Cross-end traffic identification and user behavior attribution analysis method and system

Abstract

The invention relates to a cross-end traffic identification and user behavior attribution analysis method and system. The method comprises the following steps: obtaining a user behavior data set according to application store advertisement click data and landing page downloading data; matching the user behavior data set through a preset equipment fingerprint database to obtain a global user identifier and real-time behavior data corresponding to the user behavior data set; classifying the users through a sliding window algorithm according to the global user identifier and the real-time behavior data to obtain new users and active users; attributing the newly-added users and the active users to obtain attribution results of the newly-added users and attribution results of the active users; and integrating the attribution result of the new user, the attribution result of the active user and the historical new attribution rule calculation result to generate the real-time inquirable channel efficiency, so that the method has the advantages of effectively avoiding the settlement misjudgment and realizing the channel efficiency optimization under the complex business scene of the specific financial field.

Description

Technical Field

[0001] This invention relates to the field of Internet technology applications, and in particular to a method and system for cross-platform traffic identification and user behavior attribution analysis. Background Technology

[0002] The current channel distribution attribution system of internet finance platforms faces two core challenges:

[0003] Channel status perception lag: The current system relies on a fixed period (e.g., 7 days) to update channel quality scores, but it cannot capture sudden changes in channel health in real time. For example, if channel A experiences a sudden surge in traffic due to a promotional activity (the number of applications in a single day exceeds 10 times the historical peak), the system will still push the channel to customers according to the regular traffic allocation rules, causing channel overload.

[0004] Channel data silo effect: Attribution of new users is based solely on the first contact channel (e.g., a user is permanently marked as a channel B user upon first registration), while active user attribution does not incorporate interactions from the most recent 7 days (e.g., a user completes a replacement document through channel C but is ultimately converted by channel D), leading to an excessively high bias rate in channel quality assessment. When channel E intercepts a customer due to anti-fraud measures, its risk is not linked to the customer acquisition quality score of channel F, resulting in a continuous flow of similar customers from inferior channels.

[0005] Typical channel distribution methods based on rule engines suffer from the following business rule deficiencies:

[0006] Static attribution rules are ineffective: The system adopts single-point attribution based on the first-touch channel (e.g., if a user clicks on the channel G link for the first time, it is marked as channel G), but dynamic attribution rules for multiple touchpoints of active users have not been established (e.g., if the last 3 logins involve channels H / I / J, attribution needs to be re-weighted according to the interaction frequency).

[0007] Missing status response rules: When the API response time P99 > 15s due to system maintenance in channel K, the traffic circuit breaker rule based on the time window (such as a timeout rate > 20% for 30 consecutive minutes) was not triggered, resulting in a 22% increase in customer churn rate;

[0008] Rigid constraints on settlement rules: Channel L pays according to a fixed profit-sharing ratio, but has not established differentiated settlement rules for new users and active users, resulting in excessive packaging of the registered traffic to an excessively high proportion.

[0009] Industry practices reveal deep-seated business contradictions:

[0010] Touchpoint attribution fragmentation: New user attribution is only associated with the first download event (such as registration via channel M QR code), but the subsequent supplementary purchases by active users through channel N and signing up through channel O are not included in the attribution chain, resulting in the conversion contribution of high-quality channel P being underestimated;

[0011] Lack of risk feedback: When the customer base delivered by Channel Q is blocked by risk control (e.g., the trigger rate of anti-fraud rule R3 is 21%), there is no traffic quality deduction rule for Channel Q (e.g., automatically reducing its traffic weight to 50% after the rule is triggered), which leads to the continuous delivery of high-risk customer base by similar channels.

[0012] Settlement cost mismatch: The platform uses channel settlement or manual settlement, and the existing settlement rules do not establish accurate rules for the investment costs of high-quality and low-quality channels, lacking an efficient and accurate attribution system to support it.

[0013] Currently, no effective solutions have been proposed to address the issues of error-prone channel attribution and complex channel settlement in related technologies, which lead to a lack of rapid and accurate attribution methods for channel cooperation. Summary of the Invention

[0014] The purpose of this invention is to address the shortcomings of existing technologies by providing a method and system for cross-platform traffic identification and user behavior attribution analysis, thereby solving the technical problems of error-prone channel attribution, complex channel settlement, and the lack of fast and accurate attribution methods for channel cooperation.

[0015] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0016] This invention provides a system for cross-platform traffic identification and user behavior attribution analysis, comprising: a cross-platform data fusion module for collecting app store ad click data and landing page download data, and obtaining a user behavior dataset based on the app store ad click data and landing page download data; a data collection and device identification association module for matching the user behavior dataset with a preset device fingerprint database to obtain the global user identifier and real-time behavior data corresponding to the user behavior dataset; a dynamic attribution calculation module for classifying users according to the global user identifier and real-time behavior data using a sliding window algorithm to obtain new users and active users; attributing new users and active users respectively to obtain the attribution results for new users and active users; a big data calculation module for executing historical new user attribution rules in batches on a T+1 basis according to pre-stored historical behavior data to obtain the historical new user attribution rule calculation results; and a big data query engine module for integrating the attribution results for new users, the attribution results for active users, and the historical new user attribution rule calculation results to generate real-time queryable channel performance.

[0017] Optionally, the cross-platform data fusion module includes: an app store ad click unit, a landing page download unit, and a fusion unit. The app store ad click unit is used to collect and integrate ad click data from various app stores to obtain app store ad click data. The app store ad click data is used to collect user download behavior data on the landing page to obtain landing page download data. The fusion unit is used to associate the app store ad click data and landing page download data with device feature information to obtain a user behavior dataset.

[0018] Furthermore, optionally, the fusion unit is also used to obtain a user behavior dataset composed of feature information, account, and user behavior by associating it with an account, based on the feature information of at least two types of terminals in the app store ad click data and landing page download data; wherein, the feature information is classified according to the device type to obtain the identifier of the smart device, the open anonymous device identifier, or the International Mobile Equipment Identity Code.

[0019] Optionally, the data collection and device identification association module is also used to build a device fingerprint database before the cross-platform data fusion module collects application store ad click data and landing page download data. The device fingerprint database includes at least one or a combination of the following: device model, operating system version, browser fingerprint, and network environment parameters.

[0020] Optionally, the dynamic attribution calculation module is also used to determine whether a user is a new user or an active user by monitoring the user's active behavior in real time based on global user identifiers and real-time behavior data; when the user is an active user, the user behavior data of the active user is filtered using a sliding window algorithm, and the attribution result of the active user is calculated by the dynamic attribution algorithm based on the user's most recent interaction channel; when the user is a new user, the attribution is performed based on the user's first contact channel to obtain the attribution result of the new user.

[0021] Optionally, the big data computing module is also used to perform batch execution of historical new attribution rules in parallel by multiple computing nodes through a distributed architecture based on the user's pre-stored historical behavior data, and obtain the calculation results of historical new attribution rules, where T is the current date, 1 is the adjacent date, and T+1 is the time period for the iteration between the current date and the adjacent date.

[0022] Optionally, the big data query engine module is also used to integrate the attribution results of new users, the attribution results of active users, and the calculation results of historical new attribution rules through data indexing and caching mechanisms to generate real-time queryable channel performance.

[0023] This invention provides a method for cross-platform traffic identification and user behavior attribution analysis, comprising: collecting app store ad click data and landing page download data; obtaining a user behavior dataset based on the app store ad click data and landing page download data; matching the user behavior dataset with a preset device fingerprint database to obtain global user identifiers and real-time behavior data corresponding to the user behavior dataset; classifying users according to the global user identifiers and real-time behavior data using a sliding window algorithm to obtain new users and active users; attributing new users and active users respectively to obtain attribution results for new users and attribution results for active users; executing historical new user attribution rules in batches on a T+1 basis according to pre-stored historical behavior data to obtain historical new user attribution rule calculation results; and integrating the attribution results for new users, the attribution results for active users, and the historical new user attribution rule calculation results to generate real-time queryable channel performance.

[0024] Optionally, collecting app store ad click data and landing page download data, and obtaining a user behavior dataset based on the app store ad click data and landing page download data includes: collecting and integrating ad click data from major app stores to obtain app store ad click data; collecting user download behavior data on landing pages to obtain landing page download data; and associating the app store ad click data and landing page download data with device feature information to obtain a user behavior dataset. Specifically, associating the app store ad click data and landing page download data with device feature information to obtain a user behavior dataset includes: based on at least two types of terminal feature information in the app store ad click data and landing page download data, and associating them with accounts to obtain a user behavior dataset composed of feature information, accounts, and user behavior; the feature information is classified according to device type to obtain a smart device identifier, an open anonymous device identifier, or an International Mobile Equipment Identity (IMEI).

[0025] Optionally, before collecting app store ad click data and landing page download data, the method further includes: constructing a device fingerprint database, wherein the device fingerprint database includes at least one or a combination of: device model, operating system version, browser fingerprint, and network environment parameters.

[0026] This invention employs the above technical solution, collecting app store ad click data and landing page download data to obtain a user behavior dataset. The user behavior dataset is then matched against a pre-set device fingerprint database to obtain the corresponding global user identifier and real-time behavior data. Users are categorized using a sliding window algorithm based on the global user identifier and real-time behavior data to obtain new users and active users. Attribution is performed on new users and active users respectively to obtain attribution results for new users and active users. Historical new user attribution rules are executed in batches (T+1) based on pre-stored historical behavior data to obtain historical new user attribution rule calculation results. The attribution results for new users, active users, and historical new user attribution rule calculation results are integrated to generate real-time queryable channel performance. Compared with existing technologies, this invention has the following technical effects: effectively avoiding settlement misjudgments and optimizing channel performance in complex business scenarios within specific financial fields. Attached Figure Description

[0027] Figure 1 This is a schematic diagram of a system for cross-terminal traffic identification and user behavior attribution analysis according to Embodiment 1 of the present invention;

[0028] Figure 2 This is a schematic diagram of each layer in a cross-terminal traffic identification and user behavior attribution analysis system according to Embodiment 1 of the present invention;

[0029] Figure 3 This is a schematic diagram of a new user attribution analysis method in a cross-terminal traffic identification and user behavior attribution analysis system according to Embodiment 1 of the present invention;

[0030] Figure 4 This is a schematic diagram of a user activity attribution analysis method in a cross-terminal traffic identification and user behavior attribution analysis system according to Embodiment 1 of the present invention;

[0031] Figure 5 This is a schematic diagram of a big data computing module in a cross-terminal traffic identification and user behavior attribution analysis system according to Embodiment 1 of the present invention;

[0032] Figure 6 This is a flowchart illustrating a method for cross-terminal traffic identification and user behavior attribution analysis according to Embodiment 2 of the present invention. Detailed Implementation

[0033] To make the objectives, technical solutions, and advantages of this application clearer, the application is described and illustrated below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application. All other embodiments obtained by those skilled in the art based on the embodiments provided in this application without inventive effort are within the scope of protection of this application.

[0034] Obviously, the accompanying drawings described below are merely some examples or embodiments of this application. Those skilled in the art can apply this application to other similar scenarios based on these drawings without any inventive effort. Furthermore, it is understood that although the efforts made in this development process may be complex and lengthy, for those skilled in the art related to the content disclosed in this application, any changes to design, manufacturing, or production based on the technical content disclosed in this application are merely conventional technical means and should not be construed as insufficient disclosure of the content of this application.

[0035] In this application, the reference to "embodiment" means that a specific feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places in the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment that is mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described in this application may be combined with other embodiments without conflict.

[0036] Unless otherwise defined, the technical or scientific terms used in this application shall have the ordinary meaning understood by one of ordinary skill in the art to which this application pertains. The terms “a,” “an,” “an,” “the,” and similar words used in this application do not indicate quantity limitation and may indicate singular or plural. The terms “comprising,” “including,” “having,” and any variations thereof used in this application are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or device that includes a series of steps or units (elements) is not limited to the listed steps or units, but may also include steps or units not listed, or may include other steps or units inherent to these processes, methods, products, or devices. The terms “connected,” “linked,” “coupled,” and similar words used in this application are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. The terms “multiple” / “several” used in this application refer to two or more. “And / or” describes the relationship between related objects, indicating that three relationships may exist; for example, “A and / or B” can indicate: A alone, A and B simultaneously, or B alone. The character " / " generally indicates that the preceding and following objects are in an "or" relationship. The terms "first," "second," and "third" used in this application are merely to distinguish similar objects and do not represent a specific ordering of the objects.

[0037] Example 1

[0038] An illustrative embodiment of the present invention, such as Figure 1 As shown, Figure 1 This is a schematic diagram of a system for cross-platform traffic identification and user behavior attribution analysis according to Embodiment 1 of the present invention. The system for cross-platform traffic identification and user behavior attribution analysis provided in this application includes:

[0039] The cross-platform data fusion module 10 is used to collect app store ad click data and landing page download data, and obtain a user behavior dataset based on the app store ad click data and landing page download data; the data collection and device identification association module 12 is used to match the user behavior dataset with a preset device fingerprint database to obtain the global user identifier and real-time behavior data corresponding to the user behavior dataset; the dynamic attribution calculation module 14 is used to classify users according to the global user identifier and real-time behavior data using a sliding window algorithm to obtain new users and active users; attribution is performed on new users and active users respectively to obtain the attribution results of new users and active users; the big data calculation module 16 is used to execute historical new user attribution rules in batches on T+1 according to pre-stored historical behavior data to obtain the historical new user attribution rule calculation results; the big data query engine module 18 is used to integrate the attribution results of new users, the attribution results of active users and the historical new user attribution rule calculation results to generate real-time queryable channel performance.

[0040] Optionally, the cross-platform data fusion module 10 includes: an app store ad click unit, a landing page download unit, and a fusion unit. The app store ad click unit is used to collect and integrate ad click data from various app stores to obtain app store ad click data. The app store ad click data is used to collect user download behavior data on the landing page to obtain landing page download data. The fusion unit is used to associate the app store ad click data and the landing page download data with the device's feature information to obtain a user behavior dataset.

[0041] Furthermore, optionally, the fusion unit is also used to obtain a user behavior dataset composed of feature information, account, and user behavior by associating it with an account, based on the feature information of at least two types of terminals in the app store ad click data and landing page download data; wherein, the feature information is classified according to the device type to obtain the identifier of the smart device, the open anonymous device identifier, or the International Mobile Equipment Identity Code.

[0042] Optionally, the data acquisition and device identification association module 12 is also used to build a device fingerprint database before the cross-end data fusion module 10 collects application store ad click data and landing page download data. The device fingerprint database includes at least one or a combination of the following: device model, operating system version, browser fingerprint and network environment parameters.

[0043] Optionally, the dynamic attribution calculation module 14 is also used to determine whether a user is a new user or an active user by monitoring the user's active behavior in real time based on the global user identifier and real-time behavior data; when the user is an active user, the user behavior data of the active user is filtered by the sliding window algorithm, and the attribution result of the active user is calculated by the dynamic attribution algorithm based on the user's most recent interaction channel; when the user is a new user, the attribution is performed based on the user's first contact channel to obtain the attribution result of the new user.

[0044] Optionally, the big data computing module 16 is also used to perform batch execution of historical new attribution rules in a distributed architecture by multiple computing nodes based on the user's pre-stored historical behavior data, and obtain the calculation results of the historical new attribution rules, where T is the current date, 1 is the adjacent date, and T+1 is the time period for the iteration between the current date and the adjacent date.

[0045] Optionally, the big data query engine module 18 is also used to integrate the attribution results of new users, the attribution results of active users, and the calculation results of historical new attribution rules through data indexing and caching mechanisms to generate real-time queryable channel performance.

[0046] In conclusion, Figure 2 This is a schematic diagram of each layer in a cross-platform traffic identification and user behavior attribution analysis system according to Embodiment 1 of the present invention. In this embodiment, the application store ad click unit, landing page download unit, data collection and device identification association module 12, dynamic attribution calculation module 14, big data calculation module 16, and big data query engine module 18 are configured in the application layer of the cross-platform traffic identification and user behavior attribution analysis system. Figure 2 As shown, it is represented as: App Store Ad Click Unit (BCS), Landing Page Download Unit and Fusion Unit (MDS), Data Collection and Device Identification Association Module (CIS), Dynamic Attribution Calculation Module (TRS), Big Data Calculation Module (Spark), and Big Data Query Engine Module (Hover); In addition, Figure 2 The document also shows relational databases (MySQL) and non-relational databases (HBase).

[0047] In this embodiment of the application, the cross-platform data fusion module 10 includes: an app store ad click unit, a landing page download unit, and a fusion unit. Figure 2 (not shown in the image), where,

[0048] The App Store Ad Click Unit is responsible for collecting and integrating ad click data from various app stores. This data includes: the time a user clicks an ad, the ad's unique identifier, and basic information about the user's device. This data is acquired in real-time through API integration with the ad platforms and stored in the system's data warehouse.

[0049] The landing page download unit is used to collect user download behavior data on the landing page. This download behavior data includes information such as the time the user spends on the landing page, the specific time of the download operation, and the network environment of the user's device. By embedding specific tracking code on the landing page, user download behavior can be accurately captured, and this data can be correlated with ad click data.

[0050] In this embodiment, the data acquisition and device identification association module 12 uses device feature mapping technology to achieve unified association of user identities across multiple channels. A device fingerprint feature library is constructed, containing key features such as device model, operating system version, browser fingerprint, and network environment parameters. Based on these features, a unique code is generated for each device.

[0051] When a user performs actions on different devices, the system uses the device's unique identifier to link the user's behavior across those devices. For example, a user might click on an advertisement on their phone and then complete a download on their tablet. By associating these actions through device identifiers, the system can identify that they belong to the same user.

[0052] Specifically, in this embodiment, based on account system association, a data fusion system allows users to register or log in to applications using the same mobile phone number. During user registration or login, the device's OAID, IMEI, and DUID are bound to the user account corresponding to that mobile phone number.

[0053] Specifically, ID-Mapping technology is used to associate identifiers on different devices. By combining various relationship pairs between accounts and device models, as well as user data such as device usage patterns, and applying rules and patterns, stable ID relationship pairs are output, and a unique user identifier (UID) is generated as a unique identifier for that object.

[0054] By analyzing user login time, usage frequency, and operational behavior across different devices, the application can determine whether these devices belong to the same user. Regardless of which device a user logs in with the same account, the application can identify the same user through the account and associate it with the corresponding device identifier.

[0055] The DUID, or unique identifier for smart devices, typically refers to a 32-bit hexadecimal number used to uniquely identify smart devices such as televisions and smart speakers. It's used by applications for device identification and data statistics. However, its meaning can vary depending on the context. For example, a DUID for storage devices identifies the storage device itself, while a DUID in a distributed system addresses the unique identification problem within the system. Generally generated by the device manufacturer or related systems according to certain rules, it is unique, usually related to the device's hardware information or specific configuration, and remains unchanged throughout the device's lifecycle.

[0056] The Open Anonymous Device Identifier (OAID) is based on the Open Anonymous Device Identifier and is primarily used for Android devices, especially in the Chinese market. It's used to provide personalized advertising to users while protecting their personal data privacy and security, and also facilitates conversion attribution analysis for advertisers through third-party monitoring platforms. Generated using a device-specific algorithm, it's a device-level identifier; different apps on the same device will obtain the same OAID value. It's a non-permanent device identifier, generated upon the system's first boot. Its value is affected by the application's "cross-application association access permission" setting and may change in certain scenarios, such as user resets or device factory resets.

[0057] The International Mobile Equipment Identity (IMEI) is used by all devices that support cellular networks, including mobile devices such as smartphones and tablets. It serves as a unique identifier for these devices within the mobile phone network, much like a mobile phone's ID card. Typically programmed into the device by the manufacturer during production, it cannot be modified by the user and remains unique and unchanged throughout the device's lifespan. It consists of 15 digits: the first 6 digits are the type approval number, the next 2 are the factory assembly code, the following 6 are the serial number, and the last digit is a spare code.

[0058] In a preferred example, when a user operates on different terminals (such as mobile phones, tablets, computers, etc.), the system first extracts the device's characteristic information, such as the device model and operating system version. Then, based on the aforementioned characteristic information, the system associates the device with the user's identity.

[0059] For example, when a user clicks on an advertisement on their mobile phone, the system records the phone's device characteristics and binds them to the user's identity. When the user completes a download on a tablet, the system identifies the same user's behavior based on the device characteristics and links the two (i.e., the process of obtaining the user behavior dataset in this application embodiment).

[0060] This application uses various device identifiers, such as the identifier DUID for smart devices, the open anonymous device identifier OAID, and the International Mobile Equipment Identity (IMEI). These identifiers are unique and stable, and can effectively identify and associate user behavior on different devices.

[0061] For example, DUID is typically used to identify smart devices such as TVs and smart speakers; OAID is mainly used for Android devices, providing personalized advertising and conversion attribution analysis while protecting user privacy; IMEI is applicable to all devices that support cellular networks and is a unique identifier for the device in the mobile phone network.

[0062] In this embodiment, the dynamic attribution calculation module 14 performs real-time dynamic attribution analysis based on the real-time behavior of target users at each business stage, distinguishing between attribution rules for newly added users and active users on the same day. For example, for newly added users, the system will focus on analyzing the channel source of their initial registration; while for active users, it will consider their multi-touchpoint interaction channels in the recent period.

[0063] By employing real-time attribution algorithms, the system can accurately calculate attribution results shortly after a user's action occurs. This allows channel performance evaluations to be reflected in the system promptly, providing a basis for subsequent channel optimization and resource allocation.

[0064] In a preferred example, users are identified as new or active users by monitoring their activity in real time. For active users, a sliding window algorithm (time window ≥ 7 days) is used to filter continuously active user behavior, preventing it from being incorrectly counted as new user acquisition value for new channels.

[0065] For example, if a user logs into the app every day for the past 7 days, that user is considered an active user, and their influence on the value of acquiring new users through new channels is excluded during attribution.

[0066] The system dynamically calculates attribution results based on users' real-time and historical behavior data. It comprehensively considers multiple factors, such as the user's initial contact channel and most recent interaction channel, and uses a complex attribution algorithm to calculate the most appropriate attribution result.

[0067] For example, if a user registers for the first time through channel A, completes the document replacement process on channel B, and finally completes the contract signing on channel C, the system will calculate the contribution of each channel to the user based on the weight of these actions and make reasonable attributions.

[0068] Attribution path models are built separately for new users and active users. For new users, the model will focus on analyzing their registration channels and first download behavior; for active users, it will consider their multi-touchpoint interaction channels in recent times.

[0069] By establishing separate attribution path models, the effectiveness of each channel can be evaluated more accurately, providing a scientific basis for channel optimization and resource allocation.

[0070] In a preferred example Figure 3 This is a schematic diagram of a user addition attribution analysis method in a cross-terminal traffic identification and user behavior attribution analysis system according to Embodiment 1 of the present invention, as shown below. Figure 3 As shown, the details are as follows:

[0071] First, identify the new user and trigger the recognition of the user's new product;

[0072] Second, determine if there are any new registrations in the first 7 days (first time); if so, determine the registration channel and record the attribution type: registration channel 0-7; if not, determine if there are any new landing page download clicks in the first 3 days (recent);

[0073] Third, if the download clicks are from the landing page in the last 3 days (recent), then determine the download click channel and record the attribution type: download clicks 0-3; if they are not from the landing page in the last 3 days (recent), then determine whether the login package channel is not the app store.

[0074] Fourth, if the login package channel is not an app store, then determine the registration channel and record the attribution type: historical registration; if there is no login package channel other than an app store, then determine whether it is a new registration within the previous 8-30 days (first time);

[0075] Fifth, if the registration is new and within the previous 8-30 days (first time), then confirm the registration and record the attribution type 8-30; if it is not new and within the previous 8-30 days (first time), then determine whether it is a new landing page download click (recent) within the previous 4-30 days.

[0076] Sixth, if the download clicks are from the landing page within the last 4-30 days (recent), then determine the download click channel and record the attribution type: download clicks 4-30; if the download clicks are not from the landing page within the last 4-30 days (recent), then determine whether the download clicks are from a previous registration (first time).

[0077] Seventh, if it is a new registration before registration (first time), then determine the registration channel and record the attribution type: historical registration; if it is not a new registration before registration (first time), then the login package channel and record the attribution type: app store package.

[0078] In a preferred example Figure 4This is a schematic diagram of a user activity attribution analysis method in a cross-terminal traffic identification and user behavior attribution analysis system according to Embodiment 1 of the present invention, as shown below. Figure 4 As shown, the details are as follows:

[0079] First, identify the user as an active user and trigger the user activity recognition (first time of the day);

[0080] Second, determine if the user was added within the previous 7 days; if so, determine if the user was added on the current day; if so, identify the user's added channel and record the attribution type: source channel number for the added day (e.g., ...). Figure 4 If it is not the same day as the new user, then determine the new user channel and record the attribution type: channel number used for new users added within 7 days; if it is not the 7 days before the new user was added, then determine whether there was no activity in the 7 days before the most recent download click record and whether the most recent download click record was within the last 3 days.

[0081] Third, if there was no activity in the 7 days prior to the most recent download click record and the most recent download click record was within the last 3 days, then determine whether the activity was in the last 3 days; if it was active in the last 3 days, then determine the download click channel and record the attribution type: the channel number for activation on the current day; if it was inactive in the last 3 days, then determine the download click channel and record the attribution type: the channel number for activation in the last 7 days.

[0082] Fourth, if there was no activity in the 7 days prior to the most recent download / click record and the most recent download / click record was within the last 3 days, then determine whether the user was effectively reactivated in the past; if not, then determine the new user channel and record the attribution type: no reactivation, continue using channel number.

[0083] Fifth, if the previous activation was successfully reactivated, determine if it was within 7 days of the last successful activation date; if so, identify the last active attribution channel and record the attribution type: 7-day activation channel number; if not, identify the last active attribution channel and record the attribution type: continued activation channel number.

[0084] In this embodiment, the big data computing module 16 adopts a T+1 batch computing method to calculate attribution rules for target historical behavior. This means that the system will process the user behavior data of the previous day in batches at fixed times every day to calculate the attribution results of historically added users.

[0085] In particular, the batch processing method can effectively handle massive amounts of historical data, ensuring the accuracy and completeness of attribution results. It also provides data support for long-term channel performance evaluation.

[0086] In this application, the big data computing module 16 is a core software system specifically designed for large-scale data computation and analysis. It primarily addresses the demands of processing massive amounts of data (such as TB / PB-level data) that traditional single-machine computing cannot handle. The core capability of the big data computing module 16 is to decompose complex computational tasks into parallel processing across multiple computing nodes, improving computational efficiency through a distributed architecture, while simultaneously supporting diverse data processing scenarios (offline computing, real-time computing, interactive analysis, etc.). Figure 5 As shown, Figure 5 This is a schematic diagram of a big data computing module in a cross-terminal traffic identification and user behavior attribution analysis system according to Embodiment 1 of the present invention.

[0087] The big data computing module in this embodiment is as follows:

[0088] Data sources cover multiple data source inputs, mainly including log types: including event tracking, application / service logs, and service management logs; structured data: provided by the service database; and streaming data: accessed through message middleware and services.

[0089] Data extraction is used to deploy multiple components to complete data collection and preprocessing, real-time collection of log data, support for capturing existing and incremental data, and ensure database data synchronization; all extracted data flows to the Kafka cluster.

[0090] Message queues, serving as a data buffer and flow hub, achieve high-throughput, low-latency data storage and distribution through multiple Kafka middleware nodes, decoupling data production and consumption, and supporting the asynchronous requirements of real-time stream processing.

[0091] The computing engine, based on a stream-batch integrated computing framework, builds the core computing layer and features flexible job submission and execution modes: Submission mode: supports software package and SQL script submission to adapt to different development habits; Execution mode: covers local debugging, independent cluster, and resource scheduling to meet the needs of development and production environments.

[0092] Data storage (for) Figure 5 It supports multiple storage terminals, including write-back message queues, non-relational databases, traditional databases, columnar storage, retrieval, relational databases, and analytical databases, and can persist or further transfer calculation results as needed.

[0093] Cluster management / monitoring provides cluster resource management and job scheduling capabilities, supporting resource allocation for components such as Flink;

[0094] The architecture of the big data computing module in this embodiment achieves a closed loop of real-time data "collection-transmission-computation-storage" through componentization and layered design. It supports high-concurrency streaming data processing, adapts to diverse data sources and storage needs, and leverages the powerful computing capabilities of the big data framework to ensure the accuracy of data processing, thus facilitating the implementation of real-time business analysis, monitoring, and decision-making scenarios.

[0095] The big data computing module 16 in this application employs a distributed architecture for its big data computing engine, enabling the decomposition of complex computing tasks into multiple computing nodes for parallel processing. It supports various data processing scenarios, including offline computing, real-time computing, and interactive analysis, and can effectively handle massive amounts of data.

[0096] In this embodiment, the big data query engine module 18 provides high-performance reading capabilities for massive amounts of data to support the real-time query needs of the attribution analysis module. This module employs advanced data indexing technology and a caching mechanism, enabling it to quickly respond to user query requests. For example, when operations personnel need to view the real-time conversion rate of a specific channel, the big data query engine can return accurate data within seconds, providing timely support for decision-making.

[0097] In the embodiments of this application, the big data query engine module is a software system specifically designed for large-scale data querying and analysis. Its core function is to enable users to quickly and efficiently extract valuable information from massive amounts of data.

[0098] In this embodiment, the big data query engine module 18 employs advanced data indexing technology and a caching mechanism, enabling it to quickly respond to user query requests. The big data query engine module 18 in this embodiment supports multiple query methods, such as SQL queries and full-text search, to meet the needs of different users.

[0099] The data indexing technology in this embodiment is an auxiliary data structure built upon the original data (such as database tables, files, log streams, column storage, etc.). Its core purpose is to bypass full scans and quickly locate data that meets the query conditions. Specifically, in the query process, based on the query keywords, it quickly locates rows or data blocks that meet the conditions, reducing or even avoiding full table scans. For example, when responding to an SQL query containing the condition WHERE customer_region = 'Asia' AND order_amount > 1000, the engine first uses the hash index built on the customer_region column to quickly locate all records in the 'Asia' region, and then uses the B+ tree index built on the order_amount column to efficiently find orders with an amount greater than 1000 in that subset, significantly reducing the amount of data that needs to be scanned. This improves query efficiency and further increases the overall cluster throughput.

[0100] The caching mechanism in this application embodiment is a technique that utilizes high-speed storage media (usually memory) to temporarily store frequently accessed or computationally expensive copies of data or results. The aim is to avoid repeatedly performing costly operations (such as disk I / O, complex calculations, and remote calls), thereby significantly improving the response speed of subsequent identical or similar requests.

[0101] Role in the query process: In which stages of the query processing chain does the caching mechanism intervene and play a role?

[0102] Before the query begins: The engine checks the cache (query result cache, materialized view cache):

[0103] If a perfectly matching cached result exists and is not expired, return the cached result directly, skipping all subsequent calculations and I / O.

[0104] If a partially matching cache exists (such as intermediate results or materialized views), that cache is used to accelerate subsequent calculations.

[0105] Query in progress:

[0106] When performing an index lookup, the index cache in memory is accessed first.

[0107] When reading data, the data block cache in memory is accessed first.

[0108] When performing complex calculations, check if there are available results in the computation cache.

[0109] After the query is completed:

[0110] Based on the configured strategy, the results or key intermediate results of this query will be stored in the cache (result cache, intermediate result cache).

[0111] Update cache statistics.

[0112] In summary, the cross-terminal traffic identification and user behavior attribution analysis system provided in this application embodiment is applicable to multi-terminal collaborative business scenarios such as e-commerce finance, joint lending, and online education, and provides technical solutions for complex issues such as missing ad click loops, weak information channel attribution blind spots, misattribution of active users, secondary distribution of withdrawal value, and settlement.

[0113] In a preferred example, suppose an e-commerce financial platform uses the cross-platform traffic identification and user behavior attribution analysis system provided in this application embodiment. This e-commerce financial platform places advertisements on multiple channels, including app stores, social media, and search engines. Users may click on an advertisement on their mobile phone, then complete the download on a tablet, and finally register and transact on a computer.

[0114] The first step is data collection and association: When a user clicks on an ad on their phone, the system collects ad click data through the app store ad click module and extracts device characteristic information, such as device model and operating system version. Then, the system generates a unique code for the device and associates it with the user's identity.

[0115] When a user completes a download on a tablet, the system collects download data through the landing page download module and extracts device feature information again. By associating the device identifier, the system recognizes that this is the same user's behavior and links the two together.

[0116] The second step is user behavior identification and attribution: By monitoring user activity in real time, the system determines whether a user is a new user or an active user. If a user is registering for the first time, the system will identify them as a new user and attribute the user based on their initial contact channel.

[0117] If a user is an active user, the system will use a sliding window algorithm to filter continuously active user behavior, preventing it from being incorrectly counted as user acquisition value for new channels. Simultaneously, the system will comprehensively consider the user's most recent interaction channels and calculate the most appropriate attribution result using a dynamic attribution algorithm.

[0118] The third part, Channel Performance Evaluation and Optimization: Through the big data computing module and query engine module, the system can calculate and query channel performance indicators in real time, such as conversion rate, delinquency rate, and risk score. Operations personnel can use this data to adjust channel placement strategies and optimize resource allocation in a timely manner.

[0119] For example, if a channel has a low conversion rate but a high delinquency rate, operators can reduce spending on that channel or adjust the content and format of the ads to improve the channel's effectiveness.

[0120] As can be seen from the above examples, the cross-terminal traffic identification and user behavior attribution analysis system provided in this application embodiment can effectively solve the problems existing in channel attribution of existing credit platforms, improve the accuracy and real-time performance of channel efficiency assessment, and is applicable to channels in complex business scenarios such as e-commerce finance and joint lending.

[0121] The cross-platform traffic identification and user behavior attribution analysis system provided in this application includes channel device feature extraction and association, dynamic user activity determination, dynamic calculation of attribution logic, modeling of new and active attribution paths, and dynamic correction of channel value. It solves problems in existing credit platforms' channel attribution such as incomplete loop of ad click data, inability to associate weak information channels with user identity, incorrect attribution of active users, and the inclusion of non-target behaviors in new attribution logic. It also addresses attribution blind spots caused by fragmented channel data, effectively avoiding settlement misjudgments and resource misallocation. It is suitable for channel efficiency optimization in complex business scenarios such as e-commerce finance and joint lending. Furthermore, this solution supports rapid adaptation to multiple scenarios, helping them build a refined traffic operation system and enhance market competitiveness.

[0122] This invention employs the above technical solution, collecting app store ad click data and landing page download data to obtain a user behavior dataset. The user behavior dataset is then matched against a pre-set device fingerprint database to obtain the corresponding global user identifier and real-time behavior data. Users are categorized using a sliding window algorithm based on the global user identifier and real-time behavior data to obtain new users and active users. Attribution is performed on new users and active users respectively to obtain attribution results for new users and active users. Historical new user attribution rules are executed in batches (T+1) based on pre-stored historical behavior data to obtain historical new user attribution rule calculation results. The attribution results for new users, active users, and historical new user attribution rule calculation results are integrated to generate real-time queryable channel performance. Compared with existing technologies, this invention has the following technical effects: effectively avoiding settlement misjudgments and optimizing channel performance in complex business scenarios within specific financial fields.

[0123] Example 2

[0124] An illustrative embodiment of the present invention, such as Figure 6 As shown, Figure 6 This is a flowchart illustrating a method for cross-platform traffic identification and user behavior attribution analysis according to Embodiment 2 of the present invention, applied to the system for cross-platform traffic identification and user behavior attribution analysis in Embodiment 1. The method for cross-platform traffic identification and user behavior attribution analysis provided in this application includes:

[0125] Before collecting app store ad click data and landing page download data in step S600, the cross-terminal traffic identification and user behavior attribution analysis method provided in this application embodiment further includes: constructing a device fingerprint database, wherein the device fingerprint database includes at least one or a combination of device model, operating system version, browser fingerprint and network environment parameters.

[0126] Step S600: Collect app store ad click data and landing page download data; based on the app store ad click data and landing page download data, obtain the user behavior dataset.

[0127] Optionally, in step S600, collecting app store ad click data and landing page download data, and obtaining a user behavior dataset based on the app store ad click data and landing page download data, includes: collecting and integrating ad click data from major app stores to obtain app store ad click data; collecting user download behavior data on landing pages to obtain landing page download data; and associating the app store ad click data and landing page download data with device feature information to obtain a user behavior dataset. Specifically, associating the app store ad click data and landing page download data with device feature information to obtain a user behavior dataset includes: based on at least two types of terminal feature information in the app store ad click data and landing page download data, and associating them with accounts to obtain a user behavior dataset composed of feature information, accounts, and user behavior; the feature information is classified according to device type to obtain a smart device identifier, an open anonymous device identifier, or an International Mobile Equipment Identity (IMEI).

[0128] Step S602: Match the user behavior dataset with the preset device fingerprint database to obtain the global user identifier and real-time behavior data corresponding to the user behavior dataset;

[0129] Step S604: Based on global user identifiers and real-time behavior data, users are classified using a sliding window algorithm to obtain new users and active users; attribution is performed on new users and active users respectively to obtain attribution results for new users and active users.

[0130] Step S606: Based on the pre-stored historical behavior data, execute the newly added historical attribution rules in batches via T+1 to obtain the calculation results of the newly added historical attribution rules;

[0131] Step S608: Integrate the attribution results of new users, the attribution results of active users, and the calculation results of historical new attribution rules to generate real-time queryable channel performance.

[0132] This invention employs the above technical solution, collecting app store ad click data and landing page download data to obtain a user behavior dataset. The user behavior dataset is then matched against a pre-set device fingerprint database to obtain the corresponding global user identifier and real-time behavior data. Users are categorized using a sliding window algorithm based on the global user identifier and real-time behavior data to obtain new users and active users. Attribution is performed on new users and active users respectively to obtain attribution results for new users and active users. Historical new user attribution rules are executed in batches (T+1) based on pre-stored historical behavior data to obtain historical new user attribution rule calculation results. The attribution results for new users, active users, and historical new user attribution rule calculation results are integrated to generate real-time queryable channel performance. Compared with existing technologies, this invention has the following technical effects: effectively avoiding settlement misjudgments and optimizing channel performance in complex business scenarios within specific financial fields.

[0133] The above description is merely a preferred embodiment of the present invention and does not limit the implementation and protection scope of the present invention. Those skilled in the art should realize that any equivalent substitutions and obvious changes made based on the description and illustrations of the present invention should be included within the protection scope of the present invention.

Claims

1. A system for cross-platform traffic identification and user behavior attribution analysis, characterized in that, include: The cross-platform data fusion module is used to collect app store ad click data and landing page download data, and to obtain a user behavior dataset based on the app store ad click data and the landing page download data; The data acquisition and device identification association module is used to match the user behavior dataset with a preset device fingerprint database to obtain the global user identifier and real-time behavior data corresponding to the user behavior dataset. The dynamic attribution calculation module is used to classify users based on the global user identifier and the real-time behavior data using a sliding window algorithm to obtain new users and active users; and to perform attribution on the new users and the active users respectively to obtain the attribution results of the new users and the attribution results of the active users. The big data computing module is used to execute historical newly added attribution rules in batches on a T+1 basis based on pre-stored historical behavior data, and obtain the calculation results of historical newly added attribution rules. The big data query engine module is used to integrate the attribution results of the new users, the attribution results of the active users, and the calculation results of the historical new user attribution rules to generate real-time queryable channel performance.

2. The system for cross-terminal traffic identification and user behavior attribution analysis according to claim 1, characterized in that, The cross-platform data fusion module includes: an app store ad click unit, a landing page download unit, and a fusion unit, wherein... The app store ad click unit is used to collect and integrate ad click data from various app stores to obtain the app store ad click data. The app store ad click data is used to collect user download behavior data on the landing page to obtain the landing page download data; The fusion unit is used to associate the app store ad click data and the landing page download data with the device's feature information to obtain the user behavior dataset.

3. The system for cross-terminal traffic identification and user behavior attribution analysis according to claim 2, characterized in that, The fusion unit is further configured to obtain the user behavior dataset composed of feature information, account, and user behavior by associating it with the account, based on the feature information of at least two types of terminals in the app store ad click data and the landing page download data; The feature information is classified according to the device type to obtain the identifier of the smart device, the open anonymous device identifier, or the international mobile device identity code.

4. The system for cross-terminal traffic identification and user behavior attribution analysis according to claim 1 or 2, characterized in that, The data acquisition and device identification association module is further configured to construct the device fingerprint database before the cross-platform data fusion module collects the app store ad click data and the landing page download data. The device fingerprint database includes at least one or a combination of the following: device model, operating system version, browser fingerprint, and network environment parameters.

5. The system for cross-terminal traffic identification and user behavior attribution analysis according to claim 4, characterized in that, The dynamic attribution calculation module is further configured to determine whether a user is a new user or an active user by real-time monitoring of the user's active behavior based on the global user identifier and the real-time behavior data; when the user is an active user, the sliding window algorithm is used to filter the user behavior data of the active user, and the dynamic attribution algorithm is used to calculate the attribution result of the active user based on the user's most recent interaction channel; when the user is a new user, attribution is performed based on the user's first contact channel to obtain the attribution result of the new user.

6. The system for cross-terminal traffic identification and user behavior attribution analysis according to claim 5, characterized in that, The big data computing module is also used to perform batch execution of historical new attribution rules by multiple computing nodes in a distributed architecture based on the user's pre-stored historical behavior data, and obtain the calculation result of the historical new attribution rules, where T is the current date, 1 is the adjacent date, and T+1 is the time period for the iteration between the current date and the adjacent date.

7. The system for cross-terminal traffic identification and user behavior attribution analysis according to claim 6, characterized in that, The big data query engine module is also used to integrate the attribution results of the new users, the attribution results of the active users, and the calculation results of the historical new attribution rules through data indexing and caching mechanisms to generate real-time queryable channel performance.

8. A method for cross-platform traffic identification and user behavior attribution analysis, characterized in that, include: Collect app store ad click data and landing page download data, and obtain a user behavior dataset based on the app store ad click data and the landing page download data; The user behavior dataset is matched with a preset device fingerprint database to obtain the global user identifier and real-time behavior data corresponding to the user behavior dataset; Based on the global user identifier and the real-time behavior data, users are classified using a sliding window algorithm to obtain new users and active users; attribution is performed on the new users and the active users respectively to obtain the attribution results of the new users and the attribution results of the active users. Based on pre-stored historical behavior data, historical newly added attribution rules are executed in batches over time (T+1) to obtain the calculation results of historical newly added attribution rules. The attribution results of new users, active users, and historical new user attribution rules are integrated to generate real-time queryable channel performance.

9. The method for cross-terminal traffic identification and user behavior attribution analysis according to claim 8, characterized in that, The collection of app store ad click data and landing page download data, and the resulting user behavior dataset based on these data, include: Collect and integrate ad click data from major app stores to obtain the app store ad click data; Collect user download behavior data on the landing page to obtain the landing page download data; The user behavior dataset is obtained by associating the app store ad click data and the landing page download data with the device's feature information; The step of associating the app store ad click data and the landing page download data with the device's feature information to obtain the user behavior dataset includes: based on the feature information of at least two types of terminals in the app store ad click data and landing page download data, and by associating them with accounts, obtaining the user behavior dataset composed of feature information, accounts, and user behavior; the feature information is classified according to device type to obtain a smart device identifier, an open anonymous device identifier, or an international mobile device identity code.

10. The method for cross-terminal traffic identification and user behavior attribution analysis according to claim 8, characterized in that, Before collecting app store ad click data and landing page download data, the method further includes: Construct the device fingerprint database, wherein the device fingerprint database includes at least one or a combination of: device model, operating system version, browser fingerprint, and network environment parameters.

Images