An advertisement delivery parameter adaptation method and system fusing scene analysis

Abstract

The application relates to the field of advertisement delivery optimization, in particular to an advertisement delivery parameter adaptation method and system fusing scene analysis, obtaining multi-dimensional features of a plurality of known scenes in historical advertisement delivery data; combining a preset feature embedding model, mapping the multi-dimensional features of the plurality of known scenes into high-dimensional vectors to generate a known scene vector set, so as to construct a scene vector space; in response to receiving a new advertisement delivery request, obtaining multi-dimensional features of a new scene; based on the same feature embedding model, mapping the multi-dimensional features of the new scene into a high-dimensional vector to generate a new scene vector; determining at least one optimal associated scene from the known scenes according to the new scene vector; combining historical advertisement delivery parameters of the at least one optimal associated scene to generate initial advertisement delivery parameters of the new scene; obtaining effect feedback data of the initial advertisement delivery parameters in real time, and generating optimized advertisement delivery parameters according to the initial advertisement effect feedback data.

Description

Technical Field

[0001] This application relates to the field of advertising optimization technology, and in particular to an advertising parameter adaptation method and system that integrates scenario analysis. Background Technology

[0002] In programmatic advertising, quickly adapting precise targeting parameters to emerging advertising scenarios is key to improving ad performance and ROI. Existing technologies typically employ shallow scenario association techniques based on contextual keyword matching and static strategy migration techniques based on historical performance data.

[0003] However, existing technologies, such as shallow scenario association based on contextual keyword matching, struggle to understand the deeper semantics and emotional tendencies of scenarios, leading to serious mis-targeting. A typical example of failure is that next to a negative news report about "new energy vehicle battery safety," the system simply categorizes it as "automotive information" because it contains the keyword "new energy vehicles," and blindly targets new energy vehicle ads. This not only results in extremely low conversion rates but also severely damages brand image. The root cause is that existing technologies can only perform "superficial associations" and cannot understand the essential "strategic association" between scenarios and advertising strategies. Secondly, static strategy migration technologies based on historical performance data have a static perspective that is difficult to adapt to dynamically changing competitive environments. For example, the system might discover that a historical scenario, "users browsing travel blogs on weekend evenings," had a high click-through rate and low cost of advertising at the time. Therefore, when a new scenario exhibits similar characteristics, the system continues to use the historical low-bid strategy. However, it ignores the fact that this historical data might have been collected during the off-season, a period of mild competition; while the current new scenario might be during the peak tourist season, with extremely fierce bidding. This practice of simply applying successful strategies from the past, when competition was mild, to the current highly competitive environment leads to the failure of strategy transfer.

[0004] Therefore, existing technologies have shortcomings and need to be improved. Summary of the Invention

[0005] In order to solve one or more problems in the prior art, the main purpose of this application is to provide a method and system for adapting advertising delivery parameters by integrating scenario analysis.

[0006] To achieve the aforementioned objectives, this application proposes a method for adapting advertising delivery parameters by integrating scenario analysis, the method comprising:

[0007] Obtain multi-dimensional features from multiple known scenarios in historical advertising delivery data;

[0008] By combining a preset feature embedding model, the multi-dimensional features of the multiple known scenes are mapped into high-dimensional vectors to generate a set of known scene vectors, thereby constructing a scene vector space;

[0009] In response to receiving a new ad delivery request, obtain multi-dimensional features of the new scenario;

[0010] Based on the same feature embedding model, the multi-dimensional features of the new scene are mapped into high-dimensional vectors to generate a new scene vector;

[0011] Based on the new scene vector, determine at least one optimal associated scene from the known scenes;

[0012] By combining the historical ad delivery parameters of the at least one optimal associated scenario, the initial ad delivery parameters for the new scenario are generated;

[0013] The system acquires real-time feedback data on the performance of initial ad delivery parameters and generates optimized ad delivery parameters based on this feedback data.

[0014] This application also provides an advertising delivery parameter adaptation system that integrates scene analysis, including:

[0015] The first acquisition module is used to acquire multi-dimensional features of multiple known scenarios in historical advertising data;

[0016] The generation module is used to combine a preset feature embedding model to map the multi-dimensional features of the multiple known scenes into high-dimensional vectors, generate a set of known scene vectors, and construct a scene vector space.

[0017] The second acquisition module is used to acquire multi-dimensional features of the new scenario in response to receiving a new ad delivery request;

[0018] The mapping module is used to map the multi-dimensional features of the new scene into a high-dimensional vector based on the same feature embedding model, thereby generating a new scene vector;

[0019] The determining module is configured to determine at least one optimal associated scene from the known scenes based on the new scene vector;

[0020] The module is used to combine historical ad delivery parameters of the at least one optimal associated scenario to generate initial ad delivery parameters for the new scenario.

[0021] The optimization module is used to obtain real-time feedback data on the performance of initial ad delivery parameters and generate optimized ad delivery parameters based on the initial ad performance feedback data.

[0022] This application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of any of the methods described above.

[0023] This application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of any of the methods described above.

[0024] The advertising delivery parameter adaptation method and system based on integrated scene analysis in this application constructs a scene vector space rich in semantic information by mapping multi-dimensional features of a scene into high-dimensional vectors using a preset feature embedding model. This fundamental step enables the system to deeply understand the connotation of the scene, fundamentally overcoming the limitations of existing technologies that match based on shallow features such as keywords. It achieves a leap from "formal similarity" to "spiritual similarity," significantly improving the accuracy of scene analysis and effectively avoiding the risk of mistakenly delivering ads next to negative content and damaging brand image. Based on this, the system calculates the similarity between the new scene vector and known scene vectors, enabling it to accurately locate the optimal associated scene from the historical experience database where the strategy can be transferred, thereby generating high-quality initial delivery parameters for the new scene. This mechanism perfectly solves the cold start problem, providing a high-starting-point intelligent initialization for new scenes, greatly shortening the trial-and-error cycle and reducing early budget waste. Finally, the system forms a powerful self-evolutionary capability by collecting feedback data in real time and dynamically generating optimization parameters. This closed-loop optimization mechanism ensures that the system can quickly adapt to the characteristics of new scenes and the dynamic competitive environment, achieving continuous improvement in delivery performance. Attached Figure Description

[0025] Figure 1 This is a flowchart illustrating an embodiment of the advertising delivery parameter adaptation method based on integrated scene analysis according to this application.

[0026] Figure 2 This is a flowchart illustrating an embodiment of the advertising delivery parameter adaptation method based on integrated scene analysis according to this application.

[0027] Figure 3 This is a schematic block diagram of an advertising delivery parameter adaptation system that integrates scene analysis according to an embodiment of this application.

[0028] Figure 4 This is a schematic block diagram of the structure of a computer device according to an embodiment of this application.

[0029] The realization of the purpose, functional features and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0030] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0031] Reference Figure 1 This application provides a method for adapting advertising delivery parameters by integrating scenario analysis, the method comprising:

[0032] S1. Obtain multi-dimensional features of multiple known scenarios from historical advertising data;

[0033] S2. Combining the preset feature embedding model, the multi-dimensional features of the multiple known scenes are mapped into high-dimensional vectors to generate a set of known scene vectors, so as to construct a scene vector space;

[0034] S3. In response to receiving a new ad delivery request, obtain multi-dimensional features of the new scenario;

[0035] S4. Based on the same feature embedding model, map the multi-dimensional features of the new scene into a high-dimensional vector to generate a new scene vector;

[0036] S5. Based on the new scene vector, determine at least one optimal associated scene from the known scenes;

[0037] S6. Combine the historical ad delivery parameters of the at least one optimal associated scenario to generate the initial ad delivery parameters for the new scenario;

[0038] S7. Obtain real-time feedback data on the effect of the initial ad delivery parameters, and generate optimized ad delivery parameters based on the initial ad performance feedback data.

[0039] As described in steps S1-S3 above, "known scenarios" refer to scenarios where the effectiveness (such as click-through rate and conversion rate) and optimal parameters have been verified in historical campaigns. "Multi-dimensional features" aim to comprehensively and holistically describe a scenario, including: User characteristics: user profile (age, gender, interest tags), historical behavior (search, browsing, purchase records). Environmental characteristics: device type (mobile phone / PC), operating system, network environment (Wi-Fi / 4G), geographical location (GPS coordinates, business district). Content characteristics: text content of the currently viewed page (keywords, themes, entities, sentiment extracted through NLP), media type (text / image / video). Time characteristics: time (morning / noon / evening), day of the week, whether it is a holiday. The fusion of multi-dimensional features provides a rich data foundation for subsequent deep semantic analysis, avoiding the one-sidedness of single-dimensional judgments, and is the data cornerstone for achieving accurate scenario understanding. Feature embedding models (such as Transformer-based deep learning models) can compress unstructured, high-dimensional, sparse raw features into a low-dimensional, dense vector. In this process, the model learns from massive amounts of data, making scenes semantically or strategically similar in the original feature space closer in distance (e.g., cosine similarity) in the vector space. The "scene vector space," composed of vectors from all known scenes, becomes the system's "policy experience base." Vector representation can capture deep semantics that keyword matching cannot express (e.g., distinguishing between "negative news about new energy vehicle battery safety" and "new energy vehicle performance evaluation"). Transforming complex scene similarity comparisons into efficient vector space distance calculations makes rapid and accurate discovery of related scenes possible. New scenes are "digitized" and "stored." The processing flow is completely consistent with that of known scenes, ensuring that the vectors of new scenes and known scenes are in the same vector space and are comparable. This is a prerequisite for any meaningful comparison. This guarantees the fairness and consistency of the system in treating new scenes, making vector space-based similarity measurement meaningful and achieving a unified measurement of old and new scenes under the same standard.

[0040] As described in steps S5-S7 above, in the constructed scene vector space, algorithms such as nearest neighbor search are used to quickly find the K "known scene vectors" that are closest to the "new scene vector". These known scenes are considered "optimal associated scenes". The underlying assumption is that the more similar the scenes are in the vector space, the more similar their applicable optimal advertising strategies are. This allows a completely new scene to immediately find the most worthy "mentors" from a vast amount of historical experience, thus breaking away from the traditional cold start mode of completely random exploration or reliance on rough rules. One or more "optimal associated scenes" are used to obtain historically validated successful advertising parameters (such as bid, creative ID, target audience, etc.), and a set of "initial advertising parameters" for the new scene is generated through weighted averaging, voting, or model fusion. This provides a high-quality, high-starting-point initial solution for the new scene. This avoids budget waste or missed opportunities due to inappropriate parameters in the early stages of a cold start, ensuring that the advertisement is in a relatively optimal state from the beginning. The system collects feedback data (such as impressions, clicks, conversions, etc.) generated after using the initial parameters in the new scene in real time, and uses this data to fine-tune the parameters. This is typically achieved through online learning algorithms or reinforcement learning, enabling the system to adapt to the uniqueness of new scenarios and dynamically changing market environments, thereby continuously improving advertising effectiveness.

[0041] As described above, by utilizing a pre-defined feature embedding model to map the multi-dimensional features of a scene into high-dimensional vectors, a scene vector space rich in semantic information is constructed. This fundamental step enables the system to deeply understand the connotation of the scene, fundamentally overcoming the limitations of existing technologies that match based on shallow features such as keywords. It achieves a leap from "formal similarity" to "spiritual similarity," significantly improving the accuracy of scene analysis and effectively avoiding the risk of mistakenly placing ads next to negative content and damaging the brand image. Based on this, by calculating the similarity between the new scene vector and known scene vectors, the system can accurately locate the optimal associated scene from the historical experience database where the strategy can be transferred, thereby generating high-quality initial deployment parameters for the new scene. This mechanism perfectly solves the cold start problem, providing a high-starting-point intelligent initialization for new scenes, greatly shortening the trial-and-error cycle and reducing early budget waste. Finally, by collecting feedback data in real time and dynamically generating optimization parameters, the system forms a powerful self-evolutionary capability. This closed-loop optimization mechanism ensures that the system can quickly adapt to the characteristics of new scenes and the dynamic competitive environment, achieving continuous improvement in deployment effectiveness.

[0042] Reference Figure 2 In one embodiment, the step of generating optimized advertising delivery parameters based on the initial advertising performance feedback data includes:

[0043] S71. Real-time acquisition of multi-dimensional feedback data generated by the initial advertising delivery parameters;

[0044] S72. Perform credibility analysis on the multivariate feedback data, and generate a credibility weight for each piece of feedback data based on the analysis results.

[0045] S73. Based on the multivariate feedback data assigned the credibility weight, perform multidimensional attribution analysis to extract the optimization factors that cause fluctuations in advertising effectiveness. The optimization factors include positive driving factors and negative inhibiting factors.

[0046] S74. Input the optimization factor into a pre-trained parameter optimization prediction model, and predict the optimization direction and adjustment range of the advertising placement parameters through the parameter optimization prediction model;

[0047] S75. Based on the prediction results, generate the optimized advertising delivery parameters.

[0048] As described above, the overall effect of advertising is comprehensively captured. "Multi-dimensional feedback data" goes beyond a single click or conversion signal, including: Exposure data: measuring reach; Click data: measuring initial interest; Conversion data: measuring final value (purchase, registration, etc.); Negative feedback: such as cancel, close, hide, etc., directly reflecting user aversion; Interaction depth: such as video completion rate, page dwell time. This constructs a three-dimensional effect evaluation system, providing rich and multi-dimensional information input for subsequent analysis, avoiding the limitations of a single indicator (such as click-through rate) that could lead to a one-sided optimization direction. Credibility analysis of the aforementioned multi-dimensional feedback data is crucial for data cleaning and purification. It acknowledges that not all feedback data is equally credible. The analysis is based on: User value: feedback from high-value historical users carries higher weight; Behavioral authenticity: identifying and reducing the weight of fake feedback such as inflated clicks and bot traffic through anti-fraud models; Feedback behavior patterns: the credibility of instant clicks differs from clicks after in-depth browsing. It effectively filters noise and fraudulent data, preventing the optimization system from being "contaminated" and ensuring that subsequent analysis is based on high-quality, reliable data, thus leading to more reliable decisions. The reasoning process moves from "phenomenon" to "cause." Attribution analysis (such as using SHAP or game theory-based attribution models) aims to answer: "Which factor(s) caused the increase or decrease in advertising effectiveness?" Positive driving factors: Identifying the factors that contribute most to positive effects (e.g., "Creative A" is particularly effective for "young users" during "evening hours"). Negative inhibiting factors: Identifying bottlenecks that lead to poor performance (e.g., "Bidding strategy B" leads to insufficient exposure in a "highly competitive environment"). This makes the optimization process interpretable. The system is no longer a "black box" but clearly identifies the driving and inhibiting factors of existing effects, providing clear and direct action guidelines for precise adjustments in the next step. Predictive models (such as gradient boosting trees or lightweight neural networks) work by learning the complex mapping relationship of "optimization factor -> optimal parameter adjustment" on historical data. It receives factors from attribution analysis and outputs quantitative adjustment suggestions (e.g., for an identified "negative inhibitory factor: insufficient bid," the model might output "increase bid by 8%"). This achieves automated, quantitative, and intelligent parameter tuning. It replaces rule adjustments relying on human experience, discovering and applying complex strategies that are difficult for the human brain to summarize, making the optimization process more efficient and scientific. It specifically applies the "direction and magnitude" output by the predictive model to the current ad delivery parameters, generating a new, optimized parameter set. This completes the closed loop from analysis to action. It translates the intelligent analysis results of all the aforementioned steps into executable operational instructions, driving the ad delivery effect to continuously approach the optimal solution.

[0049] In one embodiment, the step of determining at least one optimal associated scene from the known scenes based on the new scene vector includes:

[0050] Obtain a set of known scene vectors, and calculate the spatial distance between the new scene vector and each known scene vector in the set of known scene vectors to obtain the first similarity corresponding to each known scene;

[0051] Obtain historical advertising performance data for each of the known scenarios; calculate the matching degree between the current advertising target of the new scenario and the historical advertising performance data of each of the known scenarios to obtain a second similarity corresponding to each known scenario;

[0052] For each known scene, its first similarity and second similarity are input into a preset weighted fusion function for calculation, and the comprehensive relevance of the known scene is output based on the calculation result.

[0053] Based on the overall relevance of all known scenarios, sort them and select one or more known scenarios with the highest ranking to determine the optimal relevance scenario.

[0054] As mentioned above, based on the constructed scene vector space, measures such as cosine similarity or Euclidean distance are used to quantify the similarity between the new scene and each known scene in terms of underlying features such as users, environment, and content. The underlying assumption is that scenes with similar features may have similar user intentions. This enables rapid and preliminary scene filtering. It can efficiently retrieve the most relevant candidate scenes from a massive number of known scenes, which is the fundamental guarantee for efficient matching. It focuses on whether the effects achieved by historical scenes are consistent with the goals pursued by the new scene. For example, if the goal of the new scene is to "increase conversion rate (CVR)," then a known scene with historical performance data showing "high CVR" will have a high second similarity, even if its feature vector is not very similar to the new scene. Conversely, a scene with similar features but a historical strategy goal of "brand exposure" (low CVR) will be assigned a lower second similarity. This fundamentally solves the "mistransfer" problem. It ensures that the system looks for "mentor" scenes that are strategically referable and aligned with the goals, not just scenes that "look similar." This perfectly avoids classic mistakes such as "advertising alongside negative news" or "using mild competition strategies in a highly competitive environment," and is the core of improving the accuracy of initialization parameters. A weighted fusion function combines the two. This function can be fixed or dynamic (e.g., when the advertiser's goal is "brand exposure," it emphasizes the first similarity; when the goal is "conversion," it emphasizes the second similarity). The aim is to generate a final score that simultaneously reflects both "formal similarity" and "spiritual similarity," achieving precise quantitative evaluation. It provides a unified and comprehensive evaluation score (overall relevance) for each candidate scenario, allowing all scenarios to be fairly compared under the same metric, providing a scientific and quantitative basis for the final selection. This is the final selection process. Based on the sole criterion of overall relevance, the system sorts all known scenarios and automatically selects the top-ranked scenario as the "optimal relevance scenario," completing the closed loop from "evaluation" to "selection." It outputs a strategy source that has undergone double verification and is of the highest quality, providing the most reliable basis for generating initial deployment parameters for new scenarios, thereby greatly improving the success rate and starting point of cold starts.

[0055] In one embodiment, the step of performing multi-dimensional attribution analysis based on multivariate feedback data assigned with credibility weights to extract optimization factors that cause fluctuations in advertising effectiveness includes:

[0056] Determine whether the total amount of feedback data collected so far has reached the effective analysis threshold;

[0057] If the total amount of feedback data collected so far does not reach the effective analysis threshold, the data collection time window is extended, and the historical advertising parameters of the optimal associated scenario are weighted and calculated based on the first similarity of each known scenario to generate preliminary optimization parameters.

[0058] If this has been achieved, then initiate the multidimensional attribution analysis.

[0059] As mentioned above, based on the law of large numbers and the principle of data validity in statistics, attribution analysis (such as SHAP and counterfactual reasoning) is a powerful data analysis tool, but it requires sufficient data samples to ensure the statistical significance and stability of its results. With extremely limited data, the variance of attribution analysis results is extremely high, easily swayed by individual accidental events (such as a single click), producing misleading "optimization factors." A scientific decision-making threshold has been established. This allows the system to self-diagnose whether it currently possesses the conditions for in-depth analysis, thus avoiding forced, immature, and high-risk optimization operations during the "data infancy," reflecting the system's rigor and scientific nature. A conservative and reliable optimization strategy for the "data sparsity period" has been defined. It includes two actions: extending the data collection time window: the principle is "trading time for data," accumulating more feedback samples by waiting longer, creating conditions for subsequent reliable analysis; and reverting to a strategy weighted based on first similarity: this is the core risk-avoidance mechanism. The principle is that when lacking its own data, the most reliable source of knowledge remains the "scene vector space" and "feature semantic similarity (first similarity)" relied upon during system initialization. At this point, the system no longer attempts risky "innovation," but instead consolidates and fine-tunes its initial, deep semantic-based "best guess." It again utilizes the verified similarity to weightedly fuse known strategies, generating a smoother, more robust fine-tuning parameter, ensuring the stability of the optimization direction. When deep analysis is not possible, a low-risk optimization path is provided, still able to improve parameters using the most reliable existing information (feature similarity). Once the data volume reaches the required level, the system activates its "advanced intelligence"—the complete optimization process, including credibility analysis, attribution analysis, and predictive models. This ensures that when conditions are ripe, the system can fully leverage its powerful, data-driven optimization capabilities to make precise and in-depth strategy adjustments, thereby maximizing advertising effectiveness.

[0060] In one embodiment, before the step of calculating the matching degree between the current advertising target of the new scenario and the historical advertising performance data of each known scenario, the method further includes:

[0061] Obtain the historical competitive environment intensity index corresponding to the time when the historical data of the known scenario was generated;

[0062] Obtain the current real-time competitive environment intensity index for the new scenario;

[0063] Based on the difference between the historical competitive environment intensity index and the real-time competitive environment intensity index, the historical advertising performance data is adjusted by discount compensation.

[0064] Based on the corrected historical advertising performance data, a matching degree is calculated between the data and the current advertising target of the new scenario.

[0065] As mentioned above, historical competitive environment intensity metrics are metadata that quantifies the intensity of market bidding at a given time, including: Auction participation: the average number of advertisers participating in bidding requests throughout history. Market price level: the mean or percentile of the cost per thousand impressions (eCPM) or cost per click (CPC) for the same type of ad placement during the same historical period. Budget burn rate: the rate at which advertisers' budgets were being consumed at that time. This establishes the context of historical performance. It restores an isolated historical performance data point (such as "2% click-through rate") to the specific market environment in which it occurred, allowing us to understand whether this "2%" was achieved under conditions of moderate or intense competition. This is a prerequisite for any cross-temporal comparison. The same competitive metrics as those in step 1 are obtained from the advertising exchange (ADX) via a real-time data interface, but these reflect the real-time market conditions when the new scenario emerges. This clarifies the real and specific competitive pressure faced by the new scenario, providing a benchmark for comparison with the historical environment. There is a strong correlation between advertising performance (such as cost per click and conversion rate) and competitive intensity. The more intense the competition, the higher the cost to achieve the same effect, or in other words, the worse the effect obtained with the same budget. If the real-time competition intensity is greater than the historical competition intensity, the historical performance data is determined to have been achieved "in a more relaxed environment," thus requiring downward correction (e.g., increasing historical cost data or decreasing historical conversion rate data) to "devalue" it to an expected level that matches the current intense environment. Conversely, if real-time competition is less intense, historical data can be corrected upward, indicating that the expected performance could be better. This achieves a "time-lapse filtering" of historical data. It eliminates data bias caused by changes in the market environment, allowing a "past" successful strategy, after correction, to truly reflect its expected performance in the "current" market environment. This fundamentally solves the "marking the boat to find the sword" type of strategy migration error. The system no longer uses raw, "inflated" historical performance data, but instead uses environmentally calibrated, more comparable "corrected historical advertising performance data" to calculate the second similarity. This greatly improves the accuracy and reliability of "strategy-oriented similarity (second similarity)." The system now seeks historical scenarios that "successfully achieved the goals pursued in the new scenario under similar competitive pressures as the current one." This ensures that the "optimal correlation scenario" and its strategy found are truly feasible and applicable in the current environment.

[0066] In one embodiment, the step of applying discount compensation to the historical advertising performance data based on the difference between the historical competitive environment intensity index and the real-time competitive environment intensity index includes:

[0067] The ratio of the real-time competitive environment intensity index to the historical competitive environment intensity index is calculated to obtain the rate of change of competitive intensity.

[0068] The rate of change of competition intensity is input into a preset nonlinear mapping function, and the mapping output is used to obtain the effect data discount factor; wherein, the nonlinear mapping function is configured such that when the rate of change of competition intensity is greater than 1, the output value is less than 1, and the output value decreases as the input value increases;

[0069] The corrected historical advertising performance data is obtained by multiplying the historical advertising performance data of the known scenario by the performance data discount factor.

[0070] As mentioned above, the rate of change is calculated using a ratio (rather than the absolute difference), and the formula is: Rate of Change in Competition Intensity = Real-time Indicator / Historical Indicator. When the rate of change > 1: it means that real-time competition is more intense than in historical periods. When the rate of change = 1: it means that the competitive environment has not changed. When the rate of change < 1: it means that real-time competition is more moderate than in historical periods. This condenses a multi-dimensional and complex difference in the competitive environment into a single, dimensionless scalar value. This provides a clear and unified input signal for any subsequent correction function, allowing the correction model to be unaffected by the specific indicator dimensions, thus possessing universality and scalability. Based on a deep understanding of the laws governing the advertising bidding market: a reverse correction mechanism (output value < 1): when competition intensifies (rate of change > 1), it means that the effect achieved with the same cost historically will be more difficult to achieve in the current environment. Therefore, historical performance data (such as high conversion rates and low costs) must be discounted downwards, outputting a discount factor less than 1 to reduce its value assessment. Non-linear diminishing returns: recognizing that the erosive effect of competition on effectiveness is not linear, but follows the law of diminishing marginal utility. When competition decreases from "mild" to "moderate," the decline in effectiveness may be slow; however, when competition increases from "very intense" to "extremely intense," even a slight increase in competition can drastically reduce effectiveness. An sigmoid function or an exponential decay function can effectively simulate this effect. Nonlinear mapping reflects the complexity of the real market more accurately than simple linear discounting. Even in the face of extreme changes in competition intensity, this function outputs a reasonable, smoothly changing discount factor, preventing the system from overreacting. Using multiplication, the discount factor obtained in the previous step is directly applied to the key performance indicators (KPIs) in the historical advertising performance data. For example: Corrected click-through rate = Historical click-through rate × Discount factor. Corrected conversion cost = Historical conversion cost / Discount factor (Note: For cost-related indicators, the operation may be division, but the principle remains the same—adjustment based on factors). Multiplication is computationally extremely efficient, meeting the stringent real-time requirements of advertising bidding systems. The final generated "corrected historical advertising performance data" is an environmentally calibrated estimate of the potential effectiveness of this historical strategy under the current competitive environment. This ensures that the subsequent calculation of "second similarity" is based on fairness and comparability.

[0071] In one embodiment, after the step of generating optimized advertising delivery parameters based on the initial advertising performance feedback data, the method further includes:

[0072] During the advertising campaign, negative feedback signals in the advertising performance feedback data are monitored in real time.

[0073] If the strength of the negative feedback signal exceeds the emergency threshold within a preset time, the use of the current advertising parameters will be immediately suspended, and the system will revert to the initial advertising parameters.

[0074] As mentioned above, identifying negative behavioral signals directly related to user aversion is crucial, as these signals often reflect serious strategy failures more directly and quickly than "low click-through rates." Monitored negative feedback signals include: Actively hiding / closing ads: Users explicitly express their unwillingness to see the ad. Reporting ads: Users believe the ad content is inappropriate, fraudulent, or offensive. Negative reviews / low ratings: Direct negative feedback left on the ad interaction interface. This enables real-time perception of the user experience quality and brand safety risks associated with ad placement. It expands the system's focus from simply "poor performance" to "whether it causes negative impact," achieving a dimensional upgrade from pursuing "positive gains" to preventing "negative losses." The decision-making logic is based on a simple principle: when the system receives abnormally high levels of negative feedback within a short period, it indicates a potentially serious problem with the current optimization strategy, and the expected risks of continuing the strategy far outweigh the potential benefits. "Preset time" and "emergency threshold": These two parameters together define a "risk window," ensuring that the system only triggers the circuit breaker when truly necessary (i.e., negative feedback is not an isolated incident but an emerging trend), avoiding erroneous actions due to individual noise. "Immediate Pause": This reflects the high priority and real-time nature of the decision-making, interrupting the execution of the current harmful strategy. "Revert to the Initial Ad Serving Parameters": This is the most critical design. The principle is that the system considers the generated initial serving parameters to be a relatively robust and safe "baseline strategy" validated through semantic scenario analysis. When encountering unknown risks, reverting to this known and relatively reliable baseline point is the option with the least loss. It can quickly prevent further waste of budget and continued damage to brand image, minimizing the negative impact of the problem. It provides crucial fault tolerance for the fully automated optimization system. Even if the optimization algorithm "goes astray" in some extreme cases, the system can automatically "pull it back on track," ensuring the robustness and commercial reliability of the entire solution.

[0075] Reference Figure 3 This application also provides an advertising delivery parameter adaptation system that integrates scenario analysis, including:

[0076] The first acquisition module 1 is used to acquire multi-dimensional features of multiple known scenarios in historical advertising delivery data;

[0077] The generation module 2 is used to combine a preset feature embedding model to map the multi-dimensional features of the multiple known scenes into high-dimensional vectors, generate a set of known scene vectors, and construct a scene vector space.

[0078] The second acquisition module 3 is used to acquire multi-dimensional features of the new scenario in response to receiving a new ad delivery request;

[0079] Mapping module 4 is used to map the multi-dimensional features of the new scene into a high-dimensional vector based on the same feature embedding model, thereby generating a new scene vector;

[0080] The determining module 5 is used to determine at least one optimal associated scene from the known scenes based on the new scene vector;

[0081] Module 6 is used to generate initial advertising parameters for the new scenario by combining historical advertising parameters of the at least one optimal associated scenario.

[0082] The optimization module 7 is used to obtain real-time feedback data on the effect of the initial ad delivery parameters and generate optimized ad delivery parameters based on the initial ad performance feedback data.

[0083] As described above, it is understood that each component of the advertising delivery parameter adaptation system for integrated scenario analysis proposed in this application can realize the function of any one of the advertising delivery parameter adaptation methods for integrated scenario analysis as described above, and the specific structure will not be described in detail.

[0084] Reference Figure 4 This application also provides a computer device, which may be a server, and its internal structure may be as follows: Figure 4 As shown, the computer device includes a processor, memory, network interface, and database connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and database. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The database stores monitoring data and other data. The network interface is used for communication with external terminals via a network connection. When the computer program is executed by the processor, it implements a method for adapting advertising delivery parameters based on integrated scene analysis.

[0085] The processor described above executes the advertising delivery parameter adaptation method for the fusion scenario analysis, including: acquiring multi-dimensional features of multiple known scenarios from historical advertising delivery data; mapping the multi-dimensional features of the multiple known scenarios into high-dimensional vectors using a preset feature embedding model to generate a set of known scenario vectors to construct a scenario vector space; in response to receiving a new advertising delivery request, acquiring multi-dimensional features of the new scenario; mapping the multi-dimensional features of the new scenario into high-dimensional vectors using the same feature embedding model to generate a new scenario vector; determining at least one optimal associated scenario from the known scenarios based on the new scenario vector; generating initial advertising delivery parameters for the new scenario by combining the historical advertising delivery parameters of the at least one optimal associated scenario; acquiring real-time effect feedback data of the initial advertising delivery parameters; and generating optimized advertising delivery parameters based on the initial advertising effect feedback data.

[0086] One embodiment of this application also provides a computer-readable storage medium storing a computer program thereon. When executed by a processor, the computer program implements an advertising delivery parameter adaptation method that integrates scene analysis, including the following steps: acquiring multi-dimensional features of multiple known scenes from historical advertising delivery data; mapping the multi-dimensional features of the multiple known scenes into high-dimensional vectors using a preset feature embedding model to generate a set of known scene vectors to construct a scene vector space; in response to receiving a new advertising delivery request, acquiring multi-dimensional features of the new scene; mapping the multi-dimensional features of the new scene into high-dimensional vectors using the same feature embedding model to generate a new scene vector; determining at least one optimal associated scene from the known scenes based on the new scene vector; generating initial advertising delivery parameters for the new scene by combining historical advertising delivery parameters of the at least one optimal associated scene; acquiring real-time effect feedback data of the initial advertising delivery parameters; and generating optimized advertising delivery parameters based on the initial advertising effect feedback data.

[0087] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in this application and in the embodiments can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual-speed SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0088] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, apparatus, article, or method that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, apparatus, article, or method. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, apparatus, article, or method that includes that element.

[0089] The above description is only a preferred embodiment of this application and does not limit the patent scope of this application. Any equivalent structural or procedural changes made based on the content of this application's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of this application.

Claims

1. A method for adapting advertising delivery parameters by integrating scenario analysis, characterized in that, The method includes: Obtain multi-dimensional features from multiple known scenarios in historical advertising delivery data; By combining a preset feature embedding model, the multi-dimensional features of the multiple known scenes are mapped into high-dimensional vectors to generate a set of known scene vectors, thereby constructing a scene vector space; In response to receiving a new ad delivery request, obtain multi-dimensional features of the new scenario; Based on the same feature embedding model, the multi-dimensional features of the new scene are mapped into high-dimensional vectors to generate a new scene vector; Based on the new scene vector, determine at least one optimal associated scene from the known scenes; Obtain a set of known scene vectors, and calculate the spatial distance between the new scene vector and each known scene vector in the set of known scene vectors to obtain the first similarity corresponding to each known scene; Obtain historical advertising performance data for each of the known scenarios; obtain the historical competitive environment intensity index corresponding to the time when the historical advertising performance data was generated; Obtain the current real-time competitive environment intensity index for the new scenario; Based on the difference between the historical competitive environment intensity index and the real-time competitive environment intensity index, the historical advertising performance data is adjusted by discount compensation. The ratio of the real-time competitive environment intensity index to the historical competitive environment intensity index is calculated to obtain the rate of change of competitive intensity. The rate of change of competition intensity is input into a preset nonlinear mapping function, and the mapping output is used to obtain the effect data discount factor; wherein, the nonlinear mapping function is configured such that when the rate of change of competition intensity is greater than 1, the output value is less than 1, and the output value decreases as the input value increases; The corrected historical advertising performance data is obtained by multiplying the historical advertising performance data of the known scenario by the performance data discount factor. Based on the corrected historical advertising performance data, a matching degree is calculated between the data and the current advertising delivery target in the new scenario. The matching degree is calculated between the current advertising target of the new scenario and the historical advertising performance data of each known scenario to obtain a second similarity corresponding to each known scenario; For each known scene, its first similarity and second similarity are input into a preset weighted fusion function for calculation, and the comprehensive relevance of the known scene is output based on the calculation result. Based on the overall relevance of all known scenarios, sort them and select one or more known scenarios with the highest ranking as the optimal relevance scenarios; By combining the historical ad delivery parameters of the at least one optimal associated scenario, the initial ad delivery parameters for the new scenario are generated; Get real-time feedback data on the performance of initial ad delivery parameters, and generate optimized ad delivery parameters based on the initial ad performance feedback data.

2. The advertising delivery parameter adaptation method based on integrated scene analysis according to claim 1, characterized in that, Real-time acquisition of diverse feedback data generated by the initial ad delivery parameters; A credibility analysis is performed on the multivariate feedback data, and a credibility weight is generated for each piece of feedback data based on the analysis results. Based on the multivariate feedback data assigned the credibility weight, a multidimensional attribution analysis is performed to extract the optimization factors that cause fluctuations in advertising effectiveness. The optimization factors include positive driving factors and negative inhibiting factors. The optimization factors are input into a pre-trained parameter optimization prediction model, which then predicts the optimization direction and adjustment range of the advertising placement parameters. Based on the prediction results, the optimized advertising delivery parameters are generated; The step of performing multi-dimensional attribution analysis based on the multivariate feedback data assigned with the credibility weights to extract the optimization factors that cause fluctuations in advertising effectiveness includes: Determine whether the total amount of feedback data collected so far has reached the effective analysis threshold; If the total amount of feedback data collected so far does not reach the effective analysis threshold, the data collection time window is extended, and the historical advertising parameters of the optimal associated scenario are weighted and calculated based on the first similarity of each known scenario to generate preliminary optimization parameters. If this has been achieved, then initiate the multidimensional attribution analysis.

3. The advertising delivery parameter adaptation method based on integrated scenario analysis according to claim 1, characterized in that, After the step of generating optimized ad delivery parameters based on the initial ad performance feedback data, the method further includes: During the advertising campaign, negative feedback signals in the advertising performance feedback data are monitored in real time. If the strength of the negative feedback signal exceeds the emergency threshold within a preset time, the use of the current advertising parameters will be immediately suspended, and the system will revert to the initial advertising parameters.

4. An advertising delivery parameter adaptation system integrating scenario analysis, characterized in that, include: The first acquisition module is used to acquire multi-dimensional features of multiple known scenarios in historical advertising data; The generation module is used to combine a preset feature embedding model to map the multi-dimensional features of the multiple known scenes into high-dimensional vectors, generate a set of known scene vectors, and construct a scene vector space. The second acquisition module is used to acquire multi-dimensional features of the new scenario in response to receiving a new ad delivery request; The mapping module is used to map the multi-dimensional features of the new scene into a high-dimensional vector based on the same feature embedding model, thereby generating a new scene vector; The determining module is configured to determine at least one optimal associated scene from the known scenes based on the new scene vector; Obtain a set of known scene vectors, and calculate the spatial distance between the new scene vector and each known scene vector in the set of known scene vectors to obtain the first similarity corresponding to each known scene; Obtain historical advertising performance data for each of the known scenarios; obtain the historical competitive environment intensity index corresponding to the time when the historical advertising performance data was generated; Obtain the current real-time competitive environment intensity index for the new scenario; Based on the difference between the historical competitive environment intensity index and the real-time competitive environment intensity index, the historical advertising performance data is adjusted by discount compensation. The ratio of the real-time competitive environment intensity index to the historical competitive environment intensity index is calculated to obtain the rate of change of competitive intensity. The rate of change of competition intensity is input into a preset nonlinear mapping function, and the mapping output is used to obtain the effect data discount factor; wherein, the nonlinear mapping function is configured such that when the rate of change of competition intensity is greater than 1, the output value is less than 1, and the output value decreases as the input value increases; The corrected historical advertising performance data is obtained by multiplying the historical advertising performance data of the known scenario by the performance data discount factor. Based on the corrected historical advertising performance data, a matching degree is calculated between the data and the current advertising delivery target in the new scenario. The matching degree is calculated between the current advertising target of the new scenario and the historical advertising performance data of each known scenario to obtain a second similarity corresponding to each known scenario; For each known scene, its first similarity and second similarity are input into a preset weighted fusion function for calculation, and the comprehensive relevance of the known scene is output based on the calculation result. Based on the overall relevance of all known scenarios, sort them and select one or more known scenarios with the highest ranking as the optimal relevance scenarios; The module is used to combine historical ad delivery parameters of the at least one optimal associated scenario to generate initial ad delivery parameters for the new scenario. The optimization module is used to obtain real-time feedback data on the performance of initial ad delivery parameters and generate optimized ad delivery parameters based on the initial ad performance feedback data.

5. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 3.

6. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 3.

Images