Financial marketing management service method and system based on multi-source data fusion

Abstract

The invention relates to the technical field of financial marketing management, in particular to a financial marketing management service method and system based on multi-source data fusion, and the method comprises the following steps: obtaining a transaction record, calculating an interval, generating a transaction rhythm label, screening a frequency and fluctuation rising customer, determining a high-conversion time period, according to the method, through serialization and mean fluctuation analysis of transaction behavior time intervals, transaction activity changes are dynamically captured, a delay curve is combined to determine a centralized response interval, a high-conversion time period is extracted, and a marketing management scheme is generated. The method comprises the following steps of: performing overlapping proportion analysis on a multi-contact active peak value, generating a priority contact sequence and matching a precise time point, performing multi-source interaction data normalization and sequence rearrangement, optimizing a contact and time putting sequence on a unified time axis, and improving the contact integrating degree and the conversion efficiency.

Description

Technical Field

[0001] The present invention relates to the technical field of financial marketing management, and in particular to a financial marketing management service method and system based on multi-source data fusion. Background Art

[0002] The field of financial marketing management technology refers to a comprehensive methodology system that uses information technology and data analysis to organize and implement planning, resource allocation, channel management, and customer information management involved in the marketing, customer development, relationship maintenance, and sales activities of financial products and services. This field covers core issues such as market segmentation and positioning analysis, customer behavior and value assessment, marketing campaign design and execution, channel operation and optimization management, and marketing performance monitoring and adjustment. Its overall technical development focuses on guiding financial institutions through the collection, integration, and analysis of multiple types of business data to achieve targeted marketing strategy formulation and implementation. Among them, traditional financial marketing management service methods based on multi-source data fusion refer to the unified processing of multiple types of data from different business systems, channel terminals, and external markets to support marketing strategy formulation in financial business marketing management activities. Commonly used data fusion methods include correlation analysis based on customer feature data and transaction record data, joint modeling based on market information data and customer interaction data, pattern matching analysis based on behavioral trajectory data and product attribute data, and trend analysis based on time series characteristics. The correlation between these data and their comprehensive utilization provide a data foundation for the marketing management process.

[0003] In the application of multi-source data fusion, existing technologies lack serialized modeling of the time distribution of transaction behavior, and are unable to identify changes in transaction activity based on rhythm feature dynamics. In response interval extraction, there is a lack of correlation analysis between delay features and concentrated high-conversion time periods. The correspondence between contact activity and delivery timing is not accurately portrayed, resulting in difficulty in matching the delivery sequence with the actual response peak. The time differences in multi-touch data cannot be uniformly processed, which can easily lead to a decline in contact utilization efficiency and a shift in resource allocation, affecting overall marketing accuracy and conversion performance. Summary of the Invention

[0004] In order to solve the technical problems existing in the prior art, an embodiment of the present invention provides a financial marketing management service method based on multi-source data fusion, comprising the following steps: To achieve the above objectives, the present invention adopts the following technical solution: a financial marketing management service method based on multi-source data fusion, comprising the following steps: S1: Obtain the account transaction records of the target customer group in a fixed time window, sort them in ascending order by transaction time, calculate the time intervals between adjacent transactions to form an interval sequence, calculate the mean and fluctuation value respectively, compare the mean difference of adjacent time windows with the transaction rhythm change threshold, and compare the fluctuation value change with the transaction fluctuation threshold, to generate the customer group transaction rhythm label; S2: Filter customer groups with increasing transaction frequency and fluctuation based on the customer group transaction rhythm tag, collect the time of marketing information sending and the first transaction time before the marketing launch, calculate the time difference to generate a response delay curve, compare the curve position difference to obtain the concentrated response period, count the number of transactions in the period, and generate a high-conversion transaction time period; S3: Based on the high-conversion transaction time period, collect the daily activity curves of similar customers at financial service touchpoints, calculate the overlap ratio between the activity peak and the high-conversion period, and sort them to generate a priority touchpoint delivery order; S4: Select time points according to the priority contact delivery order, calculate the time difference with the peak of the target customer delay curve, filter the matching time points, and generate a delivery time list.

[0005] As a further solution of the present invention, the customer group transaction rhythm label includes the transaction mean change range, the transaction fluctuation value change range, and the transaction rhythm type; the high conversion transaction time period includes the concentrated response period, the number of concentrated response transactions, and the transaction conversion peak period; the priority contact delivery order includes the contact overlap ratio, contact priority, and contact identification; the delivery time list includes the matching time point, the time difference range, and the priority delivery time point.

[0006] As a further solution of the present invention, the specific steps of S1 are: S101: Obtain account transaction records of a target customer group within a fixed time window, arrange them in ascending order by transaction time, extract the time intervals between adjacent transactions and integrate them into a sequence to generate a transaction time interval sequence; S102: extracting central level and stability features based on the transaction time interval sequence, and integrating the two into a feature set to obtain transaction interval statistical features; S103: Calling the transaction interval statistical features, comparing the center level and stability changes of adjacent time windows, combining the transaction rhythm and fluctuation threshold, and establishing a customer group transaction rhythm label.

[0007] As a further solution of the present invention, the specific steps of S2 are: S201: Obtain the transaction rhythm labels of the customer groups, screen the customer groups with increasing transaction frequencies and increasing fluctuations, integrate the customer group identifiers that meet the conditions into a set, and generate a target customer group set; S202: Based on the target customer group set, the marketing information sending time and the first completed transaction time of the customer group before the multi-channel marketing launch are collected, a time difference calculation is performed on the two types of time data, and the data are arranged in order by customer group to draw a response delay curve; S203: calling the response delay curve, comparing the position difference of the curve on the time axis to obtain the concentrated response period, and counting the number of transactions within the period, combining the concentrated response period and the number of transactions to establish a high conversion transaction time period.

[0008] As a further solution of the present invention, the specific steps of S3 are: S301: Based on the high-conversion transaction time period, collect the daily activity curve of the online financial service touchpoints of similar customers, identify the peak position in the curve and record the corresponding time point, integrate the peak time information according to the touchpoint sequence, and generate a touchpoint peak time set; S302: Calling the touch point peak time set, performing time interval overlap calculation on the peak time and the conversion transaction time period, calculating the touch point overlap percentage, and archiving the percentage results in the order of the touch points to obtain a touch point overlap percentage sequence; S303: According to the contact overlap ratio sequence, sort the ratio values ​​from large to small to generate a priority contact placement order.

[0009] As a further solution of the present invention, the specific steps of S4 are: S401: Calling the priority contact placement order, selecting the contacts with the highest ranking, obtaining the corresponding time point data, and arranging the contact time points into a set in chronological order to generate a priority contact active time set; S402: Based on the priority contact active time set, extract the peak position of the target customer delay curve, calculate the time difference with the time point, compare the difference result with the matching threshold, filter the time points that meet the threshold condition, and obtain the matching time point set; S403: Call the matching time point set, arrange them in chronological order, remove duplicate records, and establish a delivery time list.

[0010] As a further embodiment of the present invention, the method further comprises: S5: Based on the delivery time list, the start, peak, and end time differences of the customer touchpoint interaction time at multiple acquisition terminals are calculated and normalized to a unified time axis, the events are rearranged, and a financial marketing management service plan is generated.

[0011] As a further solution of the present invention, the financial marketing management service solution includes a multi-touch event sequence under a unified time axis, a touch interaction time distribution, and differentiated adjustment results between touch points.

[0012] As a further solution of the present invention, the specific steps of S5 are: S501: According to the delivery time list, collect customer touchpoint interaction time recorded by multiple collection terminals, obtain the start time, peak time, and end time in sequence, calculate the difference between the three types of time, classify and organize the calculation results by collection terminal, and generate a touchpoint time difference set; S502: Based on the contact time difference set, identifying time difference records with differences between the collection terminals, normalizing and adjusting the differentiated time differences according to a unified time axis reference value, and integrating the adjusted records into a unified structure to obtain a unified time axis record set; S503: Calling the unified timeline record set, rearranging the multi-touch events according to the adjusted time sequence, integrating them into executable marketing management content, and establishing a financial marketing management service solution.

[0013] A financial marketing management service system based on multi-source data fusion, the system comprising: The trading rhythm analysis module is used to perform S1: obtain the account transaction records of the target customer group in a fixed time window, sort them in ascending order by transaction time, calculate the time intervals between adjacent transactions to form an interval sequence, calculate the mean and fluctuation value respectively, compare the mean difference of adjacent time windows with the trading rhythm change threshold, and compare the fluctuation value change with the trading fluctuation threshold, to generate the customer group trading rhythm label; The frequency fluctuation screening module is used to perform S2: screening customer groups with increasing transaction frequency and fluctuation based on the transaction rhythm tag of the customer group, collecting the time of sending marketing information and the time of first transaction before marketing launch, calculating the time difference to generate a response delay curve, comparing the position difference of the curve to obtain a concentrated response period, counting the number of transactions in the period, and generating a high-conversion transaction time period; The contact point prioritization module is used to execute S3: based on the high conversion transaction time period, collect the daily activity curves of similar customers at financial service contact points, calculate the overlap ratio between the activity peak and the high conversion period, and sort and generate a priority contact point delivery order; The delivery time matching module is used to execute S4: selecting a time point according to the priority contact delivery order, calculating the time difference with the peak of the target customer delay curve, screening the matching time point, and generating a delivery time list; The multi-touch time integration module is used to execute S5: based on the delivery time list, calculate the start, peak, and end time differences of the customer touch interaction time of multiple acquisition terminals and normalize them to a unified time axis, rearrange the events, and generate a financial marketing management service plan.

[0014] Compared with the prior art, the advantages and positive effects of the present invention are: In the present invention, through the serialization of transaction behavior time intervals and mean fluctuation analysis, the changes in transaction activity are dynamically captured, the concentrated response interval is determined in combination with the delay curve, and the high conversion time period is extracted. The overlapping ratio is analyzed with the multi-touch activity peak, and a priority contact sequence is generated and matched with the precise time point. After normalization and sequence rearrangement of multi-source interaction data, the contact and time delivery order are optimized on a unified time axis to improve the touch fit and conversion efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0016] Figure 1 Schematic diagram of the steps of the present invention; Figure 2 This is a schematic diagram of the refinement of S1 of the present invention; Figure 3 This is a schematic diagram of the refinement of S2 of the present invention; Figure 4 This is a schematic diagram of the refinement of S3 of the present invention; Figure 5 This is a schematic diagram of the refinement of S4 of the present invention; Figure 6 This is a schematic diagram of the refinement of S5 of the present invention; Figure 7 It is a system module diagram of the present invention. DETAILED DESCRIPTION

[0017] The technical solution of the present invention is described below in conjunction with the accompanying drawings.

[0018] In the embodiments of the present invention, words such as "exemplarily" and "for example" are used to indicate examples, illustrations, or explanations. Any embodiment or design described as an "exemplary" in the present invention should not be interpreted as being preferred or advantageous over other embodiments or designs. Rather, the use of the word "exemplary" is intended to present concepts in a concrete manner. Furthermore, in the embodiments of the present invention, "and / or" can mean both or either of the two.

[0019] In the embodiments of the present invention, the terms "image" and "picture" may sometimes be used interchangeably. It should be noted that, when the distinction is not emphasized, the meanings they convey are the same. The terms "of," "corresponding," and "corresponding" may sometimes be used interchangeably. It should be noted that, when the distinction is not emphasized, the meanings they convey are the same.

[0020] In the embodiments of the present invention, sometimes a subscript such as W1 may be written as a non-subscript such as W1. When the difference is not emphasized, the meanings to be expressed are the same.

[0021] In order to make the technical problems, technical solutions and advantages to be solved by the present invention clearer, a detailed description will be given below with reference to the accompanying drawings and specific embodiments.

[0022] See also Figure 1 The embodiment of the present invention provides a financial marketing management service method based on multi-source data fusion, comprising the following steps: S1: Obtain the target customer group's account transaction behavior records within a fixed time window, sort the records in ascending order by transaction time, calculate the time intervals between adjacent transactions and summarize them into an interval sequence, calculate the mean and fluctuation value based on the interval sequence, compare the mean difference of adjacent time windows with the transaction rhythm change threshold, and compare the fluctuation value change with the transaction fluctuation threshold to generate the customer group's transaction rhythm label; S2: Filter customer groups with increasing transaction frequency and volatility based on their transaction rhythm tags. Collect the time marketing information was sent and the time the first transaction was completed before the multi-channel marketing campaign was launched. Calculate the time difference between the two and draw a response delay curve. Compare the position differences on the delay curves to obtain concentrated response periods. Count the number of transactions within these concentrated response periods to generate high-conversion transaction time periods. S3: Based on high-conversion transaction time periods, collect daily activity curves of online financial service touchpoints for similar customers. Calculate the overlap ratio between the activity peak position and the high-conversion transaction time period, sort the touchpoints based on the ratio, and generate a priority touchpoint delivery order. S4: Call the priority contact delivery order to select the time point corresponding to the top-ranked contact, calculate the time difference between the time point and the peak position of the target customer's delay curve, select the matching time points based on the difference, and generate a delivery time list; S5: Calculate the difference between the start, peak, and end times of customer touchpoint interaction times recorded by multiple acquisition terminals based on the delivery time list, normalize the time differences of differentiated acquisition terminals to a unified time axis, rearrange multiple touchpoint events according to the adjusted time, and generate a financial marketing management service plan.

[0023] The transaction rhythm labels of the customer group include the change range of the transaction mean, the change range of the transaction fluctuation value, and the transaction rhythm type. The high-conversion transaction time period includes the concentrated response period, the number of concentrated response transactions, and the transaction conversion peak period. The priority contact delivery sequence includes the contact overlap ratio, contact priority, and contact identification. The delivery time list includes the matching time point, time difference range, and priority delivery time point. The financial marketing management service plan includes the multi-touch event sequence under the unified time axis, the contact interaction time distribution, and the differentiated adjustment results between contacts.

[0024] See also Figure 2 , the specific steps of S1 are: S101: Obtain account transaction records of a target customer group within a fixed time window, arrange them in ascending order by transaction time, extract the time intervals between adjacent transactions and integrate them into a sequence to generate a transaction time interval sequence; When obtaining the account transaction records of the target customer group within a fixed time window, first clarify the scope of the time window, for example, set it to 24 hours from 0:00 to 24:00 every day, and then retrieve the transaction details of the target customer group within these hours from the transaction database. The details should include fields such as customer identification, transaction time, transaction amount, and transaction type. Convert the transaction time into a unified second timing format for easy comparison. Then, sort the transaction records in each time window in ascending order of seconds. After the sorting is completed, read the time difference between two adjacent transactions in sequence. For example, the interval between 09:10:15 and 09:25:45 is 930 seconds, and the interval between 09:25:45 and 09:45:30 is 1185 seconds. Seconds, in this way, the interval time of all adjacent transactions of the customer in the current window is calculated one by one, and these intervals are arranged in sequence. If a customer has only one transaction in the time window, a null value or zero value is directly recorded to ensure the integrity of the data structure. The interval sequences of all customers are aggregated into a two-dimensional list by customer ID. Each row represents all intervals of a customer in the time window, and each column represents the interval value at a different position. During the generation process, it is necessary to eliminate duplicate transaction records and records that exceed the time window range, such as records with timestamps before or after the start or end of the window, and clean up abnormal timestamps (such as records with timestamps less than 0 or exceeding 86400 seconds). Finally, the sequence of adjacent transaction time intervals calculated and arranged in ascending order by transaction time is obtained.

[0025] S102: Based on the transaction time interval sequence, extract the central level and stability features, and integrate the two into a feature set to obtain the transaction interval statistical features; When extracting the central level and stability features based on the transaction time interval sequence, first calculate the central level for each customer's interval sequence in the current window. The central level can be directly expressed as the average value of all intervals. For example, if the intervals of a customer in a certain window are 930 seconds, 1185 seconds, and 870 seconds, the central level is 995 seconds. Then calculate the stability. The stability is expressed by the degree of interval fluctuation. It can be obtained by taking the absolute value of the difference between each interval and the central level and then calculating the average value. For example, the difference between the aforementioned interval and the central level is 65 seconds, 190 seconds, and 125 seconds, respectively. The average value is 126.67 seconds, and the stability value of the customer in the window is obtained. It is necessary to divide the central level and stability into intervals. The interval division is set based on the statistical results of historical trading behavior. For example, customers with a central level between 0 and 600 seconds are classified as high-frequency trading, between 600 and 1800 seconds are classified as medium-frequency trading, and greater than 1800 seconds are classified as low-frequency trading. Stability between 0 and 100 seconds is judged as low volatility, between 100 and 500 seconds is judged as medium volatility, and greater than 500 seconds is judged as high volatility. These intervals can be set through distribution analysis of historical data. For example, the percentile points of all customer interval data in the past year are selected as interval dividing points. The extracted central level and stability values ​​are combined into a set of features and stored in the feature set for subsequent analysis.

[0026] S103: Calling the statistical features of transaction intervals, comparing the center level and stability changes of adjacent time windows, combining the transaction rhythm and fluctuation threshold, and establishing a transaction rhythm label for the customer group; When calling the transaction interval statistical feature to compare the changes in the center level and stability of adjacent time windows, it is necessary to first read the center level value and stability value of each customer in two consecutive time windows, and then subtract the value of the previous window from the value of the latter window to obtain the change. For example, the center level of customer A in the previous window is 995 seconds, and the next window is 1200 seconds, then the center level change is 205 seconds. Its stability is 126.67 seconds in the previous window and 200 seconds in the next window, then the stability change is 73.33 seconds. Then compare the change with the preset rhythm change threshold. The setting of the threshold can refer to the distribution of historical data changes. For example, the 75th percentile value of the center level change in the past year is taken as the threshold. Assuming that the center level threshold is 300 seconds and the stability threshold is 150 seconds, a change greater than the threshold is considered a significant change, otherwise it is considered no significant change. Labels are then divided according to the combination of changes. For example, if both the center level and stability changes exceed the threshold, it is marked as "rapid rhythm fluctuation", if the center level change does not exceed the threshold but the stability change exceeds the threshold, it is marked as "stable rhythm fluctuation", if the center level change exceeds the threshold but the stability change does not exceed the threshold, it is marked as "fast and stable rhythm", and if both do not exceed the threshold, it is marked as "stable and steady rhythm". Finally, the label corresponding to each customer in each time window is recorded in the transaction rhythm label set to form complete rhythm change classification data.

[0027] See also Figure 3 , the specific steps of S2 are: S201: Obtain customer group transaction rhythm labels, screen customer groups with increasing transaction frequency and increasing volatility, integrate customer group identifiers that meet the conditions into a set, and generate a target customer group set; When obtaining the transaction rhythm tag of a customer group, it is necessary to first read the tag value of each customer from the generated transaction rhythm tag set according to the customer ID, and then judge the tag content of each customer. The screening conditions are set as increased transaction frequency and increased volatility. The method for judging the increase in transaction frequency is that the change in the center level value of two adjacent time windows is less than zero. For example, the center level of the previous window is 1200 seconds, and the center level of the next window is 900 seconds. The change is 300 seconds. If it is less than zero, it is judged that the frequency has increased. The method for judging the increase in volatility is that the change in stability is greater than zero. For example, if the previous window is stable, the center level is 1200 seconds, and the center level of the next window is 900 seconds. The change is 300 seconds. If it is less than zero, it is judged that the frequency has increased. The method for judging the increase in volatility is that the change in stability is greater than zero. The duration is 100 seconds, the next window is 150 seconds, and the change is 50 seconds. If it is greater than zero, it is determined that the fluctuation has increased. When screening, both conditions must be met at the same time to select the target customer group. In addition, a change validity threshold must be set to prevent interference due to slight fluctuations. The threshold can be set according to the change distribution of historical transaction data for one year. For example, the absolute value of the change greater than 60 seconds is considered a valid change, and less than 60 seconds is considered no change. The threshold value can be determined by rounding the standard deviation of the historical sample change. Customer identifiers that meet the conditions need to be deduplicated and stored in a new set, which is the target customer group set.

[0028] S202: Based on the target customer group set, the marketing information sending time and the first completed transaction time of the customer group before the multi-channel marketing launch are collected, a time difference calculation is performed on the two types of time data, and the data are arranged in order by customer group to form a response delay curve; When collecting marketing information sending time and first completed transaction time for each target customer group before multi-channel marketing is launched based on a target customer group, it is necessary to first extract the marketing sending time for each target customer from the multi-channel sending records. The multi-channel range includes SMS, in-site messages, push notifications, etc. For each customer, the earliest sending time within the analysis period is selected as the customer's marketing sending time. Then, the first transaction time of the customer within the analysis period is selected from the transaction records, ensuring the time accuracy to the second. Then, the marketing sending time is subtracted from the first transaction time to obtain the response delay value. For example, if the marketing sending time is 09:00:00 and the first transaction time is 09:45:00, the delay time is 2700 seconds. The delay times of all customers are arranged in ascending order according to customer ID to form a delay time series. This series is then mapped to curve coordinates, with the horizontal axis representing the customer sequence number and the vertical axis representing the delay time in seconds. The points are connected to form a response delay curve, where each point on the curve represents the response delay data of a customer.

[0029] S203: Calling the response delay curve, comparing the position difference of the curve on the time axis to obtain the concentrated response period, and counting the number of transactions within the period. The concentrated response period and the number of transactions are combined to establish a high conversion transaction time period; When calling the response delay curve to compare the position difference of the curve on the time axis, first scan the curve data to identify the interval where the delay time values ​​are concentrated, and set the judgment threshold for concentrated response to the proportion of customers in this interval to the total number of customers reaching or exceeding 20%. This proportion threshold can be determined based on the proportion of concentrated periods in historical activity data. For example, after multiple activity analyses, it is found that the average proportion of concentrated periods is around 18%. The threshold can be set to 20%. During the scanning process, if it is found that the proportion of customers with delay times between 1800 seconds and 3600 seconds reaches 25%, then this interval is determined to be a concentrated response period. Then, the number of transactions in this period is counted. For example, there are 50 customers who complete transactions in this time period, and some customers have multiple transactions. The total number of transactions is counted as 65 times. The start delay time, end delay time and number of transactions of the concentrated response period are combined into a set of data records, and multiple such records are generated in sequence to eventually form a set of high-conversion transaction time periods.

[0030] See also Figure 4 , the specific steps of S3 are: S301: Collect daily activity curves of online financial service touchpoints for similar customers based on high-conversion transaction time periods, identify peak positions in the curves and record corresponding time points, integrate peak time information by touchpoint sequence, and generate a touchpoint peak time set; When collecting the daily activity curve of online financial service touchpoints of similar customers based on the high-conversion transaction time period, it is necessary to first count the number of active times of the target customer group in different time periods of the day for each touchpoint. The statistics of the active times are accurate to minutes or seconds, and the time and active times are recorded to generate curve data. Then, the activity values ​​of the touchpoints within a day are arranged in chronological order to form an activity curve array. For example, the activity values ​​of touchpoint A from 0 seconds to 3600 seconds are 5, 12, 18, 10, and 7, respectively. The peak activity is 18, corresponding to the time point The activity curve of the contact is traversed to find all peak points, that is, the current activity value is greater than the activity value of the previous moment and the next moment. If there are multiple peaks, all peak positions and corresponding time points are recorded. Then the peak time point information of all contacts are integrated in the order of contacts. Each contact corresponds to one or more peak time points. For example, the peak value of contact A is 1800 seconds, the peak values ​​of contact B are 2400 seconds and 4200 seconds, and the peak value of contact C is 3000 seconds. These time points are combined and stored to form a contact peak time set.

[0031] S302: Calling the touch point peak time set, calculating the time interval overlap between the peak time and the conversion transaction time period, calculating the touch point overlap percentage, and archiving the percentage results in the order of the touch points to obtain a touch point overlap percentage sequence; The calculation formula for the percentage of contact overlap is as follows: ; in, represents the percentage of overlap of the j-th contact point, represents the number of peak time points of the j-th contact, Indicates the duration of the conversion transaction period at the i-th peak time of the j-th touchpoint. represents the time value of the i-th peak time point of the j-th contact, Indicates the time value of the center point of the conversion transaction period that matches the i-th peak time point of the j-th touchpoint. Indicates the discriminant factor for whether the peak time of the j-th contact i and its matching time period overlap effectively. Take 1, otherwise take 0, represents the confidence weight coefficient of the i-th peak time point of the j-th contact, Represents the sum of the durations of all conversion transaction periods matched to the j-th touchpoint.

[0032] The operation logic in this formula is as follows: First, the absolute value Characterizes the time distance of the peak from the center position, and then uses this value from the duration Deduct from the middle to reflect the effective overlap time, and then multiply by the confidence level and discriminant factors , filter out invalid or large offset peak contributions, and finally accumulate the effective overlap of all peaks and then add them to the total duration of all segments Compare the values ​​to get the overlap ratio, and finally convert it into a percentage.

[0033] Collection and calculation examples: Taking contact B as an example, the peak activity time points collected on that day are as follows: The first peak: Second; The second peak: Second; The third peak: Second; These peaks correspond to high-converting transaction periods: Time period 1: start and end , 、 Second; Time period 2: start and end , 、 Second; Time period 3: start and end , 、 Second; According to the number of user interactions within the ±300-second window of the peak, the interaction volume of each peak neighborhood is obtained as follows: Peak 1: 110 times (maximum); Peak 2: 95 times; Peak 3: 80 times; From this we can conclude that: , , ; Then calculate the peak offset: ; ; ; Substitute each valid overlap value into the calculation: Item 1: ; Item 2: ; Item 3: ; The total value of the numerator is: ; Total duration : ; Substituting into the formula we get: ; Table 1 Contact B parameter collection and formula calculation table

[0034] As shown in Table 1, all three peaks overlap effectively, with a total confidence-weighted contribution of 1403.20 seconds. Using a normalized denominator of 2400 seconds, the final calculated overlap ratio is 58.47%.

[0035] Comparison and Conclusion: Assumed baseline value , which is set based on the median overlap ratio of the 12 historical marketing campaigns. , it indicates that the contact is slightly lower than the benchmark, and its priority may be ranked below the median in the subsequent contact sorting.

[0036] Formula innovation description: The benefit of the formula is that by introducing confidence weights Achieved quantitative expression of peak behavior intensity, combined with offset absolute value calculation and discriminant factor By controlling the timing and effectiveness of the peak, and ultimately deriving a unified comparison benchmark by normalizing the total duration, the active conversion overlap capability of multiple touchpoints can be quantified and evaluated. This design provides a clear numerical basis for the subsequent ranking of delivery strategies.

[0037] S303: Based on the contact overlap ratio sequence, sort the values ​​from large to small according to the ratio values ​​to generate a priority contact placement order; When composing a sequence based on the contact overlap ratio, the contacts in the sequence must first be compared according to the overlap ratio values, and the values ​​must be rearranged from largest to smallest. The comparison process is to take out the overlap ratios of two contacts in turn. If the ratio of the first contact is greater than that of the second contact, the first contact is placed in front. When the ratios are equal, the order can be determined based on the lexicographic order of the contact identifiers or randomly. For example, if the overlap ratio of contact A in the sequence is 80%, contact B is 66.67%, and contact C is 50%, then the sorting results are A, B, and C. After the sorting is completed, the new order is recorded as the priority contact delivery order. Each element in the order list is a contact identifier, and the position represents the priority, and finally the priority contact delivery order is generated.

[0038] See also Figure 5 , the specific steps of S4 are: S401: Call the priority contact placement order, select the contacts with the highest ranking, obtain the corresponding time point data, and organize the contact time points into a set in chronological order to generate a priority contact active time set; When calling the priority contact delivery order, first read the generated priority contact delivery order list, and select the contacts with the highest ranking starting from the first place in the list, for example, select the top 3 contacts A, B and C, and then enter the activity database or log file to obtain the time point data of these contacts one by one. The time points must be recorded in seconds and ensure time accuracy. For example, the time points of contact A are 1200 seconds, 2400 seconds, and 3600 seconds, contact B is 1800 seconds and 3000 seconds, and contact C is 1500 seconds, 2700 seconds, and 3900 seconds. Then, these time points are rearranged from small to large according to the time value. During the arrangement process, the contact identification information corresponding to the time point must be retained for subsequent tracing. At the same time, check whether there are duplicate time points and merge them. Finally, the sorted time point data is stored in a set in chronological order. This set is the priority contact active time set.

[0039] S402: Based on the priority touchpoint active time set, extract the target customer's delay curve peak position, calculate the time difference with the time point, compare the difference result with the matching threshold, filter the time points that meet the threshold condition, and obtain the matching time point set; When extracting the peak position of the target customer's delay curve based on the priority contact active time set, it is necessary to first identify the peak point from the delay curve data. The peak judgment condition is that the delay value at this time point is greater than the delay values ​​of the two adjacent time points. After identifying all peak positions, the peak time points are extracted according to the customer or group dimension. For example, the peak time of the target customer's delay curve is 1300 seconds, 2500 seconds, and 3700 seconds. Then, the difference between each peak time point and each time point in the priority contact active time set is calculated. The calculation method is to take the absolute difference. For example, the difference between the peak 1300 seconds and the active point 1200 seconds is 10 0 seconds, the difference between the peak time of 2500 seconds and the active point of 2400 seconds is 100 seconds, and the difference between the peak time of 3700 seconds and the active point of 3600 seconds is 100 seconds. After the calculation is completed, the difference is compared with the matching threshold. The matching threshold can be set based on the synchronization distribution of historical delays and active times. For example, if the median distribution of the difference between 50% of the peak times and active points in past activities is 150 seconds, the threshold is set to 150 seconds. Points with a difference less than or equal to the threshold are determined to be matching points, and the time points that meet the conditions are recorded in the matching time point set. Each element in the set contains the matching peak time and the corresponding active time.

[0040] S403: Call the matching time point set, arrange them in chronological order, remove duplicate records, and create a delivery time list; When calling the matching time point set, all matching time points in the set must first be re-sorted from small to large by time value. During the sorting process, the time point values ​​are directly compared. For example, if the time points in the set are 1200 seconds, 1300 seconds, 2500 seconds, 2400 seconds, and 3700 seconds, then after sorting they are 1200 seconds, 1300 seconds, 2400 seconds, 2500 seconds, and 3700 seconds. After the sorting is completed, it is necessary to check whether there are duplicate records. The duplication judgment condition is that the time point values ​​are the same and the corresponding peak values ​​are consistent with the active point sources. If there are duplicates, one record is retained and the redundant records are deleted. The deduplicated time points are stored in the delivery time list in sequence according to the sorting results. Each time point in the list corresponds to a specific delivery opportunity, and finally a complete delivery time list is obtained.

[0041] See also Figure 6 , the specific steps of S5 are: S501: Based on the delivery time list, collect customer touchpoint interaction time recorded by multiple collection terminals, obtain the start time, peak time, and end time in sequence, and calculate the difference between the three types of time. The calculation results are sorted by collection terminal to generate a touchpoint time difference set; When collecting customer touchpoint interaction time recorded by multiple collection terminals according to the delivery time list, it is necessary to first read the time points in the delivery time list one by one, and use the time point as the index to extract the interaction records of the corresponding touchpoints from the interaction logs of multiple collection terminals. Each record must contain the start time, peak time and end time. The start time is the time when the touchpoint first triggers the interaction, the peak time is the time when the activity reaches the maximum value during the interaction process, and the end time is the time when the touchpoint interaction is completed or interrupted. For example, the start time of a customer in collection terminal 1 is 1200 seconds, the peak time is 1500 seconds, and the end time is 1800 seconds. The start time of the same touchpoint in collection terminal 2 is 1200 seconds, the peak time is 1500 seconds, and the end time is 1800 seconds. The start time is 1250 seconds, the peak time is 1550 seconds, and the end time is 1820 seconds. Then, the start time and peak time of each record in the same collection end are subtracted to obtain the peak delay, the peak time and end time are subtracted to obtain the end delay, and the start time and end time are subtracted to obtain the total interaction duration. For example, the peak delay of collection end 1 is 300 seconds, the end delay is 300 seconds, and the total duration is 600 seconds. These difference results are sorted according to the collection end, and the difference results of each collection end are stored separately. Finally, a contact time difference set is generated. This set is based on the collection end as the dimension, and each dimension contains three types of time difference data for all contacts in the end.

[0042] S502: Based on the contact time difference set, identify time difference records with differences between the collection terminals, normalize and adjust the differentiated time differences according to the unified time axis reference value, and integrate the adjusted records into a unified structure to obtain a unified time axis record set; When identifying time difference records with differences between collection terminals based on the contact time difference set, first compare the time differences of the same contacts in different collection terminals one by one. The comparison method is to directly calculate the absolute difference. For example, the peak delays of collection terminals 1 and 2 on the same contact are 300 seconds and 350 seconds respectively, and the difference is 50 seconds. If the difference exceeds the preset difference threshold, it is determined that there is a difference. The difference threshold can be set by analyzing the consistency of historical multi-terminal records. For example, the average and standard deviation of all historical differences are calculated, and the average plus the standard deviation are rounded as the threshold. Assuming it is 40 seconds, 50 seconds in this example has exceeded the threshold. The values ​​are judged to be difference records. For records judged to be difference, they need to be normalized and adjusted according to the unified time axis benchmark value. The benchmark value can be set to the average of the earliest start time in all collection end records. For example, after counting the earliest start time of multiple contacts at different ends, the average value is 1000 seconds. Then the time difference of each end is proportionally adjusted to the time axis with 1000 seconds as the starting point. The proportional relationship between the time differences needs to be retained during the adjustment. The adjusted data is archived by contact and integrated into data records with a unified structure. Finally, a unified time axis record set is obtained. All contact time data in this record set have been aligned under the same benchmark.

[0043] S503: Calling the unified timeline record set, rearranging the multi-touch events according to the adjusted time sequence, integrating them into executable marketing management content, and establishing a financial marketing management service plan; When calling the unified timeline record set, multiple touch events need to be rearranged in the adjusted time order. Specifically, first extract the start time of all touch points in the record set and sort them in ascending order by time value. When the start time is the same, sort by peak time. If the peak time is also the same, sort by end time. For example, if the start time of touch point A is 1100 seconds, that of touch point B is 1150 seconds, and that of touch point C is 1130 seconds, then the sorting results are A, C, and B. After the sorting is completed, the detailed information of these touch points are integrated in order, including start time, peak time, end time, and time difference data. The time relationship between adjacent touch points is retained to form a continuously executable event chain. Finally, these sequentially arranged touch event sets are organized into structured marketing management content. Each record contains a touch point identifier and its corresponding three types of time information, ultimately establishing a complete financial marketing management service solution.

[0044] See also Figure 7 , a financial marketing management service system based on multi-source data fusion, including: The trading rhythm analysis module is used to perform S1: obtain the account transaction records of the target customer group in a fixed time window, sort them in ascending order by transaction time, calculate the time intervals between adjacent transactions to form an interval sequence, calculate the mean and fluctuation value respectively, compare the mean difference of adjacent time windows with the trading rhythm change threshold, and compare the fluctuation value change with the trading fluctuation threshold, to generate the customer group trading rhythm label; The frequency fluctuation screening module is used to perform S2: screening customer groups with increasing transaction frequency and fluctuation based on their transaction rhythm tags. The module collects the time of marketing information delivery and the time of the first transaction before the marketing launch, calculates the time difference to generate a response delay curve, compares the position difference of the curve to obtain the concentrated response period, counts the number of transactions during the period, and generates a high-conversion transaction time period. The touchpoint prioritization module is used to perform S3: Based on the high-conversion transaction time period, collect the daily activity curves of similar customers at financial service touchpoints, calculate the overlap ratio between the activity peak and the high-conversion period, and sort them to generate a priority touchpoint delivery order; The delivery time matching module is used to perform S4: select time points according to the priority contact delivery order, calculate the time difference with the peak of the target customer's delay curve, select matching time points, and generate a delivery time list; The multi-touchpoint time integration module is used to execute S5: Based on the delivery time list, the start, peak, and end time differences of the customer touchpoint interaction time of multiple acquisition terminals are calculated and normalized to a unified time axis, events are rearranged, and a financial marketing management service plan is generated.

[0045] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any modifications or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in the present invention should be included in the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be based on the scope of protection of the claims.

Claims

1. A financial marketing management service method based on multi-source data fusion, characterized in that: The following steps are involved: S1: Obtain the account transaction records of the target customer group in a fixed time window, sort them in ascending order by transaction time, calculate the time intervals between adjacent transactions to form an interval sequence, calculate the mean and fluctuation value respectively, compare the mean difference of adjacent time windows with the transaction rhythm change threshold, and compare the fluctuation value change with the transaction fluctuation threshold, to generate the customer group transaction rhythm label; S2: Filter customer groups with increasing transaction frequency and fluctuation based on the customer group transaction rhythm tag, collect the time of marketing information sending and the first transaction time before the marketing launch, calculate the time difference to generate a response delay curve, compare the curve position difference to obtain the concentrated response period, count the number of transactions in the period, and generate a high-conversion transaction time period; S3: Based on the high-conversion transaction time period, collect the daily activity curves of similar customers at financial service touchpoints, calculate the overlap ratio between the activity peak and the high-conversion period, and sort them to generate a priority touchpoint delivery order; S4: Select time points according to the priority contact delivery order, calculate the time difference with the peak of the target customer delay curve, filter the matching time points, and generate a delivery time list.

2. The financial marketing management service method based on multi-source data fusion according to claim 1 is characterized in that: The customer group transaction rhythm label includes the transaction mean change range, the transaction fluctuation value change range, and the transaction rhythm type; the high conversion transaction time period includes the concentrated response period, the number of concentrated response transactions, and the transaction conversion peak period; the priority contact delivery order includes the contact overlap ratio, contact priority, and contact identification; the delivery time list includes the matching time point, time difference range, and priority delivery time point.

3. The financial marketing management service method based on multi-source data fusion according to claim 1 is characterized in that: The specific steps of S1 are: S101: Obtain account transaction records of a target customer group within a fixed time window, arrange them in ascending order by transaction time, extract the time intervals between adjacent transactions and integrate them into a sequence to generate a transaction time interval sequence; S102: extracting central level and stability features based on the transaction time interval sequence, and integrating the two into a feature set to obtain transaction interval statistical features; S103: Calling the transaction interval statistical features, comparing the center level and stability changes of adjacent time windows, combining the transaction rhythm and fluctuation threshold, and establishing a customer group transaction rhythm label.

4. The financial marketing management service method based on multi-source data fusion according to claim 1 is characterized in that: The specific steps of S2 are: S201: Obtain the transaction rhythm labels of the customer groups, screen the customer groups with increased transaction frequency and increased fluctuation, integrate the customer group identifiers that meet the conditions, and generate a target customer group set; S202: Based on the target customer group set, the marketing information sending time and the first completed transaction time of the customer group before the multi-channel marketing launch are collected, a time difference calculation is performed on the two types of time data, and the data are arranged in order by customer group to draw a response delay curve; S203: calling the response delay curve, comparing the position difference of the curve on the time axis to obtain the concentrated response period, and counting the number of transactions within the period, combining the concentrated response period and the number of transactions to establish a high conversion transaction time period.

5. The financial marketing management service method based on multi-source data fusion according to claim 1 is characterized in that: The specific steps of S3 are: S301: Based on the high-conversion transaction time period, collect the daily activity curve of the online financial service touchpoints of similar customers, identify the peak position in the curve and record the corresponding time point, integrate the peak time information according to the touchpoint sequence, and generate a touchpoint peak time set; S302: Calling the touch point peak time set, performing time interval overlap calculation on the peak time and the conversion transaction time period, calculating the touch point overlap percentage, and archiving the percentage results in the order of the touch points to obtain a touch point overlap percentage sequence; S303: According to the contact overlap ratio sequence, sort the ratio values ​​from large to small to generate a priority contact placement order.

6. The financial marketing management service method based on multi-source data fusion according to claim 1 is characterized in that: The specific steps of S4 are: S401: Calling the priority contact placement order, selecting the contacts with the highest ranking, obtaining the corresponding time point data, and arranging the contact time points into a set in chronological order to generate a priority contact active time set; S402: Based on the priority contact active time set, extract the peak position of the target customer delay curve, calculate the time difference with the time point, compare the difference result with the matching threshold, filter the time points that meet the threshold condition, and obtain the matching time point set; S403: Call the matching time point set, arrange them in chronological order, remove duplicate records, and establish a delivery time list.

7. The financial marketing management service method based on multi-source data fusion according to claim 1 is characterized in that: The method further comprises: S5: Based on the delivery time list, the start, peak, and end time differences of the customer touchpoint interaction time at multiple acquisition terminals are calculated and normalized to a unified time axis, the events are rearranged, and a financial marketing management service plan is generated.

8. The financial marketing management service method based on multi-source data fusion according to claim 7 is characterized in that: The financial marketing management service solution includes a multi-touch event sequence under a unified timeline, touch interaction time distribution, and differentiated adjustment results between touch points.

9. The financial marketing management service method based on multi-source data fusion according to claim 7, characterized in that: The specific steps of S5 are: S501: According to the delivery time list, collect customer touchpoint interaction time recorded by multiple collection terminals, obtain the start time, peak time, and end time in sequence, calculate the difference between the three types of time, classify and organize the calculation results by collection terminal, and generate a touchpoint time difference set; S502: Based on the contact time difference set, identifying time difference records with differences between the collection terminals, normalizing and adjusting the differentiated time differences according to a unified time axis reference value, and integrating the adjusted records into a unified structure to obtain a unified time axis record set; S503: Calling the unified timeline record set, rearranging the multi-touch events according to the adjusted time sequence, integrating them into executable marketing management content, and establishing a financial marketing management service solution.

10. The financial marketing management service system based on multi-source data fusion is characterized by: The system is used to implement the financial marketing management service method based on multi-source data fusion according to any one of claims 1 to 9, and the system includes: The trading rhythm analysis module is used to perform S1: obtain the account transaction records of the target customer group in a fixed time window, sort them in ascending order by transaction time, calculate the time intervals between adjacent transactions to form an interval sequence, calculate the mean and fluctuation value respectively, compare the mean difference of adjacent time windows with the trading rhythm change threshold, and compare the fluctuation value change with the trading fluctuation threshold, to generate the customer group trading rhythm label; The frequency fluctuation screening module is used to perform S2: screening customer groups with increasing transaction frequency and fluctuation based on the transaction rhythm tag of the customer group, collecting the time of sending marketing information and the time of first transaction before marketing launch, calculating the time difference to generate a response delay curve, comparing the position difference of the curve to obtain a concentrated response period, counting the number of transactions in the period, and generating a high-conversion transaction time period; The contact point prioritization module is used to execute S3: based on the high conversion transaction time period, collect the daily activity curves of similar customers at financial service contact points, calculate the overlap ratio between the activity peak and the high conversion period, and sort and generate a priority contact point delivery order; The delivery time matching module is used to execute S4: selecting a time point according to the priority contact delivery order, calculating the time difference with the peak of the target customer delay curve, screening the matching time point, and generating a delivery time list; The multi-touch time integration module is used to execute S5: based on the delivery time list, calculate the start, peak, and end time differences of the customer touch interaction time of multiple acquisition terminals and normalize them to a unified time axis, rearrange the events, and generate a financial marketing management service plan.

Images