Promoting data response health status monitoring method, device and server

Abstract

The present specification provides a health state monitoring method, device and server for promotion data response, which can be used in the field of Internet technology. First, a first time sequence about the filling rate parameter of the promotion data is constructed according to the filling rate parameter of the promotion data at a current time point and the filling rate parameter of the promotion data at a current time period; then, a second time sequence about the fluctuation parameter of the filling rate parameter of the promotion data is constructed according to the first time sequence; and finally, a target dynamic threshold interval at the current time point is determined according to the first time sequence and the second time sequence based on a preset processing rule; and the health state of the promotion data response at the current time point is determined according to the target dynamic threshold interval at the current time point and the filling rate parameter of the promotion data at the current time point. Thus, the normal fluctuation and the real anomaly of the filling rate parameter of the promotion data on the network platform can be effectively distinguished, and the health state of the promotion data response at each time point can be accurately determined.

Description

Technical Field

[0001] This manual belongs to the field of Internet technology, and in particular relates to methods, devices and servers for health status monitoring based on promotional data responses. Background Technology

[0002] To promote their products or services, businesses often choose to place promotional data, such as advertisements, on online platforms. The online platform, based on the agreement with the business, then uses its backend system and a corresponding delivery mechanism to push this promotional data to users. Specifically, for example, when the backend system detects that a user has triggered a promotional request for certain promotional data on the online platform, it will respond to the request and push and display the corresponding promotional data to that user.

[0003] However, large online platforms need to respond to and process massive amounts of promotional requests every day; moreover, they also need to deal with various complex business scenarios, which makes it difficult to accurately distinguish between normal fluctuations and real anomalies in promotional data responses based on existing methods. This can easily lead to errors when monitoring the health status of promotional data responses, thereby affecting the effectiveness of promotional data delivery.

[0004] There is currently no effective solution to the above problems. Summary of the Invention

[0005] This specification provides a method, device, and server for monitoring the health status of promotional data response, which can effectively distinguish between normal fluctuations and real anomalies in the fill rate parameter of promotional data on the network platform, and accurately determine the health status of promotional data response at each time point.

[0006] This manual provides a method for monitoring health status by promoting data response, including:

[0007] Get the fill rate parameter of the promotion data at the current time point, and the fill rate parameter of the promotion data for the current time period;

[0008] Based on the fill rate parameter of the promotion data at the current time point and the fill rate parameter of the promotion data in the current time period, a first time series of the fill rate parameter of the promotion data is constructed.

[0009] Based on the first time series, a second time series of fluctuation parameters of the fill rate parameter of the promotion data is constructed;

[0010] According to the preset processing rules, the target dynamic threshold range at the current time point is determined using the first time series and the second time series.

[0011] The health status of the promotion data response at the current time point is determined based on the target dynamic threshold range at the current time point and the fill rate parameter of the promotion data at the current time point.

[0012] In one embodiment, constructing a second time series of fluctuation parameters for the fill rate parameter of the promotion data based on the first time series includes:

[0013] Based on the first time series, multiple fluctuation parameters are obtained by calculating the difference between the fill rate parameters of two adjacent time points;

[0014] Arrange the multiple fluctuation parameters in chronological order to obtain a second time series of fluctuation parameters related to the fill rate parameter of the promotion data.

[0015] In one embodiment, determining the target dynamic threshold range for the current time point using the first time series and the second time series according to preset processing rules includes:

[0016] Based on the first time series, the first stable trend center of the fill rate parameter of the promotion data for the current time point is determined;

[0017] Based on the second time series, the second stable trend center of the fluctuation parameter of the fill rate parameter of the promotion data for the current time point is determined;

[0018] The target dynamic threshold range for the current time point is determined by determining the interval center based on the first stable trend center and the interval boundary based on the second stable trend center.

[0019] In one embodiment, determining the health status of the promotion data response at the current time point based on the target dynamic threshold range at the current time point and the fill rate parameter of the promotion data at the current time point includes:

[0020] Detect whether the fill rate parameter of the promotion data at the current time point exceeds the target dynamic threshold range at the current time point;

[0021] When the fill rate parameter of the promotion data at the current time point exceeds the target dynamic threshold range at the current time point, it is determined that there is an abnormal risk in the health status of the promotion data response at the current time point.

[0022] In one embodiment, if it is determined that there is an abnormal risk in the health status of the promotion data response at the current time point, the method further includes:

[0023] Query historical records to obtain the target dynamic threshold range for the current time period;

[0024] Based on the target dynamic threshold range at the current time point and the target dynamic threshold range for the current time period, construct the first stable trend center change curve and the second stable trend center change curve respectively.

[0025] The first stable trend center change curve and the second stable trend center change curve are processed using a preset anomaly cause classification model to obtain the target classification result; wherein, the preset anomaly cause classification model is a model trained based on a graph autoencoder; the graph autoencoder includes at least an encoder, and the encoder includes at least a sequentially connected GCN layer and a TCN layer;

[0026] Based on the target classification results, identify the target anomalies that cause the health status of the promotion data response at the current time point to be at risk of being abnormal.

[0027] In one embodiment, the cause of the target anomaly includes at least one of the following: backend system anomaly, change in promotion data delivery strategy, the current time point being a special time point, or change in the inventory data of promotion data.

[0028] In one embodiment, after determining the target anomaly cause that leads to an abnormal risk in the health status of the promotion data response at the current time point based on the target classification results, the method further includes:

[0029] Based on the cause of the target anomaly, query the preset processing strategy library to determine the matching target processing strategy;

[0030] Based on the target processing strategy, the target processing terminal corresponding to the target anomaly cause is determined; and target prompt information about the target anomaly cause is generated; wherein, the target prompt information carries at least the target processing strategy;

[0031] The target processing strategy is sent to the target processing terminal.

[0032] In one embodiment, obtaining the fill rate parameter of the promotion data at the current time point and the fill rate parameter of the promotion data for the current time period includes:

[0033] Get the fill rate parameter of the promotion data at the current time point, the fill rate parameter of the promotion data at the nearest reference time point, and historical records of abnormal health status;

[0034] Based on the fill rate parameter of the promotion data at the current time point and the fill rate parameter of the promotion data at the adjacent reference time points, determine the data change information of the fill rate parameter of the promotion data at the current time point;

[0035] Based on the data changes in the fill rate parameter of the promotion data at the current time point and the historical records of abnormal health status, the matching current time period is determined.

[0036] This manual also provides a health status monitoring device for promoting data response, including:

[0037] The acquisition module is used to acquire the fill rate parameters of the promotion data at the current time point, as well as the fill rate parameters of the promotion data for the current time period.

[0038] The first construction module is used to construct a first time series of the fill rate parameters of the promotion data based on the fill rate parameters of the promotion data at the current time point and the fill rate parameters of the promotion data in the current time period.

[0039] The second construction module is used to construct a second time series of fluctuation parameters of the fill rate parameter of the promotion data based on the first time series;

[0040] The first determining module is used to determine the target dynamic threshold range at the current time point based on the first time series and the second time series according to the preset processing rules.

[0041] The second determining module is used to determine the health status of the promotion data response at the current time point based on the target dynamic threshold range at the current time point and the fill rate parameter of the promotion data at the current time point.

[0042] This specification also provides a server, including a processor and a memory for storing processor-executable instructions, wherein the processor executes the instructions to implement the steps of the health status monitoring method for the promotion data response.

[0043] This specification also provides a computer-readable storage medium having computer instructions stored thereon, which, when executed by a processor, implement the steps of the health status monitoring method for the promotion data response.

[0044] This specification also provides a computer program product comprising a computer program that, when executed by a processor, implements the steps of the health status monitoring method for the promotion data response.

[0045] Based on the health status monitoring method, device, and server for promotional data response provided in this specification, a first time series of the promotional data fill rate parameters can be constructed by first acquiring and using the fill rate parameters of the promotional data at the current time point and the promotional data fill rate parameters for the current time period. Then, based on the first time series, a second time series of the fluctuation parameters of the promotional data fill rate parameters can be constructed. Next, according to preset processing rules, the target dynamic threshold range for the current time point is determined using the first and second time series. Finally, based on the target dynamic threshold range and the promotional data fill rate parameters at the current time point, the health status of the promotional data response at the current time point is determined. Thus, by introducing a dynamically configured target dynamic threshold range, normal fluctuations and genuine anomalies in the fill rate parameters of promotional data on the network platform can be effectively and accurately distinguished, accurately determining the health status of the promotional data response at each time point, reducing monitoring errors, and thereby helping to improve the effectiveness of promotional data delivery. Attached Figure Description

[0046] To more clearly illustrate the embodiments of this specification, the accompanying drawings used in the embodiments will be briefly introduced below. The drawings described below are only some embodiments recorded in this specification. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0047] Figure 1 This is a flowchart illustrating a health status monitoring method for promotional data response provided in one embodiment of this specification;

[0048] Figure 2 This is a schematic diagram of one embodiment of the health status monitoring method for promotional data response provided in the embodiments of this specification, applied in a scenario example.

[0049] Figure 3 This is a schematic diagram of one embodiment of the health status monitoring method for promotional data response provided in the embodiments of this specification, applied in a scenario example.

[0050] Figure 4 This is a schematic diagram of one embodiment of the health status monitoring method for promotional data response provided in the embodiments of this specification, applied in a scenario example.

[0051] Figure 5 This is a schematic diagram of the structural composition of a server provided in one embodiment of this specification;

[0052] Figure 6 This is a schematic diagram of the structural composition of a health status monitoring device for promoting data response provided in one embodiment of this specification. Detailed Implementation

[0053] To enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this specification, and not all embodiments. Based on the embodiments in this specification, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of this specification.

[0054] It should be noted that the information and data related to users involved in the embodiments of this specification are all information and data authorized by the user or fully authorized by the relevant parties. Furthermore, the collection, storage, use, processing, transmission, provision, disclosure, and application of the relevant data all comply with relevant laws, regulations, and standards, and necessary confidentiality measures have been taken. They do not violate public order and good morals, and corresponding operation entry points are provided for users or relevant parties to choose to authorize or refuse.

[0055] It should also be noted that in the embodiments of this specification, certain software, components, models and other existing solutions in the industry may be mentioned. These should be regarded as exemplary and are only intended to illustrate the feasibility of implementing the technical solution of this application. However, it does not mean that the applicant has used or necessarily used the solution.

[0056] See Figure 1 As shown in the embodiments of this specification, a method for monitoring the health status of promotion data response is provided, wherein the method is specifically applied to the server side. In specific implementation, the method may include the following:

[0057] S101: Obtain the fill rate parameter of the promotion data at the current time point, and the fill rate parameter of the promotion data for the current time period;

[0058] S102: Based on the fill rate parameter of the promotion data at the current time point and the fill rate parameter of the promotion data in the current time period, construct a first time series about the fill rate parameter of the promotion data;

[0059] S103: Based on the first time series, construct a second time series of fluctuation parameters of the fill rate parameter of the promotion data;

[0060] S104: Based on the preset processing rules, the target dynamic threshold range at the current time point is determined using the first time series and the second time series.

[0061] S105: Determine the health status of the promotion data response at the current time point based on the target dynamic threshold range at the current time point and the fill rate parameter of the promotion data at the current time point.

[0062] The aforementioned promotional data can be understood as business data used to promote products or services. Specifically, this promotional data can include, for example, page advertisements, promotional videos, promotional text messages, etc.

[0063] The aforementioned fill rate parameter for promotional data (e.g., denoted as r) can be understood as an indicator parameter reflecting the status of the backend system's response to promotional data. Specifically, the fill rate parameter for promotional data can be the ratio of the number of promotional data responses filled (or the number of response pushes) by the backend system per unit time to the number of promotional requests received by the backend system per unit time.

[0064] The aforementioned current time point can be any time point of interest. The aforementioned current time period is associated with the current time point and can specifically include multiple adjacent historical time points before (excluding) the current time point. For example, if the current time point is 14:00, the aforementioned current time period could include 13:59, 13:58, 13:57, 13:56, etc. It should be noted that the aforementioned current time period can correspond to different durations and contain different numbers of historical time points for different current time points. The specific method for determining the current time period will be explained in detail later.

[0065] Specifically, the backend system can collect and record the number of response pushes for promotional data at a preset time interval (e.g., 1 minute), as well as the number of promotional requests received by the backend system.

[0066] The aforementioned target dynamic threshold range can be understood as a threshold range that is dynamically set after taking into account the actual changing trend of the fill rate parameter of the current promotion data, and is used to monitor the health status of the promotion data response at the current time point.

[0067] In practice, when it is necessary to detect and judge the health status of the promotion data response at the current time point, the processing log of the promotion data in the backend system can be queried to obtain the current time point, as well as the number of response pushes and the number of received promotion requests for the promotion data at each time point in the current time period; then, based on the obtained current time point, as well as the number of response pushes and the number of received promotion requests for the promotion data at each time point in the current time period, the fill rate parameter of the promotion data at the current time point and the fill rate parameter of the promotion data in the current time period can be calculated.

[0068] In practice, the fill rate parameters of the promotional data at the current time point and the fill rate parameters of the promotional data in the current time period can be arranged in chronological order to obtain the first time series of the fill rate parameters of the promotional data. For example, the first time series can be represented as: R = {r1, r2, ..., rt}, where rt is the fill rate parameter of the promotional data at the current time point; r1, r2, ..., rt-1 are the fill rate parameters of the promotional data at each time point in the current time period before the current time point.

[0069] In practice, based on the first time series, the difference in fill rate parameters between two adjacent time points can be calculated in reverse chronological order (for example, it can be denoted as tr, or fluctuation amplitude). Then, multiple differences can be arranged in reverse chronological order to obtain the second time series. For example, the above second time series can be represented as: TR = {tr1,tr2, ..., trt}, where trt is the difference between the fill rate parameter of the promotion data at the current time point (t) and the fill rate parameter of the promotion data at the previous time point (t-1).

[0070] In practice, based on preset processing rules, by combining the first and second time series mentioned above, a target dynamic threshold range can be determined for the current time point, taking into full account the actual changing trend of the fill rate parameter of the current promotion data, which can accurately guide the judgment of whether the health status of the promotion data response at the current time point is abnormal.

[0071] In practice, based on the target dynamic threshold range and the fill rate parameter of the promotion data at the current time, if the fill rate parameter of the promotion data at the current time falls within the target dynamic threshold range, it can be determined that the health status of the promotion data response at the current time is normal, meaning that the backend system is delivering promotion data normally according to the demand side's needs at the current time. Conversely, if the fill rate parameter of the promotion data at the current time exceeds the target dynamic threshold range, it can be determined that the health status of the promotion data response at the current time has an abnormal risk, meaning that the backend system faces an abnormal risk while delivering promotion data normally according to the demand side's needs at the current time. Consequently, corresponding prompts can be generated to provide timely risk warnings.

[0072] The aforementioned demand side can be specifically understood as a business entity that needs to distribute promotional data, such as product manufacturers and service providers.

[0073] Based on the above embodiments, by first acquiring and constructing a first time series and a second time series based on the fill rate parameter of the promotion data at the current time point and the fill rate parameter of the promotion data in the current time period; then by combining the first time series and the second time series, analyzing and utilizing the actual changing trend of the fill rate parameter of the current promotion data, dynamically setting a target dynamic threshold range for the current time point, and then using the target dynamic threshold range to effectively distinguish between normal fluctuations and real anomalies of the fill rate parameter of the promotion data on the network platform, and accurately determining the health status of the promotion data response at each time point.

[0074] In some embodiments, the construction of a second time series of fluctuation parameters of the fill rate parameter of the promotion data based on the first time series may include the following:

[0075] S1: Based on the first time series, multiple fluctuation parameters are obtained by calculating the difference between the fill rate parameters of two adjacent time points;

[0076] S2: Arrange the multiple fluctuation parameters in chronological order to obtain the second time series of fluctuation parameters of the fill rate parameter of the promotion data.

[0077] Based on the above embodiments, a second time series that meets the requirements can be quickly constructed using the first time series.

[0078] In some embodiments, see Figure 2 As shown, the above-mentioned determination of the target dynamic threshold range at the current time point based on preset processing rules, using the first time series and the second time series, includes:

[0079] S2-1: Based on the first time series, determine the first stable trend center of the fill rate parameter of the promotion data for the current time point;

[0080] S2-2: Based on the second time series, determine the second stable trend center of the fluctuation parameter of the fill rate parameter of the promotion data for the current time point;

[0081] S2-3: By determining the interval center based on the first stable trend center and the interval boundary based on the second stable trend center, the target dynamic threshold interval at the current time point is determined.

[0082] In specific implementation, the first stable trend center of the fill rate parameter of the promotion data for the current time point, determined based on the first time series, may include the following:

[0083] S1: Based on the data from other time points in the first time series, excluding the fill rate parameter of the promotion data at the current time point, calculate the reference stable trend center at the current time point (e.g., );

[0084] S2: Based on the fill rate parameter of the promotion data at the current time point and data from other time points, determine the smoothing factor for the current time point through data statistics (e.g., );

[0085] S3: Based on the fill rate parameter of the promotion data at the current time point, the reference stable trend center at the current time point, and the smoothing factor at the current time point, determine the first stable trend center of the fill rate parameter of the promotion data at the current time point (e.g., ).

[0086] Specifically, the first stable trend center can be determined using the following formula:

[0087]

[0088] in, The fill rate parameter for the promotion data at the current time point (which can also be denoted as rt) This represents the first stable trend center of the fill rate parameter for the promotional data at the current time point. As a reference stable trend center at the current time point, This is a smoothing factor.

[0089] When determining the specific smoothing factor, first determine the current time period and the total number of time points included in the current time point (e.g., N); construct an initial smoothing factor based on the total number of time points, for example, 2 / (N+1); simultaneously, calculate the mean of the fill rate parameter based on the fill rate parameter of the promotion data at the current time point and the data at other time points; calculate the individual closeness between the fill rate parameter of the promotion data at the current time point and the mean of the fill rate parameter, and the overall closeness between the data at other time points and the mean of the fill rate parameter; adjust the initial smoothing factor based on the individual closeness and the overall closeness to obtain a smoothing factor that meets the requirements.

[0090] Following the method described above for determining the first stable trend center, the second stable trend center of the fluctuation parameter of the fill rate parameter of the promotion data at the current time point can be determined based on the second time series described above. For example... For details, please refer to the above-described embodiments regarding the determination of the first stable trend center; these will not be repeated here.

[0091] For specific implementation, please refer to Figure 3As shown, the first stable trend center (or EMA) can be used as the center of the threshold interval. Then, based on the second and first stable trend centers, the boundaries of the threshold interval (including the upper and lower boundaries, or upper and lower rails) can be determined. For example, the difference between the first and second stable trend centers can be used as the lower boundary; simultaneously, the sum of the first and second stable trend centers can be used as the upper boundary. This allows the construction of the target dynamic threshold interval for the current time point.

[0092] In practice, the fluctuation range and speed of the data fluctuations in the first time series and the second time series can be combined with the sensitivity requirements to determine a matching fluctuation coefficient. Then, based on the fluctuation coefficient, the second stable trend center used can be adjusted to construct a more accurate target dynamic threshold range for the current time point.

[0093] Based on the above embodiments, the first time series and the second time series can be fully utilized to accurately construct the target dynamic threshold interval that is effective for the current time point.

[0094] In some embodiments, determining the health status of the promotion data response at the current time point based on the target dynamic threshold range at the current time point and the fill rate parameter of the promotion data at the current time point may specifically include the following:

[0095] S1: Detect whether the fill rate parameter of the promotion data at the current time point exceeds the target dynamic threshold range at the current time point;

[0096] S2: When the fill rate parameter of the promotion data at the current time point exceeds the target dynamic threshold range at the current time point, it is determined that there is an abnormal risk in the health status of the promotion data response at the current time point.

[0097] Conversely, when the fill rate parameter of the promotion data at the current time point does not exceed the target dynamic threshold range at the current time point, the health status of the promotion data response at the current time point is determined to be normal.

[0098] It should be noted that when the fill rate parameter of the promotion data at the current time point exceeds the target dynamic threshold range at the current time point, it only indicates that the health status of the promotion data response at the current time point is at risk of being abnormal, and does not necessarily mean that there is a real abnormality. For example, sometimes some normal operations (such as the demand side modifying the delivery strategy) can also cause the fill rate parameter to exceed the target dynamic threshold range for a short period of time during the promotion data response processing. Therefore, further analysis of the cause of the abnormality is needed to distinguish whether it is a real abnormality.

[0099] Based on the above embodiments, the target dynamic threshold range at the current time point and the fill rate parameter of the promotion data at the current time point can be used to accurately identify and discover abnormal risks in the health status of the promotion data response at the current time point.

[0100] In some embodiments, if it is determined that there is an abnormal risk in the health status of the promotion data response at the current point in time, see [reference needed]. Figure 4 As shown, in specific implementations, the method may also include the following:

[0101] S4-1: Query historical records to obtain the target dynamic threshold range for the current time period;

[0102] S4-2: Based on the target dynamic threshold range at the current time point and the target dynamic threshold range for the current time period, construct the first stable trend center change curve and the second stable trend center change curve respectively;

[0103] S4-3: Process the first stable trend center change curve and the second stable trend center change curve using a preset anomaly cause classification model to obtain the target classification result; wherein, the preset anomaly cause classification model is a model trained based on a graph autoencoder; the graph autoencoder includes at least an encoder, and the encoder includes at least a sequentially connected GCN layer and a TCN layer;

[0104] S4-4: Based on the target classification results, identify the target anomaly causes that lead to an abnormal risk in the health status of the promotion data response at the current time point.

[0105] The graph autoencoder (GAE) mentioned above is an extension of the autoencoder for graph-structured data, specifically designed to handle graph topologies composed of nodes and edges. Unlike traditional autoencoders that process vector or grid data (such as images), the encoder of a graph autoencoder typically employs architectures such as graph convolutional networks (GCNs), and uses the graph's adjacency matrix (representing the topological structure) and node features for joint encoding to learn the embedded representations of nodes; the decoder then reconstructs the adjacency matrix through inner product operations to achieve graph structure reconstruction, enabling better collaboration for predictive analysis.

[0106] Furthermore, the graph autoencoder in the aforementioned pre-defined anomaly cause classification model includes at least an encoder, and the encoder includes at least a sequentially connected GCN layer and a TCN layer.

[0107] In practical implementation, the pre-defined anomaly cause classification model is based on a GCN (Graph Convolutional Neural Network) layer. It can introduce data constraints between the first and second stable trend centers at the same time point through a graph structure, and embed these data constraints using graph convolution. When the model processes relevant feature parameters for cause classification, it can be associated with these data constraints to effectively couple the correlation between the changing features of the two different stable trend centers for classification prediction. Simultaneously, based on a TCN (Temporal Convolutional Neural Network) layer, corresponding temporal information can be extracted through causal dilation convolution stacking to extract the temporal dependencies of the first and second stable trend centers based on time changes. Furthermore, the model can integrate the correlation between the changing features of the two different stable trend centers, the temporal dependencies of the first and second stable trend centers based on time changes, and perform deep-level classification prediction to accurately determine and output the corresponding anomaly cause classification results.

[0108] Among them, the aforementioned historical records at least contain the target dynamic threshold range corresponding to each time point used when monitoring the health status of the promotion data response at each historical time point.

[0109] In practice, historical records can be queried to obtain the target dynamic threshold range for each time point within the current time period. Then, based on the target dynamic threshold ranges for each time point within the current time period and the target dynamic threshold range for the current time point, the first stable trend center and the second stable trend center for multiple time points are calculated. Based on the first stable trend centers for multiple time points, a change curve for the first stable trend center is constructed through data fitting. Simultaneously, a change curve for the second stable trend center for multiple time points is constructed through data fitting. The change curves for the first and second stable trend centers are then projected onto a single coordinate system to obtain a joint curve graph. Finally, a pre-defined anomaly cause classification model is used to process the joint curve graph, analyze and utilize the corresponding image feature information, and perform classification prediction to obtain the corresponding target classification result.

[0110] The above target classification results may include probability values ​​for multiple abnormal causes.

[0111] Correspondingly, based on the target classification results, the anomaly cause with the highest probability value can be identified as the target anomaly cause that leads to an abnormal risk in the health status of the promotion data response at the current time point.

[0112] Based on the above embodiments, when it is determined that there is an abnormal risk in the health status of the promotion data response at the current time point, by introducing and utilizing a preset abnormal cause classification model that includes at least an autoencoder, the target abnormal cause causing the abnormal risk can be accurately and efficiently automatically analyzed and determined by processing the first stable trend center change curve and the second stable trend center change curve. This allows for subsequent targeted processing based on the target abnormal cause, different scenarios can be distinguished.

[0113] In some embodiments, the TCN layer described above may specifically include multiple dilated convolutional modules; each dilated convolutional module corresponds to at least one time scale. Specifically, each dilated convolutional module may include multiple dilated convolutional layers, and different dilated convolutional layers within the same dilated convolutional module are configured with different dilation rates.

[0114] The number of dilated convolutional layers in each dilated convolutional module, and the dilation rate of each layer, can be determined based on the length of the corresponding time scale and the temporal receptive field requirements of the dilated convolutional network. With a fixed kernel size, by employing an exponentially increasing dilation rate, the multi-layer dilated convolutional network gradually expands its temporal receptive field with a relatively small number of layers, thereby covering the data changes of the first and second stable trend centers within the corresponding time scale.

[0115] In this way, the TCN layer can be used to extract the temporal dependencies of multiple first stable trend centers based on time changes and multiple second stable trend centers based on time changes through multiple different dilated convolution modules, so as to integrate more time scales and perform more comprehensive analysis and inference of relevant image feature information.

[0116] Furthermore, the aforementioned pre-defined anomaly cause classification model may also include: a fully connected layer, used to comprehensively utilize the correlation between the change features of two different stable trend centers, the time-varying temporal dependency of the first stable trend center, and the time-varying temporal dependency of the second stable trend center to perform anomaly cause classification prediction.

[0117] In addition, the aforementioned pre-defined anomaly cause classification model may also include: a gated network structure based on a self-attention mechanism, consisting of fully connected layers, which can be used to analyze and dynamically weight the different feature information output by dilated convolutional modules at different time scales according to the overall trend of change.

[0118] In some embodiments, before implementation, a preset anomaly cause classification model can be trained as follows: Experimental test data is obtained through simulated testing; based on the experimental test data, the first stable trend center, the second stable trend center, and experimental labels at each time point are obtained and combined; wherein, the experimental labels are used to indicate the existence of anomaly risk and the anomaly cause leading to that risk; the first stable trend center and the second stable trend center at the same time point are combined to obtain multiple sample data; the sample data are labeled according to whether experimental labels exist at the time points corresponding to the sample data, and when experimental labels exist, the indication information of the experimental labels, to obtain labeled sample data; an initial anomaly cause classification model is constructed; wherein, the initial anomaly cause classification model includes at least an autoencoder, the autoencoder including at least an encoder containing sequentially connected GCN layers and TCN layers; the initial anomaly cause classification model is trained using the labeled sample data to obtain a preset anomaly cause classification model that meets the requirements.

[0119] In some embodiments, the specific reasons for the target anomaly may include at least one of the following: backend system anomaly, change in promotion data delivery strategy, the current time point being a special time point, or changes in the inventory data of promotion data.

[0120] The aforementioned backend system anomaly can be understood as a failure to respond to promotion requests and push promotion data due to bugs or errors in the system's own backend, or due to external attacks. This is a genuine anomaly.

[0121] The aforementioned change in the promotional data delivery strategy can be understood as the demand side adjusting its promotional data delivery strategy due to changes in business needs, resulting in significant fluctuations in the promotional data response within a short period of time, which is unlikely to be a genuine anomaly.

[0122] The aforementioned special time points can specifically refer to holidays or special offer days, etc. The current time point being considered a special time point specifically means that due to a sudden surge in user access to the network platform at that particular time, the limited computing resources of the backend system cause significant fluctuations in promotional data response within a short period, which is unlikely to be a genuine anomaly.

[0123] The aforementioned changes in the inventory data of promotional data can be understood as changes in the inventory data of promotional data provided by the demand side that can be used for campaign placement. This leads to a mismatch with the original campaign placement strategy set by the demand side, resulting in significant fluctuations in the promotional data response within a short period of time, which is unlikely to be a genuine anomaly.

[0124] Of course, it should be noted that the anomalies listed above are merely illustrative. In actual implementation, depending on the specific application scenario and processing requirements, the aforementioned anomalies may also include other related anomalies such as changes in the bidding price of promotional data. This manual does not limit the scope of such anomalies.

[0125] In practice, when the cause of the target anomaly is determined to be a backend system malfunction, it can be directly confirmed that a genuine anomaly exists, namely, the health status of the promotion data response at the current time point is abnormal. In this case, according to the corresponding handling strategy, the maintenance terminal responsible for backend system maintenance can be directly designated as the target processing terminal; and a corresponding prompt message can be sent to this target processing terminal to prompt it to promptly detect and repair the backend system, thereby improving the effectiveness of the promotion data delivery.

[0126] When the cause of a target anomaly is determined to be a change in the promotional data delivery strategy, a specific time point, or changes in the promotional data inventory, it's not possible to directly determine if a genuine anomaly exists. In this case, it's necessary to obtain relevant data related to the cause of the target anomaly, such as the current promotional data delivery strategy, current time information, or current promotional data inventory. Further investigation is then conducted based on this relevant data to obtain the investigation results. Finally, based on the investigation results, a determination is made as to whether it is a genuine anomaly.

[0127] Based on the investigation results, if it is determined that the anomaly is not a genuine anomaly, it can be ignored and no action taken. Conversely, if it is determined to be a genuine anomaly, the target processing terminal (demand-side terminal and / or operations and maintenance terminal) corresponding to the cause of the anomaly can be identified according to the appropriate handling strategy. A target prompt message regarding the cause of the anomaly can be generated, and the corresponding prompt message can be sent to the target processing terminal to remind it to check and make adjustments in a timely manner to improve the effectiveness of the promotional data delivery.

[0128] In some embodiments, after determining the target anomaly cause that leads to an abnormal risk in the health status of the promotion data response at the current time point based on the target classification results, the method may further include the following:

[0129] S1: Based on the cause of the target anomaly, query the preset processing strategy library to determine the matching target processing strategy;

[0130] S2: Based on the target processing strategy, determine the target processing terminal corresponding to the target anomaly cause; and generate target prompt information about the target anomaly cause; wherein, the target prompt information carries at least the target processing strategy;

[0131] S3: Send the target processing strategy to the target processing terminal.

[0132] The preset processing strategy library can contain multiple preset processing strategies, and each processing strategy corresponds to at least one abnormal cause.

[0133] Before implementation, a large number of historical anomaly cause handling records can be collected; from these records, successfully handled historical anomaly cause handling records can be selected as sample handling records; relevant handling operation information can be extracted from the sample handling records and combined to construct multiple sample handling feature vectors; then, multiple sample handling feature vectors can be clustered to obtain clustering results; based on the clustering results, multiple preset handling strategies can be constructed; and by combining multiple preset handling strategies, a preset handling strategy library can be obtained.

[0134] Based on the above embodiments, a matching target processing strategy can be accurately determined according to a preset processing strategy library. Then, the corresponding target prompt information and target processing strategy are sent to the target processing terminal together, so that the target processing terminal can perform timely and targeted operation processing according to the target prompt information and target processing strategy, so as to quickly eliminate anomalies, maintain the healthy state of promotion data response, and ensure the delivery effect of the demand side's promotion data.

[0135] In some embodiments, obtaining the fill rate parameter of the promotion data at the current time point and the fill rate parameter of the promotion data for the current time period may specifically include the following:

[0136] S1: Obtain the fill rate parameter of the promotion data at the current time point, the fill rate parameter of the promotion data at the neighboring reference time points at the current time point, and historical records of abnormal health status;

[0137] S2: Determine the data change information of the promotion data fill rate parameter at the current time point based on the fill rate parameter of the promotion data at the current time point and the fill rate parameter of the promotion data at the adjacent reference time points of the current time point;

[0138] S3: Based on the data changes in the fill rate parameter of the promotion data at the current time point and the historical abnormal health status records, determine the matching current time period.

[0139] Specifically, the aforementioned adjacent reference time points can be understood as a small number of representative time points further selected from multiple time points included in the current time period.

[0140] In practice, a specified number of time points can be randomly selected from multiple time points included in the current time period as neighboring reference time points.

[0141] In practice, the magnitude and rate of change of the fill rate parameter of the promotional data at the current time point can be determined based on the fill rate parameter of the promotional data at the current time point and the fill rate parameter of the promotional data at the adjacent reference time points, serving as data change information. Then, based on historical abnormal health status records, the abnormal frequency of the current time period can be determined. Finally, by combining the magnitude and rate of change of the fill rate parameter of the promotional data at the current time point with the abnormal frequency of the current time period, the number of time points to be included in the current time period can be determined, thus obtaining the matching current time period for further processing.

[0142] Specifically, when the change rate of the promotion data at the current time point is relatively gradual and the frequency of anomalies in the current time period is low, the number of time points to be included in the current time period can be adaptively reduced to reduce the amount of subsequent data processing and improve the overall processing efficiency.

[0143] Conversely, when the change in the fill rate parameter of the promotion data at the current time point is drastic and the frequency of anomalies in the current time period is high, the number of time points to be included in the current time period can be adaptively increased so that subsequent analysis can be more comprehensive and detailed, thereby improving the overall processing accuracy.

[0144] Based on the above embodiments, by dynamically adjusting and determining the matching current time period, it is possible to effectively balance the efficiency and accuracy of subsequent processing, so as to efficiently and accurately determine the health status of the promotion data response.

[0145] In some embodiments, the fill rate parameter of the promotion data at the current time point and the fill rate parameter of the promotion data for the current time period may also carry a corresponding demand side identifier.

[0146] Accordingly, after obtaining the fill rate parameters of the promotion data at the current time point and the fill rate parameters of the promotion data for the current time period, these parameters can be further divided into multiple parameter groups based on the demand side identifier. Each parameter group corresponds to a demand side. Following this method, the health status of the promotion data response of each demand side at the current time point can be precisely determined. This allows for accurate monitoring and maintenance of the promotion data delivery for each demand side, ensuring that each demand side achieves good promotion results.

[0147] In some embodiments, after obtaining the first time series and the second time series, it is also possible to detect whether there is obvious erroneous data in the first time series and the second time series based on historical abnormal health status records, and promptly remove the aforementioned erroneous data, as well as the fill rate parameter or the difference between the fill rate parameters at the same time point corresponding to the erroneous data, to avoid misleading or interfering with subsequent data processing and further improve the accuracy of data processing.

[0148] As can be seen from the above, the health status monitoring method for promotional data response provided in this specification first acquires and constructs a first time series of the promotional data fill rate parameters based on the fill rate parameters of the promotional data at the current time point and the fill rate parameters of the promotional data for the current time period. Based on the first time series, a second time series of the fluctuation parameters of the promotional data fill rate parameters is constructed. Then, according to preset processing rules, the target dynamic threshold range at the current time point is determined using the first and second time series. Finally, based on the target dynamic threshold range at the current time point and the fill rate parameters of the promotional data at the current time point, the health status of the promotional data response at the current time point is determined. This effectively distinguishes between normal fluctuations and genuine anomalies in the fill rate parameters of promotional data on the online platform, accurately determines the health status of the promotional data response at each time point, reduces monitoring errors, and thus helps improve the effectiveness of promotional data delivery.

[0149] This specification provides an embodiment of a server, see below. Figure 5 As shown. The server includes a network communication port 501, a processor 502, and a memory 503. These structures are connected by internal cables so that they can perform specific data interaction.

[0150] Specifically, the network communication port 501 can be used to obtain the fill rate parameter of the promotion data at the current time point, as well as the fill rate parameter of the promotion data for the current time period.

[0151] The processor 502 can be specifically configured to: construct a first time series of the fill rate parameters of the promotional data based on the fill rate parameters of the promotional data at the current time point and the fill rate parameters of the promotional data in the current time period; construct a second time series of the fluctuation parameters of the fill rate parameters of the promotional data based on the first time series; determine the target dynamic threshold range at the current time point using the first and second time series according to preset processing rules; and determine the health status of the promotional data response at the current time point based on the target dynamic threshold range at the current time point and the fill rate parameters of the promotional data at the current time point.

[0152] The memory 503 can be used to store the corresponding instruction program and related intermediate data.

[0153] Based on the above method, the relevant structural performance of the server can be effectively utilized to improve the data processing speed of electronic devices and efficiently realize the data processing of health status monitoring for promotion data response.

[0154] In this embodiment, the network communication port 501 can be a virtual port bound to different communication protocols, thereby enabling the sending or receiving of different data. For example, the network communication port can be a port responsible for web data communication, a port responsible for FTP data communication, or a port responsible for email data communication. Furthermore, the network communication port can also be a physical communication interface or communication chip. For example, it can be a wireless mobile network communication chip, such as GSM or CDMA; it can also be a Wi-Fi chip; or it can be a Bluetooth chip.

[0155] In this embodiment, the processor 502 can be implemented in any suitable manner. For example, the processor can take the form of a microprocessor or processor and a computer-readable medium storing computer-readable program code (e.g., software or firmware) executable by the (micro)processor, logic gates, switches, application-specific integrated circuits (ASICs), programmable logic controllers, and embedded microcontrollers, etc. This specification is not limiting.

[0156] In this embodiment, the memory 503 may include multiple layers. In a digital system, anything that can store binary data can be a memory. In an integrated circuit, a circuit with storage function but no physical form is also called a memory, such as RAM, FIFO, etc. In a system, a storage device with a physical form is also called a memory, such as a memory stick, TF card, etc.

[0157] This specification also provides a computer-readable storage medium for monitoring the health status of promotional data response based on the above-described method. The computer-readable storage medium stores computer program instructions that, when executed, implement the following: obtaining the fill rate parameter of promotional data at the current time point and the fill rate parameter of promotional data for the current time period; constructing a first time series of the fill rate parameter of the promotional data based on the fill rate parameter of the promotional data at the current time point and the fill rate parameter of the promotional data for the current time period; constructing a second time series of the fluctuation parameter of the fill rate parameter of the promotional data based on the first time series; determining a target dynamic threshold range at the current time point using the first and second time series according to preset processing rules; and determining the health status of the promotional data response at the current time point based on the target dynamic threshold range and the fill rate parameter of the promotional data at the current time point.

[0158] In this embodiment, the storage medium includes, but is not limited to, Random Access Memory (RAM), Read-Only Memory (ROM), Cache, Hard Disk Drive (HDD), or Memory Card. The memory can be used to store computer program instructions. The network communication unit can be an interface configured according to standards specified in the communication protocol for network connection communication.

[0159] In this embodiment, the specific functions and effects implemented by the program instructions stored in the computer-readable storage medium can be explained in comparison with other embodiments, and will not be repeated here.

[0160] This specification also provides a computer program product, comprising at least a computer program, which, when executed by a processor, implements the following method steps: obtaining the fill rate parameter of promotional data at the current time point and the fill rate parameter of promotional data for the current time period; constructing a first time series of the fill rate parameter of the promotional data based on the fill rate parameter of the promotional data at the current time point and the fill rate parameter of the promotional data for the current time period; constructing a second time series of the fluctuation parameter of the fill rate parameter of the promotional data based on the first time series; determining a target dynamic threshold range at the current time point using the first time series and the second time series according to a preset processing rule; and determining the health status of the promotional data response at the current time point based on the target dynamic threshold range and the fill rate parameter of the promotional data at the current time point.

[0161] See Figure 6As shown in the embodiments of this specification, a health status monitoring device for promoting data response is also provided. This device may specifically include the following structural modules:

[0162] The acquisition module 601 can be used to acquire the fill rate parameter of the promotion data at the current time point, as well as the fill rate parameter of the promotion data for the current time period.

[0163] The first construction module 602 can be specifically used to construct a first time series about the fill rate parameter of the promotion data based on the fill rate parameter of the promotion data at the current time point and the fill rate parameter of the promotion data in the current time period.

[0164] The second construction module 603 can be specifically used to construct a second time series of fluctuation parameters of the fill rate parameter of the promotion data based on the first time series;

[0165] The first determining module 604 can be used to determine the target dynamic threshold range at the current time point based on the first time series and the second time series according to the preset processing rules.

[0166] The second determining module 605 can be used to determine the health status of the promotion data response at the current time point based on the target dynamic threshold range at the current time point and the fill rate parameter of the promotion data at the current time point.

[0167] In some embodiments, when the second construction module 603 is specifically implemented, a second time series of fluctuation parameters of the fill rate parameter of the promotion data can be constructed according to the first time series in the following manner: based on the first time series, multiple fluctuation parameters are obtained by calculating the difference between the fill rate parameters of two adjacent time points; the multiple fluctuation parameters are arranged in chronological order to obtain a second time series of fluctuation parameters of the fill rate parameter of the promotion data.

[0168] In some embodiments, when the first determining module 604 is specifically implemented, it can determine the target dynamic threshold range for the current time point according to the preset processing rules using the first time series and the second time series in the following manner: based on the first time series, determine the first stable trend center of the fill rate parameter of the promotion data for the current time point; based on the second time series, determine the second stable trend center of the fluctuation parameter of the fill rate parameter of the promotion data for the current time point; by determining the interval center based on the first stable trend center and determining the interval boundary based on the second stable trend center, the target dynamic threshold range for the current time point is determined.

[0169] In some embodiments, when the second determining module 605 is specifically implemented, it can determine the health status of the promotion data response at the current time point according to the target dynamic threshold range at the current time point and the fill rate parameter of the promotion data at the current time point in the following manner: detecting whether the fill rate parameter of the promotion data at the current time point exceeds the target dynamic threshold range at the current time point; when the fill rate parameter of the promotion data at the current time point exceeds the target dynamic threshold range at the current time point, it is determined that there is an abnormal risk in the health status of the promotion data response at the current time point.

[0170] In some embodiments, when it is determined that the health status of the promotion data response at the current time point has an abnormal risk, the device may further be used to: query historical records to obtain the target dynamic threshold range for the current time period; construct a first stable trend center change curve and a second stable trend center change curve based on the target dynamic threshold range at the current time point and the target dynamic threshold range for the current time period, respectively; process the first stable trend center change curve and the second stable trend center change curve using a preset anomaly cause classification model to obtain a target classification result; wherein, the preset anomaly cause classification model is a model trained based on a graph autoencoder; the graph autoencoder includes at least an encoder, and the encoder includes at least a sequentially connected GCN layer and a TCN layer; and determine the target anomaly cause that leads to the abnormal risk of the health status of the promotion data response at the current time point based on the target classification result.

[0171] In some embodiments, the specific reasons for the target anomaly may include at least one of the following: backend system anomaly, change in promotion data delivery strategy, the current time point being a special time point, changes in the inventory data of promotion data, etc.

[0172] In some embodiments, after determining the target anomaly cause that leads to an abnormal risk in the health status of the promotion data response at the current time point based on the target classification result, the device may further be used to: query a preset processing strategy library based on the target anomaly cause to determine a matching target processing strategy; determine the target processing terminal corresponding to the target anomaly cause based on the target processing strategy; and generate target prompt information about the target anomaly cause; wherein the target prompt information carries at least the target processing strategy; and send the target processing strategy to the target processing terminal.

[0173] In some embodiments, when the acquisition module 601 is specifically implemented, it can acquire the fill rate parameter of the promotion data at the current time point and the fill rate parameter of the promotion data in the current time period in the following manner: acquire the fill rate parameter of the promotion data at the current time point, the fill rate parameter of the promotion data at the adjacent reference time point of the current time point, and historical health status abnormal records; determine the data change information of the fill rate parameter of the promotion data at the current time point based on the fill rate parameter of the promotion data at the current time point and the fill rate parameter of the promotion data at the adjacent reference time point of the current time point; determine the matching current time period based on the data change information of the fill rate parameter of the promotion data at the current time point and the historical health status abnormal records.

[0174] It should be noted that the units, devices, or modules described in the above embodiments can be implemented by computer chips or physical entities, or by products with certain functions. For ease of description, the above devices are described by dividing them into various modules according to their functions. Of course, in implementing this specification, the functions of each module can be implemented in one or more software and / or hardware, or the module that implements the same function can be implemented by a combination of multiple sub-modules or sub-units, etc. The device embodiments described above are merely illustrative. For example, the division of units is only a logical functional division, and there may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection between the devices or units shown or discussed can be through some interfaces, and the indirect coupling or communication connection between devices or units can be electrical, mechanical, or other forms.

[0175] As can be seen from the above, the health status monitoring device for promotion data response provided in the embodiments of this specification can effectively distinguish between normal fluctuations and real anomalies in the fill rate parameter of promotion data on the network platform, and accurately determine the health status of promotion data response at each time point.

[0176] In a specific scenario example, the health status monitoring method for promotional data response provided in this specification can be applied to adaptively monitor abnormal fluctuations in the fill rate of advertisements (a type of promotional data). The specific implementation process may include the following:

[0177] In this example scenario, ad fill rate is a crucial metric for measuring the health of advertising operations and the overall system flow within a programmatic advertising system (e.g., a backend system). Ad fill rate is a real-time monitoring indicator for the advertising backend. It is calculated as: Ad Fill Rate = Ads Filled Per Second / Ad Requests Received Per Second. For example, if the advertising backend receives 10,000 ad requests (e.g., promotional requests) in 1 second, and only returns ads (e.g., promotional data) to 2,000 of these requests, then the ad fill rate for that second is 20%. Concatenating the fill rates for each second yields the real-time fill rate curve for the advertising backend. Monitoring this metric reflects the health of the advertising backend system. Specifically, when the advertising backend receives an ad request, it determines whether to return an ad to that request. If the system returns an ad, it indicates that the request has been filled with ads.

[0178] Under normal circumstances, fill rate fluctuates relatively smoothly, without sudden increases or decreases. When the fill rate is too high, the advertising budget may be consumed quickly or even over-consumed, resulting in poor advertising performance. When the fill rate is too low, the budget may not be used effectively, thus failing to achieve advertising and marketing goals.

[0179] As advertising system architectures become more complex and the bidding ecosystem becomes more volatile, traditional anomaly detection methods based on fixed thresholds or year-on-year / month-on-month comparisons are gradually revealing their limitations: they cannot automatically adapt to actual business fluctuations and have problems such as high maintenance costs and high false alarm / false negative rates.

[0180] To address the aforementioned issues and their root causes, this scenario example considers establishing a monitoring system capable of adapting to dynamic fluctuations, learning upper and lower bounds, and automatically identifying anomalies in a real-time changing bidding environment. This system employs an adaptive monitoring method for indicator anomalies, automatically adjusting monitoring boundaries by learning the central trend and fluctuation amplitude of indicators in real time. This achieves high sensitivity, low false alarm rate, and adaptive monitoring of abnormal fluctuations in the fill rate. Specific implementation may include the following:

[0181] (1) Data input: The system collects time series data of ad fill rate in real time (e.g., first time series, second time series TR).

[0182] (2) Trend line calculation: Extract the stable trend of the current indicator through the exponential smoothing rule and use it as the dynamic center line (e.g., the first stable trend center EMA).

[0183] (3) Fluctuation range estimation: Calculate the fluctuation range (e.g., TR) based on the difference between adjacent time points and smooth it to obtain the natural fluctuation range of the indicator (e.g., the second stable trend center ATR).

[0184] (4) Dynamic range (e.g., target dynamic threshold range) generation: Based on the trend line and fluctuation range, adaptively generate the upper and lower boundary ranges for monitoring (e.g., upper rail and lower rail). See details in [link to documentation]. Figure 3 .

[0185] The upper and lower rails can be the upper and lower boundaries of the monitoring. The upper and lower rails can be used to define the normal fluctuation range of the fill rate. If the range is exceeded, it is judged as abnormal.

[0186] Specifically, the upper boundary (e.g., upper_band) and lower boundary (e.g., lower_band) can be calculated based on the first and second stable trend centers using the following formulas:

[0187] ema_center = EMA

[0188] upper_band = ema_center + channel_multiplier * atr

[0189] lower_band = ema_center - channel_multiplier * atr.

[0190] Where ema_center is the center of the interval, channel_multiplier is the volatility coefficient, EMA is the dynamic center line, and atr is the natural fluctuation range.

[0191] (5) Anomaly detection: When the actual ad fill rate is detected to exceed the upper and lower limits of the range, it is determined to be an anomaly and an alarm event is triggered.

[0192] The causes of anomalies can be varied, including: system anomalies (e.g., backend system anomalies), where the program introduces new bugs that cause anomalies within the backend system, potentially leading to a sudden increase or decrease in the fill rate; and changes in data, such as budget contraction or adjustments to advertising operation and placement strategies due to e-commerce promotions, which can significantly affect the ad fill rate.

[0193] In practice, the above methods can be used to monitor whether there are any abnormalities in the ad fill rate at various time points within a given period. Specific monitoring results are shown in Table 1.

[0194] Table 1

[0195]

[0196] The above scenario examples verify the health status monitoring method for promotional data response provided in this manual. On the one hand, it can adaptively adjust the monitoring threshold without the need for manual threshold configuration, thus achieving intelligent monitoring. On the other hand, it can also automatically learn the fluctuation amplitude, distinguish between natural fluctuations and abnormal fluctuations, identify real anomalies, achieve dynamic noise reduction, and improve monitoring accuracy.

[0197] While this specification provides the steps of operation for the methods described in the embodiments or flowcharts, more or fewer steps may be included based on conventional or non-inventive means. The order of steps listed in the embodiments is merely one possible order of execution among many steps and does not represent the only possible order. In actual device or client product execution, the methods shown in the embodiments or drawings may be executed sequentially or in parallel (e.g., in a parallel processor or multi-threaded processing environment, or even a distributed data processing environment). The terms "comprising," "including," or any other variations thereof are intended to cover a non-exclusive inclusion, such that a process, method, product, or apparatus 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, method, product, or apparatus. Without further limitations, the presence of other identical or equivalent elements in a process, method, product, or apparatus that includes said elements is not excluded. The terms "first," "second," etc., are used to denote names and do not indicate any particular order.

[0198] Those skilled in the art will also know that, besides implementing the controller using purely computer-readable program code, the same functions can be achieved by logically programming the method steps, making the controller function as logic gates, switches, application-specific integrated circuits (ASICs), programmable logic controllers (PLCs), and embedded microcontrollers. Therefore, such a controller can be considered a hardware component, and the devices within it used to implement various functions can also be considered structures within that hardware component. Alternatively, the devices used to implement various functions can be considered as both software modules implementing the method and structures within a hardware component.

[0199] This specification can be described in the general context of computer-executable instructions that are executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, classes, etc., that perform a specific task or implement a specific abstract data type. This specification can also be practiced in distributed computing environments, where tasks are performed by remote processing devices connected via a communication network. In distributed computing environments, program modules can reside in local and remote computer-readable storage media, including storage devices.

[0200] As can be seen from the above description of the embodiments, those skilled in the art can clearly understand that this specification can be implemented by means of software plus necessary general-purpose hardware platforms. Based on this understanding, the technical solutions of this specification can essentially be embodied in the form of a software product. This computer software product can be stored in a storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, mobile terminal, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments of this specification.

[0201] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to interchangeably. Each embodiment focuses on its differences from other embodiments. This specification can be used in numerous general-purpose or special-purpose computer system environments or configurations. Examples include: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable electronic devices, network PCs, minicomputers, mainframe computers, and distributed computing environments including any of the above systems or devices, etc.

[0202] Although this specification has been described by way of examples, those skilled in the art will recognize that many variations and modifications are possible without departing from the spirit of this specification, and it is intended that the appended claims cover such variations and modifications without departing from the spirit of this specification.

Claims

1. A method for monitoring health status by promoting data response, characterized in that, include: Get the fill rate parameter of the promotion data at the current time point, and the fill rate parameter of the promotion data for the current time period; Based on the fill rate parameter of the promotion data at the current time point and the fill rate parameter of the promotion data in the current time period, a first time series of the fill rate parameter of the promotion data is constructed. Based on the first time series, a second time series of fluctuation parameters of the fill rate parameter of the promotion data is constructed; According to the preset processing rules, the target dynamic threshold range at the current time point is determined using the first time series and the second time series. The health status of the promotion data response at the current time point is determined based on the target dynamic threshold range at the current time point and the fill rate parameter of the promotion data at the current time point.

2. The method according to claim 1, characterized in that, The step of constructing a second time series of fluctuation parameters for the fill rate parameter of the promotion data based on the first time series includes: Based on the first time series, multiple fluctuation parameters are obtained by calculating the difference between the fill rate parameters of two adjacent time points; Arrange the multiple fluctuation parameters in chronological order to obtain a second time series of fluctuation parameters related to the fill rate parameter of the promotion data.

3. The method according to claim 1, characterized in that, The step of determining the target dynamic threshold range for the current time point using the first and second time series according to preset processing rules includes: Based on the first time series, the first stable trend center of the fill rate parameter of the promotion data for the current time point is determined; Based on the second time series, the second stable trend center of the fluctuation parameter of the fill rate parameter of the promotion data for the current time point is determined; The target dynamic threshold range for the current time point is determined by determining the interval center based on the first stable trend center and the interval boundary based on the second stable trend center.

4. The method according to claim 1, characterized in that, The step of determining the health status of the promotion data response at the current time point based on the target dynamic threshold range at the current time point and the fill rate parameter of the promotion data at the current time point includes: Detect whether the fill rate parameter of the promotion data at the current time point exceeds the target dynamic threshold range at the current time point; When the fill rate parameter of the promotion data at the current time point exceeds the target dynamic threshold range at the current time point, it is determined that there is an abnormal risk in the health status of the promotion data response at the current time point.

5. The method according to claim 4, characterized in that, If the health status of the promotion data response at the current point in time is determined to be abnormally risky, the method further includes: Query historical records to obtain the target dynamic threshold range for the current time period; Based on the target dynamic threshold range at the current time point and the target dynamic threshold range for the current time period, construct the first stable trend center change curve and the second stable trend center change curve respectively. The first stable trend center change curve and the second stable trend center change curve are processed using a preset anomaly cause classification model to obtain the target classification result; wherein, the preset anomaly cause classification model is a model trained based on a graph autoencoder; the graph autoencoder includes at least an encoder, and the encoder includes at least a sequentially connected GCN layer and a TCN layer; Based on the target classification results, identify the target anomalies that cause the health status of the promotion data response at the current time point to be at risk of being abnormal.

6. The method according to claim 5, characterized in that, The reasons for the target anomaly include at least one of the following: backend system anomaly, change in promotion data delivery strategy, the current time point being a special time point, and changes in the inventory data of promotion data.

7. The method according to claim 6, characterized in that, After determining the target anomaly cause that leads to an abnormal risk in the health status of the promotion data response at the current time point based on the target classification results, the method further includes: Based on the cause of the target anomaly, query the preset processing strategy library to determine the matching target processing strategy; Based on the target processing strategy, the target processing terminal corresponding to the target anomaly cause is determined; and target prompt information about the target anomaly cause is generated; wherein, the target prompt information carries at least the target processing strategy; The target processing strategy is sent to the target processing terminal.

8. The method according to claim 1, characterized in that, The process of obtaining the fill rate parameter of the promotion data at the current time point and the fill rate parameter of the promotion data for the current time period includes: Get the fill rate parameter of the promotion data at the current time point, the fill rate parameter of the promotion data at the nearest reference time point, and historical records of abnormal health status; Based on the fill rate parameter of the promotion data at the current time point and the fill rate parameter of the promotion data at the adjacent reference time points, determine the data change information of the fill rate parameter of the promotion data at the current time point; Based on the data changes in the fill rate parameter of the promotion data at the current time point and the historical records of abnormal health status, the matching current time period is determined.

9. A health status monitoring device that promotes data response, characterized in that, include: The acquisition module is used to acquire the fill rate parameters of the promotion data at the current time point, as well as the fill rate parameters of the promotion data for the current time period. The first construction module is used to construct a first time series of the fill rate parameters of the promotion data based on the fill rate parameters of the promotion data at the current time point and the fill rate parameters of the promotion data in the current time period. The second construction module is used to construct a second time series of fluctuation parameters of the fill rate parameter of the promotion data based on the first time series; The first determining module is used to determine the target dynamic threshold range at the current time point based on the first time series and the second time series according to the preset processing rules. The second determining module is used to determine the health status of the promotion data response at the current time point based on the target dynamic threshold range at the current time point and the fill rate parameter of the promotion data at the current time point.

10. A server, characterized in that, It includes a processor and a memory for storing processor-executable instructions, wherein the processor, when executing the instructions, implements the steps of the method according to any one of claims 1 to 8.

11. A computer-readable storage medium, characterized in that, It stores computer instructions that, when executed by a processor, implement the steps of the method according to any one of claims 1 to 8.

12. A computer program product, characterized in that, It includes a computer program that, when executed by a processor, implements the steps of the method according to any one of claims 1 to 8.

Images