Channel management system and method for interactive process tracking

Abstract

The invention discloses a channel management system and method for interactive process tracking, and the system comprises an API docking module which is used for building API interfaces with a plurality of channels through a unified data interface protocol, and obtaining the original data of the plurality of channels in real time based on a message queue; the data cleaning engine is used for cleaning the original data of the multiple channels by adopting a rule-based abnormal data detection algorithm to obtain cleaned multi-source channel data; the data standardization module is used for carrying out standardization processing on the cleaned multi-source channel data to obtain standardized multi-source channel data; and the channel combination determination module is used for performing quantitative analysis on the standardized multi-source channel data through a multi-objective optimization algorithm to obtain an optimal channel combination. According to the method, efficient integration, cleaning, standardized processing and optimization analysis of multi-channel data are realized, and the optimal channel combination is generated, so that the efficiency of channel management and the decision-making precision are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing and channel management, and in particular to a channel management system and method for interactive process tracking. Background Art

[0002] With the development of internet technology and e-commerce, businesses and organizations need to conduct business operations and provide customer service through multiple channels (such as online platforms, offline stores, and third-party partner platforms). However, significant differences in data formats, interface protocols, and management methods exist between these channels, making direct data integration difficult and increasing the complexity of management and decision-making. Existing technologies lack an efficient system that can uniformly acquire and standardize data from multiple channels.

[0003] The collection of multi-channel data is often accompanied by issues such as missing data, redundancy, inconsistency, and outliers. Traditional data cleaning methods rely primarily on fixed rules and manual intervention, which are inefficient and difficult to adapt to complex and dynamic data environments. Furthermore, the diverse formats of multi-channel data (such as date formats and currency units) further complicate data standardization. Efficiently cleaning and standardizing data during multi-source data integration is a major challenge in existing technologies.

[0004] Optimizing the channel portfolio is a key task in channel management decisions. Existing channel optimization methods often lack multi-objective comprehensive analysis capabilities and only optimize a single indicator, resulting in irrational allocation of channel resources. Summary of the Invention

[0005] In view of this, an embodiment of the present invention provides a channel management system and method for interactive process tracking to solve the above technical problems.

[0006] To achieve the above objectives, in a first aspect, a channel management system for interactive process tracking is provided, which includes: API docking module, used to establish API interfaces with multiple channels through a unified data interface protocol, and obtain raw data from multiple channels in real time based on message queues; A data cleaning engine, configured to clean the raw data of the plurality of channels using a rule-based abnormal data detection algorithm to obtain cleaned multi-source channel data; The data standardization module is used to standardize the cleaned multi-source channel data to obtain standardized multi-source channel data; The channel combination determination module is used to perform quantitative analysis on the standardized multi-source channel data through a multi-objective optimization algorithm to obtain the optimal channel combination.

[0007] In a second aspect, a channel management method for interactive process tracking is provided, which includes: Establish API interfaces with multiple channels through a unified data interface protocol, and obtain raw data from multiple channels in real time based on message queues; Cleaning the raw data of the multiple channels using a rule-based abnormal data detection algorithm to obtain cleaned multi-source channel data; Performing standardization processing on the cleaned multi-source channel data to obtain standardized multi-source channel data; The standardized multi-source channel data is quantitatively analyzed through a multi-objective optimization algorithm to obtain the optimal channel combination.

[0008] According to a third aspect, an electronic device is provided, comprising: one or more processors; a storage device for storing one or more programs, When the one or more programs are executed by the one or more processors, the one or more processors implement a channel management method for interactive process tracking as described in the second aspect.

[0009] In a fourth aspect, a computer-readable storage medium is provided, on which a computer program is stored. When the program is executed by a processor, the channel management method for interactive process tracking as described in the second aspect is implemented.

[0010] The above technical solution has the following beneficial technical effects: The embodiment of the present invention realizes the real-time collection and efficient connection of multi-channel data through a unified data interface protocol and message queue, and adopts a rule-based abnormal data detection algorithm to clean the original data to ensure data quality; eliminates the differences in multi-source data formats through standardization processing, and improves the compatibility and availability of data; finally, quantitatively analyzes the standardized data through a multi-objective optimization algorithm to generate the optimal channel combination, thereby improving the efficiency of multi-channel management and the scientific nature of decision-making. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] The accompanying drawings are provided for a better understanding of the present invention and are not intended to limit the present invention. Figure 1 This is a functional block diagram of a channel management system for interactive process tracking according to an embodiment of the present invention; Figure 2 This is a functional block diagram of an API docking module according to an embodiment of the present invention; Figure 3 is a functional block diagram of a data cleaning engine according to an embodiment of the present invention; Figure 4 is a functional block diagram of a data completion submodule according to an embodiment of the present invention; Figure 5 is a functional block diagram of a data standardization module according to an embodiment of the present invention; Figure 6 is a functional block diagram of a channel combination determination module according to an embodiment of the present invention; Figure 7 This is a functional block diagram of a channel characteristic determination submodule according to an embodiment of the present invention; Figure 8 is a functional block diagram of a commission calculation module according to an embodiment of the present invention; Figure 9 This is a flow chart of a channel management method for interactive process tracking according to an embodiment of the present invention; Figure 10 Schematic diagram of the structure of a computer system according to an embodiment of the present invention. DETAILED DESCRIPTION

[0012] The following description of exemplary embodiments of the present invention is made in conjunction with the accompanying drawings, in which various details of the embodiments of the present invention are included to facilitate understanding. These details should be considered as merely exemplary. Therefore, it should be appreciated by those skilled in the art that various changes and modifications may be made to the embodiments described herein without departing from the scope and spirit of the present invention. Similarly, for the sake of clarity and conciseness, descriptions of well-known functions and structures are omitted in the following description.

[0013] Example 1 like Figure 1 As shown, this embodiment provides a channel management system for interactive process tracking, which includes: API docking module, used to establish API interfaces with multiple channels through a unified data interface protocol, and obtain raw data from multiple channels in real time based on message queues; A data cleaning engine, configured to clean the raw data of the plurality of channels using a rule-based abnormal data detection algorithm to obtain cleaned multi-source channel data; The data standardization module is used to standardize the cleaned multi-source channel data to obtain standardized multi-source channel data; The channel combination determination module is used to perform quantitative analysis on the standardized multi-source channel data through a multi-objective optimization algorithm to obtain the optimal channel combination.

[0014] Specifically, the API docking module can adopt a RESTful architecture design, configure a unified OAuth 2.0 authentication protocol, and support multi-protocol adaptation (including HTTP / HTTPS, WebSocket, etc.). Specifically, through pre-configured message queue middleware (such as Apache Kafka or RabbitMQ), real-time data channels can be established with multiple channels such as e-commerce, social media, and advertising platforms. The API docking module has a built-in connection pool management mechanism that can dynamically adjust the number of connections to ensure that more than 10,000 concurrent requests can be processed per second, and has fault tolerance capabilities such as automatic reconnection and exception retry. During the data acquisition process, independent timeouts and retry strategies are set for each channel interface to maximize data real-time and integrity.

[0015] Specifically, the data cleansing engine, based on a hybrid architecture of machine learning and rule engines, first quickly filters out obvious anomalous data using predefined data validation rules (such as type verification, range restriction, and format matching). For complex potential anomalies, unsupervised learning algorithms (such as Isolation Forest and Local Outlier Factor (LOF)) are used for multi-dimensional anomaly detection. The data cleansing engine supports dynamic rule configuration, allowing for flexible adjustment of cleaning strategies based on the characteristics of different channels. The cleansing process records detailed data processing logs, enabling traceability analysis. Data identified as anomalous is flagged but not discarded, instead entering a manual review process. Specifically, the dynamic rule configuration mechanism of the data cleansing engine allows the system to flexibly adjust cleaning strategies based on the characteristics of different channels. For example, for e-commerce channels, stricter validation rules for price and inventory data can be configured: price ranges can be limited to 0.01-100,000 yuan, inventory quantities cannot be negative, and transaction amounts must be greater than 0 and less than a specific threshold. For social media channels, cleaning rules focus more on user behavior data, including user age range verification, user activity thresholds, and content length and type checks. In terms of specific implementation, the system provides a visual rule configuration interface, where administrators can customize cleaning rules for each channel separately.

[0016] Specifically, the data standardization module utilizes metadata mapping and conversion technology to predefine a unified data model and standard dictionary across channels. Through configuration-driven mapping rules, data from different channels is converted into a homogeneous standard format. Standardization processing includes, but is not limited to, field mapping, unit unification, data type conversion, and missing value filling. The data standardization module features a built-in configurable standardization rules engine that supports dynamic expansion and rapid adaptation to new channels. A multi-level validation mechanism is provided for key metrics (such as user ID and transaction amount) to ensure data consistency and accuracy.

[0017] Specifically, the channel portfolio determination module constructs a channel effectiveness evaluation model based on multi-objective programming and evolutionary algorithms. This algorithm utilizes a modified NSGA-II (non-dominated sorting genetic algorithm), taking into account multiple metrics such as channel conversion rate, customer acquisition cost, and user profile matching. The algorithm dynamically adjusts weights and, through iterative optimization, determines the optimal channel portfolio strategy. The module supports visual decision analysis and provides detailed portfolio recommendation reports, including the contribution rate of each channel and potential risk assessment. Furthermore, the channel portfolio determination module possesses real-time learning capabilities, enabling continuous adjustment of optimization strategies based on new data.

[0018] In some embodiments, the mathematical modeling of the objective function of the channel combination determination module is as follows: The optimization objective function can be expressed as: Max Z = W1 Channel Conversion Rate + W2 Channel Coverage - W3 Channel Customer Acquisition Cost + W4 Channel Response Speed ​​+ W5 Channel Processing Capacity + W6 Channel Customer Retention Rate - W7 Channel Risk. Z is the comprehensive optimization target value, which measures the overall performance of a channel portfolio. This objective function, by introducing weight coefficients W1-W7, assigns different weights to different indicators, allowing companies to flexibly adjust their channel portfolio strategies based on their specific circumstances and market environment. The objective function incorporates positive indicators such as conversion rate, coverage, and response speed, as well as negative indicators such as customer acquisition cost and channel risk, creating a comprehensive evaluation system.

[0019] To ensure the feasibility of the channel portfolio, a series of strict constraints were set as the technical boundaries of the mathematical model: the total channel investment ratio must be equal to 100%, the investment in a single channel must not exceed 40%, the overall risk of the channel portfolio must be controlled within an acceptable range, and the diversity of channel types must be maintained to avoid excessive concentration.

[0020] Employ advanced multi-objective optimization algorithms, such as genetic algorithms, particle swarm optimization, or simulated annealing. These intelligent algorithms efficiently search through vast solution spaces to find near-optimal solutions. By building a comprehensive data collection system, establishing a dynamically updated channel evaluation model, and incorporating machine learning techniques, we achieve intelligent and continuous optimization of our channel portfolio.

[0021] Taking e-commerce platforms as an example, an optimized channel mix might look like this: WeChat mini-programs account for 35% (high conversion rates and quick response), Douyin e-commerce accounts for 25% (reaching young users and having strong processing capabilities), Taobao accounts for 20% (wide coverage and high stability), offline stores account for 10% (high customer retention and personalized service), and Weibo accounts for 10% (serving as a supplementary channel to reach specific demographics). This combination reflects a multi-dimensional, differentiated channel strategy.

[0022] The channel management system for interaction flow tracking provided by the embodiment of the present application realizes efficient and real-time collection of multi-channel data through an API interfacing module, improves the comprehensiveness and timeliness of data acquisition; an advanced anomaly detection algorithm is adopted in the data cleaning engine to effectively filter and process data noise, thereby improving data quality and reliability; the data standardization module realizes unified mapping and conversion of cross-channel data, solves the semantic difference and format inconsistency between heterogeneous data sources, and provides a standardized and structured data basis for subsequent analysis; the channel combination determination module accurately quantifies the value of each channel through a multi-objective optimization algorithm, realizes intelligent configuration and dynamic optimization of channel resources, improves the accuracy and management efficiency of channel management, and can also help enterprises to achieve more accurate user acquisition, lower customer acquisition cost and higher channel conversion rate.

[0023] In some embodiments, the raw data of the plurality of channels includes: historical transaction data including cumulative transaction amount, cumulative transaction times, transaction type, historical order completion rate, number of customers in the channel service coverage area; channel data including channel identifier, channel category, channel source, real-time channel activity, total number of target market customers; user data including user identifier, user category, user registration time, real-time user activity, number of active customers in a specified time period, initial number of customers; order processing data including order request initiation time, order response completion time, total number of processed orders, total processing time; real-time dynamic data including real-time order generation, real-time transaction amount, real-time order completion rate, real-time transaction times, real-time access times, real-time click record, real-time operation time; product data including product identifier, product category, product price, real-time inventory quantity.

[0024] Specifically, the historical transaction data is the data basis for the system to track channel performance. The cumulative transaction amount is calculated by aggregating historical transaction records, reflecting the overall economic value of the channel; the cumulative transaction times record the interaction frequency of customers with the channel; the transaction type is subdivided into online transactions, offline transactions, wholesale transactions, etc., providing multi-dimensional analysis of transaction scenarios; the historical order completion rate calculates the proportion of historical successful order quantity to total order quantity, directly reflecting the channel performance; the number of customers in the channel service coverage area statistics the actual service range and market penetration depth of the channel, providing an important basis for channel strategy optimization.

[0025] Specifically, channel data provides the structural characteristics and dynamic performance of the channel itself. Channel identifiers serve as unique identifiers, ensuring accurate data tracking. Channel categories include e-commerce platforms, social media, advertising channels, and offline stores, supporting multi-dimensional channel classification. Channel sources track the acquisition and initial establishment of channels. Real-time channel activity dynamically reflects the current status of the channel through recent interaction indicators (such as visit volume and engagement rate). The total number of customers in the target market depicts the potential market size of the channel, providing a reference for channel strategy formulation.

[0026] Specifically, user data focuses on individual customer characteristics and behavioral trajectories. User IDs uniquely identify users, supporting personalized analysis. User categories are segmented into new, existing, and high-value customers, facilitating differentiated operations. User registration time marks the beginning of a user's lifecycle. Real-time user activity reflects the intensity of current user interactions. The number of active customers within a specified time period indicates the dynamics of the user base. The number of initial customers reflects the channel's original user base and is a key indicator for evaluating a channel's user acquisition capabilities.

[0027] Specifically, order processing data focuses on the time dimension and efficiency of the transaction process. The order request initiation time records the transaction start point; the order response completion time marks the transaction closure time; the total number of processed orders reflects the transaction volume of the channel; and the total processing time, which measures the average time from order initiation to completion, directly reflects the channel's service response speed and operational efficiency, and is a key indicator for measuring user experience.

[0028] Specifically, the real-time order generation volume reflects the level of transaction activity; the real-time transaction amount indicates the real-time economic value; the real-time order completion rate dynamically reflects the fulfillment capability; the real-time transaction times, real-time visit times, and real-time click records constitute an instant portrait of user behavior; the real-time operation time provides the behavior time trajectory, supporting fine-grained user behavior analysis and real-time decision-making.

[0029] Specifically, product data provides key information at the product level. Product identification serves as a unique identifier for a product; product categories support multi-dimensional classification of products; product prices reflect the market value of products; and real-time inventory quantities dynamically reflect the supply status of products, serving as a crucial basis for inventory management and sales strategies.

[0030] like Figure 2As shown, the API docking module specifically includes: an interface protocol definition submodule, which is used to predefine a unified data interface protocol to support data interaction with different channels; a channel docking submodule, which is used to establish a dynamic API interface connection with multiple channels according to the data interface protocol; a data transmission submodule, which is used to obtain the original data of multiple channels in real time based on the message queue; a data verification submodule, which is used to perform integrity verification and consistency verification on the original data obtained from multiple channels; an exception handling submodule, which is used to detect whether an exception occurs in the API interface communication, and re-request the original data of multiple channels when an exception is detected; the exception includes: network interruption, network timeout or data missing.

[0031] Specifically, the interface protocol definition submodule is responsible for predefining a standardized, universal data exchange protocol. This submodule designs unified data exchange standards, including data formats (such as JSON and XML), transmission protocols (such as HTTPS), authentication mechanisms (such as OAuth 2.0), and secure encryption schemes. By establishing this standard protocol, the system can shield technical differences between different channels and achieve seamless integration with heterogeneous data sources. The protocol definition covers data field mapping, transmission encryption, error code specifications, and other aspects, ensuring that data from all channels can be exchanged in a standardized manner according to unified standards.

[0032] Specifically, the channel docking submodule dynamically establishes API interface connections with multiple channels based on predefined interface protocols. This submodule adopts a plug-in and configurable design architecture to support rapid access to different types of channel data sources, such as e-commerce platforms, social media, advertising systems, etc. By configuring the channel's identity authentication information, interface address, request method and other parameters, the system can automatically generate a customized connection adapter. The channel docking submodule also supports dynamic token management, which can automatically handle interface authorization and refresh credentials to ensure the real-time validity of API connections. For complex channels, customized connection strategies are also provided, such as incremental synchronization and full synchronization.

[0033] Specifically, the data transmission submodule leverages high-performance message queues (such as Apache Kafka and RabbitMQ) to enable real-time access to raw data from multiple channels. This submodule utilizes a distributed, highly concurrent data transmission architecture capable of simultaneously processing data streams from multiple channels. Message queues provide key features such as decoupling, peak shaving, and asynchrony to ensure stable and real-time data transmission. The submodule supports configuration of parameters such as data concurrency, retry mechanisms, and timeout control, enabling customized transmission strategies based on the data characteristics of different channels. Furthermore, built-in optimization mechanisms such as data compression and serialization improve data transmission efficiency and network utilization.

[0034] Specifically, the data verification submodule performs strict integrity and consistency checks on the acquired multi-channel raw data. The data verification submodule is preconfigured with multi-dimensional data quality check rules, including field integrity checks, data type verification, value range checks, and logical consistency checks. For missing or abnormal data, the data verification submodule supports multiple processing strategies, such as direct filtering, filling with default values, and marking abnormalities. The verification process uses a configurable rule engine, allowing users to customize individualized verification logic based on the characteristics of different channels.

[0035] Specifically, the exception handling submodule is a fault-tolerant and self-healing mechanism for the API interface module, focusing on detecting and handling abnormal situations in API interface communication. The exception handling submodule designs a comprehensive exception monitoring system that can capture network interruptions, network timeouts, data missing, and other exceptions in real time. When an exception is detected, the exception handling submodule will automatically initiate a data re-request according to the pre-set retry strategy, supporting advanced retry algorithms such as Exponential Backoff, to avoid data loss caused by transient faults. For persistent exceptions, the exception handling submodule also provides an alarm notification mechanism that can notify the operation and maintenance personnel through SMS, email, instant messaging, and other means. Exception handling not only focuses on technical fault tolerance, but also includes detailed exception log recording and root cause analysis, providing data support for long-term optimization of the system.

[0036] In further embodiments, the exception handling submodule builds a comprehensive exception monitoring system that can capture and classify a variety of exceptions in real time, including network interruptions, timeouts, data missing, and data packet loss. Through a dynamic exception classification model, the system can update the exception classification criteria based on real-time features, improving the accuracy of exception identification. At the same time, based on the time series prediction model of historical exception data, the exception monitoring system can predict the likelihood of exception occurrence in advance, optimize monitoring parameters, and reduce the impact of sudden exceptions on system operation. In addition, the exception monitoring system supports a scene adaptive mechanism that dynamically adjusts exception monitoring thresholds and sampling frequencies based on the interface characteristics of different channels (such as response time and data volume), achieving efficient resource management.

[0037] In further embodiments, the abnormality processing submodule introduces a multi-dimensional adaptive exponential backoff algorithm, dynamically adjusting the growth factor and maximum waiting time of exponential backoff by combining the abnormality type (such as short timeout or long-term network interruption), historical retry success rate and system load. According to the retry history record and real-time feedback, the retry strategy is optimized to avoid unnecessary repeated requests. The retry mechanism adopts an intelligent hierarchical strategy, setting retry priorities for different channels or abnormal types to ensure that requests for critical channels are restored first. For high-frequency repeated abnormalities, the system integrates a fast-fail mechanism, skipping low-success-rate retry operations to improve overall system efficiency.

[0038] In further embodiments, the abnormality processing submodule designs a precise abnormality alarm and response mechanism, generating high-value alarm information through context information analysis (such as abnormal source, duration and impact range) to avoid the problem of excessive and redundant alarms in traditional monitoring. The alarm system supports multi-channel hierarchical notification, including SMS, instant messaging tools or dashboard pop-up windows, and the alarm frequency is dynamically adjusted according to the severity of the abnormality. To further optimize the response efficiency of operation and maintenance, the system integrates an alarm deduplication algorithm, which combines repeated alarms through abnormal pattern matching and provides abnormal log summaries and suggested handling measures in the alarm notification.

[0039] In further embodiments, the abnormality processing submodule comprehensively records detailed information of each abnormality, including timestamp, abnormal source, context data and processing status, generates a structured log index library for quick retrieval and analysis. The system combines data mining techniques (such as clustering analysis, causal inference models) for intelligent root cause analysis of abnormal logs, automatically identifying potential causes of abnormalities and generating abnormal propagation path diagrams.

[0040] As shown in Figure 3 the data cleaning engine specifically comprises: an abnormality detection and processing submodule for identifying abnormal values in the original data using a dynamic detection range adaptive algorithm, performing deletion processing, context relationship-based correction processing or data interpolation replacement processing according to the type and importance weight of the abnormal values to obtain the preliminary cleaned data; wherein the dynamic detection range adaptive algorithm is used to dynamically generate an abnormality detection range by calculating the mean, standard deviation or quantile of the original data in real time, for identifying abnormal values outside the detection range; A duplicate data filtering submodule is configured to perform multi-dimensional duplicate detection on the initially cleaned data, including field-level matching and data pattern similarity matching, and remove duplicate data based on a similarity threshold for duplicate records, while marking suspected duplicate data to generate non-duplicate intermediate cleaned data; wherein, the field-level matching includes comparing the complete consistency of the primary key or unique identifier field in the record; the data pattern similarity matching includes calculating the similarity between different records using a hash algorithm or a distance metric method; A data completion submodule is used to complete the missing fields in the non-repetitive intermediate cleaned data by combining predefined rules and machine learning prediction models to obtain complete cleaned data; The format normalization submodule is used to perform format normalization processing on the complete cleaned data, which includes time format standardization, unit conversion and encoding consistency conversion, and uniformly converts it into a predefined data format to generate cleaned multi-source channel data.

[0041] In this embodiment, the anomaly detection and processing submodule uses a dynamic detection range adaptive algorithm to identify and process outliers in raw data. This submodule calculates data statistical features, such as mean, standard deviation, and quantiles, in real time, dynamically generating anomaly detection ranges. For example, for transaction amount data, the 3σ principle (±3 standard deviations above the mean) or the interquartile range (IQR) method can be used to identify outliers. The module provides various processing strategies for different types and importance of outliers: extreme outliers can be directly deleted; minor outliers can be corrected based on context, such as using the mean or median of nearby data; and outliers in key fields can be replaced with data interpolation, such as linear interpolation or machine learning predictive interpolation. This dynamic and intelligent outlier processing approach ensures data quality and reliability.

[0042] In this embodiment, the duplicate data filtering submodule ensures the uniqueness and integrity of the data through a multi-dimensional and multi-level duplicate detection mechanism. This submodule first performs field-level matching, strictly comparing the primary key or unique identification field in the record, such as user ID, order number, etc., to ensure that records with complete consistency are identified as duplicates. On this basis, the module further adopts data pattern similarity matching technology, using hash algorithms (such as SimHash) or distance measurement methods (such as edit distance, cosine similarity) to calculate the similarity between different records. For records whose similarity exceeds a preset threshold, the system marks them as suspected duplicate data. Based on the similarity and business rules, you can choose to retain the most complete record or merge data according to a specific priority. This multi-dimensional duplicate detection method can not only accurately identify completely duplicate records, but also discover approximately duplicate data at the semantic level.

[0043] In this embodiment, the data completion submodule provides an intelligent completion strategy for missing fields in non-repeated intermediate cleaning data. This submodule combines predefined business rules and machine learning prediction models to achieve intelligent data completion. For simple missing fields, preset default values ​​or fixed rules can be used to fill them, such as setting the missing gender field to "unknown" by default. For complex key fields, machine learning prediction models will be used for completion, such as using random forests, gradient boosting trees or deep learning models to predict missing values ​​based on the features of other fields. The data completion submodule also supports multiple completion strategies, such as mean filling, median filling, nearest neighbor filling, etc., and dynamically selects the best completion method based on the distribution characteristics of the data and the business scenario. Through this combination of rules and machine learning completion method, the integrity and accuracy of the data can be guaranteed to the greatest extent.

[0044] In this embodiment, the format normalization submodule implements unified format standardization of multi-source channel data. This submodule includes three core functions: time format standardization, unit conversion, and encoding consistency conversion. Time format standardization converts time representations across different channels (such as "2024-12-01" and "01 / 01 / 2024") into a standard format (such as ISO 8601 format). Unit conversion ensures consistent measurement units across different channels, such as converting US dollars and euros into RMB or converting miles into kilometers. Encoding consistency conversion addresses differences in character encoding, data types, and other factors, ensuring data compatibility during subsequent processing. The format normalization submodule also supports custom conversion rules, allowing for the configuration of personalized formatting strategies based on specific business needs. Through this comprehensive format normalization process, the system ultimately generates standard, consistent multi-source channel data.

[0045] like Figure 4 As shown, the data completion submodule specifically includes: a missing field identification unit, which is used to scan the non-repetitive intermediate cleaning data based on field-level integrity rules and mark missing fields; a predefined rule completion unit, which is used to complete fixed pattern fields through static rules; a machine learning prediction completion unit, which is used to extract feature vectors and input them into the machine learning prediction model to generate dynamic completion values ​​for fields that cannot be completed through the static rules based on other field values ​​of the records where the missing fields are located, historical similar record data features or time series trends; a result verification and recording unit, which is used to verify the accuracy and consistency of the completion results generated by the predefined rule completion unit or the machine learning prediction completion unit, and generate a completion log.

[0046] In this embodiment, the missing field identification unit performs a comprehensive scan of non-duplicate intermediate cleansed data using sophisticated field-level integrity rules. This unit pre-sets multi-dimensional field integrity check rules, including null value detection, specific format verification, and business logic integrity verification. For example, for user information data, non-null constraints can be set for key fields such as name, contact information, and address; for transaction data, the integrity of fields such as transaction amount, time, and product information can be checked. The identification process not only focuses on simple null values, but also includes the identification of special values ​​(such as "-" and "N / A"). For each identified missing field, a detailed missing tag will be generated to record the location, type, and contextual information of the missing field, providing important metadata support for subsequent completion.

[0047] In this embodiment, the predefined rule completion unit uses static rules to achieve fast and accurate data completion for fields with fixed patterns and clear completion logic. The unit maintains a comprehensive rule base that covers field completion strategies for various business scenarios. For example, for the user's gender field, a default completion rule can be set: when the gender is empty, the gender is inferred based on the name characteristics or other related information; for address information, the province, city, and district code completion rule can be used to deduce the complete address based on the known partial address information. The rule base supports a multi-level matching mechanism and can dynamically select the most matching completion rule based on the specific context of the data. The unit also supports dynamic updating and configuration of rules, allowing users to flexibly adjust the completion strategy according to business changes.

[0048] In this embodiment, the machine learning prediction and completion unit focuses on processing complex fields that cannot be completed using static rules, using advanced machine learning techniques to achieve intelligent data completion. The unit first performs feature engineering, extracting feature vectors from other fields in the record containing the missing field and from similar historical records. Feature vector extraction methods include one-hot encoding, numerical normalization, and embedding techniques to ensure that the features input to the model are highly expressive and discriminative. The unit has built-in multiple machine learning models, such as random forest, gradient boosting tree (XGBoost), and deep neural networks, and can dynamically select the optimal model based on data characteristics and the complexity of the completion task. For time series data, time series prediction models such as ARIMA (Autoregressive Integrated Moving Average) and Prophet are also integrated to capture temporal trends and seasonal characteristics of the data. Pre-trained on large-scale historical data, the model can generate highly accurate dynamic completion values.

[0049] In this embodiment, the result verification and recording unit is responsible for strictly verifying the accuracy and consistency of the completion results. The unit first implements a multi-dimensional verification strategy, including: statistical feature consistency check (for example, whether the mean and variance are within a reasonable range), logical consistency check (checking whether the completion value complies with business rules), and cross-reference verification (consistency with other related fields). For the results of machine learning prediction completion, the unit also calculates the prediction confidence. For completion results with a confidence lower than the threshold, they can be marked for manual review. The verification process generates a detailed completion log, recording key information such as the completion method, completion source, and confidence level of each field. The log not only supports data traceability, but also provides important feedback for data quality improvement. For completion results with obvious anomalies, the unit supports an automatic early warning mechanism, which can notify data management personnel through emails, messages, etc.

[0050] like Figure 5 As shown, the data standardization module specifically includes: a data mapping submodule, which is used to map the fields in the cleaned multi-source channel data to a unified data model based on predefined data mapping rules. The unified data model is a predefined standardized structured model, including field names, field types and field relationships, to obtain preliminary standardized data after mapping; a data normalization verification submodule, which is used to verify the mapped preliminary standardized data using a data normalization algorithm to obtain verified standardized data; and a standardization result generation submodule, which is used to output the verified standardized data as final standardized multi-source channel data.

[0051] In this embodiment, the data mapping submodule realizes the precise conversion of multi-source channel data to a unified data model through predefined data mapping rules. This submodule maintains a mapping rule library that covers the field characteristics, naming conventions and data types of different data sources. The unified data model is a pre-designed standardized structured model, which defines in detail the field names (for example, using camel case naming conventions), field types (for example, strictly defined integers, floating point types, strings, etc.) and field relationships (for example, primary and foreign key relationships, and association relationships between entities). The mapping process adopts a multi-layer matching strategy: first, an exact match is performed, and completely identical fields are directly mapped; for fields with similar semantics but different names, a semantic similarity algorithm (such as cosine similarity) is used for intelligent matching; for fields that cannot be directly mapped, an interface for manual configuration and custom conversion rules is provided. For example, the "product price" field of different e-commerce platforms is mapped to the "productPrice" field of the unified model, and the consistency of data type and precision is ensured.

[0052] In this embodiment, the data normalization verification submodule uses advanced data normalization algorithms to perform comprehensive and rigorous quality verification on the mapped preliminary standardized data. This submodule integrates multi-dimensional verification algorithms, including: data type consistency verification (ensuring that each field conforms to the predefined data type), value range verification (verifying whether the data is within a reasonable value range), format regularity verification (using regular expressions to check fields in specific formats, such as phone numbers and email addresses), and logical consistency verification (checking whether the logical relationship between fields is reasonable). For complex verification scenarios, the submodule supports custom verification rules, allowing users to configure personalized verification logic based on specific business needs. For example, for the user age field, a reasonable range of 0-120 can be set; for the transaction amount, it can be verified that it is non-negative and within a reasonable range. For abnormal data discovered during the verification process, the module provides multiple processing strategies: you can choose to directly mark, correct or eliminate it, and generate a detailed verification report to record the specific circumstances of each abnormal data.

[0053] In this embodiment, the standardization result generation submodule is responsible for converting the mapped and verified data into the final standardized multi-source channel data. This submodule provides flexible data output methods and supports multiple target formats, such as relational database tables, JSON, CSV, Parquet, etc., to meet the data access requirements of different downstream systems. The output process is not just a simple format conversion, but also includes: data compression (reducing storage space), metadata injection (adding data source information, standardized timestamps, etc.), and permission marking (setting access control tags based on data sensitivity). For large-scale data, the module supports distributed output, which can split large data sets into multiple files or partitions to improve output efficiency. At the same time, the module also generates a detailed log of the standardization processing, recording key metadata such as data source information, mapping rules, and verification results, providing an important basis for subsequent data tracing and quality analysis. The output standardized data is not only unified and standardized in format, but also reaches highly consistent standards in semantics and quality.

[0054] like Figure 6 As shown, the channel combination determination module specifically includes: a channel characteristic determination submodule, configured to determine a plurality of channel characteristic indicators related to the channel from the standardized multi-source channel data, wherein the plurality of channel characteristic indicators include: a plurality of channel type, channel market coverage, channel response speed, channel processing capability, and channel customer retention rate; A weight setting submodule, configured to assign weights to the channel characteristic indicators according to predefined rules or user input, so as to determine the relative importance of each channel characteristic indicator in the optimization goal; An optimization objective definition submodule is used to determine the optimization objectives of a multi-objective optimization function, wherein the optimization objectives of the multi-objective optimization function include: maximizing channel resource utilization, maximizing channel coverage, minimizing channel operating costs, and minimizing channel performance fluctuations; The optimization calculation submodule is used to perform quantitative analysis on the extracted multiple channel characteristic indicators based on the multi-objective optimization algorithm, and obtain the optimal channel combination by combining the weight and optimization target of each channel characteristic indicator. The multi-objective optimization algorithm includes a genetic algorithm, a particle swarm optimization algorithm or a Pareto optimization algorithm.

[0055] In this embodiment, the channel characteristics determination submodule serves as the foundation for channel portfolio optimization. Through in-depth analysis of standardized multi-source channel data, it comprehensively extracts key channel characteristic indicators. This submodule designs a multi-dimensional channel characteristics evaluation system, extracting key indicators from standardized data, including but not limited to: channel type (e.g., online, offline, telephone, social media, etc.), market coverage (the proportion of target users reached by the channel), response speed (the average time from user request to channel response), processing capacity (the number of requests a channel can handle per unit time), and customer retention rate (the percentage of long-term customers who use the channel). The extraction process utilizes multi-level data mining techniques, including statistical analysis and machine learning feature extraction, to ensure the accuracy and comprehensiveness of the indicators. For example, for e-commerce platform channels, the market coverage and customer retention rate of each channel can be calculated by analyzing transaction data; for customer service channels, response speed and processing capacity can be calculated using response logs.

[0056] In this embodiment, the weight setting submodule is responsible for assigning relative importance to channel-specific indicators, providing a flexible weight configuration mechanism and supporting two weight setting methods: predefined rules and user-defined. Predefined rules, based on industry big data analysis and statistical models, provide default weight configuration templates for different industries and business scenarios. For example, weight distribution is determined through regression analysis of historical channel performance data. User-defined mode allows users to fine-tune the weights of each indicator based on specific business needs through an interactive interface or configuration file, thereby enhancing flexibility and applicability. Weight setting comprehensively considers multiple technical dimensions, including: the channel's historical key indicator performance (such as order volume, conversion rate, ROI, etc.), dynamically adjusting the weight distribution through weighted calculations based on historical data; channel resource utilization and cost parameters (such as traffic acquisition cost, operating and maintenance costs, and average unit sales cost) to balance the cost-effectiveness of different channels; time periods and fluctuation characteristics (such as quarterly or monthly sales fluctuations and seasonal peaks) to dynamically optimize the weight distribution; and industry-specific indicators (such as customer acquisition cost, channel share, and market growth rate), by optimizing the default weight template to adapt to industry needs. The comprehensive application of these multi-dimensional technical parameters enables the weighting submodule to effectively improve the scientific and applicable nature of channel management decisions, meeting the complex needs of multiple scenarios and industries. Weight calculation supports a variety of methods, such as the Analytic Hierarchy Process (AHP) and the Entropy Weight Method, to ensure the scientific and rationality of weight allocation. For example, for emerging companies pursuing rapid market expansion, the weights of market coverage and response speed can be increased; for mature companies pursuing cost-effectiveness, the weights of indicators related to operating costs and resource utilization can be increased.

[0057] In this embodiment, the optimization objective definition submodule comprehensively balances the complex performance indicators of the channel portfolio by constructing a multi-objective optimization function. This submodule defines four core optimization objectives: maximizing channel resource utilization (ensuring that each channel's resources are fully utilized), maximizing channel coverage (expanding channel reach in the target market), minimizing channel operating costs (reducing the overall cost of channel maintenance and operations), and minimizing channel performance fluctuations (maintaining channel performance stability). The optimization objectives are constructed using mathematical modeling methods, converting qualitative indicators into quantifiable objective functions. Each optimization objective has a detailed calculation method and boundary conditions, taking into account interdependencies and constraints between channels. For example, resource utilization can be measured as the ratio of actual processing capacity to theoretical maximum processing capacity; operating costs can include multiple dimensions such as labor costs, technology investment, and maintenance expenses. The submodule also supports trade-offs and dynamic adjustment between objectives, allowing for flexible optimization of objective weights based on business changes.

[0058] In this embodiment, the optimization calculation submodule uses advanced multi-objective optimization algorithms to achieve intelligent optimization of channel portfolios. This submodule incorporates multiple advanced optimization algorithms, including genetic algorithms, particle swarm optimization algorithms, and Pareto optimization algorithms, enabling it to effectively handle complex multi-objective optimization problems. The optimization calculation process begins by constructing feature vectors, converting channel characteristic indicators and weights into standardized numerical inputs. The genetic algorithm simulates the natural selection process to iteratively generate and select the optimal channel portfolio; the particle swarm optimization algorithm uses a swarm of intelligent particles to collaboratively search for the optimal solution; and the Pareto optimization algorithm focuses on the overall balance between multiple objectives to find the optimal compromise. The optimization calculation considers complex constraints, such as channel resource limitations and budget constraints, to ensure that the output channel portfolio meets both optimization objectives and actual business constraints. The algorithm also supports incremental learning and adaptive adjustment, enabling continuous optimization of the channel portfolio strategy as data accumulates. The resulting output channel portfolio is not only optimal in terms of quantity but also achieves an optimal balance between quality and efficiency.

[0059] Specifically, the following is a detailed description of the working process of the optimization calculation submodule: During the data preprocessing step, the optimization calculation submodule first normalizes the channel characteristic indicators obtained from the channel characteristic determination submodule. Because different indicators have different dimensions and numerical ranges, minimum-maximum normalization or Z-score normalization is used to convert each indicator to a dimensionless value between 0 and 1. The normalization formula is: X_normalized = (X - X_min) / (X_max - X_min). This ensures comparability between different indicators in subsequent calculations and eliminates the influence of dimensions and data scale.

[0060] In the weight integration step, a weighted scoring matrix is ​​constructed based on the channel characteristic indicator weights provided by the weight setting submodule. For each channel, a dot product is performed between the standardized indicator values ​​and the corresponding weights to generate a composite score. For example, if the channel type weight is 0.2, the channel market coverage weight is 0.3, the channel response speed weight is 0.2, the channel processing capacity weight is 0.15, and the channel customer retention weight is 0.15, then the composite score for each channel will reflect the weighted combination of these indicators, reflecting the relative contribution of each indicator to the channel's overall performance.

[0061] In the multi-objective optimization function construction step, a multi-objective optimization function is constructed based on the optimization objectives determined in the optimization objective definition submodule. This function integrates multiple objectives, such as channel resource utilization, channel coverage, channel operating costs, and channel performance fluctuation. Specifically, these objectives can be expressed as an objective function under constraints, for example, maximizing (channel coverage · resource utilization) - (operating costs · performance fluctuation coefficient). By transforming multiple objectives into a single, comprehensive optimization objective, a clear evaluation criterion is provided for subsequent optimization algorithms.

[0062] During the optimization algorithm selection and parameter setting steps, select an appropriate multi-objective optimization algorithm based on system characteristics and computational complexity. For channel combination optimization problems, a genetic algorithm or particle swarm optimization algorithm is recommended. Using the genetic algorithm as an example, key parameters such as population size (50-200), crossover rate (0.6-0.9), and mutation rate (0.01-0.1) need to be set. The initial population is randomly generated, with each individual representing a possible channel combination. The fitness function evaluates the performance of each individual, which is directly mapped to the multi-objective optimization function constructed previously.

[0063] During the iterative optimization and population evolution steps, multiple generations of the evolutionary algorithm are executed, continuously updating the population through selection, crossover, and mutation operations in each generation. The selection phase uses a roulette wheel or tournament selection strategy to retain individuals with high fitness; the crossover operation randomly combines the channel characteristics of outstanding individuals; and the mutation operation introduces a small amount of random perturbation to prevent the algorithm from becoming trapped in a local optimum. Termination conditions can be set, such as reaching a preset maximum number of iterations, population diversity falling below a threshold, or convergence to the optimal solution.

[0064] During the optimal solution extraction and verification step, the individuals with the highest fitness are selected from the final evolved population as the optimal channel combination. This selected channel combination undergoes secondary verification, including recalculating various metrics, simulating performance in actual business scenarios, and cross-validating whether multiple optimization objectives meet expectations. If necessary, small-scale pilots can be conducted to ultimately determine the optimal channel combination. At the same time, alternative solutions are retained to provide reference for subsequent dynamic adjustments.

[0065] In an alternative embodiment, the channel characteristic index may further include: at least one of: channel interaction entropy and channel emotional resonance coefficient.

[0066] Channel interaction entropy is a metric based on the concept of entropy from information theory. It is used to quantify the uncertainty and complexity of user-channel interactions to assess interaction quality. The calculation process first collects data on the complete user-channel interaction trajectory, including interaction steps (specific actions or state changes performed by users during the interaction process, such as browsing, clicking, and paying), interaction time, action type, and conversion path (the complete sequence of user interaction behaviors from the initial touchpoint to the final goal, such as "browsing homepage → product details → adding to cart → confirming order → completing payment," also known as the interaction path). The probability of each interaction path is calculated using the Shannon entropy formula H = -∑(Pi · log2(Pi)), where Pi represents the probability of a specific interaction path. Higher interaction entropy values ​​indicate more complex and unpredictable user interactions; lower interaction entropy values ​​indicate clearer and smoother interaction paths. Furthermore, by introducing additional weighting factors, the importance of different interaction steps or paths in the entropy calculation can be adjusted, resulting in a comprehensive channel interaction complexity metric. This metric can help companies identify interaction bottlenecks or optimization points within the channel, improving user experience and channel efficiency.

[0067] The channel emotional resonance coefficient uses natural language processing and machine learning technology to deeply analyze the emotional dimension of user-channel interactions. The calculation process is as follows: First, we collect complete text data from user interactions with the channel, including user comments, message content, and feedback logs, to serve as the foundation for sentiment analysis. We then clean and preprocess this text data, such as removing stop words, performing word segmentation, and correcting spelling, to ensure data quality and consistency. Based on this, we construct a multidimensional sentiment feature vector, which includes language sentiment polarity (identifying positive or negative sentiment), sentiment intensity (measuring the strength of emotional expression), tone variation (analyzing fluctuations in sentiment across interactions), and semantic depth (capturing implicit emotional themes within the text). This provides feature support for subsequent sentiment signal identification.

[0068] Pre-trained sentiment analysis models (such as BERT or RoBERTa) are used to perform deep feature extraction on user text data. By analyzing contextual relationships, they identify implicit emotional signals in user text, such as joy, dissatisfaction, or concern. These models combine semantic understanding and contextual sensitivity to extract complex emotional features from text and generate multi-dimensional sentiment feature vectors, providing high-quality input for further quantitative sentiment scoring.

[0069] Based on the extracted sentiment features, each user interaction with the channel is assigned a sentiment score on a scale of 0-1, with 0 indicating extremely negative sentiment and 1 indicating extremely positive sentiment. The scoring process incorporates multiple weighting factors, including interaction frequency (higher interaction frequency indicates greater engagement) and interaction duration (longer interaction duration reflects deeper emotional engagement), to improve the accuracy of the sentiment score and its relevance to user behavior.

[0070] A time decay function is introduced to assign different weights to interactions in different time periods, giving more weight to the emotions of recent interactions, thereby more accurately reflecting the timeliness of user emotions. The system also integrates emotional trends from historical user interactions, calculates long-term emotional accumulation, and comprehensively describes the emotional changes between users and channels, forming a comprehensive model of short-term and long-term emotional impact.

[0071] Using machine learning algorithms (such as support vector machines, random forests, or neural networks), the model comprehensively processes discrete sentiment data by integrating multi-dimensional input features such as single-interaction sentiment scores, interaction frequency, time decay, and long-term sentiment accumulation. Ultimately, the model generates a continuous sentiment resonance coefficient ranging from 0 to 1, uniformly quantifying the user's sentiment intensity and loyalty level. The weighting ensures that the model's results comprehensively consider different sentiment dimensions.

[0072] Ultimately, the system outputs an emotional resonance coefficient, ranging from 0 to 1, which intuitively reflects users' emotional perception and loyalty to the channel. A coefficient closer to 1 indicates a more positive user sentiment and loyalty to the channel; a coefficient closer to 0 indicates a more negative user sentiment or lower loyalty. This result serves as a key basis for channel optimization and user sentiment management, helping companies improve user experience and emotional resonance.

[0073] like Figure 7 As shown, the channel characteristic determination submodule specifically includes: A channel type extraction unit is used to extract the channel type through the channel attribute field directly recorded in the channel management system. The channel types include: online channel, offline channel, agency channel, and direct sales channel; A channel market coverage ratio determination unit is configured to obtain the number of customers within the channel service coverage area from customer sales data, which is obtained through statistics of customer purchase records or channel transaction data; obtain the total number of customers in the target market from target market data, which may be obtained from market research reports, industry data, or a customer management system; and determine the channel market coverage ratio based on the ratio of the number of customers within the channel service coverage area to the total number of customers in the target market; a channel response speed analysis unit configured to extract the request initiation time and response completion time of each order request from the order management system; determine the response time of each order request based on the difference between the response completion time and the request initiation time; sum the response times of all order requests to obtain an overall response time; count the total number of all order requests; and calculate the channel response speed based on the ratio of the overall response time to the total number of order requests; a channel processing capacity evaluation unit configured to calculate the total number of processed orders from order processing records, where the total number of processed orders refers to all orders successfully processed within a specified time period; calculate the total processing time of order processing, where the total processing time refers to the total time from the start of processing the first order to the completion of processing the last order; and obtain the channel processing capacity based on the ratio of the total number of processed orders to the total processing time; The channel customer retention rate calculation unit is used to count the number of active customers in a specified time period from the customer management system or sales records; obtain the initial number of customers registered or associated at the beginning of the specified time period; and obtain the channel customer retention rate based on the ratio of the number of active customers in the specified time period to the initial number of customers.

[0074] In a specific embodiment, the channel type extraction unit focuses on accurately extracting channel type information from the channel management system. This unit has designed a standardized channel type classification system, including online channels (such as e-commerce platforms and social media marketing), offline channels (physical stores and exhibition promotions), agency channels (distributors and agents), and direct channels (self-built sales teams). The extraction process uses automated data matching and standardization technology to ensure the accuracy and consistency of channel types. The system automatically maps the raw data in the channel management system to standardized channel types through preset channel attribute mapping rules. For example, for e-commerce platforms, the system can automatically identify and mark them as online channels; for franchise stores, they can be marked as agency channels. This automated classification method improves the efficiency and accuracy of channel type extraction.

[0075] In a specific embodiment, the channel market coverage determination unit accurately calculates the channel's market penetration capability through multi-dimensional data integration. The unit first extracts the number of customers within the channel service coverage area from customer sales data, primarily by analyzing customer purchase records and channel transaction data. At the same time, the total number of customers in the target market is obtained from multi-source market data (e.g., market research reports, industry databases, and customer management systems). The coverage rate is calculated using a standardized formula: Channel market coverage rate = Number of customers within the channel service area / Total number of customers in the target market. The calculation process takes into account the timeliness and representativeness of the data and supports multi-dimensional market segmentation, such as by region, customer type, etc. For example, for a regional e-commerce platform, its market coverage rate in a specific province or city can be calculated.

[0076] In a specific embodiment, the channel response speed analysis unit quantifies the service response efficiency of the channel through sophisticated timestamp analysis. The unit extracts the request initiation time and response completion time of each order from the order management system and calculates the response time of a single order. The response time calculation takes into account the entire process from the customer initiating the request to the channel's complete response, including order acceptance, processing, and feedback. The overall response time is obtained by summing the response times of all order requests; at the same time, the total number of order requests is counted. The formula for calculating the channel response speed is: Channel response speed = overall response time / total number of order requests. The unit supports multi-dimensional response speed analysis and can perform fine-grained performance evaluation by channel type, order type, time period, etc. For example, the response speed differences between online and offline channels can be compared to provide data support for channel optimization strategies.

[0077] In a specific embodiment, the channel processing capacity evaluation unit quantifies the business processing efficiency of the channel through an in-depth analysis of the order processing records. The unit counts the total number of processed orders within a specified time period from the order processing records, that is, all orders successfully processed. The processing time is calculated as the total time from the start of processing the first order to the completion of the last order. The calculation formula for channel processing capacity is: Channel processing capacity = total number of processed orders / total processing time. The calculation process takes into account factors such as order complexity and processing difficulty, and supports differentiated analysis by dimensions such as order type and channel type. For example, the ability of different channels to process standard orders and complex orders can be compared. This sophisticated processing capacity evaluation helps to identify channel bottlenecks and optimize resource allocation and process design.

[0078] In a specific embodiment, the channel customer retention rate calculation unit evaluates the customer stickiness and service quality of the channel by dynamically tracking customer activity. The unit counts the number of active customers in a specified time period from the customer management system or sales records, and obtains the number of initial registered or associated customers at the beginning of the time period. The calculation formula for the channel customer retention rate is: Channel customer retention rate = number of active customers in a specified time period / number of initial customers. The calculation process takes into account the multi-dimensional definition of customer activity, such as purchasing behavior, interaction frequency, service usage, etc. The unit supports retention rate analysis by dimensions such as channel type and customer segmentation, which helps to identify customer churn risks and service quality of different channels. For example, the retention rate performance of online channels and offline channels in different customer groups can be compared to provide a basis for channel strategy optimization.

[0079] like Figure 8 As shown, the system further includes: A commission calculation module, configured to calculate the commission ratio of each channel based on the historical transaction data of each channel and the real-time dynamic data, as well as preset commission rules, through a rule engine and / or a real-time data stream processing framework; The real-time data stream processing framework includes Apache Flink.

[0080] The commission calculation module specifically includes: a data extraction module, configured to extract historical transaction data of each channel from the standardized multi-source channel data; A rule matching module, configured to perform rule matching analysis on the historical transaction data according to a preset commission sharing rule to determine a basic commission sharing ratio; A real-time data processing module, configured to process and analyze the real-time dynamic data based on a real-time data stream processing framework to obtain a dynamic commission ratio, wherein the real-time dynamic data is obtained in real time by connecting to APIs of multiple channels and is generated after being processed by the data cleaning engine and the data standardization module; The commission ratio calculation module is used to generate a final commission ratio through weighted calculation based on the basic commission ratio and the dynamic commission ratio.

[0081] In some embodiments, the commission calculation module calculates the commission ratio of each channel in an intelligent and dynamic manner. This module uses multi-dimensional data analysis and real-time computing technology, comprehensively considering historical transaction data and real-time dynamic data to ensure the fairness and flexibility of the commission strategy. The module's architecture is based on a rule engine and the Apache Flink real-time data stream processing framework, which can quickly and accurately process complex commission calculation logic. The system has designed a dynamic weight calculation mechanism that can adjust the commission ratio in real time according to channel performance, effectively incentivize high-performance channels, and optimize the overall channel ecosystem. For example, for e-commerce platforms, the commission ratio can be dynamically adjusted based on multi-dimensional indicators such as channel sales, customer satisfaction, and service quality.

[0082] In some embodiments, the data extraction module forms the foundation for commission calculations and is responsible for extracting historical transaction data for each channel from standardized multi-source channel data. This module incorporates efficient data extraction algorithms and a multi-data source compatibility mechanism, enabling it to extract standardized transaction information from various channels (e.g., online platforms, offline stores, and agent systems). The extraction process includes key steps such as data source connection, incremental synchronization, and historical data backtracking. For example, for e-commerce channels, historical transaction data can be extracted, including sales, order volume, and average order value. For offline channels, sales reports and transaction details for physical stores can be extracted. The module also incorporates built-in data validation and deduplication mechanisms to ensure the accuracy and completeness of the extracted data.

[0083] In some embodiments, the rule matching module conducts an in-depth analysis of historical transaction data through preset commission rules to determine the basic commission ratio. The module adopts a flexible rule engine architecture to support complex commission rule configuration and dynamic update. The rule matching process includes multi-dimensional data label matching, historical performance evaluation, channel classification, etc. For example, you can set a commission rule based on sales volume ladder: if the monthly sales volume is less than 100,000 yuan, the commission is 5%, 100,000-500,000 yuan, the commission is 7%, and more than 500,000 yuan, the commission is 10%. The module also supports differentiated channel commission strategies, and can set personalized commission rules based on channel type (online / offline), business type, customer attributes, etc. The result of rule matching will be used as the basic weight for commission calculation.

[0084] In some embodiments, the real-time data processing module is based on the Apache Flink real-time data stream processing framework to perform high-performance, low-latency processing and analysis of real-time dynamic data. The module obtains dynamic data from multiple channels in real time through a standardized API interface, and generates real-time indicators that can be used for commission calculation after processing by the data cleaning engine and data standardization module. The processing flow includes data access, real-time ETL, indicator calculation, anomaly detection, etc. For example, dynamic indicators such as sales growth rate, customer satisfaction score, service response speed, etc. of each channel can be obtained in real time. The Flink framework supports complex streaming computing logic and can complete data processing and indicator calculation at the millisecond level, providing technical support for real-time adjustment of the commission ratio.

[0085] In some embodiments, the commission ratio calculation module uses a weighted algorithm to combine the basic commission ratio and the dynamic commission ratio to generate the final channel commission ratio. The module is designed with a flexible weight configuration mechanism, which can dynamically adjust the ratio of basic weight and dynamic weight according to business needs. The calculation process adopts a weighted average algorithm to ensure that the commission results reflect the historical performance of the channel and incentivize real-time performance. For example, the basic commission ratio can be set to 60% and the dynamic commission ratio can be set to 40%. The calculation formula can be expressed as: Final commission ratio = basic commission ratio·0.6+dynamic commission ratio·0.4. The module also supports customized commission calculation rules, and can flexibly configure commission strategies according to different channels and different business scenarios. Through this dynamic and intelligent commission mechanism, the system can effectively balance channel incentives and cost control.

[0086] Example 2 like Figure 9 As shown, this embodiment provides a channel management method for interactive process tracking, which includes the following steps: S10: Establish API interfaces with multiple channels through a unified data interface protocol, and obtain raw data from multiple channels in real time based on message queues; S20: Cleaning the raw data of the multiple channels using a rule-based abnormal data detection algorithm to obtain cleaned multi-source channel data; S30: performing standardization processing on the cleaned multi-source channel data to obtain standardized multi-source channel data; S40: Quantitatively analyzing the standardized multi-source channel data through a multi-objective optimization algorithm to obtain an optimal channel combination.

[0087] In some embodiments, the raw data obtained in step S10 includes: Historical transaction data, including cumulative transaction amount, cumulative number of transactions, transaction type, historical order completion rate, and number of customers in the channel service coverage area; Channel data, including channel identification, channel category, channel source, real-time channel activity, and total number of customers in the target market; User data, including user ID, user category, user registration time, real-time user activity, number of active customers within a specified time period, and initial number of customers; Order processing data, including order request initiation time, order response completion time, total number of processed orders, and total processing time; Real-time dynamic data, including real-time order generation volume, real-time transaction amount, real-time order completion rate, real-time transaction times, real-time visit times, real-time click records, and real-time operation time; Product data, including product identification, product category, product price, and real-time inventory quantity.

[0088] In some embodiments, step S10 specifically includes: S11: Predefine a unified data interface protocol to support data interaction with different channels; S12: Establishing dynamic API interface connections with multiple channels according to the data interface protocol; S13: Obtain raw data from multiple channels in real time based on the message queue; S14: Perform integrity check and consistency check on the original data obtained from multiple channels; S15: Detect whether an abnormality occurs in the API interface communication, and re-request the original data of multiple channels when an abnormality is detected; the abnormality includes: network interruption, network timeout or data loss.

[0089] In some embodiments, step S20 specifically includes: S21: using a dynamic detection range adaptive algorithm to identify outliers in the original data, and performing deletion processing, context-based correction processing, or data interpolation and substitution processing according to the type and importance weight of the outliers to obtain preliminarily cleaned data; S22: Perform multi-dimensional duplicate detection on the data after the preliminary cleaning, including field-level matching and data pattern similarity matching, remove duplicate data based on a similarity threshold of duplicate records, and mark suspected duplicate data to generate intermediate cleaned data without duplication; S23: Completing the missing fields in the non-repeated intermediate cleaned data by combining predefined rules and a machine learning prediction model to obtain complete cleaned data; S24: performing format normalization processing on the complete cleaned data, which includes time format standardization, unit conversion, and encoding consistency conversion, converting the data into a predefined data format to generate cleaned multi-source channel data.

[0090] In some embodiments, step S30 specifically includes: S31: Based on predefined data mapping rules, mapping the fields in the cleaned multi-source channel data to a unified data model to obtain mapped preliminary standardized data; the unified data model is a predefined standardized structured model including field names, field types, and field relationships; S32: Using a data normalization algorithm to verify the mapped preliminary normalized data to obtain verified normalized data; S33: Outputting the verified standardized data as final standardized multi-source channel data.

[0091] In some embodiments, step S40 specifically includes: S41: determining a plurality of channel characteristic indicators related to the channel from the standardized multi-source channel data, wherein the plurality of channel characteristic indicators include: a plurality of channel type, channel market coverage, channel response speed, channel processing capability, and channel customer retention rate; S42: assigning weights to the channel characteristic indicators according to predefined rules or user input to determine the relative importance of each channel characteristic indicator in the optimization goal; S43: Determine an optimization objective of a multi-objective optimization function, wherein the optimization objectives of the multi-objective optimization function include: maximizing channel resource utilization, maximizing channel coverage, minimizing channel operating costs, and minimizing channel performance fluctuations; S44: Quantitatively analyze the extracted multiple channel characteristic indicators based on the multi-objective optimization algorithm, and obtain the optimal channel combination by combining the weight and optimization target of each channel characteristic indicator. The multi-objective optimization algorithm includes a genetic algorithm, a particle swarm optimization algorithm, or a Pareto optimization algorithm.

[0092] In some embodiments, step S41 specifically includes: S411: extracting the channel type through the channel attribute field directly recorded in the channel management system, where the channel type includes: online channel, offline channel, agency channel, and direct sales channel; S412: Obtain the number of customers within the channel service coverage area from the customer sales data, which is obtained through customer purchase records or channel transaction data; obtain the total number of customers in the target market from the target market data, which can be obtained from market research reports, industry data, or a customer management system; and determine the channel market coverage rate based on the ratio of the number of customers within the channel service coverage area to the total number of customers in the target market. S413: Extracting the request initiation time and response completion time of each order request from the order management system; determining the response time of each order request based on the difference between the response completion time and the request initiation time; summing the response times of all order requests to obtain an overall response time; counting the total number of all order requests; and calculating the channel response speed based on the ratio of the overall response time to the total number of order requests; S414: Counting the total number of processed orders from the order processing records, where the total number of processed orders refers to all orders successfully processed within a specified time period; and calculating the total processing time of order processing, where the total processing time refers to the total time from the start of processing the first order to the completion of processing the last order; and obtaining the channel processing capacity based on the ratio of the total number of processed orders to the total processing time. S415: Count the number of active customers in a specified time period from the customer management system or sales records; obtain the number of initial customers registered or associated at the beginning of the specified time period; and obtain the channel customer retention rate based on the ratio of the number of active customers in the specified time period to the number of initial customers.

[0093] In some embodiments, the method further comprises: S50: Calculating the commission ratio of each channel through a rule engine and / or a real-time data stream processing framework based on the historical transaction data of each channel and the real-time dynamic data, as well as the preset commission distribution rules.

[0094] In some embodiments, step S50 specifically includes: S51: extracting historical transaction data of each channel from the standardized multi-source channel data; S52: performing rule matching analysis on the historical transaction data according to the preset commission sharing rules to determine the basic commission sharing ratio; S53: Processing and analyzing the real-time dynamic data based on a real-time data stream processing framework to obtain a dynamic commission ratio; wherein the real-time dynamic data is obtained in real time by connecting to APIs of multiple channels and is generated after being processed by the data cleaning engine and the data standardization module; S54: generating a final commission ratio through weighted calculation according to the basic commission ratio and the dynamic commission ratio.

[0095] In some embodiments, the real-time data stream processing framework includes Apache Flink.

[0096] Those skilled in the art can clearly understand that, for the convenience and brevity of description, only the division of the above-mentioned functional units and modules is used as an example for illustration. In actual applications, the above-mentioned functions can be distributed and completed by different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiment can be integrated into one processing unit, or each unit can exist physically alone, or two or more units can be integrated into one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of software functional units. In addition, the specific names of the functional units and modules are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of this application. The specific working process of the units and modules in the above-mentioned system can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here.

[0097] An embodiment of the present invention further provides a computer-readable storage medium having a computer program stored thereon, wherein the program implements any one of the above methods when executed by a processor.

[0098] If the integrated modules / units are implemented as software functional units and sold or used as standalone products, they can be stored in a computer-readable storage medium. Based on this understanding, the present invention can also implement all or part of the process steps in the above-mentioned method embodiments by using a computer program to instruct the relevant hardware. The computer program can be stored in a computer-readable storage medium. When executed by a processor, the computer program can implement the steps of each of the above-mentioned method embodiments. The computer program includes computer program code, which can be in source code form, object code form, executable file, or some intermediate form. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording medium, USB flash drive, mobile hard drive, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signal, telecommunication signal, and software distribution medium. Of course, there are other types of readable storage media, such as quantum memory and graphene memory. It should be noted that the content contained in the computer-readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practices in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practices, computer-readable media do not include electrical carrier signals and telecommunication signals.

[0099] The present invention also provides an electronic device. The electronic device according to an embodiment of the present invention includes: one or more processors; and a storage device for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the method provided by the present invention.

[0100] Reference below Figure 10 , which shows a schematic structural diagram of a computer system 600 of an electronic device suitable for implementing an embodiment of the present invention. Figure 10 The electronic device shown is only an example and should not limit the functions and scope of use of the embodiments of the present invention.

[0101] like Figure 10 As shown, computer system 600 includes a central processing unit (CPU) 601, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 602 or a program loaded from a storage unit 608 into a random access memory (RAM) 603. Various programs and data required for the operation of computer system 600 are also stored in RAM 603. CPU 601, ROM 602, and RAM 603 are connected to each other via a bus 604. An input / output (I / O) interface 605 is also connected to bus 604.

[0102] The following components are connected to the I / O interface 605: an input section 606 including a keyboard, a mouse, and the like; an output section 607 including devices such as a cathode ray tube (CRT), a liquid crystal display (LCD), and speakers; a storage section 608 including devices such as a hard disk; and a communication section 609 including a network interface card such as a LAN card or a modem. The communication section 609 performs communication processing via a network such as the Internet. A drive 610 is also connected to the I / O interface 605 as needed. Removable media 611, such as a magnetic disk, an optical disk, a magneto-optical disk, or a semiconductor memory, is installed in the drive 610 as needed, so that computer programs read from the removable media can be installed in the storage section 608 as needed.

[0103] In particular, according to embodiments disclosed herein, the processes described in the main step diagrams above can be implemented as computer software programs. For example, embodiments of the present invention include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for executing the methods shown in the main step diagrams. In the above embodiments, the computer program can be downloaded and installed from a network via the communication section 609 and / or installed from a removable medium 611. When the computer program is executed by the central processing unit 601, the above-described functions defined in the system of the present invention are performed.

[0104] It should be noted that the computer-readable medium described in the present invention may be a computer-readable signal medium or a computer-readable storage medium, or any combination thereof. Computer-readable storage media may include, for example, but not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, or components, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to, an electrical connection having one or more conductors, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In the present invention, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, device, or component. In the present invention, a computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such a propagated data signal may take a variety of forms, including, but not limited to, electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium may also be any computer-readable medium other than a computer-readable storage medium that can transmit, propagate, or transport a program for use by or in conjunction with an instruction execution system, apparatus, or device. Program code embodied on a computer-readable medium may be transmitted using any suitable medium, including but not limited to wireless, wireline, optical cable, RF, or any suitable combination thereof.

[0105] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present invention. In this regard, each box in the flow chart or block diagram can represent a module, program segment or a part of code, and the part of the above-mentioned module, program segment or code comprises one or more executable instructions for realizing the logical function of the specification. It should also be noted that in some alternative implementations, the function marked in the box can also occur in a different order than that marked in the accompanying drawings.

[0106] The above specific embodiments do not limit the scope of protection of the present invention. Those skilled in the art will appreciate that various modifications, combinations, sub-combinations, and substitutions may occur depending on design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention are intended to be included within the scope of protection of the present invention.

Claims

1. A channel management system for interactive process tracking, characterized in that: include: API docking module, used to establish API interfaces with multiple channels through a unified data interface protocol, and obtain raw data from multiple channels in real time based on message queues; A data cleaning engine, configured to clean the raw data of the plurality of channels using a rule-based abnormal data detection algorithm to obtain cleaned multi-source channel data; The data standardization module is used to standardize the cleaned multi-source channel data to obtain standardized multi-source channel data; The channel combination determination module is used to perform quantitative analysis on the standardized multi-source channel data through a multi-objective optimization algorithm to obtain the optimal channel combination.

2. The channel management system for interactive process tracking according to claim 1, characterized in that: The raw data from the multiple channels include: Historical transaction data, including cumulative transaction amount, cumulative number of transactions, transaction type, historical order completion rate, and number of customers in the channel service coverage area; Channel data, including channel identification, channel category, channel source, real-time channel activity, and total number of customers in the target market; User data, including user ID, user category, user registration time, real-time user activity, number of active customers within a specified time period, and initial number of customers; Order processing data, including order request initiation time, order response completion time, total number of processed orders, and total processing time; Real-time dynamic data, including real-time order generation volume, real-time transaction amount, real-time order completion rate, real-time transaction times, real-time visit times, real-time click records, and real-time operation time; Product data, including product identification, product category, product price, and real-time inventory quantity.

3. The system according to claim 1, wherein: The API docking module specifically includes: The interface protocol definition submodule is used to predefine a unified data interface protocol to support data interaction with different channels; A channel docking submodule, used to establish dynamic API interface connections with multiple channels according to the data interface protocol; The data transmission submodule is used to obtain the original data of multiple channels in real time based on the message queue; The data verification submodule is used to perform integrity verification and consistency verification on the original data obtained from multiple channels; The exception handling submodule is used to detect whether an exception occurs in the API interface communication, and re-request the original data of multiple channels when an exception is detected.

4. The system according to claim 1, wherein: The data cleaning engine specifically includes: The anomaly detection and processing submodule is used to identify outliers in the original data using a dynamic detection range adaptive algorithm, and perform deletion processing, context-based correction processing, or data interpolation and substitution processing based on the type and importance weight of the outliers to obtain preliminary cleaned data; A duplicate data filtering submodule is used to perform multi-dimensional duplicate detection on the data after the preliminary cleaning, including field-level matching and data pattern similarity matching, and remove duplicate data based on the similarity threshold of duplicate records, while generating markers for suspected duplicate data to generate duplicate-free intermediate cleaned data; A data completion submodule is used to complete the missing fields in the non-repeated intermediate cleaned data by combining predefined rules and machine learning prediction models to obtain complete cleaned data; The format normalization submodule is used to perform format normalization processing on the complete cleaned data, which includes time format standardization, unit conversion and encoding consistency conversion, and uniformly converts it into a predefined data format to generate cleaned multi-source channel data.

5. The system according to claim 1, wherein: The data standardization module specifically includes: The data mapping submodule is used to map the fields in the cleaned multi-source channel data to a unified data model based on predefined data mapping rules to obtain the mapped preliminary standardized data; A data normalization verification submodule is used to verify the mapped preliminary normalized data using a data normalization algorithm to obtain verified normalized data; The standardization result generation submodule is used to output the verified standardized data as final standardized multi-source channel data.

6. The system according to claim 1, wherein: The channel combination determination module specifically includes: a channel characteristic determination submodule, configured to determine a plurality of channel characteristic indicators related to the channel from the standardized multi-source channel data, wherein the plurality of channel characteristic indicators include: a plurality of channel type, channel market coverage, channel response speed, channel processing capability, and channel customer retention rate; A weight setting submodule, configured to assign weights to the channel characteristic indicators according to predefined rules or user input, so as to determine the relative importance of each channel characteristic indicator in the optimization goal; An optimization objective definition submodule is used to determine the optimization objectives of a multi-objective optimization function, wherein the optimization objectives of the multi-objective optimization function include: maximizing channel resource utilization, maximizing channel coverage, minimizing channel operating costs, and minimizing channel performance fluctuations; The optimization calculation submodule is used to perform quantitative analysis on the extracted multiple channel characteristic indicators based on the multi-objective optimization algorithm, and obtain the optimal channel combination by combining the weight and optimization target of each channel characteristic indicator. The multi-objective optimization algorithm includes a genetic algorithm, a particle swarm optimization algorithm or a Pareto optimization algorithm.

7. The system according to claim 6, characterized in that The channel characteristic determination submodule specifically includes: A channel type extraction unit is used to extract the channel type through the channel attribute field directly recorded in the channel management system; a channel market coverage ratio determination unit configured to obtain the number of customers within the channel service coverage area from customer sales data; obtain the total number of customers in the target market from target market data; and determine the channel market coverage ratio based on the ratio of the number of customers within the channel service coverage area to the total number of customers in the target market; a channel response speed analysis unit configured to extract the request initiation time and response completion time of each order request from the order management system; determine the response time of each order request based on the difference between the response completion time and the request initiation time; sum the response times of all order requests to obtain an overall response time; count the total number of all order requests; and calculate the channel response speed based on the ratio of the overall response time to the total number of order requests; a channel processing capacity evaluation unit configured to calculate the total number of processed orders from order processing records, where the total number of processed orders refers to all orders successfully processed within a specified time period; calculate the total processing time of order processing; and obtain the channel processing capacity based on the ratio of the total number of processed orders to the total processing time; The channel customer retention rate calculation unit is used to count the number of active customers in a specified time period from the customer management system or sales records; obtain the initial number of customers registered or associated at the beginning of the specified time period; and obtain the channel customer retention rate based on the ratio of the number of active customers in the specified time period to the initial number of customers.

8. The system according to claim 1, wherein: Also includes: A commission calculation module, configured to calculate the commission ratio of each channel based on the historical transaction data of each channel and the real-time dynamic data, as well as preset commission rules, through a rule engine and / or a real-time data stream processing framework; The real-time data stream processing framework includes Apache Flink.

9. The system according to claim 8, characterized in that The commission calculation module specifically includes: a data extraction module, configured to extract historical transaction data of each channel from the standardized multi-source channel data; A rule matching module, configured to perform rule matching analysis on the historical transaction data according to a preset commission sharing rule to determine a basic commission sharing ratio; A real-time data processing module, configured to process and analyze the real-time dynamic data based on a real-time data stream processing framework to obtain a dynamic commission ratio, wherein the real-time dynamic data is obtained in real time by connecting to APIs of multiple channels and is generated after being processed by the data cleaning engine and the data standardization module; The commission ratio calculation module is used to generate a final commission ratio through weighted calculation based on the basic commission ratio and the dynamic commission ratio.

10. A channel management method for interactive process tracking, characterized in that: include: Establish API interfaces with multiple channels through a unified data interface protocol, and obtain raw data from multiple channels in real time based on message queues; Cleaning the raw data of the multiple channels using a rule-based abnormal data detection algorithm to obtain cleaned multi-source channel data; Performing standardization processing on the cleaned multi-source channel data to obtain standardized multi-source channel data; The standardized multi-source channel data is quantitatively analyzed through a multi-objective optimization algorithm to obtain the optimal channel combination.

Images