Driver intention construction method and device, equipment and storage medium

Abstract

The application discloses a driver intention construction method and device, equipment and a storage medium, and belongs to the technical field of data processing. The method comprises the following steps: obtaining an internal event object containing multiple types of behavior events, constructing intentions in three branches, generating a vehicle preference intention according to a first driver list behavior event object and combining offline vehicle preference data; generating a route intention according to a second driver list behavior event object and a driver order event object; generating a find goods intention according to a third driver list behavior event object and an end intelligent event object; and outputting multiple intention calculation results in parallel when multiple types of events coexist. The application is based on real-time behavior triggering calculation of a driver, overcomes the defect of offline label updating lag, filters invalid repeated operations, and reduces system computing power consumption.

Description

Technical Field

[0001] This application belongs to the field of data processing technology, and in particular relates to a method, apparatus, device and storage medium for constructing driver intent. Background Technology

[0002] With the rapid development of online freight platforms, these platforms need to push relevant information to drivers based on their intentions.

[0003] Currently, the industry generally uses offline tagging solutions to characterize drivers' transportation preferences. Existing offline tags rely on big data platforms to collect drivers' historical data in batches, and can only update drivers' intentions once a day. This results in a significant data update lag and makes it impossible to capture real-time changes in drivers' behavior. Summary of the Invention

[0004] The purpose of this application is to provide a method, apparatus, device, and storage medium for constructing driver intent. This method constructs intent by acquiring various internal event objects generated from the driver's behavior. When a new internal event object is acquired, its difference is calculated compared with the most recently stored internal event object of the same type. If the difference is greater than or equal to a preset difference threshold, intent is constructed based on the new internal event object of the same type to obtain a new intent calculation result. This solves the problem of delayed driver intent updates and enables real-time construction of driver intent.

[0005] To achieve the above objectives, the solution proposed in this application is: In a first aspect, embodiments of this application provide a method for constructing driver intent, including: Obtain internal event objects, which include at least one of driver list behavior event objects, driver order event objects, and terminal intelligence event objects. The driver list behavior event objects include at least one of first driver list behavior event objects, second driver list behavior event objects, and third driver list behavior event objects. The intent is constructed based on the internal event object to obtain the intent calculation result; Intents are constructed based on internal event objects to obtain intent calculation results, including: When the internal event object includes the first driver list behavior event object, obtain the pre-statistical offline vehicle preference data; The vehicle preference intent is constructed based on the behavioral event objects in the first driver list and the pre-statistical offline vehicle preference data. When the internal event object includes the second driver list behavior event object and the driver order event object, the intent is constructed based on the second driver list behavior event object and the driver order event object to obtain the route intent; When the internal event object includes the third driver list behavior event object and the terminal intelligence event object, the intent is constructed based on the third driver list behavior event object and the terminal intelligence event object to obtain the cargo search intent. When there are multiple internal event objects and the conditions for constructing multiple intents are met, intents are constructed based on the multiple internal event objects, and multiple intent calculation results are obtained respectively. The internal event objects used to construct the intent are stored according to their type; When a new internal event object of the same type is obtained, the most recently stored internal event object of the same type is retrieved. Calculate the difference between the new internal event object of the same type and the most recently stored internal event object of the same type; If the difference is greater than or equal to the preset difference threshold, the intent is constructed based on the new internal event object of the same type, and a new intent calculation result is obtained.

[0006] The method described in the embodiments of this application may also have the following additional technical features: Furthermore, before retrieving the internal event object, the driver intent construction method also includes: Obtain a first original message, which includes at least one of driver list behavior message, driver order message and terminal intelligent message, and the driver list behavior message includes at least one of first driver list behavior message, second driver list behavior message and third driver list behavior message; The first raw message is transformed to obtain an internal event object; The first raw message is transformed to obtain an internal event object, including: When the first original message includes the first driver list behavior message, the first driver list behavior message is transformed to obtain the first driver list behavior event object; When the first original message includes the second driver list behavior message and the driver order message, the second driver list behavior message and the driver order message are transformed to obtain the second driver list behavior event object and the driver order event object. When the first original message includes the third driver list behavior message and the terminal intelligence message, the third driver list behavior message and the terminal intelligence message are transformed to obtain the third driver list behavior event object and the terminal intelligence event object. When there are multiple first original messages and the conditions for multiple transformation processes are met, transformation processes are performed based on the multiple first original messages to obtain multiple internal event objects.

[0007] Furthermore, the first driver list behavior event objects include vehicle data corresponding to drivers' telephone behavior events and click behavior events within a first preset time period; Based on the behavioral event objects in the first driver list and pre-statistical offline vehicle preference data, intents are constructed to obtain vehicle preference intents, including: The first real-time intent is calculated based on the vehicle data corresponding to the driver's telephone behavior events and click behavior events within the first preset time period. Use pre-statistical offline vehicle preference data as the first offline intent; The vehicle preference intent is obtained by weighting the first real-time intent and the first offline intent.

[0008] Furthermore, internal event objects also include user center event objects, which include vehicle data registered by drivers; When the pre-statistical offline vehicle preference data is empty, the driver intent construction method also includes: Get the preset vehicle configuration; The driver's registered vehicle data is matched with the preset vehicle configuration to obtain the second offline intent; The vehicle preference intent is obtained by weighting the first real-time intent and the second offline intent.

[0009] Furthermore, the second driver list behavior event objects include line data corresponding to the driver's telephone behavior events and line data corresponding to the click behavior events within the second preset time period; The driver order event object includes the route data corresponding to the driver's completed order action; Intents are constructed based on the second driver list behavior event objects and driver order event objects to obtain the route intent, including: Based on the line data corresponding to the telephone behavior event, obtain the latitude and longitude of the first origin and the first destination. Based on the route data corresponding to the click event, obtain the latitude and longitude of the second departure point and the second destination. Based on the route data corresponding to the driver's completed order, obtain the latitude and longitude of the third origin and the third destination; Statistical calculations were performed on the latitude and longitude of the first, second, and third departure points to obtain the latitude and longitude of the departure point with the highest weight. Statistical calculations were performed on the latitude and longitude of the first destination, the second destination, and the third destination to obtain the latitude and longitude of the destination with the highest weight. The route intention is derived from the latitude and longitude of the origin and the destination, which have the highest weights.

[0010] Furthermore, when the origin latitude and longitude with the highest weight and / or the destination latitude and longitude with the highest weight are empty, the driver intent construction method also includes: Obtain the second original message, which includes user center messages and map navigation messages; The user center messages and map navigation messages are converted and processed to obtain user center event objects and map navigation event objects; Among them, the user center event objects include the vehicle length data registered by the driver; the map navigation event objects include the latitude and longitude data of the driver's current location; Acquire offline feature data of big data. The offline feature data of big data includes multiple vehicle length groups. Each vehicle length group stores a set of destination latitude and longitude coordinates for each city. The driver's current city is determined based on the driver's current location latitude and longitude data; Among multiple vehicle length groups, identify the target vehicle length group that matches the vehicle length data registered by the driver; Within the target vehicle length group, retrieve the set of latitude and longitude coordinates of the current city's destination; Using the driver's current location latitude and longitude data as the fourth departure point latitude and longitude, and using at least one destination latitude and longitude from the current city's destination latitude and longitude set as the fourth destination latitude and longitude, at least one route intention is obtained based on the fourth departure point latitude and longitude and the fourth destination latitude and longitude.

[0011] Furthermore, the third driver list behavior event objects include behavior data corresponding to driver search behavior events, behavior data corresponding to click behavior events, and behavior data corresponding to page refresh behavior events; End-to-end intelligent event objects include behavioral data corresponding to end-to-end intelligent trigger events; The intent to find goods includes the intent to leave the machine running, the intent to urgently find goods, and the intent to normally find goods. Based on the third-party driver list behavior event objects and the terminal intelligent event objects, the intent to find goods is constructed, including: If the behavior data corresponding to the page refresh event and the behavior data corresponding to the terminal intelligent trigger event are empty, and if the behavior data corresponding to the search event and the behavior data corresponding to the click event are both empty within the third preset time period, then the intention to find goods is determined to be the intention to leave the machine idle. The third driver list behavior event object also includes behavior data corresponding to telephone behavior events; Weighted behavior data is obtained by weighting the behavior data corresponding to search behavior events, click behavior events, and telephone behavior events. If the weighted behavior data is greater than or equal to the preset behavior threshold, the intention to find goods is determined to be an urgent intention to find goods. If the weighted behavior data is less than the preset behavior threshold, the search intent is determined to be a normal search intent.

[0012] Furthermore, before constructing the intent based on the internal event object, the driver intent construction method also includes: Frequency control is applied to internal event objects, with different preset frequency control thresholds for different internal event objects; Intents are constructed based on internal event objects to obtain intent calculation results, which also include: Intent is constructed based on the internal event object after frequency control processing, and the intent calculation result is obtained.

[0013] Secondly, embodiments of this application provide a driver intent construction apparatus, comprising: The data acquisition module is used to acquire internal event objects, which include at least one of driver list behavior event objects, driver order event objects, and terminal intelligent event objects. The driver list behavior event objects include at least one of first driver list behavior event objects, second driver list behavior event objects, and third driver list behavior event objects. The first construction module is used to construct intents based on internal event objects and obtain intent calculation results. Intents are constructed based on internal event objects to obtain intent calculation results, including: When the internal event object includes the first driver list behavior event object, obtain the pre-statistical offline vehicle preference data; The vehicle preference intent is constructed based on the behavioral event objects in the first driver list and the pre-statistical offline vehicle preference data. When the internal event object includes the second driver list behavior event object and the driver order event object, the intent is constructed based on the second driver list behavior event object and the driver order event object to obtain the route intent; When the internal event object includes the third driver list behavior event object and the terminal intelligence event object, the intent is constructed based on the third driver list behavior event object and the terminal intelligence event object to obtain the cargo search intent. When there are multiple internal event objects and the conditions for constructing multiple intents are met, intents are constructed based on the multiple internal event objects, and multiple intent calculation results are obtained respectively. The data storage module is used to store the internal event objects intended to be constructed, based on the type of the internal event object. The data retrieval module is used to retrieve the most recently stored internal event object of the same type when a new internal event object of the same type is obtained. The difference calculation module is used to calculate the difference between new internal event objects of the same type and the most recently stored internal event objects of the same type. The second construction module is used to construct intents based on new internal event objects of the same type when the difference is greater than or equal to a preset difference threshold, and obtain new intent calculation results.

[0014] Thirdly, embodiments of this application provide a driver intent construction device, which includes a processor and a memory. The memory stores a computer program, which is loaded and executed by the processor to implement a driver intent construction method provided in the first aspect of embodiments of this application.

[0015] Fourthly, embodiments of this application provide a computer-readable storage medium storing a computer program, which, when executed by a processor, is used to implement a driver intent construction method provided in the first aspect of embodiments of this application.

[0016] The driver intent construction method provided in this application has the following advantages compared with the prior art: This application embodiment uses multiple types of internal event objects, including first, second, and third driver list behavior event objects, driver order event objects, and terminal intelligent event objects, as the basis for calculation. It uses the internal event objects corresponding to the real-time behavior of drivers to trigger intent construction in real time. Unlike the traditional offline batch statistics method, it can respond to the dynamic behavior of drivers in a timely manner.

[0017] In this application embodiment, corresponding logic is executed for different types of internal event objects. A vehicle preference intent is generated based on the first driver list behavior event object and offline vehicle preference data. A route intent is generated based on the second driver list behavior event object and driver order event object. A freight search intent is generated based on the third driver list behavior event object and terminal intelligent event object, thereby realizing the synchronous recognition of multi-dimensional intents of vehicles, routes, and freight search.

[0018] In this embodiment of the application, when there are multiple internal event objects and multiple intent construction conditions are met, multiple intent calculation results can be processed and output separately, fully covering the driver's multiple real-time needs.

[0019] This application embodiment also classifies and stores internal event objects of various types. When a new internal event object of the same type is obtained, the historical event object of the same type is retrieved to complete the difference calculation. The intent construction is re-executed only when the difference reaches a preset difference threshold to obtain a new intent calculation result. On the one hand, this realizes the dynamic capture of the driver's real-time behavioral intent, solves the defects of offline feature update lag and inability to reflect the driver's daily behavioral changes in a timely manner. On the other hand, it filters out repeated calculations without obvious behavioral changes, reduces invalid calculations, and saves system computing resources. Attached Figure Description

[0020] Figure 1 A structural block diagram of the driver intent construction system according to an embodiment of this application is shown; Figure 2 A flowchart illustrating the driver intent construction method according to an embodiment of this application is shown; Figure 3 A structural block diagram of a driver intent construction device according to an embodiment of this application is shown; Figure 4 A structural block diagram of a computer device according to an embodiment of this application is shown. Detailed Implementation

[0021] To make the above-mentioned objectives, features, and advantages of this application more apparent and understandable, the specific embodiments of this application will be described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are for illustrative purposes only and are not intended to limit the scope of this application. Furthermore, it should be noted that, for ease of description, only the parts relevant to this application are shown in the accompanying drawings, not the entire structure. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without inventive effort are within the scope of protection of this application.

[0022] The terms “comprising” and “having”, and any variations thereof, used in this application are intended to cover a non-exclusive inclusion. For example, a process, method, system, product, or apparatus that includes a series of steps or units is not limited to the steps or units listed, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to such process, method, product, or apparatus.

[0023] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may include at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0024] Please refer to Figure 1 The driver intent construction method of this application is applied to a driver intent construction system, which is divided into five logical layers from upstream to downstream: message entry layer 101, traffic splitting and processing layer 102, frequency control filtering layer 103, intent construction layer 104, and output return layer 105. Message Entry Layer 101: Accesses multi-source heterogeneous raw messages. All types of messages are uniformly accessed through message queues (such as RocketMQ and Kafka). Raw messages are divided into two categories: first raw messages and second raw messages.

[0025] Triage layer 102: Performs standardized parsing and transformation on the original message, generates internal event objects with uniform format, and triages them to the computational intent link or the non-computational intent link according to the event type.

[0026] Frequency control filtering layer 103: Executes multi-level frequency control strategies for events in the computation intent link. By using graded cooling times and special scenario frequency hopping rules, it avoids repeated calculations by the same driver in a short period of time, thereby reducing system overhead.

[0027] Intent Construction Layer 104: Determines whether recalculation is needed through a duplicate verification mechanism, assembles feature subsets as needed based on intent type, and calls corresponding strategies to generate structured intent results, including vehicle preference intent, route intent, and freight search intent. Figure 3 kind.

[0028] Output backflow layer 105: Writes the intent result to the distributed cache storage, and sends the intent change message to the downstream business system; new behavior messages generated downstream flow back to message entry layer 101, forming a self-optimizing closed loop.

[0029] Please refer to Figure 2 This application provides a method for constructing driver intent, which is applied to... Figure 1 The driver intends to build a system, and the method includes the following steps: Step 201: Obtain the internal event object.

[0030] The first raw message is obtained and standardized to obtain internal event objects in a unified format. Each internal event object includes at least three basic fields: a unique driver identifier (driver ID), a message type identifier, and an event timestamp, as well as extended fields corresponding to the business scenario.

[0031] The first original message includes at least one of driver list behavior messages, driver order messages, and terminal intelligent messages; the driver list behavior message includes at least one of the first driver list behavior message, the second driver list behavior message, and the third driver list behavior message. The conversion rules for each type of original message are as follows: Driver order messages include order ID, driver ID, order status type (completed / cancelled / arrived for unloading), cargo type, order origin latitude and longitude, and order destination latitude and longitude. After conversion and processing, driver order event objects are generated, which are used to extract features such as the driver's historical completed routes and cargo preferences.

[0032] First driver list behavior messages include driver ID, telephone behavior records within a first preset time period, click behavior records, and corresponding vehicle specification data; after conversion and processing, first driver list behavior event objects are generated, which are used to extract real-time vehicle length and model preference features.

[0033] Second driver list behavior messages include driver ID, telephone behavior records within a second preset time period, click behavior records, and corresponding route data; after conversion and processing, a second driver list behavior event object is generated, which is used to extract real-time route preference features.

[0034] Third driver list behavior messages: including driver ID, search behavior records, click behavior records, page refresh behavior records, and telephone behavior records within a third preset time period; after conversion and processing, a third driver list behavior event object is generated, which is used to extract features related to the intention to find goods.

[0035] End-to-end intelligent messages include driver ID, trigger scenario type, prediction confidence level, and prediction tag; after transformation and processing, they generate end-to-end intelligent event objects, which are used to predict changes in intent in advance and trigger pre-computation.

[0036] When there are multiple first original messages and the conditions for multiple transformation processes are met, the transformation processes are executed in parallel based on the multiple first original messages, resulting in multiple internal event objects of corresponding types. Therefore, the internal event objects include at least one of driver list behavior event objects, driver order event objects, and terminal intelligent event objects; wherein the driver list behavior event objects include at least one of first driver list behavior event objects, second driver list behavior event objects, and third driver list behavior event objects.

[0037] Step 202: Construct intent based on internal event objects to obtain intent calculation results.

[0038] This step is executed on the intent computation chain. Before intent construction, frequency control processing can be optionally performed on internal event objects. Different types of internal event objects correspond to different preset frequency control thresholds. Only events that pass the frequency control verification will enter the subsequent intent construction process.

[0039] When there are multiple internal event objects and the conditions for constructing multiple intents are met, the corresponding construction logic is executed according to the multiple internal event objects respectively, and the intent calculation results of multiple dimensions are obtained in parallel.

[0040] The three core intents are constructed as follows: Vehicle preference intent construction: When the internal event object includes the first driver list behavior event object, pre-statistical offline vehicle preference data is obtained. Intent is constructed based on the first driver list behavior event object and the pre-statistical offline vehicle preference data to obtain the vehicle preference intent. The vehicle preference intent is used to determine the required vehicle length (e.g., 4.2 meters, 6.8 meters, 9.6 meters, etc.) and vehicle type (e.g., high-sided, flatbed, van, etc.) for the preferred cargo, and is a core filtering condition in vehicle-cargo matching.

[0041] Specifically, calculate the first real-time intent: The first driver list includes vehicle data corresponding to driver phone call events and click events within a first preset time period. In this embodiment, the first preset time period is preferably 24 hours.

[0042] The number of phone calls and clicks for each vehicle length and model are counted separately, and a real-time preference score is calculated according to preset weights.

[0043] in, The number of calls corresponding to the vehicle length / model This represents the number of clicks for the corresponding vehicle length / model. Weighting of telephone behavior, This is used to assign weight to click behavior. In this embodiment... Preferably 3, The preferred value is 1.

[0044] In one optional implementation, when a driver hits a grayscale experimental group, vehicle lengths and vehicle types with fewer than a preset threshold (e.g., 3 times) are filtered out to reduce low-frequency noise interference. The grayscale experimental group is determined using a driver ID hashing method to progressively verify the effectiveness of the filtering strategy.

[0045] Determine offline intent: The pre-statistical offline vehicle preference data is used as the first offline intent, denoted as... This data is calculated daily by the big data platform based on drivers' historical orders and long-term browsing behavior across the entire platform. It is stored in an offline data warehouse and represents drivers' long-term stable vehicle size preferences.

[0046] When the pre-statistical offline vehicle preference data is empty, execute the cold start fallback logic: Obtain the vehicle data registered by the driver in the user center event object, match it with the preset vehicle configuration table, and obtain the second offline intent.

[0047] The preset vehicle configuration is a static vehicle compatibility matching rule table pre-entered in the platform backend and stored in the global configuration center. It records the adaptation and connection relationships between different vehicle lengths and models and is used for cold start calculations for drivers without historical behavior.

[0048] The total preference score is obtained by weighting the first real-time intent and the first offline intent (or the second offline intent):

[0049] in, The real-time feature weight coefficient is preferably 2 in this embodiment to highlight the preference orientation of recent behavior. After fusion, the data is sorted in descending order of total preference score, and a truncation process is performed: candidates with scores greater than a preset score threshold are retained, and a maximum of [number missing] are output. The combination of vehicle length and model is used as the final vehicle preference intention result. In the fallback calculation scenario, the boundary values ​​of vehicle length are downweighted to reduce the probability of preference misjudgment.

[0050] Route intent construction: When the internal event object includes the second driver list behavior event object and the driver order event object, the intent is constructed based on the second driver list behavior event object and the driver order event object to obtain the route intent.

[0051] Route intent is used to determine the driver's desired transportation route (origin and destination), and is the core basis for accurate cargo delivery.

[0052] This embodiment sets up two calculation branches: a precise route and a regular route. The branch is automatically selected for execution based on the amount of behavioral data, and a fallback mechanism is provided.

[0053] Specifically, the second driver list behavior event objects include the route data corresponding to the driver's telephone behavior events and click behavior events within the second preset time period; the driver order event objects include the route data corresponding to the driver's completed order behavior.

[0054] Based on the route data corresponding to the telephone behavior event, extract the latitude and longitude of the first departure point and the first destination; based on the route data corresponding to the click behavior event, extract the latitude and longitude of the second departure point and the second destination; based on the route data corresponding to the driver's completed order behavior, extract the latitude and longitude of the third departure point and the third destination.

[0055] Statistical calculations are performed on the latitude and longitude of the first, second, and third departure points to obtain the latitude and longitude of the departure point with the highest weight; statistical calculations are also performed on the latitude and longitude of the first, second, and third destination points to obtain the latitude and longitude of the destination with the highest weight; the route intention is obtained based on the latitude and longitude of the departure point with the highest weight and the latitude and longitude of the destination with the highest weight.

[0056] When the number of valid behavior samples within the second preset time period (preferably 1 hour in this embodiment) is greater than or equal to the preset sample threshold, precise route calculation is performed: The origin latitude and longitude coordinates of all behaviors are aggregated by weighted voting. Telephone behavior has the highest weight, followed by click behavior, and transaction behavior has the highest weight. The latitude and longitude coordinates with the highest cumulative weight are selected as the optimal origin latitude and longitude.

[0057] Destination clustering: The KD tree spatial clustering algorithm is used to spatially cluster the latitude and longitude of all behaviors. Destinations with Euclidean distance within a preset clustering threshold (preferably 30km in this embodiment) are grouped into the same cluster. The weighted center coordinates of samples in each cluster are calculated, and the center coordinates of the cluster with the highest total sample weight are selected as the optimal destination latitude and longitude.

[0058] The optimal origin latitude and longitude are combined with the optimal destination latitude and longitude to generate a precise route intention and output the corresponding confidence score.

[0059] When the number of valid behavior samples within the second preset time period is less than the preset sample threshold, perform regular route calculation and output a set of multiple candidate routes: Candidate origin cities: The origin city ID corresponding to each behavior is rolled up to the city dimension. For the same city, the original score of the origin city is calculated by weighting the number of behaviors and the behavior type. Sort by score in descending order and take the top score. One city as a candidate set of departure points; Destination Candidates: The original score of the destination city is calculated by weighting the number of times and the type of behavior for each behavior. Sort by score in descending order and take the top score. 10 cities were selected as a candidate destination.

[0060] Perform a Cartesian product on the candidate sets of origin and destination, and calculate the overall score for each combined route:

[0061]

[0062]

[0063]

[0064] in, The original score for the city of origin. The original score for the destination city; The origin ranking decay index, The origin ranking number (starting from 0); Destination ranking decay index, The destination is ranked by its index (starting from 0); , and These are the origin score weight, destination score weight, and product term weight, respectively.

[0065] In this embodiment, The preferred value is 0.97. The preferred value is 0.95. The preferred value is 0.6. The preferred value is 0.4. The preferred value is 0.05.

[0066] All lines Divide by the maximum value and normalize to the [0,1] interval; after sorting by score in descending order, retain lines with scores greater than a preset threshold, and retain at most [number missing] lines. Each item serves as the final route intention result set.

[0067] When the origin latitude and longitude with the highest weight and / or the destination latitude and longitude with the highest weight are empty, and neither the precise route nor the regular route can output valid results, a fallback calculation is performed: Obtain the second raw message, which includes user center message and map navigation message; perform conversion processing on the user center message and map navigation message to obtain user center event object and map navigation event object; wherein the user center event object includes the vehicle length data registered by the driver, and the map navigation event object includes the latitude and longitude data of the driver's current location.

[0068] Obtain offline feature data from big data. The offline feature data from big data includes multiple vehicle length groups, and each vehicle length group stores a set of destination latitude and longitude coordinates for each city.

[0069] This data is generated offline by a big data platform, which groups vehicles by length and uses cities as the dimension to collect the high-frequency destination latitude and longitude coordinates of each vehicle type in each city. The data is then synchronized to the online cache daily. The data uses a two-dimensional hierarchical storage structure: the first layer is the primary key for vehicle length grouping (4.2 meters, 6.8 meters, 9.6 meters, etc.), and the second layer is the sub-dimension for cities. Each combination has a pre-stored set of historical high-frequency unloading destination latitude and longitude coordinates for the corresponding size of truck in that city.

[0070] The driver's current city is determined based on the driver's current location latitude and longitude data; among multiple vehicle length groups, the target vehicle length group that matches the driver's registered vehicle length data is identified; within the target vehicle length group, the destination latitude and longitude set of the current city is retrieved.

[0071] Using the driver's current location latitude and longitude data as the fourth departure point latitude and longitude, and using at least one destination latitude and longitude from the current city's destination latitude and longitude set as the fourth destination latitude and longitude, at least one route intention is obtained based on the fourth departure point latitude and longitude and the fourth destination latitude and longitude.

[0072] Building the search intent: When the internal event object includes the third driver list behavior event object and the terminal intelligence event object, the intent is constructed based on the third driver list behavior event object and the terminal intelligence event object to obtain the cargo search intent.

[0073] The intent to find goods includes the intent to leave the machine running, the intent to urgently find goods, and the intent to normally find goods. Figure 3 kind.

[0074] The third driver list behavior event object includes behavior data corresponding to driver search behavior events, click behavior events, page refresh behavior events, and phone call behavior events; the terminal intelligence event object includes behavior data corresponding to terminal intelligence trigger events.

[0075] The specific determination steps are as follows: Determining the intention to AFK: If the behavior data corresponding to the page refresh event and the behavior data corresponding to the terminal smart trigger event are empty, determine whether the behavior data corresponding to the search event and the behavior data corresponding to the click event are both empty within the third preset time period.

[0076] In this embodiment, the third preset duration is preferably 10 minutes.

[0077] If both are empty, the intention to find goods is determined to be an idle intention, and the calculation ends; otherwise, the urgent intention to find goods is determined.

[0078] Determining the intent to urgently find goods: We calculate the weighted behavioral data by weighting the behavioral data corresponding to search behavior events, click behavior events, and telephone behavior events:

[0079] in, , and These are the weighting coefficients for search, click, and phone call behaviors, respectively.

[0080] In this embodiment, Preferably 1, Preferably 3, The optimal value is 5, which reflects the positive correlation between the intensity of the behavior and the urgency of the intention.

[0081] If the weighted behavior data is greater than or equal to the preset behavior threshold, the intention to find goods is determined to be an urgent intention to find goods; if the weighted behavior data is less than the preset behavior threshold, the intention to find goods is determined to be a normal intention to find goods.

[0082] In some embodiments, frequency control processing can be optionally performed on internal event objects before intent construction. Different types of internal event objects correspond to different preset frequency control thresholds. Only events that pass the frequency control verification enter the subsequent intent construction process. This frequency control processing specifically includes a three-level mechanism: Level 1: MQ consumption interval frequency control sets differentiated message consumption cooldown times for different types of internal event objects. Events of the same type that arrive repeatedly within the cooldown time are discarded and do not proceed to the next step.

[0083] In one alternative implementation, the cooldown time for page refresh events is 1 minute, while there is no cooldown limit for order status change events and terminal intelligence events.

[0084] Level 2: Intent generation interval frequency control sets the minimum interval for global intent generation. The same driver can generate a complete intent result at most once within the preset global interval. Events arriving within the interval directly return the most recently cached intent result without triggering recalculation.

[0085] In this embodiment, the minimum global interval is preferably 300 seconds.

[0086] Level 3: Special Scenario Frequency Hopping Rules When an internal event object is associated with location-related data, and the spatial distance between the current location latitude and longitude and the previous location latitude and longitude is greater than a preset distance threshold, the above two levels of frequency control restrictions are skipped, and the process directly proceeds to the next step.

[0087] The spatial distance is calculated using the Haversine formula, and in this embodiment, the preset distance threshold is preferably 10km.

[0088] Step 203: Store the internal event object used for the intent construction according to the type of the internal event object.

[0089] The system maintains the latest event snapshot and intent snapshot for each driver through a distributed cache (such as Redis).

[0090] Once the intent construction is completed for the first time, the internal event object used for intent construction and the generated intent calculation result are stored according to type, and used as the comparison benchmark for the next event.

[0091] If this is the first time an internal event object of this type from the driver has been received, and there are no historical snapshots to compare, then this step of storage will be performed directly after the intent construction is completed.

[0092] Step 204: When a new internal event object of the same type is obtained, retrieve the most recently stored internal event object of the same type.

[0093] When a new internal event object of the same type is received and passes the frequency control verification, the most recently stored internal event object of the driver's type and the corresponding intent snapshot are retrieved from the distributed cache for subsequent difference verification.

[0094] Step 205: Calculate the difference between the new internal event object of the same type and the most recently stored internal event object of the same type.

[0095] Core behavioral features are extracted from new internal event objects and historically stored internal event objects, including real-time click features, call features, transaction features, and search features. The latest preset number of data (preferably 5 in this embodiment) for each feature dimension is compared.

[0096] The degree of difference is quantified by the proportion of feature differences: Difference = Number of difference feature items / Total number of feature items Among them, inconsistent content of a single feature data (such as corresponding vehicle length, route latitude and longitude, behavior type) is judged as a difference item.

[0097] If the difference is less than the preset difference threshold, it is determined that the driver's intention has not changed substantially, and the cached historical intention calculation result is returned directly without triggering recalculation; if the difference is greater than or equal to the preset difference threshold, the recalculation process is initiated.

[0098] Step 206: If the difference is greater than or equal to the preset difference threshold, construct the intent based on the new internal event object of the same type to obtain the new intent calculation result.

[0099] When the event difference in the duplicate verification meets the recalculation condition, the intent construction is re-executed based on the new internal event object of the same type according to the construction logic in step 102, the updated intent calculation result is generated, and the event snapshot and intent snapshot in the cache are updated synchronously.

[0100] Regarding the intent calculation result, this application embodiment also performs two operations simultaneously: Write to the Redis forward cache, overwriting the original intent snapshot, for subsequent queries and repeated validation calls; Generate intent change MQ messages and push them to downstream business systems (user center, terminal intelligence module, order service, recommendation system, etc.) to support personalized recommendations, product push priority adjustment and other business operations.

[0101] After the downstream system consumes a message with a changed intent, it executes business logic based on the latest intent and generates new driver behavior messages (such as end-to-end intelligent trigger messages, driver page refresh messages, order status change messages, etc.).

[0102] The new message is then accessed again at message entry layer 101 and enters the next round of intent calculation process, forming a data-driven self-optimizing closed loop.

[0103] The scope of protection of this application is not limited to the above embodiments; the following variant solutions also fall within the scope of protection of this application: Storage media replacement: Storage media intended for snapshots and frequency control counting can be replaced with local cache (Caffeine, Guava Cache), Memcached, or other distributed caching systems.

[0104] Computational model alternatives: In addition to local rule strategies, machine learning models (XGBoost, Random Forest, Logistic Regression) or deep learning models can be used as alternatives, or a hybrid architecture of rule engines and lightweight models can be adopted.

[0105] Feature source expansion: Onboard terminal OBD data, fuel consumption data, road condition information, weather information, ETC passage records, etc. can be added as supplementary feature sources to further improve the accuracy of intent.

[0106] The traffic splitting logic has been extended: computation links can be dynamically allocated based on message priority, with high-priority messages entering the computation link first and low-priority messages being queued or downgraded for processing, rather than simply using static traffic splitting rules.

[0107] Parameter configurability: All parameters such as frequency control threshold, difference threshold, weight coefficient, time window, and truncation quantity can be dynamically adjusted through the global configuration center without restarting the service, adapting to different business scenario requirements.

[0108] It should be noted that the preset threshold, weight coefficient, time window and other parameters in all the above embodiments can be determined by experimental optimization based on the actual business effect, and this application does not limit them to a single parameter.

[0109] Please refer to Figure 3 This application provides a driver intent construction device, which includes: The data acquisition module 301 is used to acquire internal event objects, which include at least one of driver list behavior event objects, driver order event objects, and terminal intelligent event objects. The driver list behavior event objects include at least one of first driver list behavior event objects, second driver list behavior event objects, and third driver list behavior event objects.

[0110] The first building module 302 is used to build intents based on internal event objects and obtain intent calculation results.

[0111] Data storage module 303 is used to store internal event objects intended for construction according to the type of the internal event object.

[0112] The data retrieval module 304 is used to retrieve the most recently stored internal event object of the same type when a new internal event object of the same type is obtained.

[0113] The difference calculation module 305 is used to calculate the difference between a new internal event object of the same type and the most recently stored internal event object of the same type.

[0114] The second construction module 306 is used to construct intents based on new internal event objects of the same type when the difference is greater than or equal to a preset difference threshold, and obtain new intent calculation results.

[0115] The driver intent construction device in this application embodiment can be a computer device or a component within the computer device, such as an integrated circuit or a chip. The computer device can be a terminal or other devices besides a terminal. For example, the computer device can be a mobile phone, tablet computer, laptop computer, PDA, in-vehicle computer device, mobile internet device (MID), ultra-mobile personal computer (UMPC), netbook, or personal digital assistant (PDA), etc., and can also be a server, network attached storage (NAS), personal computer (PC), etc. This application embodiment does not specifically limit the device.

[0116] The driver intent construction device provided in this application embodiment can achieve... Figure 2 The various processes implemented in the driver intent construction method embodiment will not be described again here to avoid repetition.

[0117] This application also provides a computer device, please refer to... Figure 4 The computer device includes a processor 401 and a memory 402. The memory 402 stores a program or instruction that can run on the processor 401. When the program or instruction is executed by the processor 401, it implements the various steps of the driver intent construction method described above and can achieve the same technical effect. To avoid repetition, it will not be described again here.

[0118] It should be noted that the computer device in the embodiments of this application includes the mobile computer device and the non-mobile computer device described above.

[0119] The memory 402 can be used to store software programs and various data. The memory 402 may primarily include a first storage area for storing programs or instructions and a second storage area for storing data. The first storage area may store the operating system, application programs or instructions required for at least one function (such as sound playback, image playback, etc.). Furthermore, the memory 402 may include volatile memory or non-volatile memory, or both. The non-volatile memory may be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory. Volatile memory can be random access memory (RAM), static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDRSDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous link dynamic random access memory (SLDRAM), and direct memory bus RAM (DRRAM). The memory 402 in this embodiment includes, but is not limited to, these and any other suitable types of memory.

[0120] Processor 401 may include one or more processing units; optionally, processor 401 integrates an application processor and a modem processor, wherein the application processor mainly handles operations involving the operating system, user interface, and applications, and the modem processor mainly handles wireless communication signals, such as a baseband processor. It is understood that the aforementioned modem processor may also not be integrated into processor 401.

[0121] This application also provides a readable storage medium storing a program or instructions. When the program or instructions are executed by a processor, they implement the various processes of the above-described driver intent construction method embodiments and achieve the same technical effects. To avoid repetition, they will not be described again here.

[0122] This application also provides a chip, which includes a processor and a communication interface. The communication interface and the processor are coupled. The processor is used to run programs or instructions to implement the various processes of the above-described driver intent construction method embodiments and can achieve the same technical effect. To avoid repetition, it will not be described again here.

[0123] It should be understood that the chip mentioned in the embodiments of this application may also be referred to as a system-on-a-chip, system chip, chip system, or system-on-a-chip, etc.

[0124] This application also provides a computer program product, which is stored in a storage medium and executed by at least one processor to implement the various processes of the driver intent construction method embodiment described above, and can achieve the same technical effect. To avoid repetition, it will not be described again here.

[0125] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element. Furthermore, it should be noted that the scope of the methods and apparatuses in the embodiments of this application is not limited to performing functions in the order shown or discussed, but may also include performing functions substantially simultaneously or in the reverse order, depending on the functions involved. For example, the described methods may be performed in a different order than described, and various steps may be added, omitted, or combined. Additionally, features described with reference to certain examples may be combined in other examples.

[0126] The embodiments of this application have been described above with reference to the accompanying drawings. However, this application is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of this application without departing from the spirit and scope of the claims, and all of these forms are within the protection scope of this application.

Claims

1. A method for constructing driver intent, characterized in that, The driver intent construction method includes: Obtain an internal event object, wherein the internal event object includes at least one of a driver list behavior event object, a driver order event object, and an end-to-end intelligent event object, and the driver list behavior event object includes at least one of a first driver list behavior event object, a second driver list behavior event object, and a third driver list behavior event object; The intent is constructed based on the internal event object to obtain the intent calculation result; The process of constructing the intent based on the internal event object to obtain the intent calculation result includes: When the internal event object includes the first driver list behavior event object, obtain the pre-statistical offline vehicle preference data; The vehicle preference intent is constructed based on the behavior event objects in the first driver list and the pre-statistical offline vehicle preference data. When the internal event object includes the second driver list behavior event object and the driver order event object, the intent is constructed based on the second driver list behavior event object and the driver order event object to obtain the route intent; When the internal event object includes the third driver list behavior event object and the terminal intelligent event object, the intent is constructed based on the third driver list behavior event object and the terminal intelligent event object to obtain the cargo search intent; Wherein, when there are multiple internal event objects and the conditions for constructing multiple intents are met, intents are constructed based on the multiple internal event objects, and multiple intent calculation results are obtained respectively; The internal event objects used for the intended construction are stored according to their type; When a new internal event object of the same type is obtained, the most recently stored internal event object of the same type is retrieved. Calculate the difference between the new internal event object of the same type and the most recently stored internal event object of the same type; If the difference is greater than or equal to a preset difference threshold, an intent is constructed based on the new internal event object of the same type to obtain a new intent calculation result.

2. The driver intent construction method according to claim 1, characterized in that, Before obtaining the internal event object, the driver intent construction method further includes: Obtain a first original message, which includes at least one of driver list behavior message, driver order message and terminal intelligent message, wherein the driver list behavior message includes at least one of first driver list behavior message, second driver list behavior message and third driver list behavior message; The first original message is transformed to obtain the internal event object; The process of transforming the first original message to obtain the internal event object includes: When the first original message includes the first driver list behavior message, the first driver list behavior message is transformed to obtain the first driver list behavior event object. When the first original message includes the second driver list behavior message and the driver order message, the second driver list behavior message and the driver order message are transformed to obtain the second driver list behavior event object and the driver order event object. When the first original message includes the third driver list behavior message and the terminal intelligence message, the third driver list behavior message and the terminal intelligence message are transformed to obtain the third driver list behavior event object and the terminal intelligence event object. Wherein, when there are multiple first original messages and the conditions for multiple transformation processes are met, transformation processes are performed on the multiple first original messages to obtain multiple internal event objects respectively.

3. The driver intent construction method according to claim 2, characterized in that, The first driver list behavior event object includes vehicle data corresponding to driver's telephone behavior events and click behavior events within a first preset time period; The process of constructing the vehicle preference intent based on the first driver list behavior event objects and pre-statistical offline vehicle preference data includes: The first real-time intent is calculated based on the vehicle data corresponding to the driver's telephone behavior events and click behavior events within the first preset time period. The pre-statistical offline vehicle preference data is used as the first offline intent; The vehicle preference intent is obtained by weighting the first real-time intent and the first offline intent.

4. The driver intent construction method according to claim 3, characterized in that, The internal event objects also include user center event objects, which include vehicle data registered by the driver; When the pre-statistical offline vehicle preference data is empty, the driver intent construction method further includes: Get the preset vehicle configuration; The vehicle data registered by the driver is matched with the preset vehicle configuration to obtain the second offline intent; The vehicle preference intent is obtained by weighting the first real-time intent and the second offline intent.

5. The driver intent construction method according to claim 2, characterized in that, The second driver list behavior event object includes line data corresponding to the driver's telephone behavior events and line data corresponding to the click behavior events within a second preset time period; The driver order event object includes the route data corresponding to the driver's completed order behavior; The step of constructing the route intent based on the second driver list behavior event object and the driver order event object includes: Based on the line data corresponding to the telephone behavior event, obtain the latitude and longitude of the first departure point and the latitude and longitude of the first destination; Based on the route data corresponding to the click event, obtain the latitude and longitude of the second departure point and the second destination. Based on the route data corresponding to the driver's completed order, obtain the latitude and longitude of the third departure point and the latitude and longitude of the third destination; Statistical calculations are performed on the latitude and longitude of the first departure point, the second departure point, and the third departure point to obtain the departure point latitude and longitude with the highest weight. Statistical calculations are performed on the latitude and longitude of the first destination, the second destination, and the third destination to obtain the latitude and longitude of the destination with the highest weight. The route intention is obtained based on the latitude and longitude of the origin with the highest weight and the latitude and longitude of the destination with the highest weight.

6. The driver intent construction method according to claim 5, characterized in that, When the latitude and longitude of the origin with the highest weight and / or the latitude and longitude of the destination with the highest weight are empty, the driver intent construction method further includes: Obtain the second raw message, which includes user center messages and map navigation messages; The user center message and the map navigation message are converted to obtain the user center event object and the map navigation event object; The user center event object includes the vehicle length data registered by the driver; the map navigation event object includes the latitude and longitude data of the driver's current location. Acquire offline feature data of big data, which includes multiple vehicle length groups, and each vehicle length group stores a set of destination latitude and longitude coordinates of each city. The driver's current city is determined based on the driver's current location latitude and longitude data; Among the multiple vehicle length groups, a target vehicle length group that matches the vehicle length data registered by the driver is determined; Within the target vehicle length group, retrieve the destination latitude and longitude set of the current city; Using the driver's current location latitude and longitude data as the fourth departure latitude and longitude, and using at least one destination latitude and longitude from the current city's destination latitude and longitude set as the fourth destination latitude and longitude, at least one of the route intentions is obtained based on the fourth departure latitude and longitude and the fourth destination latitude and longitude.

7. The driver intent construction method according to claim 2, characterized in that, The third driver list behavior event object includes behavior data corresponding to driver search behavior events, behavior data corresponding to click behavior events, and behavior data corresponding to page refresh behavior events; The edge intelligent event object includes the behavioral data corresponding to the edge intelligent trigger event; The intent to find goods includes the intent to hang up, the intent to urgently find goods, and the intent to normally find goods. The process of constructing the intent based on the third driver list behavior event object and the terminal intelligent event object to obtain the cargo search intent includes: If the behavior data corresponding to the page refresh event and the behavior data corresponding to the terminal smart trigger event are empty, and if the behavior data corresponding to the search event and the behavior data corresponding to the click event are both empty within a third preset time period, then the intention to find goods is determined to be the intention to hang up. The third driver list behavior event object also includes behavior data corresponding to telephone behavior events; Weighted behavior data is obtained by weighting the behavior data corresponding to the search behavior event, the click behavior event, and the telephone behavior event. If the weighted behavior data is greater than or equal to the preset behavior threshold, then the intention to find goods is determined to be the intention to urgently find goods. If the weighted behavior data is less than the preset behavior threshold, then the search intention is determined to be a normal search intention.

8. The driver intent construction method according to any one of claims 1 to 7, characterized in that, Prior to constructing the intent based on the internal event object, the driver intent construction method further includes: Frequency control processing is performed on the internal event objects, and different internal event objects correspond to different preset frequency control thresholds; The step of constructing the intent based on the internal event object and obtaining the intent calculation result further includes: Intent is constructed based on the internal event object after frequency control processing, and the intent calculation result is obtained.

9. A driver intent constructing device, characterized in that, The driver intent construction device includes: The data acquisition module is used to acquire internal event objects, which include at least one of driver list behavior event objects, driver order event objects, and terminal intelligent event objects. The driver list behavior event objects include at least one of first driver list behavior event objects, second driver list behavior event objects, and third driver list behavior event objects. The first construction module is used to construct intents based on the internal event objects and obtain intent calculation results. The process of constructing the intent based on the internal event object to obtain the intent calculation result includes: When the internal event object includes the first driver list behavior event object, obtain the pre-statistical offline vehicle preference data; The vehicle preference intent is constructed based on the behavior event objects in the first driver list and the pre-statistical offline vehicle preference data. When the internal event object includes the second driver list behavior event object and the driver order event object, the intent is constructed based on the second driver list behavior event object and the driver order event object to obtain the route intent; When the internal event object includes the third driver list behavior event object and the terminal intelligent event object, the intent is constructed based on the third driver list behavior event object and the terminal intelligent event object to obtain the cargo search intent; Wherein, when there are multiple internal event objects and the conditions for constructing multiple intents are met, intents are constructed based on the multiple internal event objects, and multiple intent calculation results are obtained respectively; The data storage module is used to store the internal event objects intended to be constructed according to the type of the internal event objects; The data retrieval module is used to retrieve the most recently stored internal event object of the same type when a new internal event object of the same type is obtained. The difference calculation module is used to calculate the difference between the new internal event object of the same type and the most recently stored internal event object of the same type; The second construction module is used to construct intent based on the new internal event object of the same type when the difference is greater than or equal to a preset difference threshold, and obtain a new intent calculation result.

10. A driver intent construction device, the driver intent construction device comprising a processor and a memory, the memory storing a computer program, characterized in that, The computer program is loaded and executed by the processor to implement the driver intent construction method as described in any one of claims 1 to 8.

11. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it is used to implement the driver intent construction method as described in any one of claims 1 to 8.

Images