Game advertisement putting effect evaluation system and method combined with knowledge graph

Abstract

The invention relates to the technical field of advertisement analysis, in particular to a game advertisement putting effect evaluation system and method combined with a knowledge graph, and the method comprises the following steps: S1, capturing a non-task exploration behavior sequence of an open world game, and extracting an exploration behavior feature set comprising a scene retention hotspot, environment interaction frequency and path deviation degree; s2, injecting the exploration behavior feature set into a pre-constructed advertisement knowledge point graph to generate an enhanced knowledge graph carrying an exploration weight; and S3, generating an exploration feature compensation strategy based on the enhanced knowledge graph, and outputting an effect evaluation report fused with the implicit interest mark. According to the method, the reaching rate, the conversion rate and the rejection probability of the advertisement content to players of different exploration behavior types are quantified, the behavior interpretation of an evaluation result and the adjustability of an advertisement strategy are improved, and a behavior-putting-feedback three-dimensional integrated evaluation system is helped to be realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of advertisement analysis, in particular to a game advertisement delivery effect evaluation system and method combined with a knowledge graph. BACKGROUND

[0002] With the wide popularity of open-world games, the behavior patterns of game players show a trend of high freedom, diversity and non-task orientation. Players often stay in non-mainline task areas for a long time, repeatedly collect resources, explore around the path, etc. Although such behaviors do not directly generate task benefits, they often imply individual interest preferences and interaction tendencies. Therefore, how to achieve personalized advertisement delivery based on the free exploration behavior of players has become an important direction for the development of in-game advertising (IAA, In-Game Advertising).

[0003] Most existing game advertisement delivery solutions rely on explicit task behavior, such as task completion, character attributes or prop usage frequency, etc. for modeling. Such methods cannot accurately capture the active behavior of players in non-guided areas, and are prone to cause disconnection between advertisement content and user interest, resulting in low advertisement click rate and serious immersion destruction.

[0004] On the other hand, some studies attempt to introduce a knowledge graph to structure the management of advertisement content, but the traditional graph injection method mainly relies on static rule matching, lacks perception and fusion of real-time behavior characteristics, and leads to lag in updating the graph structure, which cannot dynamically reflect the current interest state of the player. The existing methods rarely model the exploration intention of the player from the perspective of open world, and cannot quantify the influence of complex behaviors such as path deviation and environmental interaction on the value of advertisement delivery. Most solutions only focus on advertisement exposure or click rate, and fail to form a compensation feedback mechanism for player behavior response, lacking closed-loop optimization capability. SUMMARY

[0005] The present application provides a game advertisement delivery effect evaluation system and method combined with a knowledge graph, which integrates an open-world exploration behavior modeling, dynamic graph enhancement and delivery strategy evaluation comprehensive method to realize more accurate and more immersive game advertisement effect evaluation and optimization. The present application is a targeted solution proposed to fill this technical gap.

[0006] A game advertisement delivery effect evaluation method combined with a knowledge graph, comprising the following steps: S1: capturing non-task exploration behavior sequences of an open-world game, and extracting an exploration behavior feature set including scene stagnation hotspots, environmental interaction frequency and path deviation degree; S2: injecting the exploration behavior feature set into a pre-constructed advertisement knowledge point graph to generate an enhanced knowledge graph carrying exploration weights; S3: Generate an exploration feature compensation strategy based on the enhanced knowledge graph, and output an evaluation report on the effect of fusing implicit interest tags.

[0007] Optionally, the extraction of the scene retention heat zone includes: Deploy dynamic monitoring grids in the game scene to record the time players spend in non-task areas in real time; when the cumulative time spent in the same coordinate area exceeds the dwell time threshold, it is marked as a lingering hot zone and associated with the environmental element attributes of the current area.

[0008] Optionally, the extraction of the frequency of environmental interactions includes: Listen for the interaction events between the current player and destructible objects in the scene, and extract interaction logs including the number of times the object is destroyed and the collection repetition rate; assign weighting coefficients to the interaction behaviors of objects with a rarity greater than the rarity threshold to obtain the final environmental interaction frequency.

[0009] Optionally, the calculation of the path deviation includes: Obtain the coordinate set of the player's actual movement path; Generate a shortest path reference coordinate set based on the navigation system; For each coordinate point in the player's actual movement path coordinate set, find the coordinate point in the shortest path reference coordinate set that is closest to it, calculate the Euclidean distance between the two coordinate points, and average all the minimum distances to obtain the path deviation.

[0010] Optionally, the advertising knowledge point graph includes multiple types of nodes and semantically related edges, forming a directed heterogeneous graph structure. The advertising knowledge point graph includes scene nodes, interactive object nodes, material attribute nodes, advertising nodes, geographical region nodes, and behavioral feature nodes, which are used to carry the game environment, resource type, product function, advertising content, and player exploration intent, respectively. In S2, the player's "scene lingering hot zone" is mapped to scene nodes in the graph through spatial location and environmental attribute matching, and weighted exploration edges are generated; "object destruction records" are mapped to interactive object nodes through interaction count and rarity, and further associated with material attribute nodes to construct interaction intensity edges; path deviation behavior is connected to region nodes by adding new behavior nodes to establish exponential decay edges, thereby realizing the semantic injection and graph enhancement of exploration behavior features into advertising strategies.

[0011] Optionally, S2 specifically includes: S21, Feature-Knowledge Point Mapping: The hot zones in the scene are matched with the environmental element attributes to scene nodes in the advertising knowledge point graph to generate a hot zone-scene mapping table. Based on the number of times objects are destroyed in the environmental interaction frequency, the material attributes of the interactive object nodes in the advertising knowledge point graph are associated. S22, Dynamic weighted edge generation: Create an exploration weighted edge for each entry in the hot zone-scene mapping table, and generate the edge weight of the exploration weighted edge; create an interaction strength edge for the interactive object node associated with the material attribute, and generate the edge weight of the interaction strength edge; S23, Path Exploration Node Expansion: Add a path exploration node to the advertising knowledge point graph, set the node value to path deviation, establish deviation association edges between the path exploration node and the geographic region node, and generate the edge weights of the deviation association edges.

[0012] Optionally, step S2 further includes integrating the exploration weight edges, interaction strength edges, and deviation association edges to generate a weighted enhanced knowledge graph. For nodes in the enhanced knowledge graph whose cumulative incoming edge weights exceed a judgment threshold, a topology reinforcement operation is performed on the node, and an ad delivery priority subgraph is constructed during the delivery phase.

[0013] Optionally, S3 specifically includes: S31, Compensation Strategy Generation: Traverse the exploration weight edges and interaction intensity edges in the enhanced knowledge graph that are greater than the judgment threshold, and generate immersive advertisements or path compensation advertisement packages based on the edge weights and the path deviation of the path exploration nodes. S32, Implicit Interest Tagging: Analyze disruptive behavior patterns in environmental interaction frequency, tag advertising resistance index, and generate exploration tendency labels based on path deviation direction angle variance; S33, Assessment Report Construction: Overlay the distribution of retention hot zones with the location of advertising exposure, calculate the regional overlap rate, and output the compensation effectiveness matrix.

[0014] Optionally, the exploration tendency labels include linear exploration type and divergent exploration type.

[0015] A game advertising performance evaluation system combining knowledge graphs, used to implement the aforementioned game advertising performance evaluation method, includes an exploration behavior extraction module, a knowledge graph injection module, and a strategy generation and evaluation module, wherein: The exploration behavior extraction module captures non-task exploration behavior sequences in open-world games and extracts exploration behavior feature sets such as scene loitering hot zones, environmental interaction frequency, and path deviation. The graph injection module injects the exploration behavior feature set into the pre-constructed advertising knowledge point graph to generate an enhanced knowledge graph carrying exploration weights. The strategy generation and evaluation module generates an exploration feature compensation strategy based on the enhanced knowledge graph and outputs an advertising effectiveness evaluation report that integrates implicit interest tags.

[0016] The beneficial effects of this invention are: This invention introduces a behavior-driven graph enhancement mechanism and compensation strategy generation method, realizing a deep correlation modeling between player free exploration behavior and advertising response. It solves the problems of single evaluation indicators, delayed behavior response, and lack of interest feedback loop in existing solutions. By constructing an effect evaluation report that integrates exploration weights and implicit interest tags, this invention can quantify the reach, conversion rate, and resistance probability of advertising content to players with different exploration behavior types, improve the behavioral interpretability of evaluation results and the controllability of advertising strategies, and help realize a three-dimensional integrated evaluation system of behavior-placement-feedback.

[0017] This invention constructs a feature extraction mechanism for non-task exploration behavior sequences in open worlds, comprehensively considering three key dimensions: scene dwell hotspots, environmental interaction frequency, and path deviation, to form a complete "exploration behavior feature set," thereby improving the accuracy of perceiving users' true interests. Compared to traditional user modeling methods based on task behavior, this invention can capture spontaneous behavioral patterns of players in non-guided areas, avoiding misjudgment of interests and supporting more relevant advertising content matching.

[0018] This invention injects behavioral features into a pre-built advertising knowledge graph, constructing an enhanced knowledge graph carrying behavioral semantics. By introducing a dynamic edge weight calculation model (such as dwell time weight edges, interaction intensity edges, and path deviation exponential edges), and performing node topology reinforcement and delivery subgraph construction when the weight exceeds the threshold, it solves the problems of static graph structures and lack of behavioral drive in traditional graphs. This improves the accuracy of advertising reasoning graph paths and delivery efficiency.

[0019] This invention proposes an advertising compensation strategy generation mechanism based on behavioral intensity conditions, building upon an enhanced knowledge graph. This mechanism includes various strategy forms such as immersive scene advertising and path compensation advertising packages. Simultaneously, it constructs a dual interest tagging system of advertising resistance index and exploration tendency label to quantitatively identify users' potential preferences and resistant behaviors. Based on this, a compensation effectiveness matrix is ​​output, establishing a complete closed loop from behavior perception → strategy execution → effect evaluation. This supports adaptive optimization and iterative updates of the advertising system, significantly improving its responsiveness and economic efficiency in the face of complex open-world behaviors. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only for this invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 This is a schematic diagram of the evaluation method flow according to an embodiment of the present invention; Figure 2 This is a schematic diagram of the functional modules of the evaluation system according to an embodiment of the present invention. Detailed Implementation

[0022] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments. For some well-known technologies, those skilled in the art may also use other alternative methods to implement the invention. Moreover, the accompanying drawings are only for more specific description of the embodiments and are not intended to specifically limit the present invention.

[0023] like Figure 1 As shown, a method for evaluating the effectiveness of game advertising using knowledge graphs includes the following steps: S1: Capture non-task exploration behavior sequences in open-world games and extract a set of exploration behavior features, including scene dwell hotspots, frequency of environmental interactions, and path deviation.

[0024] During game execution, main / side quest paths are marked (e.g., map markers, target areas, story trigger areas). Using these tags, all behavioral events are filtered by timestamp, excluding behavioral records within the quest area or quest navigation path, and retaining the behaviors as candidates for "non-quest exploration behavior".

[0025] S11, Scene Stay Hotspot Capture: Deploy dynamic monitoring grids in the game scene to record the time players spend in non-task areas in real time; when the cumulative stay time in the same grid cell... Exceeding the threshold At that time, the unit is marked as a heat retention zone, and the environmental element attributes of that zone are associated with it. Specifically: Divide the entire open-world game map into fixed side lengths. Square grid cells form a two-dimensional monitoring grid system. ,in These are the row and column indices, and the recommended grid side length. This scale can cover a typical local interaction space (such as rock formations, bushes, the area around a teleportation point, etc.) in most 3DRPGs; Each grid cell The preset label indicates whether it is the "main task area". If it is not a task area, the lingering detection is activated. Real-time recording of player coordinates Updated once per second, determining the current grid location. And a counter for the cumulative dwell time in that grid. Add one. That is: When the player is still in the grid And there is no mission objective; among them s is the sampling period.

[0026] When the cumulative dwell time of a certain grid cell exceeds a set threshold ,Right now Then, the unit is marked as a "retention hot zone," and its environmental attributes are extracted, such as: Terrain type (e.g., snowfield, volcano, forest); Weather (such as sandstorms, rainy days); Distribution of specific items (such as rare minerals, NPCs).

[0027] threshold Based on actual gameplay, a value of 15 minutes is recommended. Analysis of multiple open-world games shows that players spend an average of 3-7 minutes at non-quest locations, with stays exceeding 10 minutes typically indicating active exploration or emotional focus. Setting the time to 15 minutes helps avoid misjudgments caused by AFK (away from keyboard) or temporary stops (such as players drinking water or organizing their inventory).

[0028] S12 performs environmental interaction frequency statistics: During game operation, it continuously monitors player interactions with interactive environmental elements in the scene, such as destroyable rocks, mining minerals, picking up plants, and opening chests, and extracts interaction logs, including the cumulative number of item destruction interactions. and rarity indicators Each environment object has its rarity identifier predefined in the map editor or runtime asset tags. , which represents the reciprocal order of the frequency of the object in the global environment.

[0029] Rarity Identifier Example Table S13, Path Deviation Calculation: Obtain the coordinate set of the player's actual movement path. Generate a shortest path reference coordinate set based on the navigation system: ; Calculate path deviation : ; in, Indicates the player's position in the game. The actual coordinates at that moment Indicates the first Shortest path reference coordinates For Euclidean distance, This represents the total number of actual path points. For the total number of reference path points, if , then it is determined to be a path with a high tendency to explore, among which This is the path deviation threshold.

[0030] S2: Inject the exploration behavior feature set into the pre-constructed advertising knowledge graph to generate an enhanced knowledge graph carrying exploration weights.

[0031] The advertising knowledge graph is a directed graph structure composed of multiple types of nodes and semantic relationship edges, used to express the relationship between advertising content and "scene-object-function-user behavior" in the game world.

[0032] The main node types in the graph include the following: Node type 1: SceneNode, which is a scene-type node in the game that has the ability to carry advertisements, such as "volcano area", "snow mountain", and "desert altar". Node type two: ObjectNode, which is an environmental object node that can be interacted with by the player, for example: "Meteorite Crystal", "Ice Crystal Ore", "Volcanic Rock"; Node type 3: MaterialAttrNode, which is a product / item attribute node associated with ObjectNode, for example: "Mining Tools", "Fireproof Equipment", "Backpack Expander". Node Type 4: AdNode, i.e., advertising target node (display content). Examples: "Fireproof Clothing Coupon", "Mining Pickaxe Special Offer Ad"; Node type five: RegionNode, which is a geographical region node in the game world, for example: "Starlight Cave", "Dragon Ridge Snow Mountain"; Node type six: PlayerActionNode, which is an abstract node that records player characteristics and behaviors. Examples include "high exploration path" and "high interaction density". The main edge types (including directions) in the graph include the following: Edge type 1: SceneNode → AdNode, that is: scene-adapted advertising content, example: "Snow Mountain" → "Cold-resistant Equipment Ad"; Edge type two: ObjectNode → MaterialAttr, which is the functional attribute label of an object, for example: "Rare Ore" → "Mining Tools"; Edge type 3: ObjectNode → AdNode, which is: advertising content related to interactive behavior, example: "Meteorite Crystal" → "Mining Tool Ad"; Edge type four: PlayerActionNode → AdNode, i.e., behavior feature-driven recommendation ads, example: "High deviation exploration path" → "Exploration gift pack ad"; Edge type 5: RegionNode → SceneNode, that is: the geographical region contains multiple scenes, for example: "Dragon Ridge Snow Mountain" → "Ice Crystal Mine".

[0033] S2 is as follows: S21, Feature-Knowledge Point Mapping: S211, map the spatial location and environmental attributes in the "scene dwell hot zone" to the corresponding scene nodes in the advertising knowledge point graph, and establish a hot zone-scene mapping table: Mapping entries: ; The mapping method is as follows: Step 1: Spatial Location - Scene Node Mapping: Map the grid coordinates of the hotspot The region is searched in the graph and the RegionNode is matched with the set of SceneNodes it contains. The corresponding scene labels (such as "Lava Valley" and "Volcanic Inner Area") are found, and the location is mapped to the scene node. Step 2: Environmental Attribute-Scene Semantic Fusion Matching: Fuzzy match the list of hot zone environmental attributes with the preset "semantic attributes" of each scene node (such as matching keywords: "high temperature", "fire", "lava", etc.); use overlap scoring (using semantic Jaccard similarity ≥ 0.6) to confirm the final SceneNode. If there are multiple matching results, prioritize the node with high environmental overlap and long player dwell time. Step 3: Output hotspot-scene mapping entries (used for subsequent generation of exploration weight edges).

[0034] S212, the object destruction records extracted from "Environmental Interaction Frequency" are mapped to interactive object nodes in the graph according to object identifier and rarity, and associated with their material attribute nodes: Interactive events: ; in, Indicates the properties of environmental elements. Indicates the first Each scene node (a node in the advertising knowledge graph). A unique identifier representing an interactive object. Indicates the first Each object node (a node in the graph that represents a resource object).

[0035] The mapping method is as follows: Step 1: Match ObjectNodes in the graph by object_id In the interaction log Compare with interactive object nodes in the graph; if a direct match cannot be found, use object tags (such as "rare ore" or "volcanic rock") for fuzzy matching. Step 2: Rarity Filtering (only processes rarest species with a rarity greater than R) th (objects) If rarity If the condition is not met, the interaction record is ignored; otherwise, it is retained and used for subsequent weighted edge calculation. Step 3: Establish the edge between object node and material attribute node: Find the MaterialAttrNode corresponding to the ObjectNode, for example, "volcanic rock" → "heat-resistant protective gear"; "meteorite crystal" → "mining tools"; establish the functional association between object and material attributes; Step 4: Output the mapping structure for generating interaction strength edges.

[0036] S22, Dynamic Weighted Edge Generation: S221, Exploration Weighted Edge (Scene Layer): Generates an exploration edge with weight value for each hotspot-scene mapping entry. The calculation is as follows: ;in, This indicates the player's cumulative time spent in this hot zone. The environmental relevance coefficient (reflecting the semantic overlap between the scene environment and the advertising theme) indicates the degree of matching between environmental attributes and the advertising usage scenario. ,but ,otherwise , This represents the baseline dwell time for this scene type, dynamically set according to regional attributes, for example: Dangerous area: ; Resource Zone: .

[0037] S222, Interaction Strength Edge (Object Layer): For each record of object destruction interactions, create an interaction strength edge with a weight value defined as follows: ;in, Used as a rarity identifier (1~5) for interactive objects. This represents the number of times the player has destroyed the object during the current session.

[0038] In open-world games, players frequently interact with various environmental objects, such as chopping trees, mining, and opening chests. However, the value of these actions varies greatly. For example, breaking a common rock (found everywhere) is usually a casual action on a quest path; repeatedly collecting rare ores (e.g., only 1-2 appearing on the map) reflects a clear exploration intention and preference. Therefore, it is necessary to incorporate the "rarity" of interactive objects into the edge weights of interaction intensity.

[0039] Rarity serves as a proxy variable for the intensity of interest. In games, each object typically has a "rarity" tag, reflecting its distribution frequency or output value on the map. For example, rare ores, special plants, or limited-time items usually have low spawn rates and may even be environmentally exclusive. Players' repeated interactions with these high-rarity objects indicate an implicit preference or long-term goal for a particular resource, possessing significant advertising potential. The more interactions, the higher the player's level of attention to that type of object. This frequency of behavior measures the "persistence" of the behavior. Therefore, combining rarity (reflecting scarcity) with the number of destructions (reflecting the intensity of intent) creates a more reasonable measure of exploration value.

[0040] S23, Path Exploration Node Expansion: This represents the path deviation in the current exploration behavior. Add a new "Path Exploration Node" with its value set to the path deviation of this behavior segment: ;in, Nodes representing path exploration are used to indicate path deviations exhibited by the player during a certain action. This represents the numerical attribute of the node; Construct deviation-related edges from path exploration nodes to geographic region nodes, with weights calculated as follows: ;in, This represents the deviation from the current path. This indicates the historical average deviation within the region. This is the regional exploration coefficient, with a range of values. This is used to control the sensitivity to changes in deviation.

[0041] S24, graph enhancement output, integrates the three types of edges generated above: Explore weighted edges ; Interaction strength edge ; Deviation from associated edge ; Construct an enhanced advertising knowledge graph with exploration weights. If the cumulative incoming edge weight of a node in the graph is greater than a judgment threshold... The system then performs topology enhancement operations on that node, constructing an ad-priority subgraph during the ad delivery phase. Threshold determination. In enhanced knowledge graphs, a threshold is used to filter high-weight nodes. When the sum of the cumulative incoming edge weights of an advertising node (or a strongly related intermediate node) exceeds this threshold, it is considered to have high relevance and reach value in the current exploration behavior. Such nodes will be included in the priority subgraph for ad placement, used for subsequent ad strategy deployment and display resource allocation (e.g., priority pop-up, priority loading, dedicated push, etc.). Simultaneously, it also serves as a trigger condition for "topology reinforcement," performing local structural enhancement on the node (e.g., aggregating upstream and downstream nodes, strengthening the reasoning path, etc.). Value range. The recommended default value is 20. Statistical analysis across multiple game scenarios reveals that the incoming edge weights of most ad nodes cluster around 5. In the 18-interval range, setting the judgment threshold to 20 can effectively filter out the top 10-15% of highly relevant nodes for targeted advertising resource allocation.

[0042] Therefore, the enhanced knowledge graph is constructed by integrating the player's exploration behavior characteristics on the basis of the pre-built advertising knowledge graph, and includes six types of core nodes and three types of enhanced weighted edges.

[0043] The node types include: SceneNode: Represents the game scene area where the player is staying; ObjectNode: Represents an object that is interacted with by the player (such as destroyed or collected); MaterialAttrNode: Represents the function or product attribute corresponding to the interactive object; AdNode: Represents the actual node where the ad content is displayed; RegionNode: Represents geographic region information; PathExploreNode: Represents the node where the player deviates from the path, with the node value set to the path deviation degree D.

[0044] Three types of behavioral feature edges are injected into this graph: Explore weighted edges: connecting SceneNode and AdNode, representing the intensity of a player's interest in an ad during their stay in a certain hot zone, and assigning edge weights; Interaction Intensity Edge: Connects ObjectNode and AdNode, representing the degree to which the frequency of player interaction with rare objects reinforces the advertising content, and is assigned an edge weight; Deviation-related edge: Connects PathExploreNode and RegionNode, representing the relationship between the degree of deviation of the player's exploration path and the value of regional exploration, and is assigned an edge weight.

[0045] S3: Generate an exploration feature compensation strategy based on the enhanced knowledge graph, and output an evaluation report on the effect of fusing implicit interest tags.

[0046] S31, Compensation Strategy Generation: For edge weights in the enhanced knowledge graph that are greater than the judgment threshold... The strategy is determined based on the exploration behavior edges (focusing on the exploration path with the highest behavioral value to avoid wasting computational resources on low-weight noisy edges): S311, Scene-based immersive advertising trigger: Traverse and explore weighted edges and interaction intensity edges, if the following triggering conditions are met: Then, immersive advertising compensation will be implemented, specifically including: Dynamically project advertising materials onto the surface of environmental elements corresponding to the hot zone; The advertising display time is calculated as follows: ; in, To explore the edge weights of weighted edges, This represents the deviation of the current path segment. The average deviation of the region. The threshold for triggering projection ads, Based on the display duration, The weight values ​​for trigger edges include exploration weight edges and interaction strength edges.

[0047] S312, Path Compensation Ad Pack Triggered: When the deviation of the path exploration node meets the following conditions: Then, a path-compensated ad package is generated and distributed according to distance gradients, with an ad density of: ;in, This represents the current path deviation. The average deviation of the region. The maximum known deviation for the region. Maximum ad density per unit path This represents the ad density for the current path segment.

[0048] Set the deviation to be greater than twice the average deviation of the region The trigger condition for path compensation ads is primarily to exclude occasional, minor path deviations, ensuring that ads are only displayed when players exhibit a clear tendency to actively explore. In open-world games, slight deviations from the recommended path are common, such as avoiding obstacles or taking brief detours; these deviations do not represent genuine exploration intent. Displaying ads for such behaviors would be disruptive and wasteful of resources. Setting the trigger threshold to twice the average deviation clearly distinguishes between "normal path disturbances" and exploratory behaviors such as "deliberate detours for gathering and exploration," allowing the system to focus on higher-value target users.

[0049] S32, Generation of latent interest tags: S321, Ad Resistance Index Calculation: The ad resistance index is defined based on the player's repeated destructive behavior during environmental interactions. ;in, Indicates the total number of acts of vandalism. As the behavioral resistance decay factor, , The higher the value, the more likely the player is to resist the advertising content. It's a monotonically increasing, exponentially decaying function used to simulate the degree of psychological resistance players may develop towards advertising content when they repeatedly engage in destructive behavior in the game. In the game, players repeatedly destroy objects related to certain advertising materials, such as resource gathering items or placed advertising decorations, not necessarily for resource acquisition, but as an expression of rejection, ignoring, or destruction. The more interactions, the more likely it represents their dissatisfaction or fatigue with this type of advertising. Therefore, the formula is set so that the resistance index gradually increases with the number of acts of destruction. (Exponential function) It has the characteristic of "rapid growth at first, followed by saturation", which is very suitable for simulating the process of player emotional change: initially, a small amount of destruction may indicate resistance, and as the number of destructions continues to increase, the resistance tends to stabilize and no longer rises linearly.

[0050] S322, Exploration Tendency Label Determination: Variance of Path Deviation Direction Angle Sequence Determine exploration style: like Marked as: "Linear Exploratory"; like Marked as: "Divergent Exploration Type".

[0051] in, This represents the statistical variance of the path direction angle sequence. This indicates the preset style classification threshold.

[0052] First, the player's movement path is obtained, consisting of a series of consecutive coordinate points, such as P1, P2, P3, etc. Each pair of adjacent points determines a movement direction. For each path segment, its direction angle is calculated, which is the turning angle of that segment relative to due east. This direction angle is obtained by calculating the angle between two adjacent points in a plane coordinate system. All direction angles are collected into a sequence and converted to angle units (range between 0 and 360 degrees). The average of these direction angles is calculated by summing all the direction angles and dividing by the number of segments. Then, the difference between each direction angle and the average is squared, and all squared differences are averaged. The final result is the variance of the path deviation from the direction angle sequence, representing the degree of fluctuation in the player's movement direction. A smaller value indicates a straighter path and less directional change, while a larger value indicates more frequent directional changes and a more divergent path. The value is 30°. The value is 100°.

[0053] S33, Assessment Report Construction: S331, Spatiotemporal Coupled Heatmap: Overlaying the distribution of lingering heat zones with the location of advertising displays generates a heatmap, and the region overlap rate is calculated. ;in, This represents the intersection area of ​​the actual advertising display area and the hot zone. This represents the total coverage area of ​​all hot zones.

[0054] S332, Compensation Performance Matrix Output: Generate a structured compensation strategy effect report: ; in, To compensate for the number of times the strategy is triggered, For clicks on contextual ads. For scene ad exposure, Number of click paths The total number of ad paths is represented, and the resulting conversion rate reflects the ad's response performance under compensation measures.

[0055] After implementing the ad compensation strategy, the performance of each strategy is recorded. Each row of this matrix represents a compensation strategy ("immersive scene advertising" or "path compensation package"), and each column represents the key metrics for that strategy, including: Compensation type: The name of the strategy, indicating the specific advertising method used; Trigger count ( or ): This indicates how many times the policy was triggered during the entire system operation; Implicit Interest Conversion Rate: For "immersive scenario advertising," the conversion rate is calculated by dividing the number of clicks on the scenario ad by the number of impressions. This indicates how many clicks were made per ad display; for the "path compensation package," the conversion rate is calculated by dividing the number of click paths by the total number of ad paths. The value indicates the percentage of ads placed along the path that are actually reached. This matrix clearly reflects the response effectiveness of two advertising strategies triggered by user behavior, thus providing a quantitative basis for subsequent strategy optimization. The higher the value, the more effective the strategy.

[0056] like Figure 2 As shown, the game advertising performance evaluation system, used to implement the above methods, includes an exploration behavior extraction module, a graph injection module, and a strategy generation and evaluation module, wherein: The exploration behavior extraction module captures non-task exploration behavior sequences in open-world games and extracts exploration behavior feature sets such as scene loitering hot zones, environmental interaction frequency, and path deviation. The graph injection module injects the exploration behavior feature set into the pre-constructed advertising knowledge point graph to generate an enhanced knowledge graph carrying exploration weights. The strategy generation and evaluation module generates an exploration feature compensation strategy based on the enhanced knowledge graph and outputs an advertising effectiveness evaluation report that integrates implicit interest tags.

[0057] This invention encompasses any substitutions, modifications, equivalent methods, and solutions made within the spirit and scope of this invention. To provide the public with a thorough understanding of this invention, specific details are described in detail in the following preferred embodiments; however, those skilled in the art will fully understand the invention even without these details. Furthermore, to avoid unnecessary misunderstanding of the essence of this invention, well-known methods, processes, procedures, components, and circuits are not described in detail.

[0058] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for evaluating the effectiveness of game advertising combined with knowledge graphs, characterized in that, Includes the following steps: S1: Capture non-task exploration behavior sequences in open-world games and extract a set of exploration behavior features, including scene loitering hotspots, frequency of environmental interactions, and path deviation. S2: Inject the exploration behavior feature set into the pre-constructed advertising knowledge graph to generate an enhanced knowledge graph carrying exploration weights; S3: Generate an exploration feature compensation strategy based on the enhanced knowledge graph, and output an evaluation report on the effect of fusing implicit interest tags.

2. The method for evaluating the effectiveness of game advertising combined with knowledge graphs according to claim 1, characterized in that, The extraction of the scene's persistent hot zone includes: Deploy dynamic monitoring grids in the game scene to record the time players spend in non-task areas in real time; when the cumulative time spent in the same coordinate area exceeds the dwell time threshold, it is marked as a lingering hot zone and associated with the environmental element attributes of the current area.

3. The method for evaluating the effectiveness of game advertising combined with knowledge graphs according to claim 1, characterized in that, The extraction of the frequency of environmental interactions includes: Listen for the interaction events between the current player and destructible objects in the scene, and extract interaction logs including the number of times the object is destroyed and the collection repetition rate; assign weighting coefficients to the interaction behaviors of objects with a rarity greater than the rarity threshold to obtain the final environmental interaction frequency.

4. The method for evaluating the effectiveness of game advertising combined with knowledge graphs according to claim 1, characterized in that, The calculation of the path deviation includes: Obtain the coordinate set of the player's actual movement path; Generate a shortest path reference coordinate set based on the navigation system; For each coordinate point in the player's actual movement path coordinate set, find the coordinate point in the shortest path reference coordinate set that is closest to it, calculate the Euclidean distance between the two coordinate points, and average all the minimum distances to obtain the path deviation.

5. The method for evaluating the effectiveness of game advertising combined with knowledge graphs according to claim 1, characterized in that, The advertising knowledge point graph includes multiple types of nodes and semantically related edges, forming a directed heterogeneous graph structure. The advertising knowledge point graph includes scene nodes, interactive object nodes, material attribute nodes, advertising nodes, geographical region nodes, and behavioral feature nodes, which are used to carry game environment, resource type, product function, advertising content, and player exploration intent, respectively.

6. The method for evaluating the effectiveness of game advertising combined with knowledge graphs according to claim 5, characterized in that, S2 specifically includes: S21, Feature-Knowledge Point Mapping: The hot zones in the scene are matched with the environmental element attributes to scene nodes in the advertising knowledge point graph to generate a hot zone-scene mapping table. Based on the number of times objects are destroyed in the environmental interaction frequency, the material attributes of the interactive object nodes in the advertising knowledge point graph are associated. S22, Dynamic weighted edge generation: Create an exploration weighted edge for each entry in the hot zone-scene mapping table, and generate the edge weight of the exploration weighted edge; create an interaction strength edge for the interactive object node associated with the material attribute, and generate the edge weight of the interaction strength edge; S23, Path Exploration Node Expansion: Add a path exploration node to the advertising knowledge point graph, set the node value to path deviation, establish deviation association edges between the path exploration node and the geographic region node, and generate the edge weights of the deviation association edges.

7. The method for evaluating the effectiveness of game advertising combined with knowledge graphs according to claim 6, characterized in that, The S2 further includes integrating the exploration weight edges, interaction strength edges, and deviation association edges to generate a weighted enhanced knowledge graph. For nodes in the enhanced knowledge graph whose cumulative inbound edge weight is greater than the judgment threshold, a topology reinforcement operation is performed on the node, and an ad placement priority subgraph is constructed during the placement phase.

8. The method for evaluating the effectiveness of game advertising combined with knowledge graphs according to claim 7, characterized in that, S3 specifically includes: S31, Compensation Strategy Generation: Traverse the exploration weight edges and interaction intensity edges in the enhanced knowledge graph that are greater than the judgment threshold, and generate immersive advertisements or path compensation advertisement packages based on the edge weights and the path deviation of the path exploration nodes. S32, Implicit Interest Tagging: Analyze disruptive behavior patterns in environmental interaction frequency, tag advertising resistance index, and generate exploration tendency labels based on path deviation direction angle variance; S33, Assessment Report Construction: Overlay the distribution of retention hot zones with the location of advertising exposure, calculate the regional overlap rate, and output the compensation effectiveness matrix.

9. The method for evaluating the effectiveness of game advertising combined with knowledge graphs according to claim 8, characterized in that, The exploration tendency labels include linear exploration type and divergent exploration type.

10. A game advertising effectiveness evaluation system combining knowledge graphs, used to implement the game advertising effectiveness evaluation method combining knowledge graphs as described in any one of claims 1-9, characterized in that, It includes an exploration behavior extraction module, a graph injection module, and a strategy generation and evaluation module, among which: The exploration behavior extraction module captures non-task exploration behavior sequences in open-world games and extracts exploration behavior feature sets such as scene loitering hot zones, environmental interaction frequency, and path deviation. The graph injection module injects the exploration behavior feature set into the pre-constructed advertising knowledge point graph to generate an enhanced knowledge graph carrying exploration weights. The strategy generation and evaluation module generates an exploration feature compensation strategy based on the enhanced knowledge graph and outputs an advertising effectiveness evaluation report that integrates implicit interest tags.

Images