Operation management system and method based on financial marketing big data

Abstract

The invention relates to the technical field of big data analysis, in particular to an operation management system and method based on financial marketing big data, and the system comprises a data standardization module, a customer portrait disambiguation module, a strategy matching module, a real-time transmission module and an operation strategy generation module. According to the method, a data table with a unified structure is constructed through combined judgment of field semantics and numerical rules, the data consistency is improved, customer information integration is executed based on field conflict time, credibility level and logic verification, the recognition accuracy is enhanced, and the matching relation is judged by combining the transaction amount, the product association degree and the interaction frequency with strategy attributes; according to the method, strategy dynamic screening is achieved, pushing content is screened according to the strategy emergency degree and the channel state, the transmission efficiency is improved, customer clicking, conversion and feedback participate in behavior trend judgment, strategy updating linkage is supported, an overall processing chain is of a closed-loop structure, and the fusion accuracy and the strategy response timeliness are improved under the conditions of data fragmentation and multi-source heterogeneous.

Description

Technical Field

[0001] The present invention relates to the field of big data analysis technology, and in particular to an operation management system and method based on financial marketing big data. Background Art

[0002] The field of big data analytics encompasses a technical system that leverages massive amounts of heterogeneous data to extract valuable insights and drive decision-making. Its core focus is on building high-precision analytical models through the collection, cleaning, and structured processing of multi-source data, and implementing data-driven strategy optimization based on business scenarios. This area systematically covers the collection, storage, calculation, and visualization of user behavior, transaction data, and market dynamics data in financial marketing scenarios. It focuses on addressing issues such as data fragmentation, analysis lags, and static strategies, providing financial institutions with real-time, dynamic marketing decision support.

[0003] The operational management system based on financial marketing big data is a technical solution for financial product and service marketing scenarios. It integrates structured correlations between user transaction records, social media interaction information, and market trend data to analyze and build customer behavior prediction models and generate dynamic marketing strategies. This system focuses on technical aspects such as cross-platform data standardization, integrating multi-dimensional customer characteristics, and exploring potential demand correlations for quantitative analysis. Specifically, it cleans unstructured data to build a customer tag system, establishes classification and prediction rules based on historical behavior data, and optimizes marketing paths by updating the strategy library using real-time data streams.

[0004] Existing technologies commonly suffer from insufficient structured entry points, static data fusion rules, and delayed policy responses during data processing. Faced with heterogeneous data from multiple sources, there's a lack of unified parsing standards for field naming and values, often leading to inconsistent behavioral data structures and hindering standard modeling. During customer profiling, conflicting data fields from different sources fail to determine trustworthiness. Field integration often relies on chronological order and lacks logical matching principles, easily leading to distortion of user characteristics. Regarding policy matching, tag binding is often based on fixed rules, failing to reflect current customer behavior. Policy push generally employs static path settings, ignoring network fluctuations and transmission resource load, leading to delayed or failed policy responses. At the feedback level, click and conversion data is used solely for statistical purposes and fails to factor into policy adjustments and optimization logic. This results in a one-way marketing process and prevents real-time policy updates from being linked to customer behavior. For example, when customer transaction volume fluctuates dramatically within a short period, existing technologies lack the ability to adjust policy recommendations based on behavioral fluctuations. This can lead to policy failures and compromise the continuity of the customer response chain. Summary of the Invention

[0005] The purpose of the present invention is to solve the shortcomings of the existing technology and propose an operation management system and method based on financial marketing big data.

[0006] In order to achieve the above objectives, the present invention adopts the following technical solutions: The operation and management system based on financial marketing big data includes: The data standardization module obtains user transaction records, social media interaction records, and market trend data in financial marketing scenarios, analyzes the semantic matching of field names, verifies whether the field distribution complies with business rules, and generates standardized customer behavior data tables; The customer profile disambiguation module, based on the standardized customer behavior data table, identifies conflicting information about the same customer in different data sources, compares data update time, data source credibility level, and field verification results, and generates a unified customer profile table; The strategy matching module calls the transaction amount, product relevance, and interaction frequency in the unified customer portrait table, combines the target product, calculates the customer value and strategy fit, and obtains a strategy matching list; The real-time transmission module screens the unique identification code of the customer to be transmitted and the policy content based on the policy urgency in the policy matching list, performs channel priority sorting according to the transmission path delay level, packet loss situation and node load status, and generates a policy transmission queue; The operation strategy generation module calls the customer click records, conversion rates and market feedback information in the strategy transmission queue, combines the return on investment and customer retention cycle, and generates a financial marketing strategy optimization plan.

[0007] As a further solution of the present invention, the standardized customer behavior data table includes field semantic mapping results, field value distribution rationality verification results, and unified field format data; the unified customer portrait table includes conflict information disambiguation results, data credibility level assessment results, and logical consistency verification results; the strategy matching list includes customer value assessment results, strategy fit score, and target strategy plan items; the strategy transmission queue includes a customer unique identification code, a strategy content summary, and a transmission path scheduling parameter; the financial marketing strategy optimization plan includes customer click behavior data, conversion effect feedback, and strategy benefit evaluation results.

[0008] As a further solution of the present invention, the data standardization module includes: The semantic recognition submodule obtains user transaction record fields, social media interaction fields, and market trend fields, calculates the degree of character similarity between field names, the number of combined occurrences of field names in the total data, and the degree of comparison with data source labels. It selects field pairs whose character similarity and combined occurrence exceed the field fit benchmark value and the combination frequency benchmark value, and generates the field semantic consistency matching degree. The numerical judgment submodule analyzes the change interval, frequency change, and data fluctuation of the field value at the difference time node based on the transaction frequency value, keyword frequency value, and trend indicator value corresponding to the field pair in the field semantic consistency matching degree, compares the overlap between the actual change interval of the field value and the logical reference interval, and obtains the field value logical consistency ratio; The behavior integration submodule calls the behavior label, time label and source label of the field based on the logical data items in the logical conformity ratio of the field value, merges the transaction frequency value, interaction frequency value and trend indicator value in the same dimension, integrates the cross-category and unifies the interval format to obtain a standardized customer behavior data table.

[0009] As a further solution of the present invention, the specific calculation formula for the change interval of the analysis field value at the difference time node is: ; in, Represents the field value in the field pair The change range indicator, Represents a field pair At the time point The field value of Represents a field pair At the time point The field value of Represents a field pair The associated keyword frequency value, Represents a field pair The transaction frequency value during the analysis period, Represents a field pair The value of the trend indicator within the analyzed period.

[0010] As a further solution of the present invention, the customer profile disambiguation module includes: The field verification submodule obtains the customer identification, contact information, and transaction information field data in the standardized customer behavior data table, calculates the format matching rate, internal code matching rate, and completeness value based on the field structure, internal code consistency, and filling frequency, and determines the difference with the field consistency benchmark value to generate the field consistency deviation rate; The data optimization submodule calls the field consistency deviation rate, combines the field update time interval, the trust level ranking value and the consistency offset value, performs sorting and screening according to the corresponding trust interval, and generates a filtered field trust level ranking value; The customer normalization submodule performs priority matching and conflict elimination on the field values ​​of the same customer under different data sources according to the trusted level ranking value of the filtered fields, merges the field values, and generates a unified customer portrait table.

[0011] As a further solution of the present invention, the strategy matching module includes: The transaction feature recognition submodule obtains the transaction amount, product relevance, and interaction frequency from the unified customer portrait table, combines the three items to perform data fusion processing in the customer dimension, forms a unified representation value by setting an order, and obtains the customer behavior intensity value; The value fit calculation submodule calls the customer behavior intensity value, combines it with the target product list in the marketing strategy library, and performs item-level screening against the target products based on the product association information in the customer profile, identifies the density of the corresponding relationship, and performs mapping conversion based on the customer behavior intensity value to obtain the strategy fit intensity value; The strategy matching list generation submodule locates the top-ranked customer identification information based on the strategy fit strength value, summarizes the target product and customer portrait number content, and combines them with the strategy fit strength value to generate a strategy matching list.

[0012] As a further solution of the present invention, the specific calculation formula for the density of the identification correspondence is: ; Among them, D represents the relationship density value between customers and target products in item-level screening, represents the customer's relevance score for the Kth product feature, represents the set weight score of the target product on the Kth product feature, represents the average relevance score of customers on all product features, It represents the average set score of the target product on all product features, n represents the total number of product features, Represents the total weighted matching value between customers and target products in all feature dimensions. It represents the product of the average customer preference value and the average product target value. Represents the square root of the product.

[0013] As a further solution of the present invention, the real-time transmission module includes: The urgency screening submodule obtains the urgency of the policies in the policy matching list and the corresponding customer unique identification codes, compares the urgency with the policy urgency threshold, screens the policies and customer identification codes with urgency higher than the threshold, and generates a high urgency policy tag set; The path evaluation submodule calls the unique customer identification code in the high urgency strategy tag set, obtains the delay level, packet loss situation and node load status of the corresponding transmission path, compares them with the delay threshold, packet loss rate standard value and node load benchmark value respectively, records the paths that meet the standards, and generates a set of available path identification codes; The queue generation submodule sorts the policies and customer identification codes in the available path identification code set by policy urgency, arranges the transmission order, binds the path identifier, policy content and customer identification code, and generates a policy transmission queue.

[0014] As a further solution of the present invention, the operation strategy generation module includes: The click behavior extraction submodule obtains customer click records in the strategy transmission queue, identifies the distribution characteristics of click time, page location, and click frequency under product categories, and generates a customer click deviation rate by combining the change in the number of clicks and the page dwell time in the same time period; The conversion rate matching calculation submodule calls the customer click deviation rate, identifies the conversion time and status type in the conversion record, extracts the temporal relationship between click behavior and conversion, and performs matching analysis with the change difference of the customer retention time range. Combined with the fluctuation of the number of conversion statuses, it derives the conversion linkage change amount; The strategy prioritization submodule calls the conversion linkage change amount, refers to the investment return rate of financial products in the current cycle, selects the product set in the high range according to the ranking of the return rate in the product set, extracts the conversion linkage change amount of related customers, identifies the value performance of customers under different strategy configurations, and arranges the strategy execution order according to the value performance to generate a financial marketing strategy optimization plan.

[0015] The operation management method based on financial marketing big data includes the following steps: S1: Obtain the field contents of transaction, interaction, and market data in financial marketing scenarios, perform consistency judgment based on the semantic matching of field names and the logical distribution of field values, and generate a standardized customer behavior data table; S2: Based on the standardized customer behavior data table, identify the same customer entries in the different sources, determine field conflict information based on data time, source level and field verification status, and generate a unified customer profile table; S3: Call the key fields in the unified customer profile table, combine them with the product tags in the policy library, determine the matching relationship between customer characteristics and policy content, and generate a policy matching list; S4: Based on the urgency level in the policy matching list, the content to be pushed is screened, and channel priority is performed in combination with path delay, node load and packet loss to generate a policy transmission queue; S5: Call the push records in the strategy transmission queue, collect click, conversion and feedback data, combine investment and retention information to determine changes in customer behavior, and generate a financial marketing strategy optimization plan.

[0016] Compared with the prior art, the advantages and positive effects of the present invention are: In the present invention, a data table with a unified structure is constructed through the joint judgment of field semantics and numerical rules to improve data consistency, and customer information integration is performed based on the time, trust level and logical verification of field conflicts to enhance recognition accuracy. The transaction amount, product relevance and interaction frequency are combined with strategy attributes to determine the matching relationship, realize dynamic strategy screening, and push content to filter paths according to the strategy urgency and channel status to improve transmission efficiency. The customer click, conversion and feedback participation behavior trends are determined to support strategy update linkage. The overall processing chain has a closed-loop structure, which improves fusion accuracy and strategy response timeliness under data fragmentation and multi-source heterogeneity conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 is a system flow chart of the present invention; Figure 2 It is a submodule flow chart of the present invention; Figure 3 The figure is a flow chart of the steps of the method of the present invention. DETAILED DESCRIPTION

[0018] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0019] In the description of the present invention, it should be understood that the terms "length," "width," "up," "down," "front," "back," "left," "right," "vertical," "horizontal," "top," "bottom," "inside," "outside," and the like, indicating positions or relationships, are based on the positions or relationships shown in the accompanying drawings and are intended only to facilitate the description of the present invention and simplify the description. They do not indicate or imply that the devices or elements referred to must have a specific orientation, be constructed, or operate in a specific orientation. Therefore, they should not be construed as limiting the present invention. Furthermore, in the description of the present invention, "plurality" means two or more, unless otherwise expressly and specifically defined.

[0020] See also Figure 1 , the operation management system based on financial marketing big data includes: The data standardization module detects user transaction records, social media interaction records, and market trend data in financial marketing scenarios, analyzes the semantic matching degree of field names, verifies whether the field value distribution conforms to business logic rules, and generates standardized customer behavior data tables; The customer profile disambiguation module, based on a standardized customer behavior data table, identifies conflicting information about the same customer across different data sources, compares data update time, data source credibility, and field logic verification results, and generates a unified customer profile table. The strategy matching module uses the transaction amount, product relevance, and interaction frequency in the unified customer profile table, combines it with the target products in the marketing strategy library, calculates the customer value and strategy fit, and generates a strategy matching list; The real-time transmission module selects the customer's unique identification code and policy content to be transmitted based on the policy urgency in the policy matching list, evaluates the transmission path's delay level, packet loss, and node load status, and generates a policy transmission queue; The operation strategy generation module calls the customer click records, conversion rates and market feedback information in the strategy transmission queue, combines the return on investment and customer retention cycle, and generates a financial marketing strategy optimization plan.

[0021] The standardized customer behavior data table includes field semantic mapping results, field value distribution rationality verification results, and unified field format data. The unified customer portrait table includes conflicting information disambiguation results, data credibility level assessment results, and logical consistency verification results. The strategy matching list includes customer value assessment results, strategy fit score, and target strategy plan items. The strategy transmission queue includes the customer's unique identification code, strategy content summary, and transmission path scheduling parameters. The financial marketing strategy optimization plan includes customer click behavior data, conversion effect feedback, and strategy benefit evaluation results.

[0022] See also Figure 2 , the data standardization module includes: The semantic recognition submodule obtains user transaction record fields, social media interaction fields, and market trend fields, calculates the degree of character similarity between field names, the number of combined occurrences of field names in the total data, and the degree of comparison with data source labels. It selects field pairs whose character similarity and combined occurrence exceed the field fit benchmark value and the combination frequency benchmark value, and generates the field semantic consistency matching degree. The semantic recognition submodule extracts required fields from multiple data sources, including fields such as transaction time and amount in user transaction records, comment content and number of reposts in social media, and popularity changes in market trend data. It then evaluates the character similarity between different field names, comparing the "transaction amount" and "transaction value" fields. By calculating the character differences, it identifies fields with a high degree of similarity. It then calculates the frequency of these field names based on whether they appear together in historical data. For example, if a combination appears more than 900 times in 10,000 data points, it is considered to have a high frequency. The field names are also compared with existing data source labels, and the comparison results must meet the set fit criteria. For example, trend fields must be accurately classified into the "market trend" label category. Field combinations that meet high character similarity, high combination frequency, and high fit are selected. For example, "transaction amount" and "popularity change" meet these three conditions and are included in the subsequent analysis set. Finally, based on each field pair's performance in terms of character similarity, combination frequency, and label fit, an overall evaluation of field semantic consistency is calculated using an average method.

[0023] The numerical judgment submodule analyzes the change range, frequency change, and data fluctuation of field values ​​at different time nodes based on the transaction frequency, keyword frequency, and trend indicator values ​​corresponding to the field pairs in the field semantic consistency matching. It compares the overlap between the actual change range of the field value and the logical benchmark range to obtain the field value logical consistency ratio. The specific calculation formula for the change range of the analysis field value at the difference time node is: ; in, Represents the field value in the field pair The change range indicator, Represents a field pair At the time point The field value of Represents a field pair At the time point The field value of Represents a field pair The associated keyword frequency value, Represents a field pair The transaction frequency value during the analysis period, Represents a field pair Trend indicator values ​​within the analyzed period; Detailed explanation of the formula and the calculation process of the formula: The formula involves the following parameters: , , , , The acquisition and calculation of each parameter are as follows: and It is directly collected through the data monitoring system and set as =100 and =150, indicating that an indicator has arrive Increased by 50 units.

[0024] The number of keyword occurrences is obtained by aggregating the data acquisition system and set as =200, based on data monitoring for the first six months.

[0025] Obtained through transaction system records, assuming the data is =75.

[0026] The slope is obtained by analyzing historical data through a specific algorithm, set as =10.

[0027] Substituting these parameters, the calculation process of the formula is as follows: calculate ; Calculating absolute values ; Calculate the denominator ≈8.064; Calculate the final formula result ≈1240.7; The results show that the field Considering the time node arrive During this period, considering the overall frequency and trend of keywords, the field value change index was 1240.7. This value reflects the significant changes in the field pair during the analysis period, providing a quantitative basis for subsequent data analysis and decision-making.

[0028] The behavior integration submodule uses the behavior tags, time tags, and source tags of the fields based on the logical data items in the field value logic compliance ratio. It then merges transaction frequency values, interaction frequency values, and trend indicator values ​​in the same dimension, integrates cross-categories, and unifies the interval format to obtain a standardized customer behavior data table. The behavior integration submodule filters data items with a high logical conformity ratio, retains data pairs that meet the threshold conditions, and adds classification labels to these data items. For example, it labels them as financial platforms or social platforms according to the data source, labels specific quarters or months according to time records, and then labels them as consumer behavior or interactive behavior according to the behavior category. On this basis, data items under the same label are merged. For example, multiple transaction frequency fields come from different sources but all belong to the "consumer behavior" category, which can be merged through the average value. If the data from a certain source is more authoritative, a higher weight can be set to calculate the merged value, and the data are further formatted in a unified manner. For example, all transaction frequencies and trend values ​​are converted to a percentage system, or the value range is compressed to between zero and one, and converted through linear conversion to ensure that various behavioral indicators have the same scale, which is convenient for subsequent horizontal comparison and comprehensive utilization, and finally a standardized behavioral data table with a clear structure and unified content is constructed.

[0029] See also Figure 2 , the customer portrait disambiguation module includes: The field validation submodule obtains data from the customer ID, contact information, and transaction information fields in the standardized customer behavior data table. Based on the field structure, internal code consistency, and filling frequency, it calculates the format matching rate, internal code matching rate, and completeness value. It then determines the difference with the field consistency benchmark value and generates the field consistency deviation rate. First, select customer record data from the past quarter, such as transaction records within 90 days, and extract key field contents such as customer identification, contact information and transaction information field by field. In the structure analysis stage, data processing tools are used to determine the field type and check whether the data value conforms to the basic format specification set for the field. For example, the mobile phone number should be an 11-digit string starting with 1, and the transaction time should follow the "year-month-day" format. If there are abnormal characters or inconsistent structures, it will be marked as format mismatched data. The internal code consistency stage is verified according to internal naming rules. For example, the customer identification should be uniformly prefixed with "CUST" and a fixed number of digits. The system compares the field value structure one by one to see if it is consistent with the standard coding template. If a field that does not conform to the standard format is found, it will be included in the deviation data set. The field integrity is verified through statistics. The number of non-empty records in each field in the data table is compared with the total number of records to obtain the filling ratio of each field in the overall data. If a field has many vacancies, it is considered to have low completeness. The format matching rate and internal code matching rate are also calculated by comparing the number of compliant records with the total number of records. The consistency benchmark values ​​of each field are set, such as format matching must not be less than 95%, completeness must not be less than 90%, and coding consistency rate must be no less than 98%. The current value of each field is compared and analyzed with the corresponding benchmark value. If the current format matching value of a field is lower than the benchmark value standard, it can be determined that there is a deviation in the field consistency. By calculating the difference amplitude of each deviation, the consistency deviation rate of the field can be obtained. After the deviation rates of all fields are summarized, they are used for data optimization and customer normalization operations in subsequent modules.

[0030] The data optimization submodule calls the field consistency deviation rate, combines the field update time interval, the trust level ranking value and the consistency offset value, sorts and filters according to the corresponding trust interval, and generates the filtered field trust level ranking value; The data optimization submodule uses the field consistency deviation rate as a reference, and conducts a comprehensive assessment and analysis based on information such as the field update interval, consistency deviation, and preliminary ranking weight. During the processing, the update time points of each field in the historical data version are counted. The update frequency is calculated by comparing the date of each update. For example, if a contact field is modified multiple times within two months, the field update frequency is high, which in turn affects the field credibility assessment. The deviation value needs to examine the degree of difference between field values ​​in different periods. For example, if there are a large number of inconsistent records between the current version of the customer name field and the previous version, the deviation value is high, which affects the overall credibility judgment. Weight parameters are set for each field. For example, the field consistency comparison result is regarded as the core indicator and given a larger proportion. Then, based on the update time frequency and value deviation, each field is scored according to the weight factor to generate a credibility ranking result. During the ranking process, fields with good consistency, reasonable update frequency, and small deviation values ​​are given higher priority, so that subsequent modules can filter and call fields. Finally, a credibility ranking sequence for each field is output for field merging and unified field optimization processing.

[0031] The customer normalization submodule performs priority matching and conflict screening on the field values ​​of the same customer under different data sources based on the trusted ranking values ​​of the filtered fields, merges the field values, and generates a unified customer profile table; For a customer with multiple field records in different business systems, the trust ranking scores corresponding to these field values ​​are first compared, and the data with higher scores are retained first. For example, when a customer's mobile phone number field has a high score in the main system but a low score in another system, the main system data is selected as the retained value. If multiple field values ​​are found to have obvious conflicts between different sources, such as inconsistencies in the address field content, the system will retain the field value with higher credibility based on the scoring results and delete or weaken the record value with lower score. At the same time, for field values ​​that exist in the system with higher scores but are missing in other systems (such as email addresses), they can be directly supplemented from the high-credibility system into a unified data structure. After the field merge is completed, the system will also perform standardization on the merged field values, such as unifying the field format specifications and removing redundant punctuation. Finally, the integrated field values ​​are written into a unified customer portrait table, so that customer records have a unified field structure and content-preferred data.

[0032] See also Figure 2 , the strategy matching module includes: The transaction feature recognition submodule obtains the transaction amount, product relevance, and interaction frequency from the unified customer profile table, combines these three items for data fusion processing at the customer dimension, and forms a unified representation value by setting an order to obtain the customer behavior intensity value; The transaction feature recognition submodule extracts three indicators, namely transaction amount, product relevance and interaction frequency, from the unified customer portrait information. The transaction amount field is aggregated by reading all transaction records of the customer within a specified period. For example, all consumption records within three consecutive months are accumulated month by month to form the total transaction amount at the customer level. Product relevance is achieved by establishing a matching model between product categories and customer usage behaviors. According to the types of products involved in the customer's historical transactions, each customer is matched and coded with the product tag library, and their preference for different products is evaluated through similarity calculation. The interaction frequency is calculated based on the customer's consultation, clicks, service use, active contact and other behaviors on the platform. The average number of interactions within three months is counted, and the behavior channels and time distribution are recorded. Different weight coefficients are set for the above three indicators, and standardized conversion is performed based on the position of the total customer transaction amount in the entire sample. For example, the highest amount is set as the upper limit and the lowest amount is set as the lower limit. The numerical values ​​are converted to a unified scoring range through linear mapping. The interaction frequency is set with the maximum value as the standard, and the number of interactions of each customer is converted into a score within this range. The product relevance is numerically evaluated based on the similarity calculation results. All scoring items are standardized and weightedly integrated with the set weight coefficients to form a unified behavior intensity score. The range of the behavior intensity score value is divided into intervals according to the fusion score, and the division results are high intensity, medium intensity, and low intensity. The behavior intensity value constitutes a representation sequence on the customer dimension and is stored in the portrait information as the basis for subsequent matching and screening.

[0033] The value fit calculation submodule uses the customer behavior intensity value, combines it with the target product list in the marketing strategy library, and performs item-level screening against the target products based on the product association information in the customer profile. It identifies the density of the corresponding relationship and performs mapping conversion based on the customer behavior intensity value to obtain the strategy fit intensity value. The specific calculation formula for the density of the identification correspondence is: ; Among them, D represents the relationship density value between customers and target products in item-level screening, represents the customer's relevance score for the k-th product feature, represents the set weight score of the target product on the kth product feature, represents the average relevance score of customers on all product features, It represents the average set score of the target product on all product features, n represents the total number of product features, Represents the total weighted matching value between customers and target products in all feature dimensions. It represents the product of the average customer preference value and the average product target value. represents the square root of the product; Detailed explanation of the formula and the process of formula calculation and derivation: Parameter definition and acquisition method: It is derived through analysis of customer behavior data, for example, monitoring the frequency of customer interaction on a certain product feature, purchase history, etc., and quantified using a standardized scoring mechanism with a scoring range of 0 to 10 points.

[0034] Determined by the product strategy team based on market research and product positioning, it reflects the importance of this feature to the overall appeal of the product, with a score range of 0 to 10.

[0035] nThe number of specific dimensions divided according to product characteristics, such as functionality, appearance design, brand influence, etc., a total of 5 items.

[0036] The calculation formula is: ; : The average set weight score of the target product on all product features. The calculation formula is: ; Parameter value setting and basis: :By monitoring and analyzing customers’ past behavior data, we obtained the following scores: Feature 1: 8 points (customers have a high frequency of interaction on this feature); Feature 2: 6 points (customers have a moderate frequency of interaction on this feature); Feature 3: 7 points (customers interact more frequently on this feature); Feature 4: 5 points (customers have an average frequency of interaction on this feature); Feature 5: 9 points (customers interact very frequently on this feature); : The product strategy team sets the following weights based on market research and product positioning: Feature 1: 7 points (this feature is important for product attractiveness); Feature 2: 8 points (this feature is important for product attractiveness); Feature 3: 6 points (this feature has moderate appeal to the product); Feature 4: 7 points (this feature is important for product attractiveness); Feature 5: 9 points (this feature is very important to the product's attractiveness); n: Total number of product features, 5 in total.

[0037] Parameter calculation process: calculate : ; calculate : ; calculate : ; calculate : ; Substitute into the formula to calculate D: ; Result analysis: The calculated D value of 25.73 indicates a high degree of match between the customer and the target product across all characteristic dimensions. A higher value indicates a higher customer preference for the target product and a stronger strategic fit. This result allows for the development of targeted marketing strategies to improve customer conversion rates and satisfaction.

[0038] The strategy matching list generation submodule locates the top-ranked customer identification information based on the strategy fit strength value, summarizes the target product and customer profile number content, and combines them with the strategy fit strength value to generate a strategy matching list; The strategy matching list generation submodule uses the strategy fit strength values ​​obtained in the previous step as the sorting basis, sorting them in descending order by value. Targets with the highest ranking are prioritized. During the screening process, a lower threshold for fit strength values ​​is set, retaining only customer identification information above this threshold. During match list generation, the selected customer's profile ID is used as the primary key field, and each product in the corresponding strategy target product list is grouped together. The customer ID information is then merged with the product list at the field level to generate a structured combination data table. For example, if a customer ID is C001, and the match list matches target products P1, P3, and P5, and their fit score is 0.72, the resulting combination structure is: [C001, P1, P3, P5, 0.72]. This process is completed in batches at the system level using data scripts. Field matching, data merging, and sorting are performed using standard data processing logic. The resulting list serves as a screening list before strategy execution. It contains the customer ID, strategy product combination, and strategy fit strength value. The data table can be updated over time to maintain a dynamic relationship between strategy and customer profile.

[0039] See also Figure 2 , the real-time transmission module includes: The urgency screening submodule obtains the urgency of the policies in the policy matching list and the corresponding customer unique identification code, compares the urgency with the policy urgency threshold, screens the policies and customer identification codes with urgency higher than the threshold, and generates a high-urgency policy tag set; First, each policy item is read from the system database, and its corresponding urgency value and unique customer identifier are extracted. For example, if a policy item is identified as 82, its corresponding customer identifier is CID-20240401. Next, based on the policy's category (such as real-time transaction, data backhaul, or video transmission), the system's urgency threshold for that policy is searched. For example, the urgency threshold for video transmission policies is set to 70, while that for real-time transaction policies is set to 90. By comparing each policy's urgency with the urgency threshold of the policy category, policies with urgency exceeding the threshold are identified and recorded along with their customer identifiers. This comparison process traverses the entire policy list and performs logical analysis on each record. Policy items and their customer identifiers that meet the criteria are extracted and formed into a data tag set. In practice, for example, when processing financial transaction policies in a bank data center, if a policy's urgency is set to 95, exceeding the threshold of 90 for its transaction policy category, its corresponding identifier, CID, is tagged along with the policy ID and included in the high-urgency policy set. When processing multiple policies, they can be read and processed in batches to form a structured table record, such as {CID, policy ID, urgency}, which is used in the subsequent path evaluation stage.

[0040] The path evaluation submodule uses the unique customer identification code in the high-urgency strategy tag set to obtain the delay level, packet loss situation, and node load status of the corresponding transmission path. It compares these with the delay threshold, packet loss rate standard value, and node load benchmark value, records the paths that meet the standards, and generates a set of available path identification codes. After receiving the high-urgency strategy tag set, the path evaluation submodule needs to call the transmission monitoring system or network status interface for each customer unique identifier in the set to obtain the current delay value, packet loss level, and load status of the path nodes of all optional transmission paths. The system sets the upper limit of delay to 80 milliseconds, the upper limit of packet loss ratio to 1%, and the node load baseline to 85%. If the obtained delay value is less than or equal to the upper limit, and the packet loss ratio does not exceed the standard value and the node load is within the baseline value, then the path can be considered to meet the usage standard. Taking customer CID-20240401 as an example, it corresponds to two paths. Path A has a delay of 60 milliseconds, a packet loss rate of 0.5%, and a node load of 68%, all of which are within the standard range; Path B has a delay of 92 milliseconds, a packet loss rate of 1.2%, and a node load of 84%. Both the delay and packet loss values ​​are out of range, so Path A is recorded as an available path. During the entire evaluation process, the system compares the performance parameters of each customer's path with the standard limits one by one. The qualified paths are organized into a set of structured data, which together with the path identifier, customer identification code and corresponding indicators constitute an available path record set, which is subsequently used in the queue scheduling process.

[0041] The queue generation submodule sorts the policies and customer identification codes in the available path identification code set by policy urgency, arranges the transmission order, binds the path identification, policy content and customer identification code, and generates a policy transmission queue; The queue generation submodule receives a set of available path identifiers and prioritizes all policies based on the urgency values ​​recorded in the policies. The sorting operation prioritizes the policies from highest to lowest urgency values. The sorted results are then bound to the path identifiers and customer identification codes obtained in the previous stage, creating a policy queue for transmission. For example, three policies: Policy 1 corresponds to customer CID-10001, with an urgency of 90 and a path identifier of P1; Policy 2 corresponds to customer CID-10002, with an urgency of 85 and a path identifier of P2; and Policy 3 corresponds to customer CID-10003, with an urgency of 70 and a path identifier of P3. The order after sorting is Policy 1, Policy 2, Policy 3. Each queue item consists of a customer identification code, a path identifier, and policy content. Once bound, they form a transmission queue structure, such as the first item {CID-10001, P1, Policy 1 content}. This queue storage structure can be implemented as a key-value table or converted into a read-write data table for invocation by downstream scheduling modules. The queue generation process is based on two input information: path availability judgment and policy urgency ranking, and the policy scheduling order is determined through a double binding mechanism.

[0042] See also Figure 2 , the operation strategy generation module includes: The click behavior extraction submodule obtains customer click records from the strategy transmission queue, identifies the distribution characteristics of click time, page location, and click frequency within the product category, and generates the customer click deviation rate by combining the changes in the number of clicks and the page dwell time within the same time period. When the click behavior extraction submodule retrieves customer click records from the policy transmission queue, it extracts and processes each record item by item. Each click record contains a customer identification code, the time of the click, the page identifier, the click area location code, the address of the page visited, the time when the page was loaded, and the total duration of the page dwell time. After extraction, all records are sorted chronologically to form a continuous click sequence. Next, clicks are categorized and aggregated based on the product tags attached to each click record, such as funds, insurance, and wealth management. The number of page clicks within each category is then counted. To further refine page click characteristics, the page can be divided into the top, content, and bottom areas, and the number of clicks and their proportion in each area are counted. For example, for a certain category of page, the top area accounts for 40% of all clicks. Based on this, the total number of clicks within each period, such as hourly or half-hourly periods, is counted, and the increase or decrease in clicks within each period is recorded. Next, the page dwell time is counted for different product categories, and the mean and median values ​​are used to assess the degree of concentration and fluctuation. For example, the average dwell time for wealth management pages is 38 seconds, with a median of 35 seconds. After integrating the above information, we construct an indicator to measure the stability and drift of customer clicks based on the change in click volume and the fluctuation in page dwell time. This indicator comprehensively considers the fluctuation in click volume and dwell time. For example, if a customer's click volume increases by 20% between 9:00 and 10:00, and the page dwell time fluctuates significantly, the drift indicator will increase accordingly, reflecting the degree of change in the customer's click behavior within the current product category.

[0043] The conversion rate matching calculation submodule uses the customer click deviation rate to identify the conversion time performance and status type in the customer conversion record, extracts the temporal relationship between click behavior and conversion performance, performs matching analysis based on the change difference between the temporal relationship and the customer retention time range, and refers to the fluctuation of the number of conversion statuses to obtain the conversion linkage change; The conversion rate matching calculation submodule first needs to extract information from existing customer conversion records, mainly including the time point when the conversion occurred and the status type of the conversion. Next, this conversion information is associated with the customer's click behavior sequence to form a click-to-conversion chain for each customer. In this chain, the time difference between the last click time point and the first conversion time point is identified. For example, if a customer's last click time is 10:20 am and the conversion time is 10:35 am, the interval time is 15 minutes. This time difference is recorded and combined with the length of time the corresponding customer has been in the system, the relative lag between click behavior and conversion behavior is calculated. For example, if the retention period is 30 days and the conversion interval is 1 day, the difference is small, indicating that the conversion behavior is relatively concentrated. Subsequently, all customers are classified by conversion status type, such as completed, processing, canceled, etc., and the aforementioned time difference data in each category is statistically analyzed to identify the differences between the types. We also need to analyze the changes in the number of each conversion status over a specific time period. For example, if there were 200 "Completed" statuses last week and 250 this week, the fluctuation value is 50. By analyzing fluctuations over multiple periods, we can construct an assessment metric that represents the magnitude of status changes. Combining these three types of information, we can assess the impact of customer conversion interactions. For example, if a customer's click behavior exhibits significant shifts, conversions occur rapidly, and the number of statuses is growing, their combined change value will be high.

[0044] The strategy prioritization submodule uses the conversion linkage change quantity, refers to the investment return rate performance of financial products in the current cycle, and delineates the product set with the return rate in the upper range based on the ranking of the return rate in the product set. It then extracts the conversion linkage change quantity of the associated customers in the set, identifies the value performance of the customers under different strategy configuration structures, arranges the strategy implementation sequence according to the order of value performance, and generates a financial marketing strategy optimization plan; The strategy prioritization submodule prioritizes financial products based on the aforementioned conversion linkage change values, taking into account their return performance within the current cycle. During execution, the current product set is first established, and each product is assigned a corresponding yield, for example, by setting an annualized return within a reference cycle. All products are then sorted from highest to lowest by yield, and a group of products with higher returns is identified, such as the top 20% of all products, as the upper interval. Next, the customer base of these products is counted, and the conversion linkage change for each customer is extracted. The changes across all customers are then summed or averaged to serve as a reference for the customer response strength of the product. For example, if a wealth management product has ten customers and the average change is 1.6, the product's linkage weight is 1.6. All products in the upper interval are ranked according to this value, with higher values ​​indicating stronger customer behavior linkage and higher priority. On this basis, the fluctuation range of the customer linkage value under each high-priority product is extracted, such as the standard deviation and skewness indicators of these values ​​are counted to determine the level of difference under different strategy configuration conditions, and then the entire strategy deployment order is rearranged according to the priority list, and finally an optimized strategy sequence structure suitable for the current product portfolio is generated.

[0045] See also Figure 3 , the operation management method based on financial marketing big data includes the following steps: S1: Obtain the field contents of transaction, interaction, and market data in financial marketing scenarios, perform consistency judgment based on the semantic matching of field names and the logical distribution of field values, and generate a standardized customer behavior data table; S2: Based on the standardized customer behavior data table, identify the same customer entries in different sources, determine field conflict information based on data time, source level and field verification status, and generate a unified customer profile table; S3: Calls the key fields in the unified customer profile table, combines them with the product tags in the policy library, determines the matching relationship between customer characteristics and policy content, and generates a policy matching list; S4: Based on the urgency level in the policy matching list, the content to be pushed is screened, and channel prioritization is performed based on path delay, node load, and packet loss to generate a policy transmission queue. S5: Call the push records in the strategy transmission queue, collect click, conversion and feedback data, combine investment and retention information to determine changes in customer behavior, and generate a financial marketing strategy optimization plan.

[0046] The above are merely preferred embodiments of the present invention and do not limit the present invention in any other form. Any technician familiar with the profession may use the technical content disclosed above to change or modify it into an equivalent embodiment with equivalent changes and apply it to other fields. However, any simple modification, equivalent change and modification made to the above embodiment based on the technical essence of the present invention without departing from the content of the technical solution of the present invention shall still fall within the scope of protection of the technical solution of the present invention.

Claims

1. An operations management system based on financial marketing big data, characterized by: The system comprises: The data standardization module obtains user transaction records, social media interaction records, and market trend data in financial marketing scenarios, analyzes the semantic matching of field names, verifies whether the field distribution complies with business rules, and generates standardized customer behavior data tables; The customer profile disambiguation module, based on the standardized customer behavior data table, identifies conflicting information about the same customer in different data sources, compares data update time, data source credibility level, and field verification results, and generates a unified customer profile table; The strategy matching module calls the transaction amount, product relevance, and interaction frequency in the unified customer portrait table, combines the target product, calculates the customer value and strategy fit, and obtains a strategy matching list; The real-time transmission module screens the unique identification code of the customer to be transmitted and the policy content based on the policy urgency in the policy matching list, performs channel priority sorting according to the transmission path delay level, packet loss situation and node load status, and generates a policy transmission queue; The operation strategy generation module calls the customer click records, conversion rates and market feedback information in the strategy transmission queue, combines the return on investment and customer retention cycle, and generates a financial marketing strategy optimization plan.

2. The operation management system based on financial marketing big data according to claim 1, characterized in that: The standardized customer behavior data table includes field semantic mapping results, field value distribution rationality verification results, and unified field format data; the unified customer portrait table includes conflict information disambiguation results, data credibility level assessment results, and logical consistency verification results; the strategy matching list includes customer value assessment results, strategy fit score, and target strategy plan items; the strategy transmission queue includes customer unique identification code, strategy content summary, and transmission path scheduling parameters; the financial marketing strategy optimization plan includes customer click behavior data, conversion effect feedback, and strategy benefit assessment results.

3. The operation management system based on financial marketing big data according to claim 1 is characterized by: The data standardization module includes: The semantic recognition submodule obtains user transaction record fields, social media interaction fields, and market trend fields, calculates the degree of character similarity between field names, the number of combined occurrences of field names in the total data, and the degree of comparison with data source labels. It selects field pairs whose character similarity and combined occurrence exceed the field fit benchmark value and the combination frequency benchmark value, and generates the field semantic consistency matching degree. The numerical judgment submodule analyzes the change interval, frequency change, and data fluctuation of the field value at the difference time node based on the transaction frequency value, keyword frequency value, and trend indicator value corresponding to the field pair in the field semantic consistency matching degree, compares the overlap between the actual change interval of the field value and the logical reference interval, and obtains the field value logical consistency ratio; The behavior integration submodule calls the behavior label, time label and source label of the field based on the logical data items in the logical conformity ratio of the field value, merges the transaction frequency value, interaction frequency value and trend indicator value in the same dimension, integrates the cross-category and unifies the interval format to obtain a standardized customer behavior data table.

4. The operation management system based on financial marketing big data according to claim 3 is characterized by: The specific calculation formula for the change interval of the analysis field value at the difference time node is: ； in, Represents the field value in the field pair The change range indicator, Represents a field pair At the time point The field value of Represents a field pair At the time point The field value of Represents a field pair The associated keyword frequency value, Represents a field pair The transaction frequency value during the analysis period, Represents a field pair The value of the trend indicator within the analyzed period.

5. The operation management system based on financial marketing big data according to claim 3 is characterized by: The customer profile disambiguation module includes: The field verification submodule obtains the customer identification, contact information, and transaction information field data in the standardized customer behavior data table, calculates the format matching rate, internal code matching rate, and completeness value based on the field structure, internal code consistency, and filling frequency, and determines the difference with the field consistency benchmark value to generate the field consistency deviation rate; The data optimization submodule calls the field consistency deviation rate, combines the field update time interval, the trust level ranking value and the consistency offset value, performs sorting and screening according to the corresponding trust interval, and generates a filtered field trust level ranking value; The customer normalization submodule performs priority matching and conflict elimination on the field values ​​of the same customer under different data sources according to the trusted level ranking value of the filtered fields, merges the field values, and generates a unified customer portrait table.

6. The operation management system based on financial marketing big data according to claim 5 is characterized by: The strategy matching module includes: The transaction feature recognition submodule obtains the transaction amount, product relevance, and interaction frequency from the unified customer portrait table, combines the three items to perform data fusion processing in the customer dimension, forms a unified representation value by setting an order, and obtains the customer behavior intensity value; The value fit calculation submodule calls the customer behavior intensity value, combines it with the target product list in the marketing strategy library, and performs item-level screening against the target products based on the product association information in the customer profile, identifies the density of the corresponding relationship, and performs mapping conversion based on the customer behavior intensity value to obtain the strategy fit intensity value; The strategy matching list generation submodule locates the top-ranked customer identification information based on the strategy fit strength value, summarizes the target product and customer portrait number content, and combines them with the strategy fit strength value to generate a strategy matching list.

7. The operation management system based on financial marketing big data according to claim 6 is characterized by: The specific calculation formula for the density of the identification correspondence is: ； Among them, D represents the relationship density value between customers and target products in item-level screening, represents the customer's relevance score for the k-th product feature, represents the set weight score of the target product on the kth product feature, represents the average relevance score of customers on all product features, It represents the average set score of the target product on all product features, n represents the total number of product features, Represents the total weighted matching value between customers and target products in all feature dimensions. It represents the product of the average customer preference value and the average product target value. Represents the square root of the product.

8. The operation management system based on financial marketing big data according to claim 6, characterized in that: The real-time transmission module includes: The urgency screening submodule obtains the urgency of the policies in the policy matching list and the corresponding customer unique identification codes, compares the urgency with the policy urgency threshold, screens the policies and customer identification codes with urgency higher than the threshold, and generates a high urgency policy tag set; The path evaluation submodule calls the unique customer identification code in the high urgency strategy tag set, obtains the delay level, packet loss situation and node load status of the corresponding transmission path, compares them with the delay threshold, packet loss rate standard value and node load benchmark value respectively, records the paths that meet the standards, and generates a set of available path identification codes; The queue generation submodule sorts the policies and customer identification codes in the available path identification code set by policy urgency, arranges the transmission order, binds the path identifier, policy content and customer identification code, and generates a policy transmission queue.

9. The operation management system based on financial marketing big data according to claim 8, characterized in that: The operation strategy generation module includes: The click behavior extraction submodule obtains customer click records in the strategy transmission queue, identifies the distribution characteristics of click time, page location, and click frequency under product categories, and generates a customer click deviation rate by combining the change in the number of clicks and the page dwell time in the same time period; The conversion rate matching calculation submodule calls the customer click deviation rate, identifies the conversion time and status type in the conversion record, extracts the temporal relationship between click behavior and conversion, and performs matching analysis with the change difference of the customer retention time range. Combined with the fluctuation of the number of conversion statuses, it derives the conversion linkage change amount; The strategy prioritization submodule calls the conversion linkage change amount, refers to the investment return rate of financial products in the current cycle, selects the product set in the high range according to the ranking of the return rate in the product set, extracts the conversion linkage change amount of related customers, identifies the value performance of customers under different strategy configurations, and arranges the strategy execution order according to the value performance to generate a financial marketing strategy optimization plan.

10. The operation management method based on financial marketing big data is characterized by: The operation management system based on financial marketing big data according to any one of claims 1 to 9 is implemented, comprising the following steps: S1: Obtain the field contents of transaction, interaction, and market data in financial marketing scenarios, perform consistency judgment based on the semantic matching of field names and the logical distribution of field values, and generate a standardized customer behavior data table; S2: Based on the standardized customer behavior data table, identify the same customer entries in the different sources, determine field conflict information based on data time, source level and field verification status, and generate a unified customer profile table; S3: Call the key fields in the unified customer profile table, combine them with the product tags in the policy library, determine the matching relationship between customer characteristics and policy content, and generate a policy matching list; S4: Based on the urgency level in the policy matching list, the content to be pushed is screened, and channel priority is performed in combination with path delay, node load and packet loss to generate a policy transmission queue; S5: Call the push records in the strategy transmission queue, collect click, conversion and feedback data, combine investment and retention information to determine changes in customer behavior, and generate a financial marketing strategy optimization plan.

Images