Financial marketing SAAS platform based on DeepSeek

Abstract

The invention relates to the technical field of user behavior analysis, in particular to a DeepSeek-based financial marketing SAAS platform, the platform carries a control system, and the system comprises a behavior feature acquisition module, a marketing strategy priority evaluation module, a window behavior analysis module, a strategy deployment adaptation module and a resource dynamic mapping module. According to the method, cross-platform time sequence alignment is completed and a cross-system behavior time sequence chain is formed by collecting user behavior event time features and transaction type identifiers and combining financial product state tags and user portrait attributes, high-frequency transaction behavior distribution features are extracted based on the behavior time sequence chain, and strategy demand urgency is dynamically corrected in combination with the product state tags. Time windows are divided according to the strategy priority sequence, trigger density characteristics and time interval fluctuation characteristics are extracted, strategy optimization is executed through a trigger density comparison bearing threshold value, calculation and storage resources are dynamically distributed, and the resource configuration efficiency and the system stability are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of user behavior analysis, and particularly to a SAAS platform for financial marketing based on DeepSeek. Background Art

[0002] The technical field of user behavior analysis includes a set of technologies for collecting, processing, and modeling the behavior data of users in financial activities to support marketing decisions. The core content of this field is to identify potential needs and behavior patterns by integrating multi-dimensional data such as user transaction records, browsing preferences, and interaction feedback, and then generate targeted marketing strategies. Its systematic technical process covers from raw data cleaning and feature extraction to behavior pattern mining and dynamic prediction, and finally forms an executable marketing plan to ensure that financial institutions can optimize resource allocation based on objective behavior data and improve service accuracy.

[0003] Among them, the SAAS platform for financial marketing based on DeepSeek refers to a technical solution that uses a specific data processing architecture to integrate cross-platform user behavior data and generates financial product recommendation strategies through a preset analysis model. The technical matters targeted by this patent theme include the standardized access process for multi-source heterogeneous behavior data, the behavior patterns based on time series features and association rules, the parsing method, and the strategy generation mechanism for financial scenarios. By constructing a distributed data storage layer, real-time collection of high-concurrency behavior logs is achieved, and a rule engine and dynamic weight calculation are used to complete the automatic mapping of user behavior tags. Combined with the strategy execution interface, the analysis results are directly output to the marketing system of financial institutions to complete strategy deployment.

[0004] In the existing technology, during the process of user behavior data collection and processing, there are problems of incomplete integration of multi-source heterogeneous data. The standardized access mechanism is limited to the unification of basic fields and is difficult to cover the alignment requirements of complex time series features, resulting in timestamp drift and behavior chain breakage of cross-platform behavior data, weakening the accuracy of behavior pattern recognition. In terms of strategy generation, the existing technology focuses on deduction based on rules or static models and lacks a dynamic risk correction mechanism, resulting in distorted strategy priority ranking and inability to adjust in time when user needs change rapidly, affecting the accuracy of marketing response. In the behavior analysis stage, the window division is rough and the granularity of behavior feature extraction is low, resulting in a lack of fine-grained behavior support basis during the strategy execution process, affecting strategy adaptation and execution effects. In terms of resource scheduling, static or estimated resource allocation methods are mostly used, ignoring the actual task volume changes and dynamic resource occupancy, which easily leads to resource waste or bottlenecks, causing system operation instability or response delay. For example, during the high-concurrency promotion period, uneven resource allocation often leads to partial task failures or delayed pushes, seriously affecting the overall marketing effect and customer experience. Summary of the Invention

[0005] The objective of the present invention is to address the drawbacks existing in the prior art, and a SAAS platform for financial marketing based on DeepSeek is proposed.

[0006] To achieve the above objective, the present invention adopts the following technical solutions: The SAAS platform for financial marketing based on DeepSeek includes: The behavior feature acquisition module calls the platform interface to obtain the time features of user behavior events and transaction type identifiers, synchronously collects financial product status tags and user portrait attributes, performs cross-platform time series alignment on the time features of behavior events, and generates a cross-system behavior time series chain. The marketing strategy priority evaluation module extracts the distribution features of high-frequency trading behaviors based on the cross-system behavior time series chain, calculates the urgency of strategy requirements in combination with financial product status tags, and dynamically corrects the urgency through user risk preference tags to generate a strategy priority sequence. The window behavior analysis module divides the cross-system behavior time series chain into time windows according to the strategy classification in the strategy priority sequence, extracts the trigger density features and time interval fluctuation features of strategy-related behaviors within the window, and generates a set of strategy behavior features. The strategy deployment adaptation module compares the trigger density features in the set of strategy behavior features with a preset load threshold. When the trigger density exceeds the load threshold, it performs strategy splitting processing. When the resource occupancy features break through the limit threshold, it performs strategy optimization processing to generate a strategy deployment plan and an optimization instruction set.

[0007] As a further solution of the present invention, the cross-system behavior time series chain includes behavior event timestamps, transaction type identifiers, financial product status tags, and user portrait attributes. The strategy priority sequence includes high-frequency trading behavior distribution features, strategy requirement urgency, and dynamic correction factors. The set of strategy behavior features includes trigger density features, time interval fluctuation features, and strategy-related behaviors. The strategy deployment plan and the optimization instruction set include strategy splitting instructions, strategy optimization instructions, and resource occupancy limit parameters.

[0008] As a further solution of the present invention, the behavior feature acquisition module includes: The behavior time extraction sub-module obtains the time features of user behavior events and transaction type identifiers returned by the platform interface, extracts the occurrence time, duration, and frequency parameters based on the event time field, corresponds the time features and transaction type identifiers in time order, and arranges the time intervals between adjacent events to generate an event time interval quantity. The product status synchronization sub-module synchronizes and collects financial product status tags and user portrait attributes based on the event time interval, filters data records consistent with the behavior event time, classifies and processes them according to the transaction type identifier and product status identifier, organizes the product status information corresponding to the transaction type, and generates the transaction product status quantity. The time series chain generation sub-module organizes the time characteristics of behavior events under the transaction type according to the transaction product status quantity, arranges the event records in chronological order, filters data segments with a time interval less than the synchronization reference value, groups them into continuous behavior chains, and generates a cross-system behavior time series chain.

[0009] As a further solution of the present invention, the marketing strategy priority evaluation module includes: The transaction behavior extraction sub-module obtains the cross-system behavior time series chain transaction data, detects the transaction timestamp, type, and amount, collects continuous transaction sequences, classifies transaction characteristics, filters high-frequency behaviors, and generates high-frequency transaction behavior distribution values. The product status evaluation sub-module, based on the high-frequency transaction behavior distribution value, calls the financial product status tag, detects the current status and changes of the product, extracts the activity interval and change range, and obtains the urgency of the strategy requirement. The priority sequence generation sub-module, according to the urgency of the strategy requirement, calls the user risk preference tag, detects the risk level and tolerance interval, sets a correction factor, adjusts the urgency measure, and sorts to generate a strategy priority sequence.

[0010] As a further solution of the present invention, the window behavior analysis module includes: The strategy classification sub-module obtains the cross-system behavior time series chain and the strategy priority sequence, assigns behaviors to corresponding strategy categories according to the priority, arranges the strategy categories in chronological order and marks the behavior nodes, and counts the occurrence times of the strategy categories in the overall time series chain to generate a strategy distribution quantity value. The time window division sub-module calls the strategy distribution quantity value, determines the time distribution characteristics of the behavior nodes according to the density change of the strategy behavior in the time series chain, calibrates the time window boundary according to the change of the time interval between the behavior nodes, and combines the cumulative situation of the strategy categories to generate a time window demarcation quantity value. The behavior feature extraction sub-module calls the time window demarcation quantity value, collects the strategy category behaviors within the time window, extracts the time fluctuation characteristics according to the behavior trigger time series, and combines the local behavior quantity change characteristics to generate a strategy behavior feature set.

[0011] As a further solution of the present invention, the specific calculation formula for counting the occurrence times of the strategy categories in the overall time series chain is: ; Wherein, Represents the weighted occurrence quantity of policy category j, Represents the global priority correction coefficient of the behavior nodes in the time series chain, Represents the original priority of the k-th behavior node attributed to policy category j, Represents the number of time unit intervals between the k-th behavior node and the end point of the current time series chain, d represents the total number of behavior nodes attributed to policy category j, Is a smoothing constant to prevent the denominator from being zero, Represents the collaborative weight factor of cross-system behavior nodes, Represents the number of behavior nodes of the m-th associated subsystem in policy category j, q represents the total number of associated subsystems.

[0012] As a further solution of the present invention, the policy deployment adaptation module includes: The density comparison sub-module obtains the trigger density feature and the preset load threshold in the policy behavior feature set, calls the trigger density feature and the preset load threshold for comparison, and filters out the policy behavior features exceeding the load threshold interval according to the comparison result to generate a trigger density deviation degree; The resource judgment sub-module obtains the resource occupancy feature and the limit threshold in the policy behavior feature set based on the trigger density deviation degree, calls the resource occupancy feature and the limit threshold for difference determination, filters out the resource occupancy features breaking through the limit threshold, and generates a resource load breakthrough rate; The solution generation sub-module calls the policy behavior features exceeding the load threshold according to the resource load breakthrough rate, divides the policy set according to the resource load breakthrough rate, calls the divided policy set for deployment parameter formulation and instruction mapping processing, and generates a policy deployment solution and an optimization instruction set.

[0013] As a further solution of the present invention, the specific calculation formula for calling the trigger density feature and the preset load threshold for comparison is: ; Wherein, H represents the trigger density deviation trend parameter, ρ represents the average trigger density value of all samples in the policy behavior feature set, ρ i Represents the density statistical value of the i-th feature dimension in the sample set in the policy behavior feature set, T represents the reference value of the preset load threshold, Represents the load threshold correction amount corresponding to the i-th feature dimension, α represents the dynamic adjustment coefficient, μ represents the harmonic mean of the historical density mean and the current density peak, γ represents the density fluctuation tolerance factor, β i Represents the non-linear deviation gain coefficient of the i-th feature dimension, η represents the threshold stability correction parameter, and n represents the total dimension number of the policy behavior features.

[0014] As a further solution of the present invention, the system further includes: Based on the task volume characteristics and optimization amplitude characteristics in the policy deployment plan and optimization instruction set, the resource dynamic mapping module dynamically allocates the computing resource capacity characteristics and storage resource quota characteristics to generate a cross-system resource configuration table; The cross-system resource configuration table includes task volume characteristics, optimization amplitude characteristics, computing resource capacity characteristics, and storage resource quota characteristics.

[0015] As a further solution of the present invention, the resource dynamic mapping module includes: The task feature extraction sub-module obtains the task volume data and task category parameters in the policy deployment plan and optimization instruction set, calls the task volume data to determine the capacity requirement value, calls the task category parameters to identify the attribute classification, and filters the high-demand task items based on the capacity requirement value and attribute classification to generate a task capacity feature value; The optimization amplitude analysis sub-module, based on the task capacity feature value, obtains the optimization target parameters and optimization priority parameters in the optimization instruction set, calls the optimization target parameters to determine the adjustment range value, calls the optimization priority parameters to arrange the application order, and filters the high-priority task items based on the adjustment range value and application order to generate a task adjustment amplitude feature value; The resource relationship generation sub-module, according to the task adjustment amplitude feature value, obtains the initial capacity parameter of the computing resource and the initial quota parameter of the storage resource, calls the initial capacity parameter of the computing resource to determine the allocation ratio, calls the initial quota parameter of the storage resource to determine the adjustment ratio, and establishes a mapping parameter set based on the allocation ratio and adjustment ratio to generate a cross-system resource configuration table.

[0016] Compared with the prior art, the advantages and positive effects of the present invention are as follows: In the present invention, by collecting the time characteristics of user behavior events and transaction type identifiers, combining with the financial product status tags and user portrait attributes, cross-platform time series alignment is completed to form a cross-system behavior time series chain. Based on the behavior time series chain, the distribution characteristics of high-frequency trading behaviors are extracted, and the urgency of the policy requirements is dynamically corrected in combination with the product status tags. According to the policy priority sequence, time windows are divided, and the trigger density characteristics and time interval fluctuation characteristics are extracted. Through the comparison of the trigger density with the bearing threshold, policy optimization is performed, and computing and storage resources are dynamically allocated to improve the resource allocation efficiency and system stability. Brief Description of the Drawings

[0017] Figure 1 is the system flow chart of the present invention; Figure 2 is the system block diagram of the present invention. Detailed Embodiment

[0018] In order to make the objectives, 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 used to limit the present invention.

[0019] In the description of the present invention, it should be understood that the orientation or positional relationship indicated by the terms "length", "width", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc. is based on the orientation or positional relationship shown in the accompanying drawings, and is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the present invention. In addition, in the description of the present invention, the meaning of "a plurality of" is two or more, unless otherwise specifically defined.

[0020] Please refer to Figure 1-2 , the SAAS platform for financial marketing based on DeepSeek includes: The behavior feature acquisition module calls the platform interface to obtain the time features of user behavior events and the transaction type identifier, synchronously collects the financial product status tags and user portrait attributes, performs cross-platform time series alignment on the time features of behavior events, and generates a cross-system behavior time series chain; The marketing strategy priority evaluation module extracts the distribution features of high-frequency trading behaviors based on the cross-system behavior time series chain, calculates the urgency of strategy requirements in combination with the financial product status tags, and dynamically corrects the urgency through the user risk preference tags to generate a strategy priority sequence; The window behavior analysis module divides the cross-system behavior time series chain according to the strategy classification in the strategy priority sequence, extracts the trigger density feature and the time interval fluctuation feature of the strategy-related behaviors within the window, and generates a strategy behavior feature set; The strategy deployment adaptation module compares the trigger density feature in the strategy behavior feature set with the preset load threshold. When the trigger density exceeds the load threshold, it performs strategy splitting processing. When the resource occupancy feature breaks through the limit threshold, it performs strategy optimization processing to generate a strategy deployment plan and an optimization instruction set; The resource dynamic mapping module dynamically allocates the computing resource capacity feature and the storage resource quota feature based on the task volume feature and the optimization amplitude feature in the strategy deployment plan and the optimization instruction set, and generates a cross-system resource configuration table.

[0021] The cross-system behavior time sequence chain includes behavior event timestamps, transaction type identifiers, financial product status tags, and user profile attributes. The policy priority sequence includes high-frequency trading behavior distribution characteristics, policy requirement urgency, and dynamic correction factors. The policy behavior feature set includes trigger density features, time interval fluctuation features, and policy-related behaviors. The policy deployment plan and optimization instruction set includes policy splitting instructions, policy optimization instructions, and resource occupancy limit parameters. The cross-system resource configuration table includes task volume features, optimization amplitude features, computing resource capacity features, and storage resource quota features.

[0022] Please refer to Figure 2 , the behavior feature acquisition module includes: The behavior time extraction sub-module obtains the user behavior event time features and transaction type identifiers returned by the platform interface, extracts the occurrence time, duration, and frequency parameters based on the event time field, corresponds the time features and transaction type identifiers according to the time sequence, and arranges the time intervals between adjacent events to generate the event time interval quantity. The behavior time extraction sub-module obtains the user behavior event time features and transaction type identifiers returned by the platform interface. First, it reads the data records returned by the platform interface, including the event time field and the transaction type identifier field. It reads each record, extracts the event time as the occurrence time, directly records the time field as the specific time point of the event occurrence, such as 10:00 on April 1, 2025. If there is an event duration field, it directly uses it. If not, it defaults that each event lasts for 5 seconds. The frequency parameter is obtained by counting the number of occurrences of the same transaction type behavior per day. The statistical method is to filter out the records of the same transaction type that occur for a certain user in a day. For example, if a user makes three payment behaviors in a day, the payment frequency is recorded as three times a day. After sorting all events in ascending order of time, when the events of the same user are arranged in sequence, the time interval between two adjacent events is calculated by directly reading the difference between the time fields of the latter and the former. For example, the first time is 10:00 and the second time is 10:05, and the time interval is 5 minutes. After sorting and collecting all time intervals, the interval range is defined. A time interval less than 5 minutes is defined as a short interval, and a time interval greater than or equal to 5 minutes is defined as a long interval. By sorting out the number of time intervals in different intervals, an overall feature description and classification record of the event time interval quantity are generated.

[0023] The product status synchronization sub-module synchronously acquires the financial product status tags and user profile attributes based on the event time interval quantity, filters the data records consistent with the behavior event time, classifies and processes them according to the transaction type identifier and product status identifier, and arranges the product status information corresponding to the transaction type to generate the transaction product status quantity. The product status synchronization sub-module synchronizes and collects financial product status tags and user portrait attributes based on the event time interval. When collecting data, it extracts the product status fields and user portrait fields from the financial product system records, and filters out the data records whose time is the same as or whose time deviation does not exceed one minute from the time when the behavior event occurs. For example, if a behavior event occurs at 10:00 sharp and the product status record time is 10:01, it is considered that the two match successfully. During the filtering process, it is judged by directly comparing whether the difference in seconds between the two time fields is less than or equal to 60 seconds. After filtering, it classifies and processes according to the transaction type of the behavior event and the corresponding product status. For example, if the transaction type is purchase and the product status is on sale, this event is classified into the set of purchased on-sale products. Subsequently, it sorts out all the classified records and counts the number of different product statuses under each transaction type. For example, in purchase transactions, the on-sale status appears 10 times and the out-of-sale status appears 2 times, so the proportion of on-sale products is 83.33%. A benchmark value is set at 95%. If the proportion is lower than this benchmark value, it is marked as the data type that needs to be synchronized and corrected, thus completing the classification processing and generating the transaction product status quantity.

[0024] The time series chain generation sub-module sorts out the time characteristics of behavior events under the transaction type according to the transaction product status quantity, arranges the event records in chronological order, filters out the data segments with time intervals less than the synchronization benchmark value, and groups them to form continuous behavior chains, generating cross-system behavior time series chains; The time series chain generation sub-module sorts out the time characteristics of the corresponding behavior events under each transaction type. First, it extracts all the behavior event records under each transaction type, such as the event record list of all payment transactions. It arranges them in ascending order according to the time field in all the records. After arranging, it calculates the time interval for each pair of adjacent events in turn. If the time interval between a certain event and the next event is less than five minutes, these two events are classified into the same continuous behavior chain. For example, the time of the first event is 10:00 sharp and the time of the second event is 10:02. The time interval is two minutes, which is less than the five-minute benchmark value. Therefore, these two payment behaviors are grouped into the same chain. If the time interval exceeds five minutes, it is divided into a new chain and grouped in turn to form continuous behavior chains. During the process of sorting out the chains, it also records the start time, end time, the transaction type to which the chain belongs, and the number of behavior events included in the chain. For example, the start time of a certain chain is 10:00 sharp, the end time is 10:02, the transaction type is payment, and it contains two behavior events. For the division standard of the chain length, a single-behavior chain is defined as a low-frequency chain, a two-behavior chain is defined as a medium-frequency chain, and a three-or-more-behavior chain is defined as a high-frequency chain. Through the above method, the generation and sorting of cross-system behavior time series chains are completed.

[0025] Please refer to Figure 2, the marketing strategy priority evaluation module includes: The transaction behavior extraction sub-module obtains cross-system behavior time-series chain transaction data, detects transaction timestamps, types, and amounts, collects continuous transaction sequences, classifies transaction characteristics, filters high-frequency behaviors, and generates high-frequency transaction behavior distribution values; The transaction behavior extraction sub-module first needs to connect to the data interfaces of multiple independent transaction systems and continuously call the data acquisition service to complete the cross-platform transaction record collection. Each record needs to include user identification, transaction time, transaction type, and transaction amount. The transaction time detection is judged based on whether the adjacent transaction time interval is less than 5 minutes. If the time difference between two adjacent transactions is less than or equal to 5 minutes, it is considered a continuous transaction. For example, the interval between a transfer of 500 yuan at 08:00 and a payment of 120 yuan at 08:03 is only 3 minutes, which meets the continuous standard. When classifying transaction types, it is matched according to the pre-established coding table. For example, a transfer is classified as code T01, and a consumption is classified as code T02. The classification of transaction amounts needs to be based on the set intervals. For example, amounts from 0 to 100 yuan are grouped into the first category, 100 to 500 yuan into the second category, and 500 to 1000 yuan into the third category. The number of transactions in each interval is accumulated. When filtering high-frequency behaviors, it is necessary to calculate the proportion of the number of transactions in each interval to the total number of transactions. When the proportion of the number of transactions in a certain amount interval exceeds 50%, it is determined as a high-frequency transaction interval. For example, in a daily record, there are 23 total transactions, and 15 transactions in the 0 to 100 yuan amount interval. The calculated proportion is 15 divided by 23, which is approximately 65%, exceeding the 50% threshold, so it is marked as a high-frequency transaction interval, and the high-frequency transaction behavior distribution value is set to 65%. During the detection of continuous transactions, the traversal comparison method can be used to calculate the difference between the current transaction time and the previous transaction time in turn. The process of classifying transaction characteristics adopts the type dictionary direct matching method and is divided into categories according to the amount. The screening process is judged based on the set frequency threshold. The process of generating the distribution value is the conversion of the ratio of the high-frequency occurrence times in each amount interval to the total number of times. Finally, the extraction and archiving of the high-frequency transaction behavior distribution value are completed.

[0026] Based on the high-frequency transaction behavior distribution value, the product status evaluation sub-module calls the financial product status label, detects the current status and changes of the product, extracts the activity interval and the change range, and obtains the urgency of the strategy requirement; After receiving the high-frequency trading behavior distribution value, the product status evaluation sub-module needs to synchronously call the latest status label of the corresponding financial product. Status labels such as active and stagnant are set by the unified standards of the financial system. When dividing the activity range, 0 to 0.3 is defined as low activity, 0.3 to 0.7 as medium activity, and 0.7 to 1.0 as high activity. Map the obtained high-frequency trading behavior distribution value to the activity range. For example, if the distribution value of a certain product is 0.652, it falls within the medium activity range. Then, when detecting changes in the product status, it is necessary to compare the difference between the activity value in the previous period and the current activity value. For example, if the activity value in the previous period is 0.45 and the current value is 0.652, the difference between the two is 0.202. This change amplitude exceeds the preset change threshold of 0.1. Therefore, it is classified as a significant change in activity. When determining the urgency of the strategy requirement, different levels can be set according to the change amplitude. For example, a change amplitude less than or equal to 0.1 is classified as low urgency, greater than 0.1 and less than or equal to 0.3 as medium urgency, and greater than 0.3 as high urgency. In this example, the change amplitude is 0.202, falling into the medium urgency range. The detection process needs to select the activity flag field from the time series trading records, and calculate the current product change status through field comparison and change amplitude calculation. The process of extracting the activity range is completed by directly comparing the activity value with the upper and lower bounds of the set range. The change amplitude is calculated by subtracting the historical value from the current value. For example, if the activity increases from 0.45 to 0.652, the change amplitude is 0.202. The change amplitude determination range is set according to business statistical standards. For example, according to the change amplitude of the activity of general financial products in the market, most changes are concentrated between 0.05 and 0.25. Therefore, the medium urgency range is reasonably set to be in the range of 0.1 to 0.3.

[0027] The priority sequence generation sub-module calls the user risk preference label according to the urgency of the strategy requirement, detects the risk level and tolerance range, sets a correction factor, adjusts the urgency measure, and sorts to generate the strategy priority sequence value; After receiving the urgency level of the policy requirement, the priority sequence generation sub-module needs to synchronously read the user risk preference tags. Risk preferences are usually divided into low risk, medium risk, and high risk. The risk tolerance range is directly associated with the risk preference tags. For example, medium-high risk preference corresponds to the high-risk range, that is, 0.6 to 1.0. When setting the urgency correction factor, the corresponding factor value is determined according to the risk preference type. For example, the factor for low risk preference is set to 0.8, the factor for medium risk is set to 1.0, and the factor for high risk is set to 1.2. If the urgency level is medium urgency and the corresponding basic urgency measurement setting value is 2, the corrected urgency measurement is obtained by multiplying the basic urgency measurement by the risk preference correction factor. For example, the medium urgency basic value 2 is multiplied by the high-risk correction factor 1.2, and the corrected urgency measurement is 2.4. Subsequently, the policies are sorted by priority according to the corrected urgency measurement value, and all policies are arranged in descending order. For example, the corrected urgency measurement of policy A is 3.0, the corrected urgency measurement of policy B is 2.4, and the corrected urgency measurement of policy C is 1.8. Finally, the policy execution order is formed. The user risk preference field needs to be directly extracted from the customer portrait system. The risk level matching is detected by comparing the risk preference with the risk level range. The setting of the correction factor is formulated with reference to the risk tolerance standard document. The adjustment of the urgency measurement is quantified by simply multiplying the basic urgency measurement value by the correction factor. The sorting process uses the descending sorting method, and the sorted priority list is used by the subsequent policy module for calling.

[0028] Please refer to Figure 2 , the window behavior parsing module includes: The policy classification sub-module obtains the cross-system behavior time series chain and the policy priority sequence, assigns the behavior to the corresponding policy category according to the priority, arranges the policy categories in chronological order and marks the behavior nodes, and counts the number of occurrences of the policy categories in the overall time series chain to generate the policy distribution quantity value; The specific calculation formula for counting the number of occurrences of the policy category in the overall time series chain is: ; Among them, represents the weighted number of occurrences of policy category j, represents the global priority correction coefficient of the behavior node in the time series chain, represents the original priority of the kth behavior node attributed to policy category j, represents the number of time unit intervals between the kth behavior node and the end point of the current time series chain, d represents the total number of behavior nodes attributed to policy category j, is a smoothing constant to prevent the denominator from being zero, represents the collaborative weight factor of the cross-system behavior node, represents the number of behavior nodes of the mth associated subsystem in policy category j, and q represents the total number of associated subsystems; Global priority correction coefficient Calculated by the standard deviation of the priority distribution of nodes in the cross-system behavior time sequence chain. The standard deviation is 0.45 and the average priority is 7.2. =1.1 is obtained from It can be seen that the correction coefficient increases with the increase of the standard deviation. The total number of behavior nodes d belonging to policy category j = 3, which is obtained by counting the actual number of behavior nodes in the time sequence chain. The original priority of the first behavior node , the second , the third , and the priority is quantified by the urgency score of task execution in the system log. The scoring range is 1 - 10. Interval time =2, =5, =0, with the unit of hour, calculated by the time stamp difference of the time sequence chain nodes. Smoothing constant =0.1, set according to the minimum time unit of 0.1 hour in the time sequence chain to prevent the denominator from being zero. Collaborative weight factor =0.9, calculated based on the cross-system interaction frequency ratio of 45%, , and the weight increases linearly with the interaction frequency. Number of subsystem behavior nodes , , counted from the subsystem activity log. The activity threshold ≥ 3 times / hour is converted into the number of nodes, and the total number of associated subsystems q = 2.

[0029] Calculation process: Calculate the absolute value term: Node 1: ; Node 2: ; Node 3: ; Sum of absolute values: 6.07 + 2.92 + 31.33 = 40.32; Calculate the collaborative term: Average of subsystem nodes: ; Value of the collaborative term: 0.9 × 4.5 = 4.05; Value of the total weighted statistic: ; The result shows that the weighted statistic value of 44.37 is the comprehensive behavior density index of policy category j in the time series chain. The numerical result quantifies the dynamic weight characteristics of the policy distribution value by superimposing the priority correction term and the subsystem collaboration term, and directly serves as the input data for generating the policy distribution value.

[0030] The time window division sub-module calls the policy distribution value. According to the density change of the policy behavior in the time series chain, it determines the time distribution characteristics of the behavior nodes, calibrates the time window boundary based on the change of the time interval between the behavior nodes, and generates the time window demarcation value in combination with the cumulative situation of the policy categories. After the time window division sub-module calls the policy distribution value, it first sets the initial time window width. For example, every 5 minutes is a basic time window. Subsequently, it analyzes the time interval between the behavior nodes. For two adjacent behavior nodes, it calculates the difference in their occurrence times. For example, behavior a1 occurs at 10:00 and b1 occurs at 10:02, and the time interval between them is 2 minutes, which is less than the set 5-minute window width. Therefore, it is determined that they belong to the same time window. When it is found that the time interval between two behavior nodes is significantly greater than the set time window width, for example, the interval is 15 minutes, far exceeding the 5-minute threshold, a new time window is split here. For the method of judging density changes, it traverses all behavior nodes and continuously calculates the time difference between each pair of adjacent nodes. When multiple large time intervals appear continuously, it indicates that the behavior density decreases, and the time window division needs to be further refined. The cumulative situation of the policy categories is also used as an auxiliary basis. The number of occurrences of the policy category is accumulated within a window. For example, within a window, the policy category 1 appears 3 times, reaching the set threshold of 3 times, then the current window boundary is confirmed. When the number of behaviors is less than the set threshold, the window is extended until the requirement is met. For judging whether to create a new time window, it uses the method of comparing the current time interval with the threshold in real time. If the time interval is greater than 5 minutes, a new window is immediately delimited. For the definition of high density and low density, a time interval less than 2 minutes is set as high density, and greater than 10 minutes is set as low density. For example, a time interval of 1 minute is judged as high density, and a time interval of 12 minutes is judged as low density. Finally, through continuous comparison and the accumulation of the number of policy categories, the time window demarcation value, that is, the start and end boundaries of each window, is generated.

[0031] The behavior feature extraction sub-module calls the time window demarcation value, collects the policy category behaviors within the time window, extracts the time fluctuation characteristics based on the behavior trigger time series, and generates the policy behavior feature set in combination with the local behavior quantity change characteristics. After the behavior feature extraction sub-module calls the time window demarcation value, it sequentially collects all policy category behaviors recorded within each time window. Subsequently, based on the trigger time points of each behavior, it extracts the time fluctuation characteristics. For any two adjacent behavior nodes, it records their time difference, and obtains the overall time change trend by summarizing all time differences. For example, if the time intervals mostly concentrate between 1 and 3 minutes, it indicates that the behaviors occur frequently and have small fluctuations. However, if the range of time intervals varies greatly, such as from 2 minutes to 20 minutes, it indicates large fluctuations. Further, in combination with the local behavior quantity change characteristics, within each time window, for example, every 5-minute window, it counts the number of behaviors. For example, if 6 behaviors occur within a 5-minute window, the behavior density is 6 times divided by 5 minutes, with a density of 1.2 times per minute. Then, it compares the behavior density changes of adjacent time windows one by one. For example, the density of the first window is 1.2 times / minute, and the density of the second window is 2 times / minute, with the density increasing by 0.8 times / minute. If the set change threshold is 0.5 times / minute, it is considered that the behavior density has a significant increase. Regarding how to set the standard for significant fluctuations, it is assumed that a change value greater than 1 time / minute is determined as a significant increase, less than -1 time / minute is determined as a significant decrease, and a change between the two is determined as general. Finally, based on the time mean, time fluctuation amplitude, local behavior quantity, and change conditions within each window, a complete set of policy behavior feature sets is extracted, providing basic data support for subsequent processing.

[0032] Please refer to Figure 2 , the policy deployment adaptation module includes: The density comparison sub-module obtains the trigger density feature and the preset bearing threshold in the policy behavior feature set, calls the trigger density feature to compare with the preset bearing threshold, and filters the policy behavior features beyond the bearing threshold interval according to the comparison result to generate the trigger density deviation degree; The specific calculation formula for calling the trigger density feature to compare with the preset bearing threshold is: ; Among them, H represents the trigger density deviation trend parameter, ρ represents the average trigger density value of all samples in the policy behavior feature set, ρ i represents the density statistical value of the i-th feature dimension in the sample set of the policy behavior feature set, T represents the reference value of the preset bearing threshold, represents the bearing threshold correction amount corresponding to the i-th feature dimension, α represents the dynamic adjustment coefficient (calculated from the geometric mean of the system load factor and the resource utilization rate), μ represents the harmonic mean of the historical density mean and the current density peak, γ represents the density fluctuation tolerance factor (calculated from the product of the historical maximum density difference and the current environment coefficient), β iThe non - linear offset gain coefficient representing the i - th feature dimension, η represents the threshold stability correction parameter (calculated from the ratio of the variance of the preset threshold to the sample size), and n represents the total number of dimensions of the strategic behavior characteristics; Parameter assignment and acquisition method: ρ i Collected through the real - time monitoring system, with a value range of 0.1 to 2.0. The measured sample data are ρ i = 0.85, ρ2 = 1.20, ρ3 = 1.50.

[0033] T: The reference value of the preset load threshold, defined by the system configuration file, with a value of T = 0.75.

[0034] The calculation method of α is the geometric mean of the system load factor (monitoring value 0.6) and the resource utilization rate (monitoring value 0.7), and the formula is .

[0035] The historical density mean of μ is 0.5 (query the mean value of the past 30 days through the historical database), and the current density peak is 0.8 (the maximum value of real - time monitoring), and the formula is .

[0036] The calculation method of γ is the product of the historical maximum density difference (the maximum fluctuation value of historical data 0.3) and the current environmental coefficient (environmental monitoring value 0.9), and the formula is ×0.9 = 0.27.

[0037] β i Output by the feature importance analysis model, with a value range of 0.1 to 0.6, taking β1 = 0.55, β2 = 0.30, β3 = 0.48.

[0038] The calculation method of η is the ratio of the variance of the preset threshold (calculated value 0.1) to the sample size (current sample size 100), and the formula is .

[0039] The measured number of dimensions n = 3.

[0040] Formula calculation and derivation process: The first term of the calculation formula: ; This result shows that the dynamic deviation between the current trigger density feature and the threshold after normalization is 0.370, indicating that the density fluctuation is within the tolerance range.

[0041] The second term of the calculation formula: Sum for the three dimensions: Dimension 1: ; Dimension 2: ; Dimension 3: ; Sum result: 11.704 + 8.049 + 14.914 = 34.667; Calculate H: H = 0.370 + 34.667 = 35.037; This result indicates that the trigger density offset trend parameter H is 35.037. When H exceeds the preset threshold (e.g., H ≥ 1.0), it is determined that the corresponding policy behavior feature exceeds the bearing threshold interval, and a trigger density offset degree needs to be generated.

[0042] Based on the trigger density offset degree, the resource judgment sub-module obtains the resource occupancy characteristics and limit thresholds in the policy behavior feature set, calls the resource occupancy characteristics and limit thresholds for difference determination, filters out the resource occupancy characteristics that exceed the limit threshold, and generates a resource load breakthrough rate; The resource judgment sub-module extracts various resource occupancy feature data from the policy behavior features marked as overloaded, such as CPU usage rate, memory usage rate, etc. A reasonable limit threshold is set for each resource. For example, the CPU usage rate limit is set to 80%, and the memory usage rate limit is set to 75%. The current resource usage of each policy behavior feature is read in sequence, and the extracted usage rate is compared and analyzed with the corresponding limit threshold. If the CPU usage rate of a certain policy behavior feature reaches 85% while the set limit is only 80%, it is considered that this feature has exceeded the CPU resource usage limit. If the memory usage rate is 70%, it is considered to be within a reasonable range. In an actual scenario, for example, for a cloud server instance, the monitored CPU resource usage rate data shows 85%, which exceeds the standard value of 80%, and it is directly classified as a resource load breakthrough. To quantify the severity of the resource breakthrough, the difference between the resource usage rate and the limit threshold is calculated, and the difference is converted into a percentage form of the limit threshold, thereby obtaining the resource load breakthrough rate. Taking the example, the degree of CPU breakthrough limit is 6.25%, and this breakthrough rate is recorded in the resource load breakthrough rate set for subsequent processing.

[0043] The solution generation sub-module calls the policy behavior features that exceed the bearing threshold according to the resource load breakthrough rate, divides the policy set based on the resource load breakthrough rate, calls the divided policy set for deployment parameter formulation and instruction mapping processing, and generates a policy deployment plan and an optimization instruction set; The solution generation sub-module further divides the policy behavior characteristics according to the resource load breakthrough rate, sets the resource breakthrough rate division criteria. For example, the breakthrough rate from 0% to 5% is classified as mild overload, 5% to 10% as moderate overload, and above 10% as severe overload. Extract the breakthrough rate values corresponding to each policy behavior characteristic and classify and organize them according to the interval where the breakthrough rate is located. For example, if the breakthrough rate of a certain policy behavior characteristic is 6.25%, it belongs to the category of moderate overload, while if the breakthrough rate of another policy behavior characteristic is 11%, it is classified into the category of severe overload. Based on the classification results, formulate targeted deployment parameters. The policy behavior characteristics of moderate overload are processed using a flow-limiting and caching mechanism, while the policy behavior characteristics of severe overload adopt more stringent isolation and circuit-breaking measures. In specific operations, the moderate overload category can set the upper limit of the number of accesses per second to 3 times and increase the cache aging time to 30 seconds; the severe overload category can enable the circuit-breaking mechanism and set that the isolation measure is triggered when there are 3 consecutive failures. Taking a microservice interface as an example, if the breakthrough rate of interface A is 6.25%, then set the flow-limiting rule for it to 3 times per second and add cache processing at the same time to ensure the smoothness of the system resource load. Finally, based on the breakthrough rate and classification of each policy behavior characteristic, generate the overall system policy deployment plan and the corresponding optimization instruction set.

[0044] Please refer to Figure 2 , the resource dynamic mapping module includes: The task feature extraction sub-module obtains the task volume data and task category parameters in the policy deployment plan and optimization instruction set, calls the task volume data to determine the capacity requirement value, calls the task category parameters to identify the attribute classification, filters out the high-demand task items based on the capacity requirement value and attribute classification, and generates the task capacity feature value; The task feature extraction sub-module extracts the task volume data and task category parameters from the policy deployment plan and the optimization instruction set. First, according to the time nodes and task distribution records in the deployment plan, it queries the task quantity in each time period. For example, the task volume record in April is 1200 pieces, and the average task inflow rate per hour is 50 pieces. In the stage of extracting task category parameters, it is divided into category A, category B, and category C according to the task type labels, and high priority, normal, and low priority weights are set respectively. The weight of category A is set to 0.8, category B is set to 0.5, and category C is set to 0.2. In the link of determining the capacity demand value, the extracted task volume is multiplied by the average inflow rate for calculation. For example, for 1200 tasks with an average inflow rate of 50 pieces per hour, it is calculated that 60000 units of processing capacity are required. Subsequently, based on the task category parameters, classification is carried out according to the set attribute classification threshold. Among them, if the capacity demand of category A tasks is greater than or equal to 20000 units, they are classified into high-demand tasks. The capacity demand of category B tasks is between 10000 and 19999 units, and the capacity demand of category C tasks is less than 10000 units. According to the above setting criteria, tasks are screened, and high-demand task items that meet the conditions are selected. For example, 500 tasks are screened out from category A, and the average capacity demand per task is 25000 units. Finally, the task capacity feature value is formed by multiplying the task quantity by the category weight. For example, the task capacity feature value of category A tasks is 500 multiplied by 0.8 to get 400. Combined with a specific system, such as in a logistics sorting system, if the average daily package processing volume is 1200 orders, then 400 units of capacity resources need to be preferentially guaranteed for pre-allocation for high-priority sorting tasks to achieve targeted and efficient resource allocation.

[0045] Based on the task capacity feature value, the optimization amplitude analysis sub-module obtains the optimization target parameters and optimization priority parameters in the optimization instruction set, calls the optimization target parameters to determine the adjustment range value, calls the optimization priority parameters to arrange the application order, and screens high-priority task items according to the adjustment range value and the application order to generate the task adjustment amplitude feature value; The optimization amplitude analysis sub-module extracts the optimization target parameters and optimization priority parameters from the optimization instruction set according to the generated task capacity characteristic values. In the stage of extracting the optimization target parameters, minimizing the processing time and optimizing the resource occupancy rate are set as the main objectives, where minimizing the processing time is given a weight of 0.7 and optimizing the resources is given a weight of 0.3. In the link of calling the optimization target parameters to determine the adjustment range value, the task capacity characteristic value is combined with the optimization target coefficient. If the task capacity characteristic value is 400 and the coefficient of minimizing the processing time is 0.9, the adjustment range value is approximately 360. In the stage of calling the optimization priority parameters to arrange the application order, they are sorted from high to low according to the weights, with minimizing the processing time placed first and resource optimization ranked second. According to the adjustment range value and the priority application order, high-priority task items are screened. The screening criterion is tasks with an adjustment range value higher than 300 units and the corresponding optimization target weight greater than 0.5. For example, 300 priority optimization tasks are screened. The adjustment amplitude characteristic value of each task is obtained by multiplying the adjustment range value by the priority weight. For example, 360 multiplied by 0.7 gives 252. Combining with the actual scenario, such as in the optimization of task scheduling in a data center, for video encoding tasks with high requirements for processing time, high-priority tasks are screened according to the above method and optimization strategies are assigned to form a specific adjustment task list, providing a clear basis for subsequent resource reconfiguration.

[0046] The resource relationship generation sub-module obtains the initial capacity parameters of computing resources and the initial quota parameters of storage resources according to the task adjustment amplitude characteristic value, calls the initial capacity parameters of computing resources to determine the allocation ratio, calls the initial quota parameters of storage resources to determine the adjustment ratio, and establishes a mapping parameter set based on the allocation ratio and the adjustment ratio to generate a cross-system resource configuration table; Based on the task adjustment amplitude eigenvalue, the resource relationship generation sub-module obtains the initial capacity parameter of computing resources and the initial quota parameter of storage resources. When extracting the initial capacity parameter of computing resources, it reads the system benchmark configuration data. For example, the initial computing resources are 80,000 core-hours. In the stage of extracting the initial quota parameter of storage resources, it reads the total storage space of the initial configuration, such as set to 100TB. In the stage of determining the allocation ratio by calling the initial capacity parameter of computing resources, it compares the task adjustment amplitude eigenvalue with the total computing capacity. For example, the task adjustment amplitude eigenvalue is 252 units. Compared with the total computing capacity of 80,000 core-hours, the allocation ratio is approximately 0.00315. In the stage of determining the adjustment ratio by calling the initial quota parameter of storage resources, it compares the task adjustment amplitude eigenvalue with the total storage capacity. If 100TB is converted to 102,400MB, the adjustment ratio is approximately 0.00246. Based on the above allocation ratio and adjustment ratio, a mapping parameter set is established to allocate clear CPU resource ratios and storage resource ratios for each task. For example, a mapping table is established to mark the resource allocation ratios corresponding to the task IDs. Task001 is allocated 0.00315 ratio of computing resources and 0.00246 ratio of storage resources. In the actual system configuration, through a resource scheduling platform such as Kubernetes, resource allocation instructions can be issued according to the above mapping data to ensure that tasks can be reasonably configured proportionally among different subsystems. For example, in a hybrid cloud environment, data backup tasks are preferentially allocated to high-capacity nodes according to this configuration to reduce overall resource conflicts.

[0047] The above are only the preferred embodiments of the present invention and do not limit the present invention in other forms. Any person skilled in the art may use the disclosed technical content to make changes or modifications into equivalent embodiments with equivalent changes and apply them to other fields. However, as long as it does not depart from the technical solution content of the present invention, any simple modification, equivalent change, and modification made to the above embodiments based on the technical essence of the present invention still fall within the protection scope of the technical solution of the present invention.

Claims

1. A SAAS platform for financial marketing based on DeepSeek, the platform is equipped with a control system, characterized in that: The system includes: The behavior feature collection module calls the platform interface to obtain the time features of user behavior events and the transaction type identifier, synchronously collects the financial product status tags and user portrait attributes, performs cross-platform time series alignment on the time features of behavior events, and generates a cross-system behavior time series chain; The marketing strategy priority evaluation module extracts the distribution features of high-frequency trading behaviors based on the cross-system behavior time series chain, calculates the urgency of strategy requirements in combination with the financial product status tags, and dynamically corrects the urgency through the user risk preference tags to generate a strategy priority sequence; The window behavior parsing module divides the cross-system behavior time series chain according to the strategy classification in the strategy priority sequence, extracts the trigger density features and time interval fluctuation features of the strategy-related behaviors within the window, and generates a strategy behavior feature set; The strategy deployment adaptation module compares the trigger density features in the strategy behavior feature set with the preset load threshold. When the trigger density exceeds the load threshold, it performs strategy splitting processing. When the resource occupancy feature breaks through the limit threshold, it performs strategy optimization processing, and generates a strategy deployment plan and an optimization instruction set.

2. The SAAS platform for financial marketing based on DeepSeek according to claim 1, wherein: The cross-system behavior time series chain includes the behavior event timestamp, transaction type identifier, financial product status tag, and user portrait attribute. The strategy priority sequence includes the distribution features of high-frequency trading behaviors, the urgency of strategy requirements, and the dynamic correction factor. The strategy behavior feature set includes trigger density features, time interval fluctuation features, and strategy-related behaviors. The strategy deployment plan and the optimization instruction set include strategy splitting instructions, strategy optimization instructions, and resource occupancy limit parameters.

3. The SAAS platform for financial marketing based on DeepSeek according to claim 1, characterized in that: The behavior feature collection module includes: The behavior time extraction sub-module obtains the time features of user behavior events and the transaction type identifier returned by the platform interface, extracts the occurrence time, duration, and frequency parameters based on the event time field, corresponds the time features and the transaction type identifier according to the time sequence, and arranges the time intervals between adjacent events to generate an event time interval quantity; The product status synchronization sub-module synchronously collects the financial product status tags and user portrait attributes based on the event time interval quantity, filters the data records consistent with the behavior event time, classifies and processes them according to the transaction type identifier and the product status identifier, and arranges the product status information corresponding to the transaction type to generate a transaction product status quantity; The time series chain generation sub-module arranges the time features of behavior events under the transaction type according to the transaction product status quantity, arranges the event records in time sequence, filters the data segments with a time interval less than the synchronization reference value, and groups them into continuous behavior chains to generate a cross-system behavior time series chain.

4. The SAAS platform for financial marketing based on DeepSeek according to claim 3, characterized in that: The marketing strategy priority evaluation module includes: The transaction behavior extraction sub-module obtains the transaction data of the cross-system behavior time series chain, detects the transaction timestamp, type, and amount, collects continuous transaction sequences, classifies transaction features, and filters high-frequency behaviors to generate a high-frequency trading behavior distribution value; The product status evaluation sub-module calls the financial product status tags based on the high-frequency trading behavior distribution value, detects the current status and changes of the product, extracts the activity interval and change range, and obtains the urgency of strategy requirements; The priority sequence generation sub-module calls the user risk preference tags according to the urgency of the policy requirements, detects the risk level and tolerance range, sets the correction factor, adjusts the urgency measure, and sorts to generate the policy priority sequence.

5. The SAAS platform for financial marketing based on DeepSeek according to claim 4, characterized in that: The window behavior analysis module includes: The policy classification sub-module obtains the cross-system behavior time series chain and the policy priority sequence, assigns the behavior to the corresponding policy category according to the priority, arranges the policy categories in chronological order and marks the behavior nodes, and counts the number of occurrences of the policy categories in the overall time series chain to generate the policy distribution value; The time window division sub-module calls the policy distribution value, determines the time distribution characteristics of the behavior nodes according to the density change of the policy behavior in the time series chain, calibrates the time window boundary based on the change of the time interval between the behavior nodes, and combines the cumulative situation of the policy categories to generate the time window demarcation value; The behavior feature extraction sub-module calls the time window demarcation value, collects the policy category behaviors within the time window, extracts the time fluctuation characteristics according to the behavior trigger time series, and combines the local behavior quantity change characteristics to generate the policy behavior feature set.

6. The SAAS platform for financial marketing based on DeepSeek according to claim 5, wherein: The specific calculation formula for counting the number of occurrences of the policy categories in the overall time series chain is: ； Among them, represents the weighted occurrence quantity of policy category j, represents the global priority correction coefficient of the behavior node in the time sequence chain, represents the original priority of the k-th behavior node attributed to policy category j, represents the number of time unit intervals between the k-th behavior node and the end point of the current time sequence chain, d represents the total number of behavior nodes attributed to policy category j, is a smoothing constant to prevent the denominator from being zero, represents the collaborative weight factor of cross-system behavior nodes, represents the number of behavior nodes of the m-th associated subsystem in policy category j, q represents the total number of associated subsystems.

7. The SAAS platform for financial marketing based on DeepSeek according to claim 5, characterized in that: The policy deployment adaptation module includes: The density comparison sub-module obtains the trigger density feature and the preset bearing threshold in the policy behavior feature set, calls the trigger density feature to compare with the preset bearing threshold, and filters the policy behavior features that exceed the bearing threshold interval according to the comparison result to generate the trigger density deviation degree; The resource judgment sub-module obtains the resource occupancy feature and the limit threshold in the policy behavior feature set based on the trigger density deviation degree, calls the resource occupancy feature to determine the difference with the limit threshold, and filters the resource occupancy features that break through the limit threshold to generate the resource load breakthrough rate; The solution generation sub-module calls the policy behavior features that exceed the bearing threshold according to the resource load breakthrough rate, divides the policy set according to the resource load breakthrough rate, calls the divided policy set for deployment parameter formulation and instruction mapping processing, and generates the policy deployment plan and the optimization instruction set.

8. The SAAS platform for financial marketing based on DeepSeek according to claim 7, characterized in that: The specific calculation formula for calling the trigger density feature to compare with the preset bearing threshold is: ； Among them, H represents the triggering density offset trend parameter, ρ represents the average triggering density value of all samples in the set of policy behavior characteristics, and ρ i represents the density statistical value of the i-th feature dimension in the set of policy behavior characteristics in the sample set, T represents the reference value of the preset bearing threshold, represents the bearing threshold correction amount corresponding to the i-th feature dimension, α represents the dynamic adjustment coefficient, μ represents the harmonic mean of the historical density mean and the current density peak value, γ represents the density fluctuation tolerance factor, and β i represents the non-linear offset gain coefficient of the i-th feature dimension, η represents the threshold stability correction parameter, and n represents the total number of dimensions of the policy behavior characteristics.

9. The SAAS platform for financial marketing based on DeepSeek according to claim 1, characterized in that: The system further includes: The resource dynamic mapping module dynamically allocates the computing resource capacity feature and the storage resource quota feature based on the task volume feature and the optimization amplitude feature in the policy deployment plan and the optimization instruction set, and generates the cross-system resource configuration table; The cross-system resource configuration table includes the task volume feature, the optimization amplitude feature, the computing resource capacity feature, and the storage resource quota feature.

10. The SAAS platform for financial marketing based on DeepSeek according to claim 9, characterized in that: The resource dynamic mapping module includes: The task feature extraction sub-module obtains the task volume data and the task category parameters in the policy deployment plan and the optimization instruction set, calls the task volume data to determine the capacity requirement value, calls the task category parameters to identify the attribute classification, and filters the high-demand task items according to the capacity requirement value and the attribute classification to generate the task capacity feature value; Based on the task capacity eigenvalue, the optimization amplitude analysis sub-module obtains the optimization target parameters and optimization priority parameters in the optimization instruction set, calls the optimization target parameters to determine the adjustment range value, calls the optimization priority parameters to arrange the application order, filters the high-priority task items according to the adjustment range value and the application order, and generates the task adjustment amplitude eigenvalue; According to the task adjustment amplitude eigenvalue, the resource relationship generation sub-module obtains the initial capacity parameter of the computing resource and the initial quota parameter of the storage resource, calls the initial capacity parameter of the computing resource to determine the allocation ratio, calls the initial quota parameter of the storage resource to determine the adjustment ratio, and establishes a mapping parameter set according to the allocation ratio and the adjustment ratio to generate a cross-system resource configuration table.

Images