing specific oil product discounts, guidance to designated gas station locations, and points redemption incentives is pushed to the target users.

[0012] As one embodiment of the present invention, obtaining the vehicle attribute data of the target user specifically includes: extracting the vehicle brand, vehicle model, engine displacement, fuel type and vehicle service life from the vehicle registration information database; obtaining the vehicle identification code from the vehicle terminal or fuel card binding information and associating it with the user's identity identifier.

[0013] As one embodiment of the present invention, the acquisition of historical refueling behavior data specifically includes: extracting from the gas station transaction system the user's refueling timestamp, refueling amount, selected oil grade, single consumption amount, payment method, whether coupons were used, and whether participation in points accumulation within a preset time window; the preset time window is the most recent twelve months.

[0014] As one embodiment of the present invention, the acquisition of geographic location trajectory data specifically includes: collecting user location coordinates every thirty minutes through the mobile terminal positioning module, and identifying the user's permanent residence area, commuting route and frequently passed gas stations in combination with map road network data; the permanent residence area includes residence, workplace and weekend activity area.

[0015] As one embodiment of the present invention, the acquisition of promotional response feedback data specifically includes: recording the user's opening behavior, clicking behavior, on-site redemption behavior, and subsequent refueling behavior of historical push information; the on-site redemption behavior refers to the user's behavior of completing a refueling transaction and using the coupon at a designated gas station within the validity period after receiving the coupon push.

[0016] As one embodiment of the present invention, the construction of a multi-dimensional user profile specifically includes: structuring and encoding vehicle attribute data to generate vehicle type tags, wherein the vehicle type tags are divided into four categories: small cars, mid-size SUVs, large trucks, and new energy hybrid vehicles; statistically analyzing historical refueling behavior data to calculate the user's average refueling cycle, standard deviation of single refueling volume, and proportion of high-octane fuel selection, thereby generating refueling frequency tags, refueling stability tags, and fuel preference tags; clustering geographic location trajectory data to identify the set of active gas stations within a five-kilometer radius of the user and marking them as geographically active area tags; and modeling the response rate of promotional response feedback data to calculate the user's redemption conversion rate for price discounts, full reductions, and points bonus promotions, thereby generating promotion sensitivity tags.

[0017] As one embodiment of the present invention, the real-time collection of current oil price fluctuation data specifically includes: obtaining the benchmark oil price adjustment range from the refined oil price adjustment announcement interface, and calculating the current retail guidance price of each oil product in combination with local surcharges; the inventory status data of each gas station is uploaded to the central management system in real time through oil tank level sensors to determine whether the remaining inventory of each oil product is lower than the safety threshold; the promotional activity configuration data is issued by the marketing management system, including the start and end time of the activity, applicable oil products, discount level, target user group and inventory quota limit.

[0018] As one embodiment of the present invention, the dynamic matching analysis specifically includes: first, determining whether the user's current location is within the range defined by their geographically active area tag; if so, filtering gas stations within that range that have sufficient inventory and are currently conducting promotional activities; second, matching applicable oil product promotional activities based on the user's oil product preference tag; third, combining the user's promotion sensitivity tag, prioritizing the promotion type with the highest historical redemption rate; and finally, considering the current oil price fluctuation trend, if the oil price is on an upward trend, pushing out marketing information that locks in a limited-time discounted price in advance.

[0019] As one embodiment of the present invention, the personalized push strategy specifically includes: determining push content elements, the elements including the target gas station name, recommended fuel grade, discounted unit price, discount validity period, expected savings, and additional points rewards; determining the push timing, the timing being based on the user's historical refueling cycle to predict the next refueling time, and triggering the push 24 hours before the predicted time; and determining the push channel, the channel including mobile application messages, SMS, and WeChat service account template messages, selecting the optimal channel based on the user's historical channel open rate.

[0020] As one embodiment of the present invention, the push marketing information specifically includes: displaying a graphic card containing a map navigation entry on the mobile terminal, the graphic card showing a real-view image of the target gas station, the current number of vehicles in the queue, discount details, and a one-click navigation button; at the same time, after the user starts navigation, the electronic coupon is automatically activated, and a redemption prompt is triggered via Bluetooth beacon or geofencing technology after arriving at the gas station.

[0021] According to another aspect of the present invention, a precise oil sales push system based on user profiles is provided, comprising:

[0022] The multi-source data acquisition module is used to acquire vehicle attribute data, historical refueling behavior data, geographical location trajectory data, and promotional response feedback data of target users;

[0023] A dedicated user profile building module is used to build a multi-dimensional user profile based on the data obtained by the multi-source data acquisition module, including vehicle type tags, refueling frequency tags, fuel preference tags, geographically active area tags, and promotion sensitivity tags.

[0024] The operational status awareness module is used to collect real-time data on current oil price fluctuations, inventory status of each gas station, promotional activity configuration, and the user's current location.

[0025] The dynamic strategy matching module is used to dynamically match and analyze the multi-dimensional user profile with the data collected by the operation status perception module to generate a personalized push strategy for the target user.

[0026] The precise information push module is used to push marketing information containing specific oil product discounts, designated gas station location guidance, and points redemption incentives to the target user according to the personalized push strategy.

[0027] In one embodiment of the present invention, the multi-source data acquisition module includes a vehicle information interface unit, a transaction log parsing unit, a location trajectory aggregation unit, and a feedback behavior tracking unit. The vehicle information interface unit is connected to a vehicle management database for extracting structured vehicle attributes. The transaction log parsing unit interfaces with a gas station POS system for parsing standardized transaction records. The location trajectory aggregation unit collects latitude and longitude sequences through a mobile terminal SDK and performs trajectory compression and stop point identification. The feedback behavior tracking unit listens to the user's operation event stream on push messages and records the complete behavior chain.

[0028] As one embodiment of the present invention, the dedicated user profile construction module includes a tag generation submodule and a profile storage submodule; the tag generation submodule embeds a rule engine and a statistical model to execute tag calculation logic; the profile storage submodule adopts a columnar storage structure, indexed by the user's unique identifier, and supports millisecond-level profile query response.

[0029] In one embodiment of the present invention, the operation status perception module includes an oil price monitoring unit, an inventory monitoring unit, an activity configuration synchronization unit, and a real-time positioning unit; the oil price monitoring unit periodically polls the official price release source; the inventory monitoring unit receives liquid level sensor data from each gas station through an IoT gateway; the activity configuration synchronization unit maintains bidirectional synchronization with the marketing management system; and the real-time positioning unit obtains the user's current location through a mobile network or satellite positioning.

[0030] In one embodiment of the present invention, the dynamic strategy matching module includes a geofencing determination unit, an inventory-activity filtering unit, a user-activity matching unit, and a strategy sorting unit; the geofencing determination unit calculates the distance between the user's current location and each gas station; the inventory-activity filtering unit excludes options with insufficient inventory or ended activities; the user-activity matching unit filters suitable activities based on profile tags; and the strategy sorting unit sorts candidate strategies in descending order according to the expected conversion probability.

[0031] In one embodiment of the present invention, the precise information push module includes a content assembly unit, a timing control unit, and a channel distribution unit; the content assembly unit fills a preset message template according to strategy parameters; the timing control unit calculates the optimal push time based on a refueling cycle prediction model; and the channel distribution unit selects a push channel based on the user's historical channel preference weights and records the delivery status.

[0032] Compared with existing technologies, the beneficial effects of this invention are as follows: This method for precise push notifications for petroleum sales based on user profiles constructs a multi-dimensional user profile system specifically for petroleum sales scenarios. It deeply integrates four core dimensions: vehicle attributes, refueling behavior, geographical location, and promotional response, overcoming the shortcomings of general advertising push models that rely on online behavioral data and are disconnected from offline refueling scenarios. This invention introduces a dynamic strategy engine, performing joint optimization matching of user profiles with real-time oil prices, gas station inventory, promotional activities, and user location under multiple constraints. This solves the technical bottleneck of existing push methods in the power or e-commerce fields, which cannot adapt to the low-frequency, high-value, strongly regional, and complex incentive response characteristics of petroleum retail. Through precise timing control based on refueling cycle prediction and an intelligent distribution mechanism based on historical channel preferences, this invention significantly improves the open rate of marketing information and the conversion rate of on-site verification. Simultaneously, the system integrates map navigation and automatic verification functions into the push content, forming a closed loop from information reach to transaction completion, effectively enhancing user stickiness and optimizing gas station inventory turnover efficiency, achieving a dual improvement in customer experience and operational efficiency. Attached Figure Description

[0033] Figure 1 This is a schematic diagram of the overall technical solution architecture of the present invention. Detailed Implementation

[0034] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0035] Please see Figure 1 This invention provides a technical solution: a method for precise push notifications for oil sales based on user profiles. It aims to address the technical problem in existing general-purpose user profile models that fail to fully integrate the multi-dimensional behavioral characteristics and operational constraints unique to the oil retail scenario, resulting in low relevance and insufficient conversion efficiency of the pushed information. This method constructs a dedicated user profile system for oil sales operations and designs a dynamic strategy engine to achieve personalized, timely, and highly effective precise push notifications of marketing information.

[0036] The method includes the following steps: S1, acquiring vehicle attribute data, historical refueling behavior data, geographic location trajectory data, and promotional response feedback data of the target user; S2, based on the vehicle attribute data, historical refueling behavior data, geographic location trajectory data, and promotional response feedback data, constructing a multi-dimensional user profile including vehicle type tags, refueling frequency tags, fuel preference tags, geographic active area tags, and promotion sensitivity tags; S3, collecting real-time data on current oil price fluctuations, inventory status data of each gas station, promotional activity configuration data, and the user's current location; S4, dynamically matching and analyzing the multi-dimensional user profile with the current oil price fluctuation data, inventory status data of each gas station, promotional activity configuration data, and the user's current location data to generate a personalized push strategy for the target user; S5, according to the personalized push strategy, pushing marketing information to the target user including specific fuel discount information, designated gas station location guidance, and points redemption incentive combinations.

[0037] In step S1, vehicle attribute data, historical refueling behavior data, geographic location trajectory data, and promotional response feedback data of the target user are acquired. This step is the data input foundation for the entire precision push process, and its data completeness and accuracy directly determine the quality of subsequent profile construction and strategy matching accuracy.

[0038] The specific operations for obtaining vehicle attribute data include: extracting vehicle brand, model, engine displacement, fuel type, and vehicle age from the vehicle registration information database; obtaining the vehicle identification number (VIN) from the on-board terminal or fuel card binding information, and uniquely associating the VIN with the user's identity identifier. The vehicle registration information database is maintained by the traffic management department, and its data structure consists of standardized fields containing information such as the vehicle's unique identification number, registration date, owner's name, and vehicle type. The on-board terminal reads the VIN via the OBD interface or CAN bus and uploads it to the user account system through a secure encrypted channel. The fuel card binding information is linked using the vehicle information submitted by the user when registering as a member at the gas station. The data from these two sources is cross-validated using the user's identity identifier. If a conflict exists, the vehicle registration information database takes precedence, triggering a manual review process.

[0039] The specific operations for obtaining historical refueling behavior data include: extracting the user's refueling timestamp, refueling volume, selected fuel grade, single transaction amount, payment method, whether a coupon was used, and whether participation in points accumulation within a preset time window from the gas station transaction system; the preset time window is the most recent twelve months. The gas station transaction system records each transaction using a unified data format, including transaction serial number, pump number, fuel type, volume in liters, unit price, total price, payment channel, coupon ID, points change value, and user membership card number. The system indexes all transaction records by user membership card number and arranges them in reverse chronological order. For cardless refueling users, the system matches historical transactions through license plate recognition or mobile phone number to ensure that behavioral data covers the entire user group. Transaction data is batch-synchronized to the central data warehouse every morning and undergoes data cleaning to remove test transactions, refund transactions, and outliers (such as single refueling volumes exceeding 100 liters for non-truck vehicles).

[0040] The specific operations for acquiring geographic location trajectory data include: collecting user location coordinates every 30 minutes via the mobile terminal positioning module, and combining this with map road network data to identify the user's frequent residence areas, commuting routes, and frequently visited gas stations; the frequent residence areas include the residence, workplace, and weekend activity areas. The mobile terminal positioning module uses a GPS and base station triangulation fusion algorithm, automatically switching to Wi-Fi fingerprint positioning indoors or in areas with weak signals. Location coordinates are stored in latitude and longitude format, with timestamps, positioning accuracy, and positioning type identifiers added. The system performs a trajectory compression algorithm on the location sequence daily, eliminating stationary and redundant points and retaining key moving nodes. Subsequently, the DBSCAN clustering algorithm is used to cluster the stop points, setting a minimum stop time of four hours and a maximum cluster radius of 500 meters to identify the user's main frequent residence areas. The commuting route is fitted by combining the shortest path connecting the residence and workplace with historical actual driving trajectories. High-frequency passing gas stations are determined by calculating the number of intersections between the user's trajectory and the gas station's geofence. If a gas station has been passed more than five times in the past thirty days and is less than 800 meters away from the trajectory point, it is marked as a high-frequency passing station.

[0041] The specific operations for obtaining promotion response feedback data include: recording user actions such as opening historical push notifications, clicking on them, redeeming them at the designated gas station, and subsequent refueling. Redeeming at the designated gas station refers to the user completing a refueling transaction and using the coupon within the validity period after receiving the coupon push notification. The system embeds a unique tracking ID in the push notification. When a user opens the message, the client reports an opening event; when they click on the discount details or navigation button, a click event is reported; when the user arrives at the gas station and completes the transaction, the system compares the coupon ID in the transaction record with the push notification ID to confirm the redemption. Subsequent refueling refers to whether the user refuels within 72 hours after redemption, used to evaluate the long-term pull effect of the promotion. All behavioral events are written to the behavior log database in real time via a message queue and indexed by user ID and activity ID, supporting millisecond-level queries.

[0042] In step S2, a multi-dimensional user profile is constructed based on the above four types of data. This profile consists of five core tags: vehicle type tag, refueling frequency tag, fuel type preference tag, geographically active area tag, and promotion sensitivity tag.

[0043] The specific operations for constructing vehicle type tags include: structuring and encoding vehicle attribute data to generate vehicle type tags, which are divided into four categories: small cars, mid-size SUVs, large trucks, and new energy hybrid vehicles. The classification rules are as follows: gasoline vehicles with an engine displacement of 1.6 liters or less and a body length of less than 4.7 meters are classified as small cars; those with an engine displacement greater than 1.6 liters and less than or equal to 2.5 liters and a body height greater than 1.7 meters are classified as mid-size SUVs; vehicles with a gross vehicle weight greater than 3.5 tons or more and three or more axles are classified as large trucks; and vehicles that use electric power and have external charging capabilities are classified as new energy hybrid vehicles. The system uses a rule engine to match each vehicle attribute field, outputting a unique tag value, which is stored in the `vehicle_type` field of the user profile table.

[0044] The specific steps for constructing refueling frequency tags include: statistically analyzing historical refueling behavior data, calculating the user's average refueling cycle, standard deviation of single refueling volume, and proportion of high-octane fuel selection, and generating refueling frequency tags, refueling stability tags, and fuel preference tags respectively. The average refueling cycle is obtained by averaging the differences between adjacent refueling timestamps, in days. If the average cycle is less than or equal to five days, the user is marked as a high-frequency user; greater than five days but less than or equal to fifteen days, a user is marked as a medium-frequency user; and greater than fifteen days, a user is marked as a low-frequency user. The standard deviation of single refueling volume reflects the regularity of refueling behavior; a standard deviation less than ten liters indicates a stable refueling pattern, otherwise it indicates a fluctuating pattern. The proportion of high-octane fuel selection refers to the percentage of 95-octane and above gasoline in the total number of refuelings. If it is greater than or equal to 70%, it is marked as a high-standard preference; less than 30%, a low-standard preference; and the remainder indicates a mixed preference. These three sub-tags together constitute a complete description of the refueling behavior dimension.

[0045] The specific steps for constructing geographic active area tags include: clustering geographic location trajectory data to identify a set of active gas stations within a five-kilometer radius of the user and marking them as geographic active area tags. The system delineates a circular area with a five-kilometer radius centered on each user's residence area and filters all gas stations within that area. For commuting routes, a strip-shaped area is formed extending two kilometers on each side of the route, and gas stations are similarly filtered. Finally, all gas stations within the areas are merged, deduplicated, and a user-specific list of active gas stations is generated. This list serves as the content of the geographic active area tag and is stored as a JSON array, containing gas station ID, name, distance, and access frequency.

[0046] The specific steps for constructing promotion sensitivity tags include: modeling response rates on promotional response feedback data, calculating user redemption conversion rates for price discounts, spending-based discounts, and points-based promotions, and generating promotion sensitivity tags. The redemption conversion rate is defined as the number of redemptions divided by the number of push notifications. The system groups and statistically analyzes promotions by type. If the conversion rate for a certain type of activity is greater than or equal to 20%, it is marked as highly sensitive; greater than or equal to 10% but less than 20%, it is marked as moderately sensitive; and less than 10%, it is marked as low sensitive. The final output is a triplet tag, such as (highly sensitive, moderately sensitive, low sensitive), corresponding to the price discount, spending-based discount, and points-based discount activities, respectively.

[0047] In step S3, real-time data on current oil price fluctuations, inventory status at each gas station, promotional activity configurations, and the user's current location are collected. This step ensures that the push strategy remains synchronized with the external operating environment.

[0048] The specific operations for collecting current oil price fluctuation data include: obtaining the benchmark oil price adjustment range from the refined oil price adjustment announcement interface, and calculating the current retail guidance price for each oil product in conjunction with local surcharges. The system polls the official interface every twelve hours, parses the price adjustment announcement in XML format, and extracts the benchmark price change values ​​for gasoline and diesel. Local surcharges include consumption tax, value-added tax, and urban construction tax, the rates of which are published by the tax authorities, and the system has a built-in fixed parameter table. Retail guidance price = benchmark price × (1 + value-added tax rate) + consumption tax + other surcharges. The calculation results are stored in the oil price cache table, valid for twenty-four hours.

[0049] The specific operations for collecting inventory status data at each gas station include: real-time uploading data to the central management system via tank level sensors to determine whether the remaining inventory of each type of fuel is below a safety threshold. A high-precision ultrasonic level gauge is installed at the bottom of each tank, sampling every five minutes. Data is uploaded to an edge gateway via a 4G IoT module and then aggregated into the central inventory database. The system sets a safety threshold for each type of fuel, typically 20% of the total capacity. If the current level is below this threshold, the fuel is marked as "low inventory" and prohibited from participating in promotional activities.

[0050] The specific operations for collecting promotional activity configuration data include: The data is distributed by the marketing management system, including the activity start and end times, applicable fuel types, discount levels, target user groups, and inventory quota limits. The marketing management system provides a RESTful API, and the dynamic strategy matching module periodically pulls the latest activity list. Each activity record includes the activity ID, start time, end time, fuel type, discount per liter, minimum refueling volume, target tag conditions (e.g., only for users with high-preference preferences), total quota, and used quota. The system caches activity data locally and verifies the activity status before each matching.

[0051] The specific operations for collecting the user's current location data include: obtaining the user's current location via mobile network or satellite positioning. This operation is performed immediately before the push notification is triggered to ensure the real-time nature of the location information. The location request is initiated by the push service, which calls the system's location service through the mobile terminal SDK to return latitude, longitude, and positioning accuracy. If positioning fails or the accuracy is greater than 500 meters, the last valid location is used as a replacement.

[0052] In step S4, multi-dimensional user profiles are dynamically matched and analyzed with real-time operational data to generate personalized push strategies. This step is the core logic of this invention.

[0053] The specific operations of dynamic matching analysis include: First, determining whether the user's current location is within the range defined by their geographic active area tag; if so, filtering gas stations within that range that have sufficient inventory and are currently running promotional activities; second, matching applicable oil product promotional activities based on the user's oil product preference tag; third, combining the user's promotion sensitivity tag, prioritizing the promotional type with the highest historical redemption rate; and finally, considering the current oil price fluctuation trend, if oil prices are on an upward trend, pushing out marketing information that locks in a limited-time discounted price in advance.

[0054] Specifically, the system calculates the distance between the user's current location and each gas station in the geographically active area. If the minimum distance is less than five kilometers, the user is considered to be within the active area. Then, all gas stations within the area are traversed, excluding those with low inventory or no promotions. For the remaining stations, their promotions are checked to see if they match the user's fuel preference tags. For example, if the user has a high preference, only promotions for 95 octane and above are matched. Next, based on the promotion sensitivity tag, candidate promotions are sorted in descending order of the user's historical redemption rate for that type of promotion. If multiple promotions of the same type exist, the one with the largest discount is selected. Furthermore, the system monitors the direction of the three most recent fuel price adjustments. If there are two consecutive increases, a "price lock" strategy is activated, emphasizing in the push notification, "Fuel prices are about to increase; lock in your current discounted price now!"

[0055] The specific steps for generating a personalized push strategy include: determining the elements of the push content, including the target gas station name, recommended fuel grade, discounted unit price, discount validity period, expected savings, and additional points rewards; determining the push timing, which is based on predicting the user's next refueling time according to their historical refueling cycle, and triggering the push 24 hours before the predicted time; and determining the push channel, which includes mobile application messages, SMS, and WeChat service account template messages, selecting the optimal channel based on the user's historical channel open rate.

[0056] The push notification content is directly populated by the strategy matching results. Estimated savings = (Original price - Discounted price) × Average refueling volume per user. Bonus points rewards are set according to the activity configuration, typically double points or a fixed point value. Push timing is calculated using a refueling cycle prediction model, which uses exponential smoothing to predict the next refueling time. Channel selection is based on the user's open rate across all channels over the past 30 days, selecting the highest rate. If the open rate of all channels is below 5%, the push is delayed and reassessed.

[0057] In step S5, marketing information is pushed to target users according to a personalized push strategy. The pushed content not only includes text information but also integrates interactive functions to improve conversion efficiency.

[0058] The specific operations for pushing marketing information include: displaying a graphic card containing a map navigation entry on the mobile terminal, the graphic card showing a real-view image of the target gas station, the current number of vehicles in the queue, discount details, and a one-click navigation button; at the same time, after the user starts navigation, the electronic coupon is automatically activated, and a redemption prompt is triggered via Bluetooth beacon or geofencing technology after arriving at the gas station.

[0059] Image and text cards are displayed through the mobile application's message center using rich media templates. Real-world images are obtained from the gas station management system and updated daily. The number of vehicles queuing is obtained in real-time via AI counting from entrance cameras. A one-click navigation button invokes the system map application, with the target gas station as the preset destination. When navigation begins, the client sends a navigation start event to the server, which changes the corresponding coupon status from "pending activation" to "activated." When a user enters the gas station's geofence (within a 100-meter radius) or detects a gas station Bluetooth beacon signal, the client displays a verification prompt, guiding the user to present the coupon code to complete the transaction.

[0060] The above-described method achieves a closed-loop process from raw data collection to final marketing outreach through rigorous data flow and logical judgment. To support the efficient execution of this method, this invention also provides a precise oil sales push system based on user profiles.

[0061] The system includes a multi-source data acquisition module, a dedicated user profile building module, an operational status perception module, a dynamic strategy matching module, and a precise information push module.

[0062] The multi-source data acquisition module is used to acquire vehicle attribute data, historical refueling behavior data, geographic location trajectory data, and promotional response feedback data of target users. This module includes a vehicle information interface unit, a transaction log parsing unit, a location trajectory aggregation unit, and a feedback behavior tracking unit. The vehicle information interface unit connects to the vehicle management database via HTTPS protocol and executes SQL queries to extract structured vehicle attributes. The transaction log parsing unit listens to the Kafka message queue of the gas station's POS system, parses JSON-formatted transaction logs, and converts them into an internal standard format. The location trajectory aggregation unit receives location reports via the mobile terminal SDK, executes the Douglas-Peucker trajectory compression algorithm, and calls the Gaode Map API for road network matching and stop point identification. The feedback behavior tracking unit integrates a data tracking SDK to capture all user interaction events with push notifications and reports them to the behavior analysis platform in real time via the gRPC protocol.

[0063] A dedicated user profile building module is used to construct multi-dimensional user profiles based on data acquired by the multi-source data acquisition module. This module includes a tag generation submodule and a profile storage submodule. The tag generation submodule embeds the Drools rule engine, loads predefined tag calculation rule files, and executes batch tag calculation tasks every day at midnight. The profile storage submodule adopts the Apache Parquet columnar storage format, shards users by unique identifier hash, and is deployed on a distributed file system, supporting the return of a complete profile within fifty milliseconds using the user ID.

[0064] The operational status awareness module collects real-time data on current oil price fluctuations, inventory status at each gas station, promotional activity configurations, and user location. This module includes an oil price monitoring unit, an inventory monitoring unit, an activity configuration synchronization unit, and a real-time location unit. The oil price monitoring unit is configured with a scheduled task to call the API and parse the response every twelve hours. The inventory monitoring unit subscribes to liquid level data topics published by the IoT gateways of each gas station via the MQTT protocol. The activity configuration synchronization unit establishes a long-lived connection with the Webhook interface of the marketing management system to receive real-time notifications of activity changes. The real-time location unit requests the mobile terminal's location service via gRPC to obtain the latest location when the push service is called.

[0065] The dynamic strategy matching module dynamically matches and analyzes multi-dimensional user profiles with operational status data to generate personalized push strategies. This module includes a geofencing determination unit, an inventory-activity filtering unit, a user-activity matching unit, and a strategy ranking unit. The geofencing determination unit uses the Haversine formula to calculate spherical distance to determine if a user is within an active area. The inventory-activity filtering unit jointly queries the inventory status table and the activity configuration table to filter invalid options. The user-activity matching unit performs tag condition matching using a Boolean expression evaluation engine. The strategy ranking unit calculates the expected conversion probability of each candidate strategy using the following formula: ,in, For users, the historical redemption rate of this type of activity, Distance from the user to the gas station (unit: kilometers). This represents the inventory adequacy level (1 indicates sufficient inventory, 0 indicates tight inventory). , , These are weighting coefficients, with values ​​of 0.6, 0.3, and 0.1 respectively. The strategy is based on... Sort in descending order and take the top 1 as the final strategy.

[0066] The precise information push module is used to push marketing information to target users based on personalized push strategies. This module includes a content assembly unit, a timing control unit, and a channel distribution unit. The content assembly unit loads a preset Freemarker template, injects strategy parameters, and generates the final message content. The timing control unit calls the refueling cycle prediction model, which uses the Holt-Winters triple exponential smoothing algorithm, with the formula: ,in, For horizontal components, As a trend component, For seasonal portions, The cycle length is 7 days. To predict the step size, the channel distribution unit queries the user channel preference table, selects the channel with the highest open rate, and sends the message through the corresponding message gateway (such as Alibaba Cloud SMS or Getui Push), while recording the delivery status for subsequent performance evaluation.

[0067] In summary, this invention, through the collaborative design of methods and systems, achieves high-precision and high-conversion marketing information delivery in the oil sales scenario, effectively solving industry-specific pain points.

Claims

1. A method for accurate pushing of oil sales based on user portraits, characterized in that, include: Acquire vehicle attribute data, historical refueling behavior data, geographic location trajectory data, and promotional response feedback data of target users; Based on the vehicle attribute data, historical refueling behavior data, geographic location trajectory data, and promotion response feedback data, a multi-dimensional user profile is constructed, including vehicle type tags, refueling frequency tags, fuel preference tags, geographic active area tags, and promotion sensitivity tags. Real-time data collection includes current oil price fluctuations, gas station inventory status, promotional activity configurations, and user location data. The multi-dimensional user profile is dynamically matched and analyzed with the current oil price fluctuation data, the inventory status data of each gas station, the promotional activity configuration data, and the user's current location data to generate a personalized push strategy for the target user. According to the personalized push strategy, marketing information containing specific oil product discounts, guidance to designated gas station locations, and points redemption incentives is pushed to the target users.

2. The user portrait-based accurate pushing method for oil sales according to claim 1, characterized in that, Obtaining vehicle attribute data for target users specifically includes: extracting vehicle brand, vehicle model, engine displacement, fuel type, and vehicle age from the vehicle registration information database; obtaining the vehicle identification code from the vehicle terminal or fuel card binding information and associating it with the user's identity identifier.

3. The method for precise oil sales targeting based on user profiles according to claim 2, characterized in that, Obtaining historical refueling behavior data specifically includes: extracting the user's refueling timestamp, refueling volume, selected fuel grade, single transaction amount, payment method, whether coupons were used, and whether points were accumulated from the gas station transaction system within a preset time window; the preset time window is the most recent twelve months.

4. The method for precise oil sales push based on user profiles according to claim 3, characterized in that, The acquisition of geographic location trajectory data specifically includes: collecting user location coordinates every thirty minutes through the mobile terminal positioning module, and combining map road network data to identify the user's permanent residence area, commuting route and frequently passed gas stations; the permanent residence area includes residence, workplace and weekend activity area.

5. The method for precise oil sales push based on user profiles according to claim 4, characterized in that, Obtaining promotional response feedback data specifically includes: recording users' opening behavior, clicking behavior, on-site redemption behavior, and subsequent refueling behavior regarding historical push notifications; the on-site redemption behavior refers to the user's behavior of completing a refueling transaction and using the coupon at a designated gas station within the validity period after receiving the coupon push notification.

6. The method for precise oil sales push based on user profiles according to claim 5, characterized in that, The construction of multi-dimensional user profiles specifically includes: structuring and encoding vehicle attribute data to generate vehicle type tags, which are divided into four categories: small cars, mid-size SUVs, large trucks, and new energy hybrid vehicles; statistically analyzing historical refueling behavior data to calculate the average refueling cycle, standard deviation of single refueling volume, and proportion of high-octane fuel selection, generating refueling frequency tags, refueling stability tags, and fuel preference tags; clustering geographic location trajectory data to identify the set of active gas stations within a five-kilometer radius of the user and marking them as geographically active areas; and modeling response rates for promotional response feedback data to calculate the redemption conversion rate of users for price discounts, full reductions, and points bonus promotions, generating promotion sensitivity tags.

7. The method for precise oil sales push based on user profiles according to claim 6, characterized in that, The real-time collection of current oil price fluctuation data specifically includes: obtaining the benchmark oil price adjustment range from the refined oil price adjustment announcement interface, and calculating the current retail guidance price of each oil product in combination with local surcharges; the inventory status data of each gas station is uploaded to the central management system in real time through oil tank level sensors to determine whether the remaining inventory of each oil product is below the safety threshold; the promotional activity configuration data is issued by the marketing management system, including the start and end time of the activity, applicable oil products, discount level, target user group, and inventory quota limit.

8. The method for precise oil sales push based on user profiles according to claim 7, characterized in that, The multi-dimensional user profile is dynamically matched and analyzed with the current oil price fluctuation data, gas station inventory status data, promotional activity configuration data, and user current location data to generate a personalized push strategy for the target user. This includes: determining whether the user's current location is within the range defined by their geographic active area tag; if so, selecting gas stations within that range with sufficient inventory and currently running promotional activities; matching applicable oil product promotional activities based on the user's oil product preference tags; prioritizing promotional types with the highest historical redemption rates based on the user's promotional sensitivity tags; and considering the current oil price fluctuation trend, if oil prices are on an upward trend, pushing out marketing information about locking in a limited-time discount price in advance.

9. A petroleum sale precision pushing system based on user portrait, characterized in that, include: The multi-source data acquisition module is used to acquire vehicle attribute data, historical refueling behavior data, geographical location trajectory data, and promotional response feedback data of target users; A dedicated user profile building module is used to build a multi-dimensional user profile based on the data obtained by the multi-source data acquisition module, including vehicle type tags, refueling frequency tags, fuel preference tags, geographically active area tags, and promotion sensitivity tags. The operational status awareness module is used to collect real-time data on current oil price fluctuations, inventory status of each gas station, promotional activity configuration, and the user's current location. The dynamic strategy matching module is used to dynamically match and analyze the multi-dimensional user profile with the data collected by the operation status perception module to generate a personalized push strategy for the target user. The precise information push module is used to push marketing information containing specific oil product discounts, designated gas station location guidance, and points redemption incentives to the target user according to the personalized push strategy.

10. The user portrait-based accurate oil selling pushing system according to claim 9, characterized in that, The multi-source data acquisition module includes a vehicle information interface unit, a transaction log parsing unit, a location trajectory aggregation unit, and a feedback behavior tracking unit. The vehicle information interface unit connects to the vehicle management database to extract structured vehicle attributes. The transaction log parsing unit interfaces with the gas station POS system to parse standardized transaction records. The location trajectory aggregation unit collects latitude and longitude sequences through the mobile terminal SDK and performs trajectory compression and stop point identification. The feedback behavior tracking unit listens to the user's operation event stream of push messages and records the complete behavior chain.

Images