Advertisement management system based on big data

Abstract

The invention relates to the technical field of data processing, in particular to an advertisement management system based on big data, which comprises an acquisition module, a prediction module, a first judgment module, a first determination module, a second judgment module, a second determination module, an adjustment module and an output module. According to the invention, the key data of the e-commerce platform advertisement is collected in real time and the advertisement putting index threshold value is predicted, so that the advertisement demand is effectively determined and the advertisement type is determined; the input risk is dynamically judged and graded based on the real-time data, the potential risk is found, the putting index threshold value is adjusted, and the advertisement putting effect is further optimized; finally, targeted management suggestions are provided, which is helpful for e-commerce enterprises to reasonably plan advertisement budget, enhance market competitiveness, realize fine management and benefit maximization of advertisement delivery, and effectively solve the problems of risk identification delay and optimization adjustment lagging caused by excessive dependence on static matching strategies and lagging feedback mechanisms.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and particularly to an advertising management system based on big data. Background Art

[0002] In today's digital age, the advertising industry faces the dual challenges of massive data and a complex market environment. With the popularization of the Internet and the widespread application of mobile devices, advertising placement channels have become increasingly diversified, and user behavior data has grown explosively. Advertisers and operators need to accurately grasp market dynamics in the massive data and adjust placement strategies in real time to maximize advertising effects. However, traditional advertising management systems often have problems such as limited data processing capabilities, difficult-to-predict placement effects, and lagging risk assessment, making it difficult to meet the complex requirements of modern advertising business. Therefore, it is particularly important to develop an advertising management system with strong adaptability.

[0003] The patent document with the publication number CN112163901A discloses an advertising management system based on big data, including a central server and an advertising input module, an identity matching module, a data storage module, a classification extraction module, an analysis arrangement module, a push statistics module, and a user feedback module connected to the central server; the central server is used to receive and process data information of each module in the system; the advertising input module is used for each merchant to provide advertising push data and send it to the central server; the identity matching module is used to perform big data identity matching on each advertising push data with the user group and push the corresponding advertising push data to the user terminal according to a predetermined matching degree; the data storage module is used to store the advertising push data of all merchants and provide data retrieval for the central server; the classification extraction module is used to separate the advertising push data of all merchants by type and extract them according to different push frequencies; the analysis arrangement module is used to sort all advertising push data according to the matching degree according to the identity matching data provided by the identity matching module; the push statistics module is used to perform push situation statistics on all the already pushed advertising push data; the user feedback module is used for users to give feedback after receiving advertising push information and store it for staff to view in real time.

[0004] Thus, the advertising management system based on big data has the following problems: over-reliance on historical data matching and static push strategies, unable to respond to changes in advertising effects in real time; lack of a prediction mechanism for advertising effects, only relying on sorting and pushing according to the matching degree, unable to predict risks or optimize placement strategies; merchants cannot perceive the actual ROI of advertising placement (such as the problem of high input and low conversion); mainly relying on user identity matching and advertising classification, ignoring the real-time effect data of the advertisement itself; relying on manual processing, and there is no direct connection between feedback and push strategy adjustment; the module highly depends on the central server, and manual intervention is required to adjust the strategy. Summary of the Invention

[0005] To this end, the present invention provides an advertising management system based on big data to overcome the problems of delayed risk identification and lagged optimization adjustment in the prior art due to over-reliance on static matching strategies and lagged feedback mechanisms through real-time data monitoring and dynamic adjustment mechanisms.

[0006] To achieve the above object, the present invention provides an advertising management system based on big data, including:

[0007] A collection module for collecting the real-time investment volume, real-time click volume, real-time transaction rate of orders, and real-time search times of the target of advertisements in the e-commerce platform;

[0008] A prediction module connected to the collection module for predicting the threshold of the placement index according to the real-time investment volume, the real-time click volume, the real-time transaction rate, the real-time search times, and a preset big data prediction model;

[0009] A first determination module respectively connected to the collection module and the prediction module for determining the existence of an advertising demand according to the real-time click volume, the real-time transaction rate, the real-time search times, and the threshold of the placement index, and forming a first determination result;

[0010] A first determination module respectively connected to the collection module and the first determination module for determining the advertising type according to the first determination result, the real-time transaction rate, and the real-time search times;

[0011] A second determination module respectively connected to the collection module and the first determination module for determining the existence of an investment risk according to the advertising type, the real-time investment volume, and the real-time transaction rate, and forming a second determination result;

[0012] A second determination module respectively connected to the collection module and the second determination module for determining the risk level of the investment risk according to the second determination result, the real-time investment volume, and the real-time transaction rate;

[0013] An adjustment module respectively connected to the second determination module and the prediction module for adjusting the threshold of the placement index according to the number of occurrences of the risk level within a preset adjustment duration, and forming an adjusted threshold of the placement index;

[0014] An output module connected to the adjustment module for providing corresponding management tips for the advertising type corresponding to the risk level determined based on the adjusted threshold of the placement index.

[0015] Further, the first determination module includes:

[0016] A click volume fluctuation calculation unit is used to calculate the standard deviation of the real-time click volume within a preset first determination duration, and form a click volume fluctuation value;

[0017] A first conversion rate fluctuation calculation unit is used to calculate the standard deviation of the real-time conversion rate within the preset first determination duration, and form a first conversion rate fluctuation value;

[0018] A first search fluctuation calculation unit is used to calculate the standard deviation of the real-time search times within the preset first determination duration, and form a first search fluctuation value;

[0019] A first determination unit, which is respectively connected to the click volume fluctuation calculation unit, the first conversion rate fluctuation calculation unit, and the first search fluctuation calculation unit, is used to determine the presence of an advertising demand according to the click volume fluctuation value, the first conversion rate fluctuation value, the first search fluctuation value, and the placement index threshold, and form a first determination result.

[0020] Further, the first determination unit includes:

[0021] A placement index calculation sub-unit is used to perform a weighted sum of the click volume fluctuation value, the first conversion rate fluctuation value, the first search fluctuation value, a preset click volume weight, a preset conversion rate weight, and a preset search times weight, and form a placement index;

[0022] A first determination sub-unit, which is connected to the placement index calculation sub-unit, is used to determine the presence of an advertising demand when the placement index is greater than the placement index threshold, and form a first determination result.

[0023] Further, the first determination module includes:

[0024] A second conversion rate fluctuation calculation unit is used to calculate the standard deviation of the real-time conversion rate within a preset determination duration, and form a second conversion rate fluctuation value;

[0025] A second search fluctuation calculation unit is used to calculate the standard deviation of the real-time search times within the preset determination duration, and form a second search fluctuation value;

[0026] A first determination unit, which is respectively connected to the second conversion rate fluctuation calculation unit and the second search fluctuation calculation unit, is used to determine the advertising type according to the second conversion rate fluctuation value and the second search fluctuation value.

[0027] Further, the first determination unit includes:

[0028] A conversion rate curve drawing sub-unit is used to draw a change curve of the second conversion rate fluctuation value within the preset determination duration, and form a conversion rate curve;

[0029] A search curve drawing subunit for drawing a change curve of the second search fluctuation value within the preset determination duration to form a search curve;

[0030] A consistency calculation subunit, which is respectively connected to the conversion rate curve drawing subunit and the search curve drawing subunit, for calculating the cosine similarity between the conversion rate curve and the search curve to form a change consistency;

[0031] A first determination subunit, which is connected to the consistency calculation subunit, for determining that the advertisement type is an event advertisement when the change consistency is greater than a preset consistency threshold; determining that the advertisement type is a brand advertisement when the change consistency is less than or equal to the preset consistency threshold, and the second conversion rate fluctuation value is less than the preset second conversion rate fluctuation threshold, but the second search fluctuation value is greater than the preset second search fluctuation threshold; determining that the advertisement type is an effect advertisement when the change consistency is less than or equal to the preset consistency threshold, and the second conversion rate fluctuation value is greater than or equal to the preset second conversion rate fluctuation threshold, but the second search fluctuation value is less than or equal to the preset second search fluctuation threshold.

[0032] Further, the second determination module includes:

[0033] A first input fluctuation value calculation unit for calculating the standard deviation of the real-time input amount within a preset second determination duration to form a first input fluctuation value;

[0034] A third conversion rate fluctuation value calculation unit for calculating the standard deviation of the real-time conversion rate within the preset second determination duration to form a third conversion rate fluctuation value;

[0035] A second determination unit, which is respectively connected to the first input fluctuation value calculation unit and the third conversion rate fluctuation value calculation unit, for determining the existence of an input risk according to the first input fluctuation value and the third conversion rate fluctuation value to form a second determination result.

[0036] Further, the second determination unit includes:

[0037] An input comparison subunit for comparing the first input fluctuation value with a preset input fluctuation value threshold to form an input comparison result;

[0038] A conversion rate comparison subunit for comparing the third conversion rate fluctuation value with a preset third conversion rate fluctuation threshold to form a conversion rate comparison result;

[0039] A second determination subunit, which is respectively connected to the input comparison subunit and the conversion rate comparison subunit, is configured to determine that there is an input risk and form a second determination result when the input comparison result is that the first input fluctuation value is greater than the preset input fluctuation value threshold, or when the conversion rate comparison result is that the third conversion rate fluctuation value is less than the preset third conversion rate fluctuation threshold.

[0040] Further, the second determination module includes:

[0041] A high input risk determination unit, which is configured to determine that the input risk is a high input risk when the input comparison result is that the first input fluctuation value is greater than the preset input fluctuation value threshold and when the conversion rate comparison result is that the third conversion rate fluctuation value is less than the preset third conversion rate fluctuation threshold.

[0042] Further, the adjustment module includes:

[0043] A quantity fluctuation value calculation unit, which is configured to calculate the standard deviation of the number of times of forming the high input risk within a preset adjustment duration to form a quantity fluctuation value;

[0044] An adjustment unit, which is connected to the quantity fluctuation value calculation unit, is configured to adjust the delivery index threshold according to the quantity fluctuation value to form an adjusted delivery index threshold.

[0045] Further, the adjustment unit includes:

[0046] A quantity fluctuation deviation calculation subunit, which is configured to calculate the relative deviation between the quantity fluctuation value and the preset quantity fluctuation threshold to form a quantity fluctuation deviation when the quantity fluctuation value is greater than the preset quantity fluctuation threshold;

[0047] An adjustment subunit, which is connected to the quantity fluctuation deviation calculation subunit, is configured to reduce the delivery index threshold according to the quantity fluctuation deviation and the preset adjustment coefficient to form an adjusted delivery index threshold when the quantity fluctuation deviation is greater than the preset quantity fluctuation deviation threshold.

[0048] Compared with the prior art, the beneficial effects of the present invention are as follows. By collecting key data of e-commerce platform advertisements in real time and combining with a big data prediction model, it is possible to accurately predict the advertising placement index threshold, thereby effectively determining the advertising demand and determining the advertising type. At the same time, by dynamically determining and grading the investment risks based on real-time data, potential risks can be discovered in a timely manner and the placement strategy can be adjusted. By adjusting the advertising placement index threshold, the advertising placement effect can be further optimized, ensuring the scientificity and efficiency of advertising placement. Finally, providing targeted management suggestions for different risk levels and advertising types helps e-commerce enterprises reasonably plan their advertising budgets, improve the return on investment of advertising placement, enhance market competitiveness, achieve refined management and maximize the benefits of advertising placement, and effectively solve the problems of delayed risk identification and lagged optimization and adjustment caused by over-reliance on static matching strategies and lagged feedback mechanisms.

[0049] Furthermore, by calculating the click volume fluctuation value, the first conversion rate fluctuation value, and the first search fluctuation value within a preset first determination duration and making a comprehensive determination in combination with the advertising placement index threshold, the advertising demand can be accurately identified. This design can capture the key data fluctuation conditions in the advertising placement process in real time, help advertisers more accurately evaluate the advertising placement effect, and adjust the advertising strategy in a timely manner, thereby improving the accuracy and efficiency of advertising placement.

[0050] Furthermore, by calculating the advertising placement index and determining that there is an advertising demand when the advertising placement index is greater than the advertising placement index threshold, the fluctuation conditions of multi-dimensional data can be comprehensively considered, avoiding the limitations of a single indicator, and thus more comprehensively and accurately identifying the actual demand for advertising placement. It not only improves the accuracy of advertising demand determination but also effectively reduces misjudgments, helps advertisers better optimize the advertising placement strategy, enhances the efficiency and effect of advertising placement, and further enhances the scientificity and economic benefits of advertising placement.

[0051] Furthermore, by calculating the second conversion rate fluctuation value and the second search fluctuation value within a preset determination duration to determine the advertising type, the key data fluctuation characteristics in the advertising placement process can be accurately captured. Based on the analysis of data fluctuations, the performance differences of different advertising types within a specific time period can be more accurately identified, thereby helping advertisers select more appropriate advertising types according to the fluctuations, optimize the advertising placement strategy, improve the accuracy and effect of advertising placement, and further enhance the scientificity and economic benefits of advertising placement.

[0052] Furthermore, by plotting the conversion rate curve and the search curve and calculating the cosine similarity between the two to obtain the consistency of changes, it is possible to accurately determine the type of advertisement according to different conditions. When the consistency of changes is greater than the preset consistency threshold, it is determined as an event advertisement because the fluctuations in its conversion rate and search volume are highly consistent, indicating that the advertisement placement has had a significant synchronous impact on user behavior. When the consistency of changes is less than or equal to the preset consistency threshold, and the conversion rate fluctuates little but the search volume fluctuates greatly, it is determined as a brand advertisement because brand advertisements focus more on enhancing brand awareness and prompting users to search actively. When the consistency of changes is less than or equal to the preset consistency threshold, and the conversion rate fluctuates greatly but the search volume fluctuates little, it is determined as a performance advertisement because performance advertisements focus more on directly promoting sales. Based on the analysis of data fluctuation characteristics and correlations, it is possible to effectively distinguish the placement effects of different types of advertisements, helping advertisers choose advertising strategies more scientifically, thereby improving the accuracy and effectiveness of advertisement placement.

[0053] Furthermore, by calculating the first input fluctuation value and the third conversion rate fluctuation value within a preset second determination duration to determine the input risk, it is possible to monitor the dynamic changes of input and return in real time during the advertisement placement process, effectively identify situations where the input fluctuates greatly but the conversion rate fluctuates little, discover potential input risks in a timely manner, help advertisers make more informed decisions during the advertisement placement process, avoid unnecessary waste of funds, and thereby improve the economic benefits and return on investment of advertisement placement.

[0054] Furthermore, by comparing the first input fluctuation value with the preset input fluctuation value threshold, and the third conversion rate fluctuation value with the preset third conversion rate threshold, it is possible to accurately identify potential input risks during the advertisement placement process. When the first input fluctuation value exceeds the threshold, it indicates that the fluctuation of advertisement input is large and there may be a risk of out-of-control costs; while when the third conversion rate fluctuation value is lower than the threshold, it indicates that the promotion effect of advertisement placement on sales is not obvious and there may be a risk of insufficient return on investment. Through the double comparison, it is possible to issue a warning in a timely manner when the input is too high or the return is insufficient, help advertisers adjust the placement strategy in advance, optimize the advertisement budget allocation, and thereby effectively reduce the input risk and improve the economic benefits and return on investment of advertisement placement.

[0055] Furthermore, through a multi-level risk determination mechanism, the system can accurately identify high-investment risks based on different volatility situations. Ads with high-investment risks may face significant capital waste or unmet revenue expectations, and timely intervention is needed to avoid greater losses. In contrast, ads with medium and low investment risks are usually in a relatively stable state, with fluctuations within a controllable range and relatively little impact on the overall advertising effect. Therefore, special prompts or interventions may not be required temporarily. By concentrating resources and attention on handling ads with high-investment risks, the system can more efficiently optimize the advertising placement strategy, ensure that limited resources are preferentially allocated to the ads that require the most attention, and thus improve the overall advertising placement efficiency and return on investment.

[0056] Furthermore, by calculating the standard deviation of the number of occurrences of high-investment risks within a preset adjustment duration, a quantity fluctuation value is obtained, and based on this quantity fluctuation value, the placement index threshold is adjusted. It is possible to automatically optimize the parameter settings of advertising placement according to the dynamic changes in the risk level. Adaptive adjustment can effectively improve the flexibility and accuracy of advertising placement, help advertisers maintain the advertising placement effect and efficiency under high-investment risks, and thus improve the overall benefit of advertising placement.

[0057] Furthermore, by calculating the relative deviation between the quantity fluctuation value and the preset quantity fluctuation threshold, and reducing the placement index threshold according to the adjustment coefficient when the deviation exceeds the preset threshold, it is possible to dynamically respond to significant changes in the risk level, timely adjust the advertising placement strategy, effectively avoid poor advertising placement effects caused by excessive risk fluctuations, and at the same time, by reducing the placement index threshold, it is possible to more cautiously control the input and return of advertising placement, thereby improving the stability and economic benefits of advertising placement, and ensuring that the advertising placement strategy maintains the optimal effect in a dynamic environment. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] Figure 1 It is a schematic diagram of the big data-based advertising management system of this embodiment;

[0059] Figure 2 It is a determination logic diagram of the first determination unit of this embodiment for determining advertising demand;

[0060] Figure 3 It is a determination logic diagram of the first determination unit of this embodiment for determining the advertising type;

[0061] Figure 4 It is a determination logic diagram of the second determination unit of this embodiment for determining the existence of investment risks. DETAILED DESCRIPTION OF THE INVENTION

[0062] In order to make the objects and advantages of the present invention more clearly understood, the present invention is further described below in conjunction with 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.

[0063] The preferred embodiments of the present invention are described below with reference to the accompanying drawings. It should be understood by those skilled in the art that these embodiments are only used to explain the technical principles of the present invention and are not intended to limit the protection scope of the present invention.

[0064] See also Figure 1 As shown, it is a schematic diagram of the advertising management system based on big data in this embodiment;

[0065] This embodiment provides an advertisement management system based on big data, including:

[0066] The collection module is used to collect the real-time input volume, real-time click volume, real-time transaction rate of orders and real-time search times of the target in the e-commerce platform;

[0067] A prediction module, which is connected to the acquisition module and is used to predict a delivery index threshold according to the real-time input amount, the real-time click volume, the real-time transaction rate, the real-time search times and a preset big data prediction model;

[0068] A first determination module, which is connected to the acquisition module and the prediction module respectively, and is used to determine whether there is an advertisement demand according to the real-time click volume, the real-time transaction rate, the real-time search times and the delivery index threshold, and form a first determination result;

[0069] a first determination module, which is connected to the acquisition module and the first determination module respectively, and is used to determine the advertisement type according to the first determination result, the real-time transaction rate and the real-time search times;

[0070] A second determination module, which is connected to the acquisition module and the first determination module respectively, and is used to determine whether there is an investment risk according to the advertisement type, the real-time investment amount and the real-time transaction rate, and form a second determination result;

[0071] A second determination module, which is connected to the acquisition module and the second determination module respectively, and is used to determine the risk level of the investment risk according to the second determination result, the real-time investment amount and the real-time transaction rate;

[0072] An adjustment module, which is connected to the second determination module and the prediction module respectively, and is used to adjust the delivery index threshold according to the number of times the risk level is formed within a preset adjustment period to form an adjusted delivery index threshold;

[0073] An output module, connected to the adjustment module, is configured to provide corresponding management prompts for the advertisement type corresponding to the risk level determined based on the adjusted delivery index threshold.

[0074] The collection module constructs a complete evaluation chain from "exposure - interest - conversion - market feedback" by real-time monitoring of advertisement input volume, click volume, conversion rate, and search times. The real-time input volume refers to the amount of money a merchant spends on advertisements within a specific time period. It reflects the merchant's investment intensity in advertisements and is the basic data for measuring advertisement effectiveness. The real-time click volume refers to the number of times an advertisement is clicked by users. The higher the click volume, the greater the attraction of the advertisement to users and the more traffic it can bring to the store. The real-time conversion rate of orders refers to the ratio of the number of orders actually completed after clicking the advertisement to the click volume within a certain period, reflecting the conversion effect of the advertisement, that is, how much of the traffic brought by the advertisement is converted into actual sales. The real-time search times of the target refer to the number of times users search for the target keyword on the platform, reflecting the market demand and popularity of the target product and providing a reference for the advertisement delivery strategy.

[0075] These data are collected in real time through API interfaces, data tracking technology, and system logs, forming a dynamic association: the input volume determines the advertisement coverage scale, the click volume reflects the content attractiveness, the conversion rate verifies the commercial value, and the search times reveal potential demands; it can not only comprehensively evaluate the performance of each link of the advertisement, avoiding the limitations of traditional single indicators, but also quickly discover problems through real-time linkage analysis, providing an accurate data basis for subsequent intelligent prediction and dynamic adjustment, thereby effectively improving the scientificity and response speed of advertisement management.

[0076] As the decision-making terminal of the system, after receiving the latest data transmitted back by the adjustment module, the output module summarizes and classifies all real-time data. For the advertisement types corresponding to the risk levels determined based on the adjusted delivery index threshold, it comprehensively analyzes the operation data of the advertisements (such as click-through rate, conversion rate, etc.) using intelligent decision-making algorithms (such as reinforcement learning or Bayesian optimization), and simulates various optimization strategies such as budget reallocation, delivery time period adjustment, material optimization, and audience targeting correction. For event advertisements, it is recommended to shorten the delivery cycle, optimize the materials to highlight the time-limited offers, and at the same time increase the risk warning mechanism; for brand advertisements, it is recommended to strengthen brand monitoring and optimize the creativity to highlight the brand value; for performance advertisements, it is recommended to optimize the bidding strategy and reduce the cost of ineffective clicks and conversions. Finally, the output module integrates the analysis results and optimization suggestions, generates intelligent marketing management tips including the real-time performance of the advertisements, risk causes, level assessment, precise optimization measures, and prediction of expected effects. All tips are based on cross-analysis of multi-dimensional data and are accompanied by references to historical adjustment effects, ensuring that merchants can quickly execute data-driven optimization decisions, minimize the risk of ineffective delivery, provide data-driven dynamic optimization solutions for advertisers, and achieve the maximization of advertisement delivery efficiency, risk controllability, and sustainable improvement of marketing ROI.

[0077] The preset big data prediction model is a model constructed based on big data analysis and machine learning technologies, used to predict the delivery index threshold according to real-time advertisement data (including advertisement investment, click-through rate, conversion rate, search times).

[0078] 1. Model design concept

[0079] Due to the strong time-series dependence and multi-variable coupling characteristics of advertisement delivery data (such as the click-through rate being affected by historical exposures and the conversion rate being related to search trends), traditional statistical models (such as ARIMA) are difficult to accurately model. Therefore, the preset big data prediction model adopts the long short-term memory network (LSTM), and its advantages include:

[0080] Time-series modeling ability: The memory unit of LSTM can capture long-term dependence relationships (such as the impact of seasonal promotions).

[0081] Multi-variable collaborative analysis: Supports simultaneous input of multi-dimensional features such as advertisement investment, click-through rate, conversion rate, and search times.

[0082] Dynamic adaptability: Fine-tunes the model through online learning to adapt to real-time data streams.

[0083] 2. Initial model: LSTM network structure

[0084] 2.1 Model architecture

[0085] Input layer: Receives multi-dimensional data within a time window ([batch_size, time_steps, input_dim]), where:

[0086] input_dim = 4 (real-time input volume, click volume, conversion rate, search times).

[0087] time_steps = 7 (default 7-day sliding window, adjustable).

[0088] LSTM layer:

[0089] Adopts a double-layer LSTM, with 128 hidden units in each layer, controlling the information flow through forget gates, input gates, and output gates.

[0090] Activation functions: tanh (memory unit), sigmoid (gating).

[0091] Fully connected layer: Maps the LSTM output to the placement index threshold.

[0092] Output layer: Linearly regresses to output the predicted value (risk_threshold).

[0093] 2.2 Core parameters

[0094] Each layer of the LSTM contains multiple parameters, which determine how to process the input data, how to maintain long-term memory, and how to output the prediction results.

[0095] batch_size: Training batch size (default 32).

[0096] time_steps: Time step (historical data window, default 7 days).

[0097] hidden_units: LSTM hidden layer dimension (default 128).

[0098] Loss: Mean squared error (MSE) + L2 regularization (to prevent overfitting).

[0099] Optimizer: Adam (initial learning rate 0.001).

[0100] Weights and biases of the LSTM layer:

[0101] The LSTM layer contains four gates: input gate, forget gate, output gate, and candidate memory unit. Each gate has a set of weights and biases:

[0102] Weight matrix (W): It includes the input gate weight (W_i), forget gate weight (W_f), output gate weight (W_o), and candidate memory cell weight (W_c). The size of each matrix is (input_dim, hidden_units), where hidden_units is the number of hidden units in the LSTM layer.

[0103] Bias vector (b): Each gate has a bias term, usually of size (hidden_units,).

[0104] Activation function:

[0105] Forget gate (f): Sigmoid function (sigmoid), and the output is between [0, 1], indicating the degree of forgetting of information in the memory cell.

[0106] Input gate (i): Sigmoid function, which determines the degree of acceptance of the input information at the current time step.

[0107] Candidate memory cell (c'): Tanh function (tanh), which is used to generate the candidate memory.

[0108] Output gate (o): Sigmoid function, which determines the hidden state at the current time step.

[0109] 2.3 Hidden layer and output layer

[0110] Hidden state (h_t): The LSTM updates the hidden state iteratively over time steps, and the hidden state stores the long-term memory of the time series.

[0111] Memory cell (C_t): The memory cell of the LSTM is also updated as time steps progress, and it stores the long-term information in the sequence.

[0112] Output layer: Finally, the output of the LSTM layer will be passed to the fully connected layer or output layer to generate the prediction result.

[0113] 3. Training process

[0114] 3.1 Data preparation

[0115] Training data: Historical advertising logs (input, click-through rate, conversion rate, search volume in the past 90 days).

[0116] Label data: Manually labeled placement index thresholds.

[0117] Data normalization: Perform Min-Max normalization on the input features (to eliminate the influence of dimensionality).

[0118] 3.2 Model training

[0119] Forward propagation: Input time series data, the LSTM layer extracts features, and the fully connected layer outputs the prediction threshold.

[0120] Loss calculation: Compare the predicted value with the true label and calculate the MSE loss.

[0121] Backward propagation: Update the weights through the Adam optimizer and adjust the LSTM gating parameters.

[0122] Early stopping mechanism: If the validation set loss does not decrease for 5 consecutive rounds, terminate the training.

[0123] 3.3 Training duration

[0124] Data scale: 100,000 time series samples (about 2 hours of training on NVIDIA T4 GPU).

[0125] Convergence criterion: Validation set MSE < 0.01.

[0126] 4. Trained model

[0127] After training, the model saves the following parameters:

[0128] Weights of the LSTM layer: Weight matrices of the input gate (W_i), forget gate (W_f), and output gate (W_o).

[0129] Parameters of the fully connected layer: Weights (W_fc) and biases (b_fc) that map the LSTM output to the threshold prediction.

[0130] Normalization parameters: Mean and variance of the input features (for real-time data normalization).

[0131] 5. Use the model to output the placement index threshold

[0132] The trained LSTM model can be used to predict the placement index threshold using the following steps:

[0133] Real-time data input: The acquisition module passes in the latest 7-day advertising data (investment, click-through rate, conversion rate, search volume).

[0134] Data preprocessing: Normalization (using the mean / variance during training), and adjust to the input dimension of (1, 7, 4).

[0135] Model inference: The LSTM layer calculates the time series features, and the fully connected layer outputs the predicted placement index threshold.

[0136] The preset big data prediction model of this embodiment is based on the LSTM network. By processing the time series information in historical data, a deep learning model that can predict the threshold of the placement index is finally obtained. The core parameters of this model include the weight matrix and bias term of the LSTM layer, as well as the optimizer and loss function used during training, etc. Through training with historical data, the model can output accurate placement index thresholds to assist advertisers in managing advertisements.

[0137] By collecting the real-time input volume, real-time click volume, real-time transaction rate of products, and real-time search times of users of advertisements on the e-commerce platform. Then, based on these real-time data and the preset big data prediction model, predict the threshold of the placement index. Next, based on the real-time click volume, real-time transaction rate, real-time search times, and the threshold of the placement index, determine whether there is an advertisement demand and obtain the first determination result. According to the first determination result, real-time transaction rate, and real-time search times, determine the advertisement type. After that, based on the determined advertisement type, real-time input volume, and real-time transaction rate, determine whether there is an input risk and obtain the second determination result. According to the second determination result, real-time input volume, real-time transaction rate, and real-time click volume, determine the risk level of the input risk. According to the formation times of the risk level within the preset adjustment duration, adjust the threshold of the placement index to obtain the adjusted threshold of the placement index. Finally, provide targeted management suggestions for the advertisement type corresponding to the risk level determined based on the adjusted threshold of the placement index.

[0138] By collecting the key data of e-commerce platform advertisements in real time and combining with the big data prediction model, it is possible to accurately predict the threshold of the advertisement placement index, thereby effectively determining the advertisement demand and determining the advertisement type. At the same time, dynamically determining and grading the input risk based on real-time data can timely discover potential risks and adjust the placement strategy. By adjusting the threshold of the placement index, further optimize the advertisement placement effect, ensure the scientificity and efficiency of advertisement placement. Finally, providing targeted management suggestions for different risk levels and advertisement types helps e-commerce enterprises reasonably plan the advertisement budget, improve the return on investment of advertisement placement, enhance market competitiveness, achieve refined management and maximum benefits of advertisement placement, and effectively solve the problems of delayed risk identification and lagged optimization adjustment caused by over-reliance on static matching strategies and lagged feedback mechanisms.

[0139] Specifically, the first determination module includes:

[0140] A click volume fluctuation calculation unit for calculating the standard deviation of the real-time click volume within a preset first determination duration to form a click volume fluctuation value;

[0141] A first transaction rate fluctuation calculation unit for calculating the standard deviation of the real-time transaction rate within the preset first determination duration to form a first transaction rate fluctuation value;

[0142] A first search fluctuation calculation unit is configured to calculate the standard deviation of the real-time search times within the preset first determination duration, and form a first search fluctuation value;

[0143] A first determination unit, which is respectively connected to the click volume fluctuation calculation unit, the first conversion rate fluctuation calculation unit, and the first search fluctuation calculation unit, is configured to determine whether there is an advertising demand according to the click volume fluctuation value, the first conversion rate fluctuation value, the first search fluctuation value, and the placement index threshold, and form a first determination result.

[0144] The preset first determination duration refers to the time range in the advertising system for calculating the fluctuation values of real-time click volume, real-time conversion rate, and real-time search times, which depends on the volatility of data, business requirements, and the response speed of the system, and is usually set between 1 minute and 10 minutes. In this embodiment, it is set to 5 minutes, which can effectively capture the fluctuation of data in a short time, avoid frequent misjudgment caused by too short a duration, and better adapt to the real-time and dynamic nature of advertising placement.

[0145] By calculating the standard deviation of the real-time click volume, the click volume fluctuation value is obtained; then, by calculating the standard deviation of the real-time conversion rate, the first conversion rate fluctuation value is obtained; then, by calculating the standard deviation of the real-time search times, the first search fluctuation value is obtained; finally, according to the click volume fluctuation value, the first conversion rate fluctuation value, the first search fluctuation value, and the placement index threshold, it is determined whether there is an advertising demand, so as to obtain the first determination result.

[0146] By calculating the click volume fluctuation value, the first conversion rate fluctuation value, and the first search fluctuation value within the preset first determination duration, and making a comprehensive determination in combination with the placement index threshold, the advertising demand can be accurately identified. This design can capture the key data fluctuations in the advertising placement process in real time, help advertisers more accurately evaluate the advertising placement effect, adjust the advertising strategy in a timely manner, and thus improve the accuracy and efficiency of advertising placement.

[0147] Please continue to refer to Figure 2 as shown, which is the determination logic diagram of the first determination unit in this embodiment determining that there is an advertising demand;

[0148] The first determination unit includes:

[0149] A placement index calculation sub-unit is configured to perform a weighted sum of the click volume fluctuation value, the first conversion rate fluctuation value, the first search fluctuation value, the preset click volume weight, the preset conversion rate weight, and the preset search times weight to form a placement index;

[0150] The first determination sub-unit, which is connected to the placement index calculation sub-unit, is used to determine that there is an advertising demand when the placement index is greater than the placement index threshold, and form a first determination result.

[0151] The preset click-through rate weight is used to measure the proportion of the importance of the click-through rate in the evaluation of the advertising placement effect. It depends on the advertising target, market competition situation, and the stage of the advertising placement, and is usually set between 0.2 and 0.5. In this embodiment, it is set to 0.35, which can more effectively reflect the attractiveness of the advertisement and the user's attention, and helps to quickly evaluate the preliminary effect of the advertising placement.

[0152] The preset conversion rate weight is used to measure the proportion of the importance of the conversion rate in the evaluation of the advertising placement effect. It depends on the conversion target of the advertisement, the product type, and the market environment, and is usually set between 0.3 and 0.6. In this embodiment, it is set to 0.5, which can more directly reflect the actual contribution of the advertising placement to sales, and helps to accurately evaluate the conversion effect of the advertisement.

[0153] The preset search times weight is used to measure the proportion of the importance of the search times in the evaluation of the advertising placement effect. It depends on the target audience of the advertisement, the brand awareness, and the market competition situation, and is usually set between 0.2 and 0.4. In this embodiment, it is set to 0.15, which can reflect the user's interest and attention to the advertisement content, and at the same time avoid over-reliance on a single indicator to ensure the balance of the comprehensive evaluation.

[0154] By performing a weighted sum of the click-through rate fluctuation value, the first conversion rate fluctuation value, the first search fluctuation value, the preset click-through rate weight, the preset conversion rate weight, and the preset search times weight, the placement index is obtained; subsequently, when the placement index is greater than the placement index threshold, it is determined that there is an advertising demand, and a first determination result is obtained.

[0155] By calculating the placement index and determining that there is an advertising demand when the placement index is greater than the placement index threshold, it is possible to comprehensively consider the fluctuations of multi-dimensional data, avoid the limitations of a single indicator, and thus more comprehensively and accurately identify the actual demand for advertising placement. This not only improves the accuracy of the advertising demand determination, but also effectively reduces misjudgments, helps advertisers better optimize their advertising placement strategies, improves the efficiency and effect of advertising placement, and further enhances the scientific nature and economic benefits of advertising placement.

[0156] Specifically, the first determination module includes:

[0157] The second conversion rate fluctuation calculation unit is used to calculate the standard deviation of the real-time conversion rate within a preset determination duration, and form a second conversion rate fluctuation value;

[0158] A second search fluctuation calculation unit is configured to calculate the standard deviation of the real-time search times within the preset determination duration, and form a second search fluctuation value;

[0159] A first determination unit, which is respectively connected to the second transaction rate fluctuation calculation unit and the second search fluctuation calculation unit, is configured to determine the advertisement type according to the second transaction rate fluctuation value and the second search fluctuation value.

[0160] The preset determination duration is the time range used to calculate the standard deviation of the real-time transaction rate and the real-time search times, which depends on the volatility of the data, business requirements, and the response speed of the system. It is usually set between 10 minutes and 1 hour. In this embodiment, it is set to 30 minutes, which can not only capture the short-term fluctuations of the transaction rate and search times in a timely manner, avoiding frequent misjudgments caused by too short a duration, but also ensure that the data has a certain degree of stability and representativeness, thereby providing a more reliable basis for accurately determining the advertisement type.

[0161] By calculating the standard deviation of the real-time transaction rate, a second transaction rate fluctuation value is obtained; at the same time, by calculating the standard deviation of the real-time search times, a second search fluctuation value is obtained. Then, according to the second transaction rate fluctuation value and the second search fluctuation value, the advertisement type is determined.

[0162] By calculating the second transaction rate fluctuation value and the second search fluctuation value within the preset determination duration to determine the advertisement type, it is possible to accurately capture the key data fluctuation characteristics in the advertisement placement process. Based on the analysis of data fluctuations, it is possible to more accurately identify the performance differences of different advertisement types within a specific time period, thereby helping advertisers select more suitable advertisement types according to the fluctuations, optimize the advertisement placement strategy, improve the accuracy and effect of advertisement placement, and further enhance the scientific nature and economic benefits of advertisement placement.

[0163] Please continue to refer to Figure 3 as shown, which is the determination logic diagram for the first determination unit in this embodiment to determine the advertisement type;

[0164] The first determination unit includes:

[0165] A transaction rate curve drawing subunit is configured to draw a change curve of the second transaction rate fluctuation value within the preset determination duration, and form a transaction rate curve;

[0166] A search curve drawing subunit is configured to draw a change curve of the second search fluctuation value within the preset determination duration, and form a search curve;

[0167] A consistency calculation subunit, which is respectively connected to the transaction rate curve drawing subunit and the search curve drawing subunit, is configured to calculate the cosine similarity between the transaction rate curve and the search curve, and form a change consistency;

[0168] The first determination subunit, which is connected to the consistency calculation subunit, is used to determine that the advertisement type is an event advertisement when the change consistency is greater than the preset consistency threshold; to determine that the advertisement type is a brand advertisement when the change consistency is less than or equal to the preset consistency threshold, and the second conversion rate fluctuation value is less than the preset second conversion rate fluctuation threshold, but the second search fluctuation value is greater than the preset second search fluctuation threshold; to determine that the advertisement type is an effectiveness advertisement when the change consistency is less than or equal to the preset consistency threshold, and the second conversion rate fluctuation value is greater than or equal to the preset second conversion rate fluctuation threshold, but the second search fluctuation value is less than or equal to the preset second search fluctuation threshold.

[0169] The preset consistency threshold is a standard value used to judge the similarity degree of the change trends of the conversion rate curve and the search curve, which depends on the correlation of data fluctuations and the requirements of the advertisement placement strategy, and is usually set between 0.7 and 0.9. In this embodiment, it is set to 0.8, which can better balance the consistency and difference of curve changes, avoid misjudgment, and at the same time ensure the effective differentiation of advertisement types.

[0170] The preset second conversion rate fluctuation threshold is a standard value used to judge whether the conversion rate fluctuation is significant, which depends on the normal range of conversion rate fluctuations and the goals of advertisement placement, and is usually set between 10% and 20%. In this embodiment, it is set to 15%, which can effectively distinguish normal fluctuations from abnormal fluctuations and avoid misjudging the advertisement type due to small fluctuations.

[0171] The preset second search fluctuation threshold is a standard value used to judge whether the fluctuation of the number of searches is significant, which depends on the normal fluctuation range of the number of searches and the requirements of the advertisement placement strategy, and is usually set between 20% and 30%. In this embodiment, it is set to 25%, which can effectively distinguish normal fluctuations from abnormal fluctuations and ensure the accurate identification of brand advertisements and effectiveness advertisements.

[0172] By plotting the conversion rate curve and the search curve within a preset determined duration and calculating the cosine similarity between the two to obtain the consistency of changes, the system can accurately determine the type of advertisement according to different conditions. When the consistency of changes is greater than the preset consistency threshold, it is determined as an event advertisement; when the consistency of changes is less than or equal to the threshold, and the second conversion rate fluctuation value is less than the preset threshold but the second search fluctuation value is greater than the preset threshold, it is determined as a brand advertisement; when the consistency of changes is less than or equal to the threshold, and the second conversion rate fluctuation value is greater than or equal to the threshold but the second search fluctuation value is less than or equal to the threshold, it is determined as an effectiveness advertisement. Event advertisements usually cause significant fluctuations in both the conversion rate and the number of searches during the placement period. Therefore, when the consistency of changes between the conversion rate curve and the search curve is greater than the preset consistency threshold, it is determined as an event advertisement. Brand advertisements mainly enhance brand awareness. Their placement may significantly increase the number of searches, but have a relatively small effect on improving the conversion rate. Effectiveness advertisements mainly directly promote sales. Their placement usually significantly improves the conversion rate, but has a limited effect on increasing the number of searches.

[0173] By plotting the conversion rate curve and the search curve and calculating the cosine similarity between the two to obtain the consistency of changes, it is possible to accurately determine the type of advertisement according to different conditions. When the consistency of changes is greater than the preset consistency threshold, it is determined as an event advertisement because the fluctuations in its conversion rate and the number of searches are highly consistent, indicating that the advertisement placement has had a synchronous significant impact on user behavior. When the consistency of changes is less than or equal to the preset consistency threshold, and the conversion rate fluctuation is small but the search number fluctuation is large, it is determined as a brand advertisement because brand advertisements pay more attention to enhancing brand awareness and prompting users to actively search. When the consistency of changes is less than or equal to the preset consistency threshold, and the conversion rate fluctuation is large but the search number fluctuation is small, it is determined as an effectiveness advertisement because effectiveness advertisements pay more attention to directly promoting sales. The analysis based on the data fluctuation characteristics and correlation can effectively distinguish the placement effects of different types of advertisements, helping advertisers select advertising strategies more scientifically, thereby improving the accuracy and effectiveness of advertisement placement.

[0174] Specifically, the second determination module includes:

[0175] A first input fluctuation value calculation unit for calculating the standard deviation of the real-time input amount within a preset second determination duration to form a first input fluctuation value;

[0176] A third conversion rate fluctuation value calculation unit for calculating the standard deviation of the real-time conversion rate within the preset second determination duration to form a third conversion rate fluctuation value;

[0177] A second determination unit, which is respectively connected to the first input fluctuation value calculation unit and the third conversion rate fluctuation value calculation unit, for determining the existence of input risks according to the first input fluctuation value and the third conversion rate fluctuation value to form a second determination result.

[0178] The preset second determination duration is a time range used to evaluate the input risk of the advertisement placement, which depends on the periodic characteristics of the advertisement placement, the stability of the data, and the timeliness requirements of the business for risk assessment. It is usually set between 15 minutes and 1 hour. In this embodiment, it is set to 30 minutes, which can effectively identify the fluctuations in input and conversion rate in the short term, and avoid frequent misjudgments caused by too short a duration, so as to more accurately evaluate the input risk.

[0179] By calculating the standard deviation of the real-time input volume, the first input fluctuation value is obtained; at the same time, the standard deviation of the real-time conversion rate is calculated to obtain the third conversion rate fluctuation value. Then, based on the first input fluctuation value and the third conversion rate fluctuation value, it is determined whether there is an input risk, and the second determination result is obtained.

[0180] By calculating the first input fluctuation value and the third conversion rate fluctuation value within the preset second determination duration to determine the input risk, it is possible to monitor the dynamic changes of input and return in real time during the advertisement placement process, effectively identify the situation where the input fluctuation is large but the conversion rate fluctuation is small, timely discover potential input risks, help advertisers make more informed decisions during the advertisement placement process, avoid unnecessary waste of funds, and thus improve the economic benefits and return on investment of the advertisement placement.

[0181] Please continue to refer to Figure 4 as shown, which is the determination logic diagram for the second determination unit in this embodiment to determine that there is an input risk;

[0182] The second determination unit includes:

[0183] An input comparison subunit, which is used to compare the first input fluctuation value with a preset input fluctuation value threshold to form an input comparison result;

[0184] A conversion rate comparison subunit, which is used to compare the third conversion rate fluctuation value with a preset third conversion rate fluctuation threshold to form a conversion rate comparison result;

[0185] A second determination subunit, which is respectively connected to the input comparison subunit and the conversion rate comparison subunit, and is used to determine that there is an input risk when the input comparison result is that the first input fluctuation value is greater than the preset input fluctuation value threshold, or when the conversion rate comparison result is that the third conversion rate fluctuation value is less than the preset third conversion rate fluctuation threshold, and form a second determination result.

[0186] The preset input fluctuation value threshold refers to a standard value used to measure the degree of fluctuation of the real-time input amount during the advertising placement process. It depends on the fluctuation of historical data, the advertising placement strategy, and the advertiser's risk tolerance, and is usually set between 10% and 30%. In this embodiment, it is set to 20%, which can effectively identify the situation of excessive input fluctuation, avoid input risks caused by excessive input, and at the same time, it will not be too sensitive to trigger alarms frequently.

[0187] The preset third conversion rate fluctuation threshold refers to a standard value used to measure the degree of fluctuation of the real-time conversion rate during the advertising placement process. It depends on the fluctuation of historical conversion rates, the advertising placement target, and the market environment, and is usually set between 5% and 15%. In this embodiment, it is set to 10%, which can effectively identify the situation of too small conversion rate fluctuation, timely adjust the advertising placement strategy, and avoid waste of investment caused by poor advertising effects.

[0188] By comparing the first input fluctuation value with the preset input fluctuation value threshold, an input comparison result is obtained; at the same time, by comparing the third conversion rate fluctuation value with the preset third conversion rate fluctuation threshold, a conversion rate comparison result is obtained. When the input comparison result shows that the first input fluctuation value is greater than the preset input fluctuation value threshold, or the conversion rate comparison result shows that the third conversion rate fluctuation value is less than the preset third conversion rate fluctuation threshold, it is determined that there is an input risk, and a second determination result is obtained.

[0189] By comparing the first input fluctuation value with the preset input fluctuation value threshold and the third conversion rate fluctuation value with the preset third conversion rate fluctuation threshold, potential input risks in the advertising placement process can be accurately identified. When the first input fluctuation value exceeds the threshold, it indicates that the fluctuation of the advertising input is large, and there may be a risk of out-of-control costs; when the third conversion rate fluctuation value is lower than the threshold, it indicates that the promotion effect of the advertising placement on sales is not obvious, and there may be a risk of insufficient return on investment. Through double comparison, early warnings can be issued in a timely manner when the input is too high or the return is insufficient, helping the advertiser to adjust the placement strategy in advance, optimize the advertising budget allocation, thereby effectively reducing the input risk and improving the economic benefits and return on investment of the advertising placement.

[0190] Specifically, the second determination module includes:

[0191] A high-input-risk determination unit, configured to determine that the input risk is a high-input risk when the input comparison result is that the first input fluctuation value is greater than the preset input fluctuation value threshold and the conversion rate comparison result is that the third conversion rate fluctuation value is less than the preset third conversion rate fluctuation threshold;

[0192] By judging the input comparison result and the conversion rate comparison result. If the first input fluctuation value is greater than the preset input fluctuation value threshold, and the third conversion rate fluctuation value is less than the preset third conversion rate fluctuation threshold, it is determined that the input risk is a high input risk.

[0193] Through a multi-level risk determination mechanism, the system can accurately identify high input risks according to different fluctuation situations. Ads with high input risks may face problems such as significant waste of funds or failure to meet expected returns, and need to be intervened in a timely manner to avoid greater losses. In contrast, ads with medium and low input risks are usually in a relatively stable state, with fluctuations within a controllable range and relatively little impact on the overall placement effect, so special prompts or interventions may not be required for the time being. By concentrating resources and attention on processing ads with high input risks, the system can more efficiently optimize the ad placement strategy, ensure that limited resources are preferentially allocated to the ads that need the most attention, and thus improve the overall ad placement efficiency and return on investment.

[0194] Specifically, the adjustment module includes:

[0195] A quantity fluctuation value calculation unit for calculating the standard deviation of the number of times of the formation of the high input risk within a preset adjustment duration to form a quantity fluctuation value;

[0196] An adjustment unit connected to the quantity fluctuation value calculation unit for adjusting the placement index threshold according to the quantity fluctuation value to form an adjusted placement index threshold.

[0197] The preset adjustment duration is the time range for evaluating the number of times of the formation of the high input risk, which depends on the volatility of the data, the response speed of the system, and business requirements, and is usually set between 1 hour and 24 hours. In this embodiment, it is set to 6 hours, which can effectively reflect the dynamic changes of the risk level and avoid frequent adjustments due to too short a duration, thereby improving the stability and reliability of the system.

[0198] By calculating the standard deviation of the number of times of the formation of the high input risk, a quantity fluctuation value is obtained; then, according to this quantity fluctuation value, the placement index threshold is adjusted to obtain an adjusted placement index threshold.

[0199] By calculating the standard deviation of the number of times of the formation of the high input risk within a preset adjustment duration, a quantity fluctuation value is obtained, and according to this quantity fluctuation value, the placement index threshold is adjusted, which can automatically optimize the parameter settings of ad placement according to the dynamic changes of the risk level. Adaptive adjustment can effectively improve the flexibility and accuracy of ad placement, help advertisers maintain the effect and efficiency of ad placement under high input risks, and thus improve the overall benefit of ad placement.

[0200] Specifically, the adjustment unit includes:

[0201] A quantity fluctuation deviation calculation subunit, configured to calculate a relative deviation between the quantity fluctuation value and a preset quantity fluctuation threshold when the quantity fluctuation value is greater than the preset quantity fluctuation threshold, so as to form a quantity fluctuation deviation;

[0202] An adjustment subunit, connected to the quantity fluctuation deviation calculation subunit, configured to reduce the delivery index threshold according to the quantity fluctuation deviation and a preset adjustment coefficient when the quantity fluctuation deviation is greater than a preset quantity fluctuation deviation threshold, so as to form an adjusted delivery index threshold.

[0203] The preset quantity fluctuation threshold is a reference value for measuring whether the quantity fluctuation value is abnormal, depending on the fluctuation characteristics of the data, business requirements, and the risk tolerance of the system, and is usually set between 10% and 30%. In this embodiment, it is set to 20%, which can not only detect abnormal fluctuations in time but also avoid frequent false alarms caused by too low a threshold.

[0204] The preset quantity fluctuation deviation threshold is a reference value for measuring whether the quantity fluctuation deviation is abnormal, depending on the fluctuation characteristics of the data, business requirements, and the adjustment sensitivity of the system, and is usually set between 5% and 15%. In this embodiment, it is set to 10%, which can not only detect situations that need adjustment in time but also avoid frequent adjustments caused by too low a threshold.

[0205] The preset adjustment coefficient is a proportional factor for adjusting the delivery index threshold, depending on the adjustment strategy of the system, business requirements, and risk preference, and is usually set between 0.5 and 1. In this embodiment, it is set to 0.8, which can moderately adjust the delivery index threshold when abnormal fluctuations are detected, neither being too aggressive nor too conservative, and helps the system maintain better adaptability in a dynamic environment.

[0206] By comparing the quantity fluctuation value within a preset adjustment duration with the preset quantity fluctuation threshold, if the quantity fluctuation value is greater than the preset quantity fluctuation threshold, then calculate the relative deviation between the quantity fluctuation value and the preset quantity fluctuation threshold to obtain the quantity fluctuation deviation. When the quantity fluctuation deviation is greater than the preset quantity fluctuation deviation threshold, reduce the delivery index threshold according to the quantity fluctuation deviation and the preset adjustment coefficient, so as to obtain the adjusted delivery index threshold.

[0207] By calculating the relative deviation between the quantity fluctuation value and the preset quantity fluctuation threshold and reducing the delivery index threshold according to the adjustment coefficient when the deviation exceeds the preset threshold, it can dynamically respond to significant changes in the risk level, timely adjust the advertising delivery strategy, effectively avoid poor advertising delivery effects caused by excessive risk fluctuations, and at the same time, by reducing the delivery index threshold, it can more cautiously control the input and return of advertising delivery, thereby improving the stability and economic benefits of advertising delivery and ensuring that the advertising delivery strategy maintains the optimal effect in a dynamic environment.

[0208] So far, the technical solution of the present invention has been described in conjunction with the preferred embodiments shown in the accompanying drawings. However, it is easily understood by those skilled in the art that the protection scope of the present invention is obviously not limited to these specific embodiments. Without departing from the principle of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the protection scope of the present invention.

Claims

1. An advertising management system based on big data, characterized in that, Including: A collection module for collecting the real-time investment volume, real-time click volume, real-time transaction rate of orders, and real-time search times of the target of advertisements in the e-commerce platform; A prediction module connected to the collection module for predicting the threshold of the placement index according to the real-time investment volume, the real-time click volume, the real-time transaction rate, the real-time search times, and a preset big data prediction model; A first determination module respectively connected to the collection module and the prediction module for determining the advertisement demand according to the real-time click volume, the real-time transaction rate, the real-time search times, and the placement index threshold, and forming a first determination result; A first determination module respectively connected to the collection module and the first determination module for determining the advertisement type according to the first determination result, the real-time transaction rate, and the real-time search times; A second determination module respectively connected to the collection module and the first determination module for determining the existence of investment risks according to the advertisement type, the real-time investment volume, and the real-time transaction rate, and forming a second determination result; A second determination module respectively connected to the collection module and the second determination module for determining the risk level of the investment risk according to the second determination result, the real-time investment volume, and the real-time transaction rate; An adjustment module respectively connected to the second determination module and the prediction module for adjusting the placement index threshold according to the formation times of the risk level within a preset adjustment duration, and forming an adjusted placement index threshold; An output module connected to the adjustment module for providing corresponding management prompts for the advertisement type corresponding to the risk level determined based on the adjusted placement index threshold.

2. The advertising management system based on big data according to claim 1, wherein The first determination module includes: A click volume fluctuation calculation unit for calculating the standard deviation of the real-time click volume within a preset first determination duration, and forming a click volume fluctuation value; A first transaction rate fluctuation calculation unit for calculating the standard deviation of the real-time transaction rate within the preset first determination duration, and forming a first transaction rate fluctuation value; A first search fluctuation calculation unit for calculating the standard deviation of the real-time search times within the preset first determination duration, and forming a first search fluctuation value; A first determination unit respectively connected to the click volume fluctuation calculation unit, the first transaction rate fluctuation calculation unit, and the first search fluctuation calculation unit for determining the advertisement demand according to the click volume fluctuation value, the first transaction rate fluctuation value, the first search fluctuation value, and the placement index threshold, and forming a first determination result.

3. The advertising management system based on big data according to claim 2, wherein, The first determination unit includes: A placement index calculation sub-unit for performing weighted summation on the click volume fluctuation value, the first transaction rate fluctuation value, the first search fluctuation value, a preset click volume weight, a preset transaction rate weight, and a preset search times weight, and forming a placement index; A first determination sub-unit connected to the placement index calculation sub-unit for determining the advertisement demand when the placement index is greater than the placement index threshold, and forming a first determination result.

4. The advertising management system based on big data according to claim 3, characterized in that, The first determination module includes: A second transaction rate fluctuation calculation unit is used to calculate the standard deviation of the real-time transaction rate within a preset determination duration to form a second transaction rate fluctuation value; A second search fluctuation calculation unit is used to calculate the standard deviation of the real-time search times within the preset determination duration to form a second search fluctuation value; A first determination unit, which is respectively connected to the second transaction rate fluctuation calculation unit and the second search fluctuation calculation unit, is used to determine the advertisement type according to the second transaction rate fluctuation value and the second search fluctuation value.

5. The advertising management system based on big data according to claim 4, wherein The first determination unit includes: A transaction rate curve drawing subunit is used to draw a change curve of the second transaction rate fluctuation value within the preset determination duration to form a transaction rate curve; A search curve drawing subunit is used to draw a change curve of the second search fluctuation value within the preset determination duration to form a search curve; A consistency calculation subunit, which is respectively connected to the transaction rate curve drawing subunit and the search curve drawing subunit, is used to calculate the cosine similarity of the transaction rate curve and the search curve to form a change consistency; A first determination subunit, which is connected to the consistency calculation subunit, is used to determine that the advertisement type is an event advertisement when the change consistency is greater than a preset consistency threshold; when the change consistency is less than or equal to the preset consistency threshold, and the second transaction rate fluctuation value is less than a preset second transaction rate fluctuation threshold, but the second search fluctuation value is greater than the preset second search fluctuation threshold, determine that the advertisement type is a brand advertisement; when the change consistency is less than or equal to the preset consistency threshold, and the second transaction rate fluctuation value is greater than or equal to the preset second transaction rate fluctuation threshold, but the second search fluctuation value is less than or equal to the preset second search fluctuation threshold, determine that the advertisement type is an effect advertisement.

6. The advertising management system based on big data according to claim 5, wherein, The second determination module includes: A first input fluctuation value calculation unit is used to calculate the standard deviation of the real-time input amount within a preset second determination duration to form a first input fluctuation value; A third transaction rate fluctuation value calculation unit is used to calculate the standard deviation of the real-time transaction rate within the preset second determination duration to form a third transaction rate fluctuation value; A second determination unit, which is respectively connected to the first input fluctuation value calculation unit and the third transaction rate fluctuation value calculation unit, is used to determine the existence of an input risk according to the first input fluctuation value and the third transaction rate fluctuation value to form a second determination result.

7. The advertising management system based on big data according to claim 6, wherein, The second determination unit includes: An input comparison subunit is used to compare the first input fluctuation value and a preset input fluctuation value threshold to form an input comparison result; A transaction rate comparison subunit is used to compare the third transaction rate fluctuation value and a preset third transaction rate fluctuation threshold to form a transaction rate comparison result; A second determination subunit, which is respectively connected to the input comparison subunit and the transaction rate comparison subunit, is used to determine the existence of an input risk when the input comparison result is that the first input fluctuation value is greater than the preset input fluctuation value threshold, or when the transaction rate comparison result is that the third transaction rate fluctuation value is less than the preset third transaction rate fluctuation threshold, to form a second determination result.

8. The advertising management system based on big data according to claim 7, characterized in that, The second determination module includes: The high investment risk determination unit is used to determine that the investment risk is a high investment risk when the investment comparison result is that the first investment fluctuation value is greater than the preset investment fluctuation value threshold, and when the transaction rate comparison result is that the third transaction rate fluctuation value is less than the preset third transaction rate fluctuation threshold.

9. The advertising management system based on big data according to claim 8, characterized in that, The adjustment module comprises: A quantity fluctuation value calculation unit, used to calculate the standard deviation of the number of times the high investment risk is formed within the preset adjustment time to form a quantity fluctuation value; An adjustment unit is connected to the quantity fluctuation value calculation unit and is used to adjust the delivery index threshold according to the quantity fluctuation value to form an adjusted delivery index threshold.

10. The advertising management system based on big data according to claim 9, characterized in that, The adjustment unit comprises: A quantity fluctuation deviation calculation subunit, for calculating a relative deviation between the quantity fluctuation value and the preset quantity fluctuation threshold value when the quantity fluctuation value is greater than a preset quantity fluctuation threshold value, to form a quantity fluctuation deviation; The adjustment subunit is connected to the quantity fluctuation deviation calculation subunit and is used to reduce the delivery index threshold according to the quantity fluctuation deviation and a preset adjustment coefficient to form an adjusted delivery index threshold when the quantity fluctuation deviation is greater than a preset quantity fluctuation deviation threshold.

Images