Advertising Placement Optimization Method and System Based on Traffic Consumption

Abstract

The present invention discloses an advertising placement optimization method and system based on traffic consumption, belonging to the technical field of advertising placement. The advertising placement optimization method and system based on traffic consumption include the following specific steps: Step 1: Real-time multi-dimensional data collection: The collection objects include: the number of requests, advertisement filling response, effective exposure volume, and previous display cost; Step 2: Sliding time window modeling: A dynamic sliding time window (default window length T = 24 hours) is adopted. The data within the window is divided into multiple sub-intervals according to time slices (each 5 minutes is a storage unit), and a ring buffer is used to store the window data. When new data enters, the oldest data block exceeding the window length is eliminated. By establishing a mathematical model between the network layer bandwidth cost and the business layer economic benefits, and at the same time proposing a dynamic regulation mechanism based on statistical process control, and designing a hybrid architecture combining SDN and big data processing, the present invention reduces the bandwidth cost.

Description

Technical Field

[0001] The present invention belongs to the technical field of advertising placement, and particularly relates to an optimization method and system for advertising placement based on traffic consumption. Background Art

[0002] The optimization method for advertising placement based on traffic consumption refers to a class of optimization strategies that take the traffic acquisition, distribution, and conversion efficiency in the advertising placement process as the core, and through data-driven and strategy adjustment, achieve "obtaining high-value traffic at lower cost, more accurately allocating traffic resources, and maximizing the conversion benefit of traffic". Its core logic is: traffic is not infinite, and it is necessary to make each traffic consumption generate higher value through refined operation.

[0003] The following technical defects exist in the current optimization methods and systems for advertising placement based on traffic consumption: the fixed traffic allocation strategy causes too much bandwidth resources to be occupied by high-request and low-conversion ad positions; the traditional CPM settlement method does not establish an association model between the request volume and the consumption amount; the existing system lacks a dynamic regulation mechanism based on economic efficiency indicators. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to overcome the above-mentioned disadvantages of the prior art and provide an optimization method and system for advertising placement based on traffic consumption.

[0005] The technical solution adopted to solve the above technical problem is: the optimization method for advertising placement based on traffic consumption includes the following specific steps:

[0006] Step 1: Real-time collection of multi-dimensional data:

[0007] The collection objects include: the number of requests, ad fill response, effective exposure volume, and previous display cost;

[0008] Step 2: Sliding time window modeling:

[0009] Adopt a dynamic sliding time window (default window length T = 24 hours), the data within the window is divided into multiple sub-intervals by time slices (each 5 minutes is a storage unit), use a ring buffer (Ring Buffer) to store the window data, when new data enters, the oldest data block beyond the window length is eliminated;

[0010] Step 3: Efficiency evaluation:

[0011] Calculate the traffic efficiency coefficient using a formula, and then perform dynamic smoothing processing on the traffic efficiency coefficient;

[0012] Step 4: Dynamic regulation:

[0013] First, construct a regulation decision matrix, then set the interval division and regulation rules, and then perform elastic bandwidth allocation;

[0014] Step Five: Real-time traffic control:

[0015] Send bandwidth quota instructions to edge nodes through the OpenFlow protocol, adjust the queue scheduling algorithm, limit the request sending rate of inefficient ad slots, reserve a bandwidth buffer for efficient ad slots, and ensure elastic expansion during burst traffic.

[0016] Through the above technical solutions, by establishing a mathematical model of network layer bandwidth cost and business layer economic benefits, while proposing a dynamic regulation mechanism based on statistical process control, and designing a hybrid architecture that combines SDN and big data processing, the bandwidth cost is reduced.

[0017] Furthermore, the calculation of the traffic efficiency coefficient adopts the following formula:

[0018] For each ad slot , calculate its traffic efficiency coefficient The specific formula is as follows:

[0019] Among them, the numerator: the total consumption amount corresponding to all exposures within the statistical window ( is the unit price of the i-th exposure, is the corresponding exposure volume);

[0020] The denominator: the total number of requests generated by ad slot k within the window period .

[0021] Furthermore, the dynamic smoothing process adopts the following formula:

[0022] Exponentially weighted moving average: Introduce a smoothing factor , and perform time series smoothing on to suppress short-term fluctuations: Among them, is the efficiency coefficient of ad slot after smoothing at time slice , is calculated based on the latest window data within time slice , and every time a time slice passes, is recalculated based on the latest window data and weighted and fused with the smoothed value of the previous period .

[0023] Furthermore, the regulation decision matrix includes the following formula:

[0024] Global statistic calculation:

[0025] System average efficiency : Statistic of all ad slots The mean value, refers to the smoothing efficiency coefficient of the advertising space ;

[0026] The standard deviation : Calculate the standard deviation of to measure the degree of dispersion of the efficiency distribution;

[0027] When , increase the traffic quota. When , maintain the current quota. When , decrease the quota in a graded manner.

[0028] Furthermore, the graded quota reduction adopts the following specific formula:

[0029] .

[0030] Furthermore, the elastic bandwidth allocation adopts the following formula:

[0031] where, is the bandwidth amount dynamically allocated to the advertisement, is the total bandwidth that the system can allocate, is the total number of advertising spaces, is the regulation intensity coefficient, strengthens the bandwidth occupancy ratio of high-efficiency advertising spaces when The denominator is the sum of the

[0032] By the above technical solution, the flexibility in bandwidth allocation can be improved.

[0033] Furthermore, the real-time traffic control adopts a feedback closed-loop, recalculating and updating the regulation decision every 30 minutes. If an advertising space is in the decreasing interval for 3 consecutive cycles, a warning is triggered and its effectiveness is manually reviewed.

[0034] By the above technical solution, the overall update speed can be improved, and the waste of traffic caused by misjudgment of the program can be avoided.

[0035] Furthermore, it includes a traffic monitoring terminal, an efficiency calculation engine, a regulation decision center, and a traffic controller. The traffic monitoring terminal is a data collection probe deployed at the CDN edge node. The efficiency calculation engine is a streaming processing cluster based on Storm. The regulation decision center includes a sliding window memory, a Bayesian dynamic parameter tuning module, and a weight allocation matrix generator. The traffic controller is a bandwidth dynamic allocation component with an SDN architecture.

[0036] The beneficial effects of the present invention are as follows: By establishing a mathematical model for the network layer bandwidth cost and the business layer economic benefits, and at the same time proposing a dynamic regulation mechanism based on statistical process control, and designing a hybrid architecture combining SDN and big data processing, the bandwidth cost is reduced. Description of the Drawings

[0037] Figure 1 is the method flow block diagram of the present invention. Detailed Implementation Modes

[0038] In order to make the objectives, technical solutions and advantages of the present invention clearer and more understandable, 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.

[0039] As Figure 1 shown, the advertising placement optimization method and system based on traffic consumption in this embodiment include the following specific steps:

[0040] Step 1: Real-time acquisition of multi-dimensional data:

[0041] The acquisition objects include: number of requests, advertisement filling response, effective exposure volume, previous display cost;

[0042] Step 2: Modeling with a sliding time window:

[0043] Adopt a dynamic sliding time window (default window length T = 24 hours). The data within the window is divided into multiple sub-intervals according to time slices (each 5 minutes is a storage unit), and a ring buffer is used to store the window data. When new data enters, the oldest data block exceeding the window length is eliminated;

[0044] Step 3: Efficiency evaluation:

[0045] Use a formula to calculate the traffic efficiency coefficient, and then perform dynamic smoothing processing on the traffic efficiency coefficient;

[0046] Step 4: Dynamic regulation:

[0047] First, construct a regulation decision matrix, then set the interval division and regulation rules, and then perform elastic bandwidth allocation;

[0048] Step 5: Real-time traffic control:

[0049] Send bandwidth quota instructions to edge nodes through the OpenFlow protocol, adjust the queue scheduling algorithm, limit the request sending rate of inefficient advertisement positions, and reserve a bandwidth buffer for efficient advertisement positions to ensure elastic expansion during burst traffic.

[0050] By establishing a mathematical model for the network layer bandwidth cost and the business layer economic benefits, while proposing a dynamic regulation mechanism based on statistical process control, and designing a hybrid architecture that combines SDN and big data processing, the bandwidth cost is reduced.

[0051] The traffic efficiency coefficient is calculated using the following formula:

[0052] For each ad slot , calculate its traffic efficiency coefficient The specific formula is as follows:

[0053] Among them, the numerator: the total consumption amount corresponding to all exposures within the statistical window ( is the unit price of the i-th exposure, is the corresponding exposure volume);

[0054] The denominator: the total number of requests generated by ad slot k within the window period .

[0055] The dynamic smoothing process uses the following formula:

[0056] Exponentially weighted moving average: Introduce a smoothing factor , and perform time series smoothing on to suppress short-term fluctuations: Among them, is the efficiency coefficient of ad slot after smoothing at time slice , is the calculated based on the latest window data within time slice . Every time a time slice passes, is recalculated based on the latest window data and weighted and fused with the smoothed value of the previous cycle.

[0057] The regulation decision matrix includes the following formula:

[0058] Global statistic calculation:

[0059] System average efficiency : Calculate the mean value of for all ad slots, refers to the smoothed efficiency coefficient of ad slot ;

[0060] Standard deviation : Calculate the standard deviation of to measure the dispersion degree of the efficiency distribution;

[0061] When , increase the traffic quota. When When, maintain the current quota, when When, the quota is decreased step by step.

[0062] The step-by-step decreasing quota adopts the following specific formula:

[0063] .

[0064] The elastic bandwidth allocation adopts the following formula:

[0065] Wherein, is the bandwidth amount dynamically allocated to the advertisement, is the total bandwidth that the system can allocate, is the total number of advertisement positions, is the regulation intensity coefficient, When, strengthen the bandwidth occupation ratio of high-efficiency advertisement positions, and the denominator is the sum of the powers of the efficiency coefficients of all advertisement positions, ensuring the normalization of the total bandwidth allocation ratio, which can improve the flexibility during bandwidth allocation.

[0066] The real-time traffic control adopts a feedback closed-loop, recalculating every 30 minutes updating the regulation decision. If an advertisement position is in the decreasing interval for 3 consecutive cycles, a warning is triggered and its validity is manually audited.

[0067] It can improve the overall update speed and avoid waste of traffic caused by misjudgment of the program.

[0068] It includes a traffic monitoring terminal, an efficiency calculation engine, a regulation decision center and a traffic controller. The traffic monitoring terminal is a data collection probe deployed at the edge node of the CDN. The efficiency calculation engine is a streaming processing cluster based on Storm. The regulation decision center includes a sliding window memory, a Bayesian dynamic parameter adjustment module and a weight allocation matrix generator. The traffic controller is a bandwidth dynamic allocation component with an SDN architecture.

[0069] As mentioned above, it is only a preferred embodiment of the present invention and is not used to limit the protection scope of the present invention.

Claims

1. An advertising delivery optimization method based on traffic consumption, characterized in that: The specific steps include: Step 1: Real-time collection of multi-dimensional data: The collection objects include: number of requests, ad filling response, effective exposure, and previous display cost; Step 2: Sliding time window modeling: A dynamic sliding time window is used, with a default window length of T = 24 hours. The data in the window is divided into multiple sub-intervals according to time slices, with each 5 minutes as a storage unit. A ring buffer is used to store window data. When new data enters, the oldest data block that exceeds the window length is eliminated. Step 3: Efficiency evaluation: The flow efficiency coefficient is calculated by using a formula, and then the flow efficiency coefficient is dynamically smoothed; Step 4: Dynamic Control: First, build a control decision matrix, then set interval division and control rules, and then perform elastic bandwidth allocation; Step 5: Real-time traffic control: Send bandwidth quota instructions to edge nodes through the OpenFlow protocol, adjust the queue scheduling algorithm, limit the request sending rate of inefficient ad slots, reserve bandwidth buffers for efficient ad slots, and ensure elastic expansion in the event of burst traffic; The flow efficiency coefficient is calculated using the following formula: For each ad slot k, calculate its traffic efficiency coefficient η k The specific formula is as follows: Among them, the numerator is the sum of the consumption amounts corresponding to all exposures in the statistical window, CPM i is the unit price of the i-th exposure, Impressions i is the corresponding exposure; Denominator: the total number of requests generated by ad slot k during the window period Rsquests k ; The dynamic smoothing process uses the following formula: Exponentially weighted moving average: Introducing a smoothing factor α to k Smooth the time series to suppress short-term fluctuations: or k '(t)=αη k (t)+(1-a)n k '(t-1) Among them, η k '(t) is the efficiency coefficient of the smoothed ad slot k in time slice t, η k (t) is η calculated based on the latest window data in time slice t k , after each time slice, η is recalculated based on the latest window data k (t), and the smoothed value η of the previous period k '(t-1) weighted fusion; The control decision matrix includes the following formula: Global statistics calculation; System average efficiency μ: Statistics of all ad slots η k 'Mean, η k ' refers to the smoothing efficiency coefficient of ad slot k; Standard deviation σ: Calculate η k ' standard deviation, which measures the dispersion of efficiency distribution; When η k '>μ+2σ, increase the traffic quota, when μ≤η k When '≤μ+2σ, maintain the current quota. k When '≤μ, the quota is reduced in stages.

2. The method for optimizing advertisement delivery based on traffic consumption according to claim 1, characterized in that: The tiered decreasing quota adopts the following specific formula:

3. The method for optimizing advertisement delivery based on traffic consumption according to claim 2, characterized in that: The elastic bandwidth allocation adopts the following formula: Among them, BW k is the amount of bandwidth dynamically allocated to advertisements, BW total is the total bandwidth that can be allocated by the system, M is the total number of ad slots, β is the regulation intensity coefficient, and when β>1, the bandwidth share of high-efficiency ad slots is strengthened. The denominator is the sum of the β-power of the efficiency coefficients of all ad slots to ensure the normalization of the total bandwidth allocation ratio.

4. The method for optimizing advertisement delivery based on traffic consumption according to claim 3, characterized in that: The real-time flow control adopts a feedback closed loop, recalculating η every 30 minutes k 'Update the control decision. If an ad slot is in a decreasing range for three consecutive cycles, a warning will be triggered and its effectiveness will be manually reviewed.

5. A system for the method for optimizing advertisement delivery based on traffic consumption as claimed in claim 4, characterized in that: It includes a traffic monitoring terminal, an efficiency calculation engine, a control decision center and a traffic controller. The traffic monitoring terminal is a data collection probe deployed at the CDN edge node. The efficiency calculation engine is based on the Storm streaming processing cluster. The control decision center includes a sliding window memory, a Bayesian dynamic parameter adjustment module and a weight allocation matrix generator. The traffic controller is a bandwidth dynamic allocation component of the SDN architecture.

Images