Simulation model-based digital platform operation monitoring optimization system and method

Abstract

The invention discloses a digital platform operation monitoring optimization system and method based on a simulation model. The method comprises the following steps: S1, constructing an operation monitoring set; s2, generating a preprocessing operation monitoring set; s3, simulating the operation state of the digital platform in real time; s4, generating macroscopic trend data of platform resource requirements; s5, event level data of the discrete event simulation model and macroscopic trend data of the system dynamics simulation model are updated in real time; s6, generating comprehensive operation strategy effect prediction data by combining the prediction results of the two; and S7, based on the operation strategy effect prediction data and the real-time operation monitoring set, carrying out adaptive adjustment on the current operation strategy, and generating a corresponding strategy optimization scheme. According to the method, the system bottleneck problem caused by insufficient resources can be effectively avoided in a high-concurrency environment, and the stability and expansibility of the platform are guaranteed.

Description

Technical Field

[0001] The present invention relates to the technical field of digital platforms, and in particular to a digital platform operation monitoring and optimization system and method based on a simulation model. Background Art

[0002] In the existing technology, the operation monitoring and resource optimization of digital platforms mainly rely on traditional data analysis and manual control. Operation monitoring usually evaluates the platform operation status by regularly analyzing the platform's user behavior, transaction records and system operation data. Existing methods are often based on offline analysis of static data, which lacks dynamics and real-time performance and cannot fully reflect the changing trend of resource demand. The static data analysis method makes it difficult for the platform to respond quickly when user traffic fluctuates drastically or system load suddenly increases, resulting in a decline in user experience and insufficient resource utilization. Adjustment of resource allocation usually requires manual intervention and lacks automation and intelligent means, and the efficiency of resource allocation is therefore limited.

[0003] At present, some platforms have begun to try to use rule engines or simple machine learning algorithms to allocate resources to cope with some changing needs. However, the configuration cost of rule engines is high and the maintenance is difficult. They are usually only suitable for specific, predefined scenarios and cannot adapt to complex operating environments. Although simple machine learning methods can make predictions based on historical data, their adaptability is limited. When faced with changing user behaviors and market demands, it is difficult to generate resource allocation plans suitable for immediate adjustments.

[0004] Another major problem with existing technologies is the lack of accuracy and flexibility in predicting and adjusting the effects of operational strategies. Most platforms still rely on historical data analysis or manual experience to formulate strategies, while ignoring the frequency of events, resource demand trends, and real-time changes in system load. Therefore, existing systems often lack data support when adjusting operational strategies and are unable to optimize strategies in a timely manner based on actual needs and changes in user behavior. Not only are they unable to accurately predict the long-term impact of different strategies on the platform, they may also cause waste of resources or a reduction in user experience.

[0005] To sum up, the existing technologies have the following main shortcomings: first, operation monitoring and resource allocation lack real-time and dynamic characteristics and cannot cope with the changing operation environment; second, traditional data analysis methods are limited to offline analysis of static data and cannot automatically adapt to the rapid changes in user behavior; finally, the lack of accuracy in predicting the effects of operation strategies cannot provide the platform with intelligent decision-making support based on future trends. Summary of the invention

[0006] One purpose of the present invention is to propose a digital platform operation monitoring and optimization system and method based on a simulation model. The present invention can effectively avoid system bottleneck problems caused by insufficient resources in a high-concurrency environment, thereby ensuring the stability and scalability of the platform.

[0007] A digital platform operation monitoring and optimization method based on a simulation model according to an embodiment of the present invention comprises the following steps:

[0008] S1. Collect user behavior data, transaction record data, and system operation log data from the digital platform to build an operation monitoring set;

[0009] S2. Perform data cleaning, deduplication processing, feature extraction and standardization preprocessing on the operation monitoring set to generate a preprocessed operation monitoring set;

[0010] S3. Based on the preprocessed operation monitoring set, a discrete event simulation model is constructed to simulate the occurrence of specific events on the digital platform, and the operation status of the digital platform is simulated in real time;

[0011] S4. Based on the pre-processed operation monitoring set and the event-level operation status data generated by the discrete event simulation model, a system dynamics simulation model is constructed to generate macro trend data of platform resource demand by comprehensively considering multiple dynamic factors such as user behavior, transaction volume changes and system load;

[0012] S5. Monitor the key operational indicators of the digital platform in real time, collect real-time monitoring data and dynamically input it into the discrete event simulation model and the system dynamics simulation model, and update the event level data of the discrete event simulation model and the macro trend data of the system dynamics simulation model in real time;

[0013] S6. Predict the effects of different operational strategies through integrated discrete event simulation models and system dynamics simulation models. The discrete event simulation model is used to predict the immediate impact of a single event or short-term operational adjustment on the digital platform, and the system dynamics simulation model is used to predict the changing trend of resource demand and user behavior. The prediction results of the two are combined to generate comprehensive operational strategy effect prediction data;

[0014] S7. Based on the operation strategy effect prediction data and real-time operation monitoring set, the current operation strategy is adaptively adjusted, the allocation of platform resources is dynamically adjusted according to the predicted effects of different strategies, the user experience and platform resource utilization are optimized in real time, and when the changes in user behavior, transaction volume or system load exceed the preset values, the corresponding strategy optimization plan is generated.

[0015] Optionally, the S1 specifically includes:

[0016] S11. Collect user behavior data from digital platformsu , the user behavior data includes the user's access frequency, stay time and page jump path;

[0017] S12. Collect transaction record data from the digital platform D t , the transaction record data includes transaction frequency, transaction amount and transaction type;

[0018] S13. Collect system operation log data from the digital platform D s , the operation log data includes system resource calls, page loading time, and system error information;

[0019] S14. Integrate the collected user behavior data, transaction record data and system operation log data to generate a comprehensive operation monitoring set:

[0020] D op =D u ∪D t ∪D s .

[0021] Optionally, the S2 specifically includes:

[0022] S21, operation monitoring set D op The data in the data is cleaned to remove missing values, outliers and redundant data, and the cleaned operation monitoring set D is generated. cl ;

[0023] S22, the cleaned operation monitoring set D cl Perform deduplication processing to remove duplicate records and obtain the deduplicated operation monitoring set D rd ;

[0024] S23, after deduplication, the operation monitoring set D rd Perform feature extraction to identify and extract features that affect key indicators of digital platform operations, including user behavior features, transaction features, and system performance features, and construct the feature extraction results into a feature data set:

[0025] D ft ={f1,f2,…,f n};

[0026] Among them, f i represents the i-th key feature extracted, and n is the number of features;

[0027] S24, feature data set D ft Standardization is performed to convert each key characteristic value D ft Convert to normalized value f i ′ , and obtain the standardized pre-processed operation monitoring set Dst .

[0028] Optionally, the S3 specifically includes:

[0029] S31, based on pre-processed operation monitoring set D st Build a discrete event simulation model:

[0030] Define the event collection:

[0031] E={E u ,E t ,E s};

[0032] Among them, E u ={e u1 ,e u2 ,…,e up} represents a set of user behavior events, e ui is the i-th user behavior event, p is the number of types of user behavior events, E t ={e t1 ,e t2 ,…,e tq} represents a transaction event set, e tj is the jth transaction event, q is the number of transaction events, E s ={e s1 ,e s2 ,…,e sr} represents a set of system operation events, e sk is the kth system operation event, r is the number of types of system operation events;

[0033] According to the pre-processed operation monitoring set D st The characteristic data in the event probability density function λ is defined e (t), used to describe the occurrence intensity of event e at time t:

[0034]

[0035] Among them, β e is the baseline incidence rate γ of event e e is the feature weight vector of event e, F e (t) is the feature vector corresponding to time t, which contains the key feature parameters related to event e;

[0036] Using the event probability density function λ e (t) Generates the occurrence time series of event e The time when the nth event occurs By solving the following integral equation:

[0037]

[0038] Among them, U is a random number between (0,1),

[0039] Define the system state transition function:

[0040] S(t)=S(t - )+ΔS e ;

[0041] Among them, S(t - ) is the system state before the event at time t occurs, ΔS e is the system state change caused by event e, ΔS e The pre-processed operation monitoring set D st The characteristic data and event type in are jointly determined;

[0042] S32. Configure user behavior event triggers: For each user behavior event Based on the preprocessed operation monitoring set D st Define the probability density function of user behavior events and generate user behavior events The occurrence time series And use the system state transfer function to update the system state and simulate the real-time behavior of users on the digital platform;

[0043] S33. Configure transaction event triggers. For each transaction event Based on the preprocessed operation monitoring set D st The transaction features in the definition of the probability density function of the transaction event, generate transaction events The occurrence time series And use the system state transfer function to update the system state and simulate the transaction process of the digital platform;

[0044] S34. Configure system operation event triggers: For each system operation event Based on the preprocessed operation monitoring set D st The system performance characteristics in the system are defined, and the probability density function of the occurrence is defined to generate system operation events. The occurrence time series And use the system state transfer function to update the system state and simulate the system operation process of the digital platform;

[0045] S35, comprehensively utilize the triggers of user behavior events, transaction events and system operation events to pre-process the operation monitoring set D st The data in the real-time simulation is used to generate a discrete event simulation data set D de :

[0046]

[0047] in, Indicates that event e is The system state change that occurs and causes, D de Used to simulate and monitor the operational status of the digital platform in real time.

[0048] Optionally, the S4 specifically includes:

[0049] S41, based on pre-processed operation monitoring set D st and discrete event simulation data set D de , define the relationship between the system resource demand variable R(t) and the dynamic factors of user behavior, transaction volume and system load to build a system dynamics simulation model, which is used to describe the differential equation of resource demand changes:

[0050]

[0051] Among them, α, β and γ represent the influence weights of user behavior factors, transaction volume factors and system load factors on resource demand changes, δ is the resource consumption coefficient, U(t), T(t) and S(t) are the dynamic values ​​of user behavior, transaction volume and system load at time t, respectively;

[0052] S42. Define a user behavior function U(t) using user behavior event data to represent the behavior trend change of the user at time t. The user behavior function introduces the fluctuation of time and event frequency through the sine term to strengthen the temporal dynamic characteristics of the user behavior:

[0053]

[0054] in, is the impact weight of the i-th user behavior event, is the basic intensity of the user behavior event at time t, and Dynamic correction parameters for user behavior, Denotes the discrete event simulation data set D de The occurrence time series of the i-th user behavior event extracted from is the number of times the event occurs in the current time period;

[0055] S43. Define a trading volume function T(t) using trading event data to represent the dynamic changes of platform trading activities at time t. The trading volume function simulates the impact of changes in the frequency of trading events on the overall trading volume through an exponential decay term:

[0056]

[0057] in, is the impact weight of the j-th transaction event, is the basic intensity of the trading event at time t, is the attenuation coefficient of the transaction event, From the discrete event simulation data set D de The time series of transaction events extracted from is the number of times the event occurs in the current time period;

[0058] S44. Define the system load function S(t) using the system operation event data to represent the load change of the platform system operation at time t. The system load function constructs the nonlinear effect of the system operation on the load through the square term, reflecting the incremental effect of frequent operation events on the digital platform system load:

[0059]

[0060] in, is the impact weight of the kth system operation event, is the basic intensity of the system operation event at time t, is the weighting coefficient of system load change, is the equilibrium constant, From the discrete event simulation data set D de The time series of system operation events extracted from D de is the number of times the event occurs in the current time period;

[0061] S45. Using the differential equations in the system dynamics simulation model, the changing trend of resource demand R(t) in different time periods is calculated by numerical integration method to generate a macro trend data set D of platform resource demand. rt ,The macro trend dataset contains the resource demand values ​​at each time step t.

[0062] Optionally, the S6 specifically includes:

[0063] S61. Predict the immediate impact of a single event or short-term operational adjustment on the digital platform and define the immediate impact function I at the event level e (t), through the state change ΔS of different events in the discrete event simulation model e Estimate the immediate impact of an event when it occurs:

[0064] I e (t)=∑ e∈E ΔS e ·g e (t);

[0065] Among them, ΔS eis the system state change caused by event e, g e (t) is the time weighting factor of event e, reflecting the dynamic weight of the event’s immediate impact at different time points;

[0066] S62. Use the system dynamics simulation model to predict the resource demand and user behavior trends of the digital platform, and define the resource demand trend function R trend (t) and user behavior trend function U trend (t):

[0067]

[0068] Among them, α1, β1, γ1 and δ1 represent the resource demand change weight parameters in the system dynamics simulation model, and λ1 is the attenuation factor;

[0069] S63, the instant impact function I of the discrete event simulation model e (t) and the resource demand trend function R of the system dynamics simulation model trend (t) and user behavior trend function U trend (t) Perform weighted fusion to generate comprehensive operational strategy effect prediction data P effect (t):

[0070] P effect (t) = w1·I e (t)+w2·R trend (t)+w3·U trend (t);

[0071] Among them, w1, w2 and w3 are weight coefficients;

[0072] S64. Based on comprehensive operational strategy effect prediction data P effect (t) By analyzing the impact of different operation strategies in the future time period, the best strategy is selected and a corresponding forecast report is generated. The report content includes the immediate impact of each strategy, changes in resource demand and user behavior trends, which serve as the basis for adjusting the operation strategy.

[0073] A digital platform operation monitoring and optimization system based on simulation model, including the following modules:

[0074] The data collection module is used to collect user behavior data, transaction record data and system operation log data from the digital platform to generate an operation monitoring set. The data collection module connects to the platform database through an interface, updates the collected data in real time, and aggregates the original data of user behavior, transaction events and system operations;

[0075] The data preprocessing module is used to perform data cleaning, deduplication processing, feature extraction and standardization preprocessing on the operation monitoring set to generate a preprocessed operation monitoring set. The data preprocessing module has a built-in cleaning algorithm to remove missing values ​​and outliers, and mark and remove duplicate data;

[0076] Discrete event simulation module, which is used to build a discrete event simulation model based on the pre-processed operation monitoring set to simulate the occurrence of specific events on the digital platform, including user behavior events, transaction events and system operation events, and generate discrete event simulation data sets. The discrete event simulation module defines the event set and estimates the event occurrence time series and immediate impact through the event occurrence probability density function;

[0077] System dynamics simulation module, which is used to build a system dynamics simulation model based on the preprocessed operation monitoring set and discrete event simulation data set, and generate a macro trend data set of platform resource demand. The system dynamics simulation module describes the changes in resource demand through differential equations and predicts the long-term trend of user behavior and system load;

[0078] Operation strategy prediction module, which is used to predict the effects of different operation strategies through integrated discrete event simulation model and system dynamics simulation model, and generate comprehensive operation strategy effect prediction data;

[0079] Strategy optimization module: used to analyze the effects of different operation strategies based on operation strategy effect prediction data, select the best strategy and generate a strategy optimization report.

[0080] The beneficial effects of the present invention are:

[0081] (1) The present invention integrates discrete event simulation models and system dynamics simulation models to collect and process user behavior, transaction frequency and system load data in real time to establish a dynamic prediction model for resource demand. Compared with the traditional static resource allocation method, the present invention can automatically adjust resource allocation when user traffic suddenly changes or system load suddenly increases, thereby reducing resource waste and improving platform response speed. Through discrete event simulation models and system dynamics simulation models, the resource utilization rate and user experience of the platform are significantly improved. In a high-concurrency environment, the system bottleneck problem caused by insufficient resources can be effectively avoided, thereby ensuring the stability and scalability of the platform.

[0082] (2) The present invention predicts the immediate impact of a single event through a discrete event simulation model, and uses a system dynamics simulation model to predict long-term resource demand and user behavior change trends. These two types of prediction data are combined to generate a comprehensive operational strategy effect prediction. Different from the traditional simple prediction based on historical data, the present invention predicts the effects of different strategies in different scenarios through real-time simulation of different strategies, and dynamically adjusts the operational strategy to ensure the effectiveness of strategy optimization. This enables the platform to quickly evaluate and adjust strategies when facing a complex and changing market environment, thereby improving the platform's responsiveness and market competitiveness.

[0083] (3) The present invention integrates user behavior, transaction records and system operation data into a unified data set for comprehensive analysis through multi-level data fusion of data acquisition module, data preprocessing module, discrete event simulation module and system dynamics simulation module. It can deeply explore user behavior characteristics, transaction trends and system load changes and display the operation status in multiple dimensions in the simulation model. The multi-level analysis method greatly enhances the platform's insight into user behavior and resource requirements, enabling the platform to accurately identify changes in key operation indicators and respond, thereby achieving higher analysis accuracy and decision support in platform operation optimization. BRIEF DESCRIPTION OF THE DRAWINGS

[0084] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0085] Figure 1 A flowchart of a digital platform operation monitoring and optimization system and method based on a simulation model proposed by the present invention;

[0086] Figure 2 This is a schematic diagram of the application of a discrete event simulation model in a digital platform operation monitoring and optimization system and method based on a simulation model proposed by the present invention. DETAILED DESCRIPTION

[0087] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, which only illustrate the basic structure of the present invention in a schematic manner, and therefore only show the components related to the present invention.

[0088] refer to Figure 1-Figure 2 , a digital platform operation monitoring and optimization method based on a simulation model, comprising the following steps:

[0089] S1. Collect user behavior data, transaction record data, and system operation log data from the digital platform to build an operation monitoring set;

[0090] S2. Perform data cleaning, deduplication processing, feature extraction and standardization preprocessing on the operation monitoring set to generate a preprocessed operation monitoring set;

[0091] S3. Based on the preprocessed operation monitoring set, a discrete event simulation model is constructed to simulate the occurrence of specific events on the digital platform, and the operation status of the digital platform is simulated in real time;

[0092] S4. Based on the pre-processed operation monitoring set and the event-level operation status data generated by the discrete event simulation model, a system dynamics simulation model is constructed to generate macro trend data of platform resource demand by comprehensively considering multiple dynamic factors such as user behavior, transaction volume changes and system load;

[0093] S5. Monitor the key operational indicators of the digital platform in real time, collect real-time monitoring data and dynamically input it into the discrete event simulation model and the system dynamics simulation model, and update the event level data of the discrete event simulation model and the macro trend data of the system dynamics simulation model in real time;

[0094] S6. Predict the effects of different operational strategies through integrated discrete event simulation models and system dynamics simulation models. The discrete event simulation model is used to predict the immediate impact of a single event or short-term operational adjustment on the digital platform, and the system dynamics simulation model is used to predict the changing trend of resource demand and user behavior. The prediction results of the two are combined to generate comprehensive operational strategy effect prediction data;

[0095] S7. Based on the operation strategy effect prediction data and real-time operation monitoring set, the current operation strategy is adaptively adjusted, the allocation of platform resources is dynamically adjusted according to the predicted effects of different strategies, the user experience and platform resource utilization are optimized in real time, and when the changes in user behavior, transaction volume or system load exceed the preset values, the corresponding strategy optimization plan is generated.

[0096] In this implementation, S1 specifically includes:

[0097] S11. Collect user behavior data from digital platforms u ,User behavior data includes user’s visit frequency, stay time and page jump path;

[0098] S12. Collect transaction record data from the digital platform D t ,Transaction record data includes transaction frequency, transaction amount and transaction type;

[0099] S13. Collect system operation log data from the digital platform D s ,Operation log data includes system resource calls, page loading time, and system error information;

[0100] S14. Integrate the collected user behavior data, transaction record data and system operation log data to generate a comprehensive operation monitoring set:

[0101] D op =D u ∪D t ∪D s .

[0102] In this implementation, S2 specifically includes:

[0103] S21, operation monitoring set D op The data in the data is cleaned to remove missing values, outliers and redundant data, and the cleaned operation monitoring set D is generated. cl ;

[0104] S22, the cleaned operation monitoring set D cl Perform deduplication processing to remove duplicate records and obtain the deduplicated operation monitoring set D rd ;

[0105] S23, after deduplication, the operation monitoring set D rd Perform feature extraction to identify and extract features that affect key indicators of digital platform operations, including user behavior features, transaction features, and system performance features, and construct the feature extraction results into a feature data set:

[0106] D ft ={f1,f2,…,f n};

[0107] Among them, f i represents the i-th key feature extracted, and n is the number of features;

[0108] S24, feature data set D ft Standardization is performed to convert each key characteristic value D ft Convert to normalized value f i ′ , and obtain the standardized pre-processed operation monitoring set D st .

[0109] In this implementation, S3 specifically includes:

[0110] S31, based on pre-processed operation monitoring set D st Build a discrete event simulation model:

[0111] Define the event collection:

[0112] E={E u ,E t ,E s};

[0113] in, Represents a collection of user behavior events. is the i-th user behavior event, p is the number of types of user behavior events, Represents a collection of transaction events. is the jth transaction event, q is the number of transaction events, Represents a collection of system operation events. is the kth system operation event, r is the number of types of system operation events;

[0114] According to the pre-processed operation monitoring set D st The characteristic data in the event probability density function λ is defined e (t), used to describe the occurrence intensity of event e at time t:

[0115]

[0116] Among them, β e is the baseline incidence rate γ of event e e is the feature weight vector of event e, F e (t) is the feature vector corresponding to time t, which contains the key feature parameters related to event e;

[0117] Using the event probability density function λ e (t) Generates the occurrence time series of event e The time when the nth event occurs By solving the following integral equation:

[0118]

[0119] Among them, U is a random number between (0,1),

[0120] Define the system state transition function:

[0121] S(t)=S(t - )+ΔS e ;

[0122] Among them, S(t - ) is the system state before the event at time t occurs, ΔS e is the system state change caused by event e, ΔS e The pre-processed operation monitoring set D st The characteristic data and event type in are jointly determined;

[0123] S32. Configure user behavior event triggers: For each user behavior event Based on the preprocessed operation monitoring set Dst Define the probability density function of user behavior events and generate user behavior events The occurrence time series And use the system state transfer function to update the system state and simulate the real-time behavior of users on the digital platform;

[0124] S33. Configure transaction event triggers. For each transaction event Based on the preprocessed operation monitoring set D st The transaction features in the definition of the probability density function of the transaction event, generate transaction events The occurrence time series And use the system state transfer function to update the system state and simulate the transaction process of the digital platform;

[0125] S34. Configure system operation event triggers: For each system operation event Based on the preprocessed operation monitoring set D st The system performance characteristics in the system are defined, and the probability density function of the occurrence is defined to generate system operation events. The occurrence time series And use the system state transfer function to update the system state and simulate the system operation process of the digital platform;

[0126] S35, comprehensively utilize the triggers of user behavior events, transaction events and system operation events to pre-process the operation monitoring set D st The data in the real-time simulation is used to generate a discrete event simulation data set D de :

[0127]

[0128] in, Indicates that event e is The system state change that occurs and causes, D de Used to simulate and monitor the operational status of the digital platform in real time.

[0129] In this implementation, S4 specifically includes:

[0130] S41, based on pre-processed operation monitoring set D st and discrete event simulation data set D de , define the relationship between the system resource demand variable R(t) and the dynamic factors of user behavior, transaction volume and system load to build a system dynamics simulation model, which is used to describe the differential equation of resource demand changes:

[0131]

[0132] Among them, α, β and γ represent the influence weights of user behavior factors, transaction volume factors and system load factors on resource demand changes, δ is the resource consumption coefficient, U(t), T(t) and S(t) are the dynamic values ​​of user behavior, transaction volume and system load at time t, respectively;

[0133] S42. Define a user behavior function U(t) using user behavior event data to represent the behavior trend change of the user at time t. The user behavior function introduces the fluctuation of time and event frequency through the sine term to strengthen the temporal dynamic characteristics of the user behavior:

[0134]

[0135] in, is the impact weight of the i-th user behavior event, is the basic intensity of the user behavior event at time t, and Dynamic correction parameters for user behavior, Denotes the discrete event simulation data set D de The occurrence time series of the i-th user behavior event extracted from is the number of times the event occurs in the current time period;

[0136] S43. Define a trading volume function T(t) using trading event data to represent the dynamic changes of platform trading activities at time t. The trading volume function simulates the impact of changes in the frequency of trading events on the overall trading volume through an exponential decay term:

[0137]

[0138] in, is the impact weight of the j-th transaction event, is the basic intensity of the trading event at time t, is the attenuation coefficient of the transaction event, From the discrete event simulation data set D de The time series of transaction events extracted from is the number of times the event occurs in the current time period;

[0139] S44. Define the system load function S(t) using the system operation event data to represent the load change of the platform system operation at time t. The system load function constructs the nonlinear effect of the system operation on the load through the square term, reflecting the incremental effect of frequent operation events on the digital platform system load:

[0140]

[0141] in, is the impact weight of the kth system operation event, is the basic intensity of the system operation event at time t, is the weighting coefficient of system load change, is the equilibrium constant, From the discrete event simulation data set D de The time series of system operation events extracted from D de is the number of times the event occurs in the current time period;

[0142] S45. Using the differential equations in the system dynamics simulation model, the changing trend of resource demand R(t) in different time periods is calculated by numerical integration method to generate a macro trend data set D of platform resource demand. rt ,The macro trend dataset contains the resource demand values ​​at each time step t.

[0143] In this implementation, S6 specifically includes:

[0144] S61. Predict the immediate impact of a single event or short-term operational adjustment on the digital platform and define the immediate impact function I at the event level e (t), through the state change ΔS of different events in the discrete event simulation model e Estimate the immediate impact of an event when it occurs:

[0145] I e (t)=∑ e∈E ΔS e ·g e (t);

[0146] Among them, ΔS e is the system state change caused by event e, g e (t) is the time weighting factor of event e, reflecting the dynamic weight of the event’s immediate impact at different time points;

[0147] S62. Use the system dynamics simulation model to predict the resource demand and user behavior trends of the digital platform, and define the resource demand trend function R trend (t) and user behavior trend function U trend (t):

[0148]

[0149] Among them, α1, β1, γ1 and δ1 represent the resource demand change weight parameters in the system dynamics simulation model, and λ1 is the attenuation factor;

[0150] S63, the instant impact function I of the discrete event simulation model e (t) and the resource demand trend function R of the system dynamics simulation modeltrend (t) and user behavior trend function U trend (t) Perform weighted fusion to generate comprehensive operational strategy effect prediction data P effect (t):

[0151] P effect (t) = w1·I e (t)+w2·R trend (t)+w3·U trend (t);

[0152] Among them, w1, w2 and w3 are weight coefficients;

[0153] S64. Based on comprehensive operational strategy effect prediction data P effect (t) By analyzing the impact of different operation strategies in the future time period, the best strategy is selected and a corresponding forecast report is generated. The report content includes the immediate impact of each strategy, changes in resource demand and user behavior trends, which serve as the basis for adjusting the operation strategy.

[0154] A digital platform operation monitoring and optimization system based on simulation model, including the following modules:

[0155] The data collection module is used to collect user behavior data, transaction record data and system operation log data from the digital platform to generate an operation monitoring set. The data collection module connects to the platform database through an interface, updates the collected data in real time, and aggregates the original data of user behavior, transaction events and system operations;

[0156] The data preprocessing module is used to perform data cleaning, deduplication processing, feature extraction and standardization preprocessing on the operation monitoring set to generate a preprocessed operation monitoring set. The data preprocessing module has a built-in cleaning algorithm to remove missing values ​​and outliers, and mark and remove duplicate data;

[0157] Discrete event simulation module, which is used to build a discrete event simulation model based on the pre-processed operation monitoring set to simulate the occurrence of specific events on the digital platform, including user behavior events, transaction events and system operation events, and generate discrete event simulation data sets. The discrete event simulation module defines the event set and estimates the event occurrence time series and immediate impact through the event occurrence probability density function;

[0158] System dynamics simulation module, which is used to build a system dynamics simulation model based on the preprocessed operation monitoring set and discrete event simulation data set, and generate a macro trend data set of platform resource demand. The system dynamics simulation module describes the changes in resource demand through differential equations and predicts the long-term trend of user behavior and system load;

[0159] Operation strategy prediction module, which is used to predict the effects of different operation strategies through integrated discrete event simulation model and system dynamics simulation model, and generate comprehensive operation strategy effect prediction data;

[0160] Strategy optimization module: used to analyze the effects of different operation strategies based on operation strategy effect prediction data, select the best strategy and generate a strategy optimization report.

[0161] Embodiment 1:

[0162] In order to verify the actual effect of the present invention in the operation monitoring and optimization of digital platforms, the experimenters used two data sets, EComm-Behavior and SysLoad-Sim, for experiments. The EComm-Behavior data set contains user behavior and transaction records of a large e-commerce platform from January 2021 to December 2021, covering visit frequency, stay duration, page jump path and order generation information; the SysLoad-Sim data set records the system operation log of the platform, including server load, resource calls, page response time and error log information. The data time span is the whole year of 2021. The experimenters divided the two data sets into training set and test set in a ratio of 8:2. The experimental process dynamically simulated and optimized the real operation data through the simulation model.

[0163] In terms of user behavior data processing, the preprocessing module is first used to clean and deduplicate the EComm-Behavior dataset. After removing invalid accesses and duplicate records and filtering out abnormal data, about 9 million user behavior records are finally retained. For system load data, the main features in SysLoad-Sim are extracted, including the number of requests per second, average CPU utilization, memory consumption, and page loading time, and the data is standardized to adapt to the input requirements of the simulation model.

[0164] During the model training process, the discrete event simulation model is first used to make short-term predictions on user behavior and system load data, simulating user visits, transactions, and system operations within a specific time period. The adaptive moment estimation optimizer is used during training, and the learning rate is set to 0.001, the mini-batch is 128, and the model is trained for 200 rounds. The discrete event simulation model is used to predict the probability of occurrence of different user behavior events, transaction events, and system operation events and their impact on resource demand. The system dynamics simulation model is used to simulate changes in resource demand based on long-term data trends to ensure that resources can be dynamically scheduled during peak periods.

[0165] During the experiment, the experimenters tested the improvements of the system of the present invention in resource allocation and user experience during the two traffic peaks of "Double Eleven" and "Mid-Year Promotion". Through real-time monitoring and dynamic scheduling, the platform entered the resource allocation optimization state 30 minutes before the start of the activity, and deployed 40% of the spare resources in advance through the prediction results of the simulation model. The specific experimental results are shown in Table 1 below:

[0166] Table 1 Experimental results and comparative analysis

[0167] index Traditional methods Method of the present invention Page load time (seconds) 5.6 2.3 System load utilization (%) 125 85 Transaction success rate (%) 82 96 User bounce rate (%) 28 14 Peak response time (seconds) 3.9 1.8 Customer satisfaction rating (out of 10) 7.2 9.3

[0168] As can be seen from Table 1 above, in terms of page loading time, under the traditional method, the page loading time increases significantly during the peak traffic period, and the average loading time reaches 5.6 seconds; while the present invention significantly reduces the loading time to 2.3 seconds by allocating resources in advance, avoiding the problem of user jump out caused by slow loading. In terms of system load utilization, in the traditional method, the platform load once reached an overload state of 125%, the server response was significantly delayed, and even the system crashed. However, the method of the present invention uses the prediction results of the simulation model to increase server resources in advance during the period with higher load, and the load utilization rate is controlled at 85%, which effectively prevents system overload. In terms of transaction success rate, on the day of "Double Eleven", the transaction success rate under the traditional method was 82%, and the transaction failure rate was higher during the period when users visited frequently; and the system of the present invention increases the transaction success rate to 96% through dynamic resource adjustment, effectively reducing the occurrence of transaction failures. In terms of user jump out rate, under the traditional method, the delay in page response time causes the user jump out rate to reach 28%; and under the optimization of the method of the present invention, the jump out rate drops to 14%, showing that users have a high recognition of the stability of the platform. In terms of peak response time, through real-time monitoring and advance deployment of simulation models, the platform's response time during peak activity is controlled within 1.8 seconds, which is 2.1 seconds shorter than traditional methods, greatly improving the user experience.

[0169] During the experiment, the experimenters constructed multiple training samples to verify the effectiveness of the present invention. The following is a specific simulation sample on the night of "Double Eleven" to demonstrate the dynamic optimization process of the platform under different event scenarios.

[0170] During the night of "Double Eleven" from 0:00 to 2:00, the platform user visits reached a peak, with about 20,000 transactions per minute. The simulation model predicted the peak of visits and transactions before the event occurred and deployed resources in advance, so that the system could maintain stable operation when the peak arrived. During this process, the model optimized and deployed for the following scenarios:

[0171] Page loading optimization: Through the prediction of the simulation model, page loading resources are allocated in advance at 0:00 to ensure that the homepage loading time is controlled within 2 seconds during the peak user access period. Under the traditional method, the peak loading time was as high as 7 seconds, but it was reduced to 1.8 seconds through advance scheduling of the method of the present invention, thereby improving the user experience.

[0172] Server load optimization: The system dynamics simulation model detected at 0:30 that the server load was close to the upper limit. Through the trend prediction of the model, backup servers were allocated to the high-load server group in advance, which effectively alleviated the server pressure. The traditional method caused the system to crash due to overload during this period, while the present invention avoided the crash through load prediction and resource scheduling.

[0173] Optimization of user behavior response: During the period of concentrated user access (22:00 to 24:00), the simulation model monitors the response time of the shopping cart and order submission process in real time, estimates the changes in the system's response time during this period in advance, predicts possible bottlenecks in order submission and allocates additional resources to ensure a smooth transaction process. Test results show that the success rate of order submission has increased by 14% through advance scheduling.

[0174] The experimental results show that the present invention has significant advantages in dealing with traffic peaks, resource allocation and improving user experience in complex operating environments. The experimental data clearly show that compared with traditional methods, the method of the present invention performs well in page loading time, system load utilization, transaction success rate and customer satisfaction. It can not only predict platform needs in advance, but also intelligently allocate resources in actual operations, thereby achieving efficient and stable operation of the platform.

[0175] The present invention integrates discrete event simulation models and system dynamics simulation models to collect and process user behavior, transaction frequency and system load data in real time to establish a dynamic prediction model for resource demand. Compared with the traditional static resource allocation method, the present invention can automatically adjust resource allocation when user traffic suddenly changes or system load suddenly increases, thereby reducing resource waste and improving platform response speed. Through discrete event simulation models and system dynamics simulation models, the resource utilization rate and user experience of the platform are significantly improved. In a high-concurrency environment, the system bottleneck problem caused by insufficient resources can be effectively avoided, thereby ensuring the stability and scalability of the platform.

[0176] The present invention predicts the immediate impact of a single event through a discrete event simulation model, and uses a system dynamics simulation model to predict long-term resource demand and user behavior change trends, and combines these two types of prediction data to generate a comprehensive operational strategy effect prediction. Different from the traditional simple prediction based on historical data, it predicts the effects of different strategies in different scenarios through real-time simulation of different strategies, and dynamically adjusts the operational strategy to ensure the effectiveness of strategy optimization, so that the platform can quickly evaluate and adjust strategies when facing a complex and changing market environment, thereby improving the platform's responsiveness and market competitiveness.

[0177] The present invention integrates user behavior, transaction records and system operation data into a unified data set for comprehensive analysis through multi-level data fusion of data acquisition modules, data preprocessing modules, discrete event simulation modules and system dynamics simulation modules. It can deeply explore user behavior characteristics, transaction trends and system load changes and display the operation status in multiple dimensions in the simulation model. The multi-level analysis method greatly enhances the platform's insight into user behavior and resource requirements, enabling the platform to accurately identify changes in key operation indicators and respond, thereby achieving higher analysis accuracy and decision support in platform operation optimization.

[0178] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A digital platform operation monitoring and optimization method based on a simulation model, characterized in that: The steps include: S1. Collect user behavior data, transaction record data, and system operation log data from the digital platform to build an operation monitoring set; S2. Perform data cleaning, deduplication processing, feature extraction and standardization preprocessing on the operation monitoring set to generate a preprocessed operation monitoring set; S3. Based on the preprocessed operation monitoring set, a discrete event simulation model is constructed to simulate the occurrence of specific events on the digital platform, and the operation status of the digital platform is simulated in real time; S4. Based on the pre-processed operation monitoring set and the event-level operation status data generated by the discrete event simulation model, a system dynamics simulation model is constructed to generate macro trend data of platform resource demand by comprehensively considering multiple dynamic factors such as user behavior, transaction volume changes and system load; S5. Monitor the key operational indicators of the digital platform in real time, collect real-time monitoring data and dynamically input it into the discrete event simulation model and the system dynamics simulation model, and update the event level data of the discrete event simulation model and the macro trend data of the system dynamics simulation model in real time; S6. Predict the effects of different operational strategies through integrated discrete event simulation models and system dynamics simulation models. The discrete event simulation model is used to predict the immediate impact of a single event or short-term operational adjustment on the digital platform, and the system dynamics simulation model is used to predict the changing trend of resource demand and user behavior. The prediction results of the two are combined to generate comprehensive operational strategy effect prediction data; S7. Based on the operation strategy effect prediction data and real-time operation monitoring set, the current operation strategy is adaptively adjusted, the allocation of platform resources is dynamically adjusted according to the predicted effects of different strategies, the user experience and platform resource utilization are optimized in real time, and when the changes in user behavior, transaction volume or system load exceed the preset values, the corresponding strategy optimization plan is generated.

2. A digital platform operation monitoring and optimization method based on a simulation model according to claim 1, characterized in that: The S1 specifically includes: S11. Collect user behavior data from digital platforms u , the user behavior data includes the user's access frequency, stay time and page jump path; S12. Collect transaction record data from the digital platform D t , the transaction record data includes transaction frequency, transaction amount and transaction type; S13. Collect system operation log data from the digital platform D s , the operation log data includes system resource calls, page loading time, and system error information; S14. Integrate the collected user behavior data, transaction record data and system operation log data to generate a comprehensive operation monitoring set: D op =D u ∪D t ∪D s 。 3. The method for optimizing digital platform operation monitoring based on simulation model according to claim 1, characterized in that: The S2 specifically includes: S21, operation monitoring set D op The data in the data is cleaned to remove missing values, outliers and redundant data, and the cleaned operation monitoring set D is generated. cl ; S22, the cleaned operation monitoring set D cl Perform deduplication processing to remove duplicate records and obtain the deduplicated operation monitoring set D rd ; S23, after deduplication, the operation monitoring set D rd Perform feature extraction to identify and extract features that affect key indicators of digital platform operations, including user behavior features, transaction features, and system performance features, and construct the feature extraction results into a feature data set: D ft ={f1,f2,…,f n }; Among them, f i represents the i-th key feature extracted, and n is the number of features; S24, feature data set D ft Standardization is performed to convert each key characteristic value D ft Convert to normalized value f i ′ , and obtain the standardized pre-processed operation monitoring set D st .

4. The method for optimizing digital platform operation monitoring based on simulation model according to claim 1, characterized in that: The S3 specifically includes: S31, based on pre-processed operation monitoring set D st Build a discrete event simulation model: Define the event collection: And={And u ,AND t ,AND s }; Among them, E u ={e u1 ,e u2 ,…,e up } represents a set of user behavior events, e ui is the i-th user behavior event, p is the number of types of user behavior events, E t ={e t1 ,e t2 ,…,e tq } represents a transaction event set, e tj is the jth transaction event, q is the number of transaction events, E s ={e s1 ,e s2 ,…,e sr } represents a set of system operation events, e sk is the kth system operation event, r is the number of types of system operation events; According to the pre-processed operation monitoring set D st The characteristic data in the event probability density function λ is defined e (t), used to describe the occurrence intensity of event e at time t: Among them, β e is the baseline incidence rate γ of event e e is the feature weight vector of event e, F e (t) is the feature vector corresponding to time t, which contains the key feature parameters related to event e; Using the event probability density function λ e (t) Generates the occurrence time series of event e The time when the nth event occurs By solving the following integral equation: Among them, U is a random number between (0,1), Define the system state transition function: S(t)=S(t - )+ΔS e ; Among them, S(t - ) is the system state before the event at time t occurs, ΔS e is the system state change caused by event e, ΔS e The pre-processed operation monitoring set D st The characteristic data and event type in are jointly determined; S32. Configure user behavior event triggers: For each user behavior event Based on the preprocessed operation monitoring set D st Define the probability density function of user behavior events and generate user behavior events The occurrence time series And use the system state transfer function to update the system state and simulate the real-time behavior of users on the digital platform; S33. Configure transaction event triggers. For each transaction event Based on the preprocessed operation monitoring set D st The transaction features in the definition of the probability density function of the transaction event, generate transaction events The occurrence time series And use the system state transfer function to update the system state and simulate the transaction process of the digital platform; S34. Configure system operation event triggers: For each system operation event Based on the preprocessed operation monitoring set D st The system performance characteristics in the system are defined, and the probability density function of the occurrence is defined to generate system operation events. The occurrence time series And use the system state transfer function to update the system state and simulate the system operation process of the digital platform; S35, comprehensively utilize the triggers of user behavior events, transaction events and system operation events to pre-process the operation monitoring set D st The data in the real-time simulation is used to generate a discrete event simulation data set D de : in, Indicates that event e is The system state change caused by the occurrence of de Used to simulate and monitor the operational status of the digital platform in real time.

5. The method for optimizing digital platform operation monitoring based on simulation model according to claim 1, characterized in that: The S4 specifically includes: S41, based on pre-processed operation monitoring set D st and discrete event simulation data set D de , define the relationship between the system resource demand variable R(t) and the dynamic factors of user behavior, transaction volume and system load to build a system dynamics simulation model, which is used to describe the differential equation of resource demand changes: Among them, α, β and γ represent the influence weights of user behavior factors, transaction volume factors and system load factors on resource demand changes, δ is the resource consumption coefficient, U(t), T(t) and S(t) are the dynamic values ​​of user behavior, transaction volume and system load at time t respectively; S42. Define a user behavior function U(t) using user behavior event data to represent the behavior trend change of the user at time t. The user behavior function introduces the fluctuation of time and event frequency through the sine term to strengthen the temporal dynamic characteristics of the user behavior: Among them, ω ui is the impact weight of the i-th user behavior event, is the basic intensity of the user behavior event at time t, and Dynamic correction parameters for user behavior, Denotes the discrete event simulation data set D de The occurrence time series of the i-th user behavior event extracted from is the number of times the event occurs in the current time period; S43. Define the trading volume function T(t) using the trading event data to represent the dynamic changes of the platform's trading activities at time t. The trading volume function simulates the impact of changes in the frequency of trading events on the overall trading volume through an exponential decay term: in, is the impact weight of the j-th transaction event, is the basic intensity of the trading event at time t, is the attenuation coefficient of the transaction event, From the discrete event simulation data set D de The time series of transaction events extracted from is the number of times the event occurs in the current time period; S44. Define the system load function S(t) using the system operation event data to represent the load change of the platform system operation at time t. The system load function constructs the nonlinear effect of the system operation on the load through the square term, reflecting the incremental effect of frequent operation events on the digital platform system load: in, is the impact weight of the kth system operation event, f sk (t) is the basic intensity of the system operation event at time t, is the weighting coefficient of system load change, η sk is the equilibrium constant, From the discrete event simulation data set D de The time series of system operation events extracted from D de is the number of times the event occurs in the current time period; S45. Using the differential equations in the system dynamics simulation model, the changing trend of resource demand R(t) in different time periods is calculated by numerical integration method to generate a macro trend data set D of platform resource demand. rt ,The macro trend dataset contains the resource demand values ​​at each time step t.

6. The method for optimizing digital platform operation monitoring based on simulation model according to claim 1, characterized in that: The S6 specifically includes: S61. Predict the immediate impact of a single event or short-term operational adjustment on the digital platform and define the immediate impact function I at the event level e (t), through the state change ΔS of different events in the discrete event simulation model e Estimate the immediate impact of an event when it occurs: I e (t)=∑ e∈E ΔS e ·g e (t); Among them, ΔS e is the system state change caused by event e, g e (t) is the time weighting factor of event e, reflecting the dynamic weight of the event’s immediate impact at different time points; S62. Use the system dynamics simulation model to predict the resource demand and user behavior trends of the digital platform, and define the resource demand trend function R trend (t) and user behavior trend function U trend (t): Among them, α1, β1, γ1 and δ1 represent the resource demand change weight parameters in the system dynamics simulation model, and λ1 is the attenuation factor; S63, the instant impact function I of the discrete event simulation model e (t) and the resource demand trend function R of the system dynamics simulation model trend (t) and user behavior trend function U trend (t) Perform weighted fusion to generate comprehensive operational strategy effect prediction data P effect (t): P effect (t)=w1·I e (t)+w2·R trend (t)+w3·U trend (t); Among them, w1, w2 and w3 are weight coefficients; S64. Based on comprehensive operational strategy effect prediction data P effect (t) By analyzing the impact of different operation strategies in the future time period, the best strategy is selected and a corresponding forecast report is generated. The report content includes the immediate impact of each strategy, changes in resource demand and user behavior trends, which serve as the basis for adjusting the operation strategy.

7. A digital platform operation monitoring and optimization system based on simulation model, characterized in that: Includes the following modules: The data collection module is used to collect user behavior data, transaction record data and system operation log data from the digital platform to generate an operation monitoring set. The data collection module connects to the platform database through an interface, updates the collected data in real time, and aggregates the original data of user behavior, transaction events and system operations; The data preprocessing module is used to perform data cleaning, deduplication processing, feature extraction and standardization preprocessing on the operation monitoring set to generate a preprocessed operation monitoring set. The data preprocessing module has a built-in cleaning algorithm to remove missing values ​​and outliers, and mark and remove duplicate data; Discrete event simulation module, which is used to build a discrete event simulation model based on the pre-processed operation monitoring set to simulate the occurrence of specific events on the digital platform, including user behavior events, transaction events and system operation events, and generate discrete event simulation data sets. The discrete event simulation module defines the event set and estimates the event occurrence time series and immediate impact through the event occurrence probability density function; System dynamics simulation module, which is used to build a system dynamics simulation model based on the preprocessed operation monitoring set and discrete event simulation data set, and generate a macro trend data set of platform resource demand. The system dynamics simulation module describes the changes in resource demand through differential equations and predicts the long-term trend of user behavior and system load; Operation strategy prediction module, which is used to predict the effects of different operation strategies through integrated discrete event simulation model and system dynamics simulation model, and generate comprehensive operation strategy effect prediction data; Strategy optimization module: used to analyze the effects of different operation strategies based on operation strategy effect prediction data, select the best strategy and generate a strategy optimization report.

Images