Data processing method and device

Abstract

The invention provides a data processing method and device. The data processing method comprises the steps of obtaining a first crowd package set and a second crowd package set; generating first feature information corresponding to each first crowd packet based on crowd packet attribute information and crowd packet historical consumption information of each first crowd packet in the first crowd packet set, and predicting predicted consumption information corresponding to the first crowd packet according to the first feature information; determining at least one associated first crowd packet corresponding to each second crowd packet in the second crowd packet set, and determining predicted consumption information of each second crowd packet according to the predicted consumption information corresponding to each associated first crowd packet; and screening a target crowd packet from the first crowd packet set and the second crowd packet set based on the predicted consumption information corresponding to each first crowd packet and each second crowd packet. According to the embodiment provided by the invention, the screening quality of the target crowd package is improved.

Description

Technical Field

[0001] This application relates to the field of computer technology, and in particular to data processing methods. This application also relates to a data processing apparatus, a computing device, a computer-readable storage medium, and a computer program product. Background Technology

[0002] With the continuous development of computer technology, people's information acquisition speed is also constantly improving. Advertising is a common practice, where advertisers typically promote their ads within selected audience segments based on predetermined promotion frequencies, ensuring ad exposure while using resources efficiently. However, because the audience segments selected by different advertisers are vast, personalized promotion to different audience segments often requires calculations across all audience segments to guarantee ad exposure, thus consuming significant computing resources.

[0003] Therefore, how to filter the audience groups selected by each advertiser and ensure that the selected audience groups are used in a reasonable manner while maintaining the exposure of advertisements has become an urgent problem to be solved. Summary of the Invention

[0004] In view of this, embodiments of this application provide a data processing method. This application also relates to a data processing apparatus, a computing device, a computer-readable storage medium, and a computer program product, to address the aforementioned problems existing in the prior art.

[0005] According to a first aspect of the embodiments of this application, a data processing method is provided, including: Obtain a first group packet set and a second group packet set, wherein the historical consumption information of the first group packet in the first group packet set is greater than a first historical consumption information threshold, and the historical consumption information of the second group packet in the second group packet set is less than the first historical consumption information threshold; Based on the population package attribute information and historical consumption information of each first population package in the first population package set, generate first feature information corresponding to each first population package, and predict the predicted consumption information corresponding to the first population package based on the first feature information; Determine at least one associated first population group corresponding to each second population group in the second population group set, and determine the predicted consumption information of each second population group based on the predicted consumption information corresponding to each associated first population group; Based on the predicted consumption information corresponding to each first group and each second group, target group groups are selected from the first group set and the second group set.

[0006] According to a second aspect of the embodiments of this application, a data processing apparatus is provided, comprising: The acquisition unit is configured to acquire a first group packet set and a second group packet set, wherein the historical consumption information of the first group packet in the first group packet set is greater than a first historical consumption information threshold, and the historical consumption information of the second group packet in the second group packet set is less than the first historical consumption information threshold. The processing unit is configured to generate first feature information corresponding to each first group of people based on the group of people attribute information and the group of people historical consumption information of each first group of people in the first group of people set, and to predict the predicted consumption information corresponding to the first group of people based on the first feature information. The prediction unit is configured to determine at least one associated first population group corresponding to each second population group in the second population group set, and to determine the prediction consumption information of each second population group based on the prediction consumption information corresponding to each associated first population group. The filtering unit is configured to filter target population packages from the first population package set and the second population package set based on the predicted consumption information corresponding to each first population package and each second population package.

[0007] According to a third aspect of the embodiments of this application, a computing device is provided, comprising: Memory and processor; The memory is used to store computer programs or instructions, and the processor is used to execute the computer programs or instructions, which, when executed by the processor, implement the steps of the above-described data processing method.

[0008] According to a fourth aspect of the embodiments of this application, a computer-readable storage medium is provided that stores a computer program or instructions which, when executed by a processor, implement the steps of the above-described data processing method.

[0009] According to a fifth aspect of the embodiments of this application, a computer program product is provided, including a computer program or instructions that, when executed by a processor, implement the steps of the above-described data processing method.

[0010] According to a data processing method provided in an embodiment of this application, a first audience group set and a second audience group set are divided based on a first historical consumption information threshold. The audience group attribute information and historical consumption information of the first audience group set are used to generate first feature information for each first audience group set. This allows the audience group attribute information and historical consumption information to be considered during the prediction of consumption information based on the first feature information. Furthermore, since the historical consumption information in the second audience group set is less than the first historical consumption information threshold, to avoid obtaining insufficient predicted consumption information by directly using the second audience group set, at least one associated first audience group set is identified within the first audience group set that is related to the second audience group set. Thus, the predicted consumption information of the second audience group set is determined using at least one associated first audience group set, thereby avoiding inaccurate predicted consumption information obtained from the second audience group set. Based on the above, target audience groups with better predicted consumption information are selected based on the predicted consumption information obtained from the first and second audience groups. This ensures ad exposure and reduces the reasonable consumption of ad recommendation resources during the process of recommending ads to the selected target audience groups. Attached Figure Description

[0011] Figure 1 A flowchart of a data processing method according to an embodiment of this application is shown; Figure 2A A schematic diagram of a timing decomposition tool according to an embodiment of this application is shown; Figure 2B A schematic diagram of a multilayer fully connected neural network according to an embodiment of this application is shown; Figure 3 This diagram illustrates an overall architecture of an LSTM model according to an embodiment of this application. Figure 4 A schematic diagram of an LSTM cell structure according to an embodiment of this application is shown. Figure 5 A schematic diagram of a target audience filtering package provided according to an embodiment of this application is shown; Figure 6 This illustration shows a schematic diagram of the structure of a data processing apparatus according to an embodiment of this application; Figure 7 A structural block diagram of a computing device according to an embodiment of this application is shown. Detailed Implementation

[0012] Many specific details are set forth in the following description to provide a full understanding of this application. However, this application can be implemented in many other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of this application; therefore, this application is not limited to the specific embodiments disclosed below.

[0013] The terminology used in one or more embodiments of this application is for the purpose of describing particular embodiments only and is not intended to limit the scope of one or more embodiments of this application. The singular forms “a,” “the,” and “the” used in one or more embodiments of this application and in the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” used in one or more embodiments of this application refers to and includes any or all possible combinations of one or more associated listed items.

[0014] It should be understood that although the terms first, second, etc., may be used to describe various information in one or more embodiments of this application, such information should not be limited to these terms. These terms are only used to distinguish information of the same type from one another. For example, first may also be referred to as second without departing from the scope of one or more embodiments of this application, and similarly, second may also be referred to as first. Depending on the context, the word "if" as used herein may be interpreted as "when," "when," or "in response to a determination."

[0015] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties. Furthermore, the collection, use and processing of the relevant data must comply with the relevant laws, regulations and standards of the relevant regions, and corresponding operation entry points are provided for users to choose to authorize or refuse.

[0016] The data processing method described in this application can be applied to scenarios involving advertising to at least one audience segment. For example, it can be applied to scenarios where high-quality audience segments are selected from those chosen by advertisers during the advertising process.

[0017] First, the terms and concepts involved in one or more embodiments of this application will be explained.

[0018] Audience segmentation: refers to tools that precisely define target audiences using user behavior and characteristic data. The audience segmentation involved in this application is applied in the process of advertising and marketing.

[0019] Advertiser: refers to a legal person, other economic organization, or individual who designs, produces, or publishes advertisements, either independently or by entrusting others, to promote a target audience or provide services.

[0020] In this context, an advertiser can select at least one audience segment from the data resources, and each audience segment can be used by at least one advertiser to deliver at least one creative.

[0021] Historical spending information: Historical spending information in this application can be understood as an audience targeting efficiency metric, a core data point used to measure the effectiveness and value of an advertising audience. It reflects the audience's performance in ad placement, and audience potential can be assessed based on historical spending information. Specifically, historical spending information in this application can be understood to include at least one of historical spending, click-through rate, and conversion rate.

[0022] Historical spending: This can be understood as the total actual cost incurred by an advertiser during a specific period of time when running ads. It is used to evaluate advertising effectiveness and serves as a basis for optimizing budget allocation.

[0023] Click-through rate (CTR): This refers to the ratio of the number of times an ad is clicked by users within a targeted audience group to the number of times the ad is displayed.

[0024] Conversion rate: refers to the percentage of people who completed the target behavior set by the advertiser out of the total number of people who clicked on the ad.

[0025] With the continuous development of computer technology, advertisers select a vast number of audience segments during ad placement. Due to limited platform resources, these audience segments are typically updated at a set frequency. However, when there are ads with a faster update frequency, to ensure the efficient use of computing resources, it is usually necessary to filter the advertiser's selected audience segments to obtain target audience segments. This allows for the promotion of the faster-updating ads within the target audience segments.

[0026] In the process of screening audience groups selected by advertisers, metrics such as ad impressions and clicks over historical periods are typically used for calculation. However, since different audience groups may be served with various creatives by multiple advertisers, using historical metrics of advertiser-selected audience groups for prediction is prone to bias. Furthermore, for audience groups with limited historical spending data, predictions of future spending are often based on relevant experience. However, this approach can easily lead to the omission of high-potential audience groups with limited historical spending data, resulting in biased predictions.

[0027] Therefore, using the methods described above, it is difficult to reasonably filter the audience segments selected by advertisers, resulting in the audience segments obtained from the above filtering being unable to accurately predict the advertisers' future spending information.

[0028] Therefore, this application provides a data processing method that divides audience groups based on historical spending information, classifying audience groups with more historical spending information into a first audience group and a second audience group. The method obtains predicted spending information for the first audience group based on its first characteristic information, and then uses this predicted spending information to predict the predicted spending information for the second audience group, thereby obtaining predicted spending information for both the first and second audience groups. Target audience groups are then selected based on these predicted spending information. This improves the accuracy of target audience group selection, enabling the selected target audience groups to accurately reflect the advertiser's future spending information. This application also relates to a data processing apparatus, a computing device, a computer-readable storage medium, and a computer program product, which will be described in detail in the following embodiments.

[0029] Figure 1 A flowchart of a data processing method according to an embodiment of this application is shown, specifically including the following steps 102-108: Step 102: Obtain the first group packet set and the second group packet set.

[0030] Wherein, the historical consumption information of the first group in the first group set is greater than the first historical consumption information threshold, and the historical consumption information of the second group in the second group set is less than the first historical consumption information threshold.

[0031] The first historical consumption information threshold can be understood as a pre-set threshold for dividing high-data-value and low-data-value user groups, used to determine whether a user group has the data basis to directly predict future consumption.

[0032] In one specific embodiment provided in this application, multiple audience groups selected by the advertiser are obtained, and the multiple audience groups are divided into a first audience group set and a second audience group set according to a first historical message threshold.

[0033] For example, from multiple audience segments selected by advertisers, relevant data information for each audience segment is obtained. This includes audience size, audience demographics, and audience delivery efficiency metrics. Based on this data, historical spending information corresponding to each audience segment is determined. To improve the accuracy of subsequent target audience segment selection, audience segments are divided based on their historical spending information and a first historical spending information threshold, resulting in a first set of audience segments and a second set of audience segments with varying levels of historical spending information. For instance, if the historical spending information exceeds the first historical spending information threshold, the current audience segment is considered to have the data basis for directly predicting future spending. If the historical spending information is less than the first historical spending information threshold, the current audience segment is considered not to have the data basis for directly predicting future spending.

[0034] It is important to understand that the consumption information in this application refers to the consumption information of advertisers' ad placements. Based on this, this application determines whether an audience segment has the potential for high consumption value by using the predicted consumption information corresponding to each audience segment; that is, whether the advertiser's consumption information is proportional to the advertiser's expected revenue when placing ads to that audience segment in the future. Therefore, this application will use audience segments with different levels of historical consumption information as the basis for selecting target audience segments.

[0035] Furthermore, this application provides an exemplary description of how to obtain the first group packet set and the second group packet set.

[0036] In one specific embodiment provided in this application, obtaining a first group packet set and a second group packet set includes: Obtain a set of audience segments, and obtain the creation time corresponding to each audience segment in the set of audience segments; Determine the target audience group, wherein the target audience group is any one of the audience group sets; If the creation time of the pending audience package is greater than the creation time threshold, and it is determined that the historical consumption information of the pending audience package is greater than the first historical consumption information threshold, the pending audience package is added to the first audience package set. If the creation time of the pending audience package is less than the creation time threshold, and the historical consumption information of the pending audience package is less than the first historical consumption information threshold, the pending audience package is added to the second audience package set.

[0037] The creation time can be understood as the specific point in time when an advertising platform generates an audience package with a specific set of users and attributes through rule configuration, algorithm mining, and tag combination. It marks the beginning of the audience package's lifecycle. For example, from the creation time of the audience package, the platform begins recording subsequent dynamic data for that audience package.

[0038] The creation time threshold can be understood as a pre-set threshold used to combine the first historical consumption information threshold to divide high-data-value audience groups into low-data-value audience groups, and is used to further determine whether the audience group has the data basis to directly predict future consumption.

[0039] In one specific embodiment provided in this application, a set of audience segments is obtained, and the creation time of each audience segment in the set is determined. For any audience segment to be processed in the audience segment set, if the creation time of the audience segment to be processed is greater than a creation time threshold, and the historical consumption information threshold of the audience segment to be processed is greater than a first historical consumption information threshold, the audience segment to be processed is added to a first audience segment set, further improving the accuracy of high data value audience segmentation. Similarly, if the creation time of the audience segment to be processed is less than a creation threshold, and the historical consumption information threshold of the audience segment to be processed is less than a first historical consumption information threshold, the audience segment to be processed is added to a second audience segment set, thereby identifying low data value audience segments.

[0040] For example, the system retrieves audience segments selected by advertisers and divides each segment based on a creation time threshold and a first historical consumption information threshold. If the creation time of an audience segment is greater than the creation time threshold, the segment has a long lifespan and relatively complete data accumulation, thus placing this type of audience segment into the first audience segment set. If the creation time of an audience segment is less than the creation time threshold, the segment has a short lifespan and less data accumulation, thus placing this type of audience segment into the second audience segment set.

[0041] According to the embodiments described above in this application, by quantifying the existence time of audience segments, associating data quality, and coordinating with a first historical consumption threshold, accurate segmentation of audience segments can be achieved. This improves the accuracy of audience segmentation.

[0042] Furthermore, after obtaining the first and second population data sets, since the population data sets in the first set have long lifecycles and relatively complete data accumulation, the predicted consumption information of each population data set in the first set can be directly predicted based on the relevant information of the first population data sets using the following method.

[0043] Step 104: Generate first feature information corresponding to each first group of people based on the group of people attribute information and the historical consumption information of each first group of people in the first group of people set, and predict the predicted consumption information corresponding to the first group of people based on the first feature information.

[0044] The first feature information can be understood as the information after mapping the crowd package attribute information and the crowd package historical consumption information into numerical features.

[0045] Predictive consumption information can be understood as the predicted resource consumption information that advertisers will need to consume when placing ads in an audience package within a set future time period.

[0046] Audience segment attribute information can be understood as information used to determine the size of the audience segment, the proportion of each user profile within that audience, and so on. Audience size information can at least determine the number of users in the audience segment. The proportion of each user profile can be understood as information such as gender, age, marital status, and region within the audience segment.

[0047] Historical spending information for audience segments can be understood as including at least one of the following: audience targeting efficiency metrics. These metrics include at least one of the following: historical spending of the audience segment, click-through rate (CTR), and conversion rate.

[0048] For example, when an advertiser is running a target ad campaign, obtaining historical spending information for the target audience can be understood as using historical spending to determine the advertiser's cost per click for the target ad. The click-through rate (CTR) is then used to determine the ratio of clicks to ad impressions. Finally, the conversion rate is combined to determine the percentage of people who completed the advertiser's target action out of the total number of people who clicked the target ad. Based on this, historical spending information for the target audience can be understood as information used to determine the revenue generated based on the final conversion rate. Therefore, in determining historical spending, advertisers can use this information to judge whether the cost per click for the ad is appropriate.

[0049] Furthermore, when historical consumption information includes historical consumption data, it can also be understood as the exposure of the target ad within the audience segment.

[0050] It should be understood that the historical consumption information involved in this application is used to determine whether a population group is a high-data-value population group or a low-data-value population group. Based on this, the specific meaning of the historical consumption information can be determined according to the actual situation, and this application does not impose any limitations on it.

[0051] In one specific embodiment of this application, the audience size information and audience profile proportion information of each first audience group in the first audience group set are obtained, and this information is determined as audience group attribute information. Historical spending, click-through rate, and conversion rate information of the audience group are also obtained, and this information is determined as audience group historical spending information. The audience group attribute information and audience group historical spending information are converted into numerical representations to generate first feature information corresponding to each first audience group. Predicted spending information corresponding to the first audience group is then predicted based on the first feature information.

[0052] Furthermore, this application explains the first feature information for generating the first group of people and the predicted consumption information for predicting the first group of people based on the first feature information in the following manner.

[0053] In one specific embodiment provided in this application, first feature information corresponding to each first group of people is generated based on the group of people attribute information and the group of people historical consumption information of each first group of people in the first group of people set, including: Based on the population attribute information of each first population packet in the first population packet set, the static feature information of the first population packet is determined; Based on the historical consumption information of each first group of people in the first group of people set, the temporal characteristic information of the first group of people is determined. Based on the static feature information and the temporal feature information, first feature information corresponding to each first group packet is generated.

[0054] Among them, time-series feature information can be understood as time-series data that reflects the dynamic behavior patterns of a group of people.

[0055] In one specific embodiment provided in this application, relevant information data of each user in the first group package is collected, such as: group size, group profile ratio (gender, age, marital status, region, etc.), and group delivery efficiency indicators (historical consumption, click-through rate, conversion rate, etc.); then, the collected data is filled with missing values ​​and outliers are removed; finally, the cleaned data is assembled and mapped into a series of identifiable data, and the identifiable data is determined as the first feature information.

[0056] According to the embodiments described above in this application, by combining static features with temporal features, a full-dimensional description of the first group of people can be achieved, from attributes to behavior and from static to dynamic, thereby improving the accuracy of subsequent prediction of consumption information.

[0057] Furthermore, based on the first feature information obtained above, the prediction consumption information corresponding to the first group of people in this application is predicted based on the first feature information, including S1042-S1046: S1042. Determine the first feature information of the first group of people to be predicted, wherein the first group of people to be predicted is any one of the first group of people sets.

[0058] In one specific embodiment provided in this application, a population packet is arbitrarily selected from the first population packet as the first population packet to be predicted, and the first feature information of the first population packet to be predicted is determined. The above operation is repeated to obtain the first feature information of each first population packet in the first population packet set.

[0059] The method of obtaining the first feature information is as described in the examples above in this application, and will not be repeated here.

[0060] As can be seen from the above, the first feature information includes the temporal feature information and static feature information of the first group of people to be predicted.

[0061] S1044. Input the temporal feature information in the first feature information into the temporal feature decomposition model to obtain the temporal decomposition features output by the temporal feature decomposition model.

[0062] In one specific embodiment provided in this application, based on the first feature information, the historical consumption information, which is a time-series feature information in the first feature information, is input into the time-series feature decomposition model to obtain the time-series decomposition features output by the time-series feature decomposition model.

[0063] Specifically, the temporal feature information from the first feature information is input into the temporal feature decomposition model to obtain the temporal decomposition features output by the temporal feature decomposition model, including: The temporal feature information in the first feature information is input into the temporal feature decomposition model; Based on the time series feature decomposition model, trend terms, periodic terms, holiday terms, autoregressive terms, and custom regression terms are obtained. The trend term, periodic term, holiday term, autoregressive term, and custom regression term are identified as the time series decomposition features output by the time series feature decomposition model.

[0064] The temporal feature decomposition model involved in this application can be understood as an improved Prophet temporal prediction model.

[0065] The terms are as follows: Trend, Trend, and Custom Regression. Trend terms reflect the long-term, non-periodic patterns of spending in the first audience segment, such as overall increases, decreases, or stabilization. Periodicity terms reflect the recurring patterns of spending in the first audience segment, typically weekly (e.g., high spending on weekdays, low spending on weekends) or monthly. Holiday terms reflect the specific impact of certain time points (holidays, major promotions) on audience segment spending, such as a surge in spending during holidays. Autoregression terms reflect the dependence of past spending on current or future spending, such as high spending on the previous N days likely leading to high spending on the N+1th day. Custom regression terms reflect the impact of "static audience segment attributes" on spending, linking static features with time-series features, such as higher spending with more advertisers.

[0066] For ease of understanding, this application explains the improved Prophet time-series feature decomposition model in the following manner.

[0067] like Figure 2A As shown, Figure 2A A schematic diagram of a timing decomposition tool provided according to an embodiment of this application is shown.

[0068] In one specific embodiment provided in this application, the time series feature information is decomposed into core time series components such as trend term, period term, and autoregressive term using the Prophet time series decomposition tool.

[0069] Building upon this foundation, to further clarify the impact of different factors on consumption information, this application, based on the core time-series components obtained from the Prophet time-series decomposition tool, adds "other items" and "custom items," such as holiday items and custom regression items, to obtain an improved Prophet time-series prediction model (AR-Prophet-ctr). This allows for the coverage of the complex impacts of ad consumption based on the newly added other and custom items. Furthermore, business features such as "whether it's a major promotion, day of the week, holiday, budget, and ad campaign (AS)" are introduced into the business feature fusion module. Through interaction between the business feature fusion module and each time-series component, the decomposition results are made more closely aligned with the advertising scenario. This achieves accurate decomposition of "long-term trends, cyclical patterns, sudden impacts, and autoregressive continuity" in historical consumption, providing structured time-series decomposition features for subsequent predictions.

[0070] Based on this, in order to facilitate the... Figure 2A The understanding of AR-Prophet-ctr involved in this application, combined with Figure 2B The content shown is for Figure 2A The content is explained in a structured manner.

[0071] In one specific embodiment provided in this application, Figure 2B This diagram illustrates a multi-layer fully connected neural network according to an embodiment of this application. In the input layer, time-series data (e.g., historical sequences of consumption, clicks, etc. over the past N days) at "relevant time P (input layer)" is received to capture dynamic correlations in short-term time series. In the hidden layer, a non-linear transformation of the input features is achieved through full connectivity of multiple neurons (weight matrix w), uncovering complex patterns that are difficult to decompose using Prophet. In the output layer, the predicted result (e.g., consumption on a future day) is output.

[0072] Building upon the above, ad spending is simultaneously influenced by linear trends (such as user growth), cyclical patterns (such as weekly / monthly ad delivery schedules), non-linear fluctuations (such as sudden ad delivery by advertisers), and business specificities (such as major promotions / holidays). This application combines an improved Prophet for structured temporal decomposition with a multi-layer fully connected neural network for non-linear feature capture. This approach retains Prophet's accurate ability to decompose "trends and cycles" while using the neural network to compensate for its predictive shortcomings under "non-linear and sudden business impacts," ultimately achieving high-precision prediction of audience spending and providing support for subsequent audience selection.

[0073] Furthermore, based on the above, the Prophet time-series prediction model is improved according to the click-through rate of the audience package within a historical time period and the regression terms. The regression terms include autoregressive terms and user-defined regression variables. The improved Prophet time-series prediction model is represented as follows:

[0074] Wherein, α is the average click-through rate of the population over the past N days, used to influence the overall trend of population consumption; g(t) is a non-periodic trend term, used to represent the non-periodic trend changes in the population consumption time series data, such as overall upward or downward trends; s(t) is a periodic term, representing data with periodic changes in the population consumption time series data; h(t) is a holiday term, used to fit the impact of certain special time nodes in the time series data on the data, such as promotional activities and holidays; ε(t) is an error term, used to fit random fluctuations; a(t) is an autoregressive term, used to represent the impact of past time points on the current prediction target; and the last custom regression term is used to reflect the impact of the current population index on future predictions.

[0075] Furthermore, regarding this application Figure 2B In the middle hidden layer It can be characterized as:

[0076] Regarding this application Figure 2B The output layer can be characterized as:

[0077] in These are the k time points related to the current time t. Consumption of the crowd at time ti; for The corresponding parameters are below; for Lower bias term; for The corresponding parameters; c is the final output layer bias term.

[0078] S1046. Input the time-series decomposition features and the static feature information in the first feature information into the target consumption prediction model to obtain the predicted consumption information of the first group package to be predicted output by the target consumption prediction model.

[0079] In one specific embodiment provided in this application, the temporal decomposition features are obtained according to the above content, and the temporal decomposition features and the static feature information in the first feature information are input into the target consumption prediction model to obtain the predicted consumption information of the first group package.

[0080] It should be understood that the target consumption prediction model involved in this application can be understood as a Long Short-Term Memory (LSTM) model.

[0081] Among them, LSTM is a variant of recurrent neural network, which is suitable for modeling and predicting time series data.

[0082] like Figure 3 As shown, Figure 3 A diagram illustrating the overall architecture of an LSTM model according to an embodiment of this application is shown.

[0083] like Figure 3 The LSTM neural network introduces three main control gate mechanisms: the Input Gate, which determines which information can enter the LSTM unit (x0, x1, ..., xt); the Forget Gate, which determines which information needs to be forgotten or discarded to facilitate the learning of long-term dependencies in the sequence (represented by A); and the Output Gate, which outputs the information from the LSTM unit to the next unit or output layer (h0, h1, h2, ..., ht). Within the LSTM unit, an intermediate storage unit called the cell state is introduced to store information and pass it to the next LSTM unit.

[0084] Furthermore, this application also provides an LSTM cell structure diagram. For example... Figure 4 As shown, Figure 4 A schematic diagram of an LSTM cell structure according to an embodiment of this application is shown.

[0085] in, The input at time t, It represents the hidden state and also serves as the output at time t.

[0086] For ease of understanding, this application adopts the following combination Figure 3 as well as Figure 4 The content shown explains the process of obtaining the predicted consumption information of the first group of people to be predicted.

[0087] In the above Figure 3 as well as Figure 4 The resulting structured formula is as follows: The formula corresponding to the input gate is:

[0088] The formula corresponding to the forgetting gate is:

[0089] Cell state The calculation formula is:

[0090] The formula for the output gate is:

[0091] The formula for calculating the output is:

[0092] in,[ h,x ] indicates that two vectors h and x Piecing them together, This is the sigmoid function.

[0093] Based on the above, if the consumption of a certain first group package on day t is 8000 yuan, then the following can be obtained by decomposing it using the AR-Prophet-ctr model: Trend term g(t) = 5500 yuan (contribution of long-term upward trend); The periodic term s(t) = 300 yuan (weekend periodic contribution); Holiday item h(t) = 1500 yuan (contribution from holiday promotions); The autoregressive term a(t) = 400 yuan (the historical dependency contribution of high consumption in the last N days); The custom regression item is 300 yuan (the advertiser's contribution).

[0094] According to the embodiments described above in this application, the original Prophet model can only decompose trend, period, holiday, and error terms, making it difficult to capture historical consumption dependencies (i.e., autoregressive terms) and static attribute influences (i.e., custom regression terms), resulting in incomplete decomposition information. Therefore, the time-series prediction model in this application adds autoregressive terms combined with custom regression terms and introduces historical click-through rate to adjust the trend term, enabling the decomposition to cover all dimensions of explicit patterns. Furthermore, since the output time-series decomposition features are structured explicit patterns, they can be directly used as input features for LSTM. Consequently, LSTM no longer needs to expend computational power to extract explicit features and can focus on capturing the implicit correlations between time-series decomposition features. The synergy between the two improves the prediction accuracy of the first group packet prediction consumption information.

[0095] Furthermore, having obtained the predicted consumption information for the first group of people based on the above, this application uses the following method to determine the predicted consumption information for the second group of people.

[0096] Step 106: Determine at least one associated first population group corresponding to each second population group in the second population group set, and determine the predicted consumption information of each second population group based on the predicted consumption information corresponding to each associated first population group.

[0097] For example, based on the population package feature information between different population packages, at least one associated first population package corresponding to each second population package is determined in the first population package, and the predicted consumption information of the second population package is determined based on the predicted consumption information of each associated first population package.

[0098] It should be understood that the associated first population group corresponding to the second population group in this application can be understood as the population group attributes in the associated first population group having a similar relationship with the population group attributes of the second population group. Based on this, this application can determine the predicted consumption information of the second population group based on the predicted consumption information of the associated first population group that has a similar relationship with the second population group.

[0099] In one specific embodiment provided in this application, determining at least one associated first group package corresponding to each second group package in the second group package set includes: A second group of people to be processed is determined, wherein the second group of people to be processed is any one of the second group of people sets.

[0100] Obtain the feature information of the second population group to be processed, and obtain the feature information of the first population group corresponding to each target first population group.

[0101] For ease of understanding, this application uses the following method to explain how to obtain the second group packet feature information and the first group packet feature message to be processed.

[0102] In one specific embodiment provided in this application, obtaining the feature information of the second group of people to be processed includes: Determine the second feature information to be processed corresponding to each second user in the second group of people to be processed; By integrating the second feature information to be processed, feature information of the second population group to be processed corresponding to the second population group to be processed is generated.

[0103] In one specific embodiment provided in this application, the user characteristic information of each second user in the second group package to be processed is determined, such as the basic attribute information of each second user, like the age, gender, and region of the second user, and the behavioral tags of the second user, to obtain the individual characteristics of the second user. The individual characteristics of each second user in the second group package to be processed are fused to determine the feature information of the second group package to be processed.

[0104] The methods for fusing the individual characteristics of each second user in the second user group to be processed include, but are not limited to, determining the mean of the individual characteristics of each second user in the second user group to be processed, thereby statistically obtaining the feature information of the second user group to be processed. Alternatively, statistically summarizing the individual characteristics of each second user and performing weighted fusion to obtain the feature information of the second user group to be processed. This application does not limit the method of obtaining the feature information of the second user group to be processed.

[0105] According to the embodiments described above in this application, by fusing the individual characteristic information of the second user to obtain the characteristic information of the second group to be processed, the characteristic information of the second group to be processed is made closer to the real group attributes while preserving the diversity of the individual characteristics of the second user. This facilitates the accurate determination of the associated first group based on the characteristic information of the second group.

[0106] Similarly, in a specific embodiment provided in this application, obtaining the first group feature information corresponding to each target first group group includes: Determine the historical consumption information of the first group packet to be processed, wherein the first group packet to be processed is any one of the first group packet sets; If the historical consumption information of the first group of people to be processed is greater than the second historical consumption information threshold, the first group of people to be processed is determined as the target first group of people, wherein the second historical consumption information threshold is greater than the first historical consumption information threshold. Obtain the first group feature information corresponding to the target first group group.

[0107] In one specific embodiment provided in this application, an arbitrary first population package is selected from the first population package set as the first population package to be processed, and the historical consumption information of the first population package to be processed is determined. The historical consumption information of the first population package to be processed is compared with a second historical consumption information threshold. If the historical consumption information of the first population package to be processed is greater than the second historical consumption information threshold, the first population package to be processed is determined as an associated first population package that has a relationship with the second population package. The predicted consumption information corresponding to the associated first population package is used as the basis for determining the predicted consumption information of the second population package.

[0108] For ease of understanding, this application provides an exemplary description of how to obtain the first group feature information of the target first group.

[0109] In one specific embodiment provided in this application, obtaining the first group feature information corresponding to the target first group group includes: Determine the first feature information to be processed corresponding to each first user in the target first group packet; By integrating the first feature information to be processed, the first group feature information corresponding to the target first group group is generated.

[0110] In one specific embodiment provided in this application, user characteristic information of each first user in the first group packet to be processed is determined, such as the basic attribute information of each first user, like the age, gender, and region of the first user, and the behavioral tags of the first user, to obtain the individual characteristics of the first user. The individual characteristics of each first user in the first group packet to be processed are fused to determine the feature information of the first group packet to be processed.

[0111] In one specific embodiment provided in this application, after obtaining the first group feature information corresponding to the first group package and the second group feature information of the second group package, the feature similarity between the feature information of the second group package to be processed and the feature information of each first group package is calculated, and at least one associated first group package corresponding to the second group package to be processed is determined based on the feature similarity.

[0112] In one specific embodiment of this application, an arbitrary population package is selected from the second population package as the second population package to be processed. The feature information of the second population package to be processed is determined, and the feature information of the first population package of the target first population package in the first population package is determined. Feature similarity (e.g., cosine similarity) is calculated between the feature information of the first population package and the feature information of the second population package to be processed. If the feature similarity between the feature information of the first population package and the feature information of the second population package to be processed meets the feature similarity condition, at least one associated first population package is determined that is related to the second population package to be processed. The above operation is repeated to obtain at least one associated first population package corresponding to each second population package in the second population package set. Based on this, the predicted consumption information of the second population package is determined according to the at least one associated first population package.

[0113] The feature similarity condition can be understood as a condition used to determine the degree of similarity between the first and second population packets. If the feature similarity condition is met, it can be understood that the predicted consumption information of the first population packet corresponding to the second population packet can be directly used to determine the predicted consumption information of the second population packet.

[0114] Furthermore, the feature similarity condition can be a threshold for determining the feature similarity between the first group and the second group.

[0115] According to the embodiments described above in this application, due to limited historical consumption information and insufficient data accumulation in the second population group, it is difficult to directly predict the consumption information of the second population group using the improved Prophet time-series prediction model. Therefore, this application uses a first population group with similar characteristics to the second population group, allowing the second population group to indirectly obtain reference data on trends, cycles, and business-related patterns. Furthermore, by using feature similarity, it further ensures that the obtained first population group has a high degree of similarity to the second population group in terms of population attributes and consumption patterns. Based on this, referring to the predicted consumption information of the first population group further reduces prediction bias caused by large differences in population demographics.

[0116] In one specific embodiment provided in this application, the predicted consumption information of each second group package is determined based on the predicted consumption information corresponding to each associated first group package, including: Determine the second group packet to be processed, wherein the second group packet to be processed is any one of the second group packets set; The predicted consumption information corresponding to each associated first group of people is weighted and summed to obtain the predicted consumption information of the second group of people to be processed.

[0117] In one specific embodiment provided in this application, any one of the second population packets in the second population packet set is determined as the second population packet to be processed. If there is at least one associated first population packet corresponding to the second population packet to be processed, the predicted consumption information corresponding to each associated first population packet is weighted and summed, and the result of the weighted summation is determined as the predicted consumption information of the second population packet to be processed.

[0118] Furthermore, this application can also determine the weight value corresponding to each associated first group package based on the feature similarity between each associated first group package and the second group package, and then perform a weighted summation of the predicted consumption information based on the weight value corresponding to each associated first group package, and determine the result of the weighted summation as the predicted consumption information of the second group package to be processed.

[0119] Step 108: Based on the predicted consumption information corresponding to each first population group and each second population group, filter the target population group from the first population group set and the second population group set.

[0120] The target audience package can be understood as a high-quality audience package selected from the first audience package set and the second audience package set.

[0121] In a specific embodiment provided in this application, a screening threshold is determined based on the predicted consumption information corresponding to each first audience group and each second audience group. At least one audience group whose predicted consumption information is higher than the screening threshold is identified as a target audience group. The screening threshold can be a threshold used to screen high-value audience groups or a threshold used to screen low-value audience groups. When the screening threshold is used to screen high-value audience groups, it can be understood that during the process of advertising to the target audience group at a future time, the advertiser's consumption information is proportional to the advertiser's expected revenue. When the screening threshold is used to screen low-value audience groups, it can be understood that the advertiser can advertise to non-target audience groups at a future time. The target audience group can be understood as the group whose consumption shows a downward trend, which can be used as a basis for subsequent screening.

[0122] In this application, the method of obtaining the target audience package is further explained using the obtained target audience package as an example.

[0123] In a specific embodiment provided in this application, based on the predicted consumption information corresponding to each first population group and each second population group, a target population group is selected from the first population group set and the second population group set, including: Obtain the predicted consumption information corresponding to each first group and each second group, sort the predicted consumption information, and obtain the sorting results; Based on preset filtering conditions and the sorting results, at least one target predicted consumption information is obtained through filtering. Based on the predicted consumption information for each target, the target audience package is determined.

[0124] In a specific embodiment provided in this application, predicted consumption information for each population package (including the first population package and the second population package) in a first population package set and a second population package set is determined. Based on this, the predicted consumption information is sorted to obtain a sorting result. According to preset filtering conditions and the sorting result, at least one target predicted consumption information is obtained. And based on each target predicted consumption information, a target population package is determined.

[0125] Furthermore, once at least one target audience group has been identified, advertisers can conduct advertising campaigns within that target audience group.

[0126] Regarding the methods involved in this application, this application will provide an exemplary description of the above data processing methods in conjunction with the following examples.

[0127] In one example, assuming a total of 70 audience groups, to ensure efficient use of computing resources, the 70 audience groups are divided into subgroups (e.g., 7 groups of 10). The data from these subgroups is then updated daily for a single target ad, with one subgroup updated daily, until all 7 subgroups are updated (i.e., a 7-day cycle). In this scenario, it can be understood that the target ad updates every 7 days for each audience group. However, if target ad B wants to increase the update frequency (e.g., every 2 days) during its promotional activities, performing a 2-day cycle on all 70 audience groups would consume computing resources equivalent to updating 35 audience groups daily. This results in high computing resource consumption. To address this scenario, and to ensure ad exposure while conserving computing resources, the data processing method described in this application selects a target number of audience groups (e.g., 20) from the 70. Only this target number of audience groups is updated at the set update frequency. This ensures efficient use of computing resources. The remaining 70-20 group packages will continue to be updated at the original update frequency.

[0128] In other words, the data processing method involved in this application is equivalent to selecting at least one target audience group that meets the criteria from multiple audience groups selected by the original advertisers when the frequency of ad updates increases. The ad is then promoted within this selected target audience group, improving the quality of the target audience selection. By promoting the ad within at least one target audience group, the method ensures both ad exposure and the rational consumption of resources.

[0129] For ease of understanding, this application combines Figure 5 The process of screening the target audience group involved in this application is illustrated by way of example. Figure 5 A schematic diagram of a target audience filtering package provided according to an embodiment of this application is shown.

[0130] like Figure 5 As shown, by mining the advertising audience that includes multi-dimensional data in the platform's audience data, it is possible to identify high-quality advertising platforms and high-quality audiences from multiple perspectives. This helps developers allocate computing resources and business operations personnel promote the target audience.

[0131] In one specific embodiment of this application, the data processing method of this application involves a feature processing stage and a model building stage.

[0132] The feature processing stage can be understood as generating the training and prediction data required for the target model during its construction. Specifically, it involves determining the population package set and, based on this set, identifying the first and second population package sets.

[0133] In one example, the data to be processed is determined based on audience size, audience profile proportion, and audience targeting efficiency metrics. Audience targeting efficiency metrics can be understood as core data used to measure the effectiveness and commercial value of an advertising audience; it reflects the efficiency of the audience's performance in ad delivery. This parameter is the basis for assessing audience potential and training predictive models. Audience targeting efficiency metrics can be determined based on parameters such as historical spending, click-through rate, and conversion rate.

[0134] Historical cost can be understood as the cost an advertiser incurs when placing an ad. Click-through rate (CTR) can be understood as the ratio of the number of clicks on an ad by the target audience to the number of ad impressions. Conversion rate can be understood as the percentage of people who completed the advertiser's target action out of the total number of people who clicked on the ad.

[0135] The various parameters used to determine audience efficiency metrics can be understood as follows: based on the revenue information obtained from the final conversion rate, it can be determined whether the advertiser's cost of placing the ad was appropriate in the historical spending.

[0136] Furthermore, the collected data to be processed in this application is divided into time-series feature information and static information. Time-series feature information can be understood as audience targeting efficiency indicators. Static information can be understood as audience size, profile proportion, number of advertisers, number of creatives, etc.

[0137] The relationship between the number of advertisers and the number of creatives is as follows: Advertiser A can have at least one creative (a creative can be understood as a buy-one-get-one-free offer on the target item, a video of the target item, a text and image of the target item, etc.).

[0138] Creative count can be understood as the number of creatives associated with a specific audience. In other words, it is the total number of different content formats of ad creatives delivered by all advertisers targeting that audience.

[0139] Based on the above, a first group packet set and a second group packet set are obtained. The specific methods for determining the first group packet set and the second group packet set are described in the above embodiments of this application and will not be repeated here.

[0140] After obtaining the first and second population data sets, the static and temporal features of each first population data set are determined, and outlier handling is performed on the temporal features. A temporal feature decomposition model is used to obtain temporal decomposition features (e.g., trend terms, periodic terms, custom terms, etc.). These temporal decomposition features are then combined with the static information and input into a target consumption prediction model (e.g., an LSTM model) to obtain the predicted consumption information for each first population data set.

[0141] For the second user group set, determine the second unprocessed feature information for each second user, and then perform weighted fusion of the second unprocessed feature information of each second user group to obtain the user group feature information for each second user group. In the first user group set, identify associated first user group sets with similar user group feature information to the second user group sets. Using the predicted consumption information of the associated first user group sets, determine the predicted consumption information for each second user group set.

[0142] Based on the predicted consumption information of each first-person group and each second-person group, the predicted consumption information is sorted to filter out the target group.

[0143] Corresponding to the above method embodiments, this application also provides data processing apparatus embodiments. Figure 6 A schematic diagram of the structure of a data processing apparatus according to an embodiment of this application is shown. Figure 6 As shown, the device includes: The acquisition unit 602 is configured to acquire a first group of people and a second group of people, wherein the historical consumption information of the first group of people in the first group of people is greater than a first historical consumption information threshold, and the historical consumption information of the second group of people in the second group of people is less than the first historical consumption information threshold.

[0144] The processing unit 604 is configured to generate first feature information corresponding to each first group of people based on the group of people attribute information and the historical consumption information of each first group of people in the first group of people set, and to predict the predicted consumption information corresponding to the first group of people based on the first feature information. The prediction unit 606 is configured to determine at least one associated first population package corresponding to each second population package in the second population package set, and to determine the prediction consumption information of each second population package based on the prediction consumption information corresponding to each associated first population package.

[0145] The filtering unit 608 is configured to filter target population packages in the first population package set and the second population package set based on the predicted consumption information corresponding to each first population package and each second population package.

[0146] Optionally, the prediction unit 606 is further configured as follows: Determine the second group packet to be processed, wherein the second group packet to be processed is any one of the second group packets set; Obtain the feature information of the second group of people to be processed, and obtain the feature information of the first group of people corresponding to each target first group of people; Calculate the feature similarity between the second group of people to be processed and the feature information of each first group of people; Based on the feature similarity, at least one associated first group package corresponding to the second group package to be processed is determined.

[0147] Optionally, the prediction unit 606 is further configured as follows: Determine the second feature information to be processed corresponding to each second user in the second group of people to be processed; By integrating the second feature information to be processed, feature information of the second population group to be processed corresponding to the second population group to be processed is generated.

[0148] Optionally, the prediction unit 606 is further configured as follows: Determine the historical consumption information of the first group packet to be processed, wherein the first group packet to be processed is any one of the first group packet sets; If the historical consumption information of the first group of people to be processed is greater than the second historical consumption information threshold, the first group of people to be processed is determined as the target first group of people, wherein the second historical consumption information threshold is greater than the first historical consumption information threshold. Obtain the first group feature information corresponding to the target first group group.

[0149] Optionally, the prediction unit 606 is further configured as follows: Determine the first feature information to be processed corresponding to each first user in the target first group packet; By integrating the first feature information to be processed, the first group feature information corresponding to the target first group group is generated.

[0150] Optionally, the processing unit 604 is further configured to: Based on the population attribute information of each first population packet in the first population packet set, the static feature information of the first population packet is determined; Based on the historical consumption information of each first group of people in the first group of people set, the temporal characteristic information of the first group of people is determined. Based on the static feature information and the temporal feature information, first feature information corresponding to each first group packet is generated.

[0151] Optionally, the processing unit 604 is further configured to: Determine the first feature information of the first group of people to be predicted, wherein the first group of people to be predicted is any one of the first group of people sets; The temporal feature information in the first feature information is input into the temporal feature decomposition model to obtain the temporal decomposition features output by the temporal feature decomposition model; The time-series decomposition features and the static feature information in the first feature information are input into the target consumption prediction model to obtain the predicted consumption information of the first group of people to be predicted output by the target consumption prediction model.

[0152] Optionally, the processing unit 604 is further configured to: The temporal feature information in the first feature information is input into the temporal feature decomposition model; Based on the time series feature decomposition model, trend terms, periodic terms, holiday terms, autoregressive terms, and custom regression terms are obtained. The trend term, periodic term, holiday term, autoregressive term, and custom regression term are identified as the time series decomposition features output by the time series feature decomposition model.

[0153] Optionally, the prediction unit 606 is further configured as follows: Determine the second group packet to be processed, wherein the second group packet to be processed is any one of the second group packets set; The predicted consumption information corresponding to each associated first group of people is weighted and summed to obtain the predicted consumption information of the second group of people to be processed.

[0154] Optionally, the acquisition unit 602 is further configured to: Obtain a set of audience segments, and obtain the creation time corresponding to each audience segment in the set of audience segments; Determine the target audience group, wherein the target audience group is any one of the audience group sets; If the creation time of the pending audience package is greater than the creation time threshold, and it is determined that the historical consumption information of the pending audience package is greater than the first historical consumption information threshold, the pending audience package is added to the first audience package set. If the creation time of the pending audience package is less than the creation time threshold, and the historical consumption information of the pending audience package is less than the first historical consumption information threshold, the pending audience package is added to the second audience package set.

[0155] Optionally, the filtering unit 608 is further configured as follows: Obtain the predicted consumption information corresponding to each first group and each second group, sort the predicted consumption information, and obtain the sorting results; Based on preset filtering conditions and the sorting results, at least one target predicted consumption information is obtained through filtering. Based on the predicted consumption information for each target, the target audience package is determined.

[0156] The above is an illustrative scheme of a data processing apparatus according to this embodiment. It should be noted that the technical solution of this data processing apparatus and the technical solution of the data processing method described above belong to the same concept. For details not described in detail in the technical solution of the data processing apparatus, please refer to the description of the technical solution of the data processing method described above.

[0157] Figure 7 A structural block diagram of a computing device according to an embodiment of this application is shown. The components of the computing device 700 include, but are not limited to, a memory 710 and a processor 720. The processor 720 is connected to the memory 710 via a bus 730, and a database 750 is used to store data.

[0158] The computing device 700 also includes an access device 740, which enables the computing device 700 to communicate via one or more networks 760. Examples of these networks include Public Switched Telephone Network (PSTN), Local Area Network (LAN), Wide Area Network (WAN), Personal Area Network (PAN), or combinations of communication networks such as the Internet. The access device 740 may include one or more of any type of wired or wireless network interface (e.g., a network interface card (NIC)), such as an IEEE 802.11 Wireless Local Area Network (WLAN) wireless interface, a Wi-MAX (Worldwide Interoperability for Microwave Access) interface, an Ethernet interface, a Universal Serial Bus (USB) interface, a cellular network interface, a Bluetooth interface, a Near Field Communication (NFC) interface, and so on.

[0159] In one embodiment of this application, the aforementioned components of the computing device 700 and Figure 7 Other components, not shown, can also be connected to each other, for example, via a bus. It should be understood that... Figure 7 The block diagram of the computing device shown is for illustrative purposes only and is not intended to limit the scope of this application. Those skilled in the art can add or replace other components as needed.

[0160] The computing device 700 can be any type of stationary or mobile computing device, including mobile computers or mobile computing devices (e.g., tablet computers, personal digital assistants, laptop computers, notebook computers, netbooks, etc.), mobile phones (e.g., smartphones), wearable computing devices (e.g., smartwatches, smart glasses, etc.) or other types of mobile devices, or stationary computing devices such as desktop computers or personal computers (PCs). The computing device 700 can also be a mobile or stationary server.

[0161] The processor 720 is used to execute the following computer program / instructions, which, when executed by the processor, implement the steps of the above-described data processing method.

[0162] The above is an illustrative scheme of a computing device according to this embodiment. It should be noted that the technical solution of this computing device and the technical solution of the data processing method described above belong to the same concept. For details not described in detail in the technical solution of the computing device, please refer to the description of the technical solution of the data processing method described above.

[0163] An embodiment of this specification also provides a computer-readable storage medium storing a computer program / instructions that, when executed by a processor, implement the steps of the above-described data processing method.

[0164] The above is an illustrative scheme of a computer-readable storage medium according to this embodiment. It should be noted that the technical solution of this storage medium and the technical solution of the data processing method described above belong to the same concept. For details not described in detail in the technical solution of the storage medium, please refer to the description of the technical solution of the data processing method described above.

[0165] An embodiment of this specification also provides a computer program product, including a computer program / instructions that, when executed by a processor, implement the steps of the above-described data processing method.

[0166] The above is an illustrative scheme of a computer program product according to this embodiment. It should be noted that the technical solution of this computer program product and the technical solution of the data processing method described above belong to the same concept. For details not described in detail in the technical solution of the computer program product, please refer to the description of the technical solution of the data processing method described above.

[0167] The foregoing has described specific embodiments of this application. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims may be performed in a different order than that shown in the embodiments and may still achieve the desired results. Furthermore, the processes depicted in the drawings do not necessarily require the specific or sequential order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0168] The computer instructions include computer program code, which may be in the form of source code, object code, executable file, or certain intermediate forms. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording media, USB flash drive, portable hard drive, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. It should be noted that the content included in the computer-readable medium may be appropriately added or removed according to the requirements of patent practice. For example, in some regions, according to patent practice, computer-readable media may not include electrical carrier signals and telecommunication signals.

[0169] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that this application is not limited to the described order of actions, as some steps may be performed in other orders or simultaneously according to this application. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions and modules involved are not necessarily essential to this application.

[0170] In the above embodiments, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0171] The preferred embodiments disclosed above are merely illustrative of this application. The optional embodiments do not exhaustively describe all details, nor do they limit the invention to the specific implementations described. Clearly, many modifications and variations can be made based on the content of this application. These embodiments are selected and specifically described in this application to better explain the principles and practical applications of this application, thereby enabling those skilled in the art to better understand and utilize this application. This application is limited only by the claims and their full scope and equivalents.

Claims

1. A data processing method, characterized in that, include: Obtain a first group packet set and a second group packet set, wherein the historical consumption information of the first group packet in the first group packet set is greater than a first historical consumption information threshold, and the historical consumption information of the second group packet in the second group packet set is less than the first historical consumption information threshold; Based on the population package attribute information and historical consumption information of each first population package in the first population package set, generate first feature information corresponding to each first population package, and predict the predicted consumption information corresponding to the first population package based on the first feature information; Determine at least one associated first population group corresponding to each second population group in the second population group set, and determine the predicted consumption information of each second population group based on the predicted consumption information corresponding to each associated first population group; Based on the predicted consumption information corresponding to each first group and each second group, target group groups are selected from the first group set and the second group set.

2. The method as described in claim 1, characterized in that, Determining at least one associated first group package corresponding to each second group package in the second group package set includes: Determine the second group packet to be processed, wherein the second group packet to be processed is any one of the second group packets set; Obtain the feature information of the second group of people to be processed, and obtain the feature information of the first group of people corresponding to each target first group of people; Calculate the feature similarity between the second group of people to be processed and the feature information of each first group of people; Based on the feature similarity, at least one associated first group package corresponding to the second group package to be processed is determined.

3. The method as described in claim 2, characterized in that, Obtaining the feature information of the second group of people to be processed includes: Determine the second feature information to be processed corresponding to each second user in the second group of people to be processed; By integrating the second feature information to be processed, feature information of the second population group to be processed corresponding to the second population group to be processed is generated.

4. The method as described in claim 2, characterized in that, Obtain the first-group feature information corresponding to each target first-group packet, including: Determine the historical consumption information of the first group packet to be processed, wherein the first group packet to be processed is any one of the first group packet sets; If the historical consumption information of the first group of people to be processed is greater than the second historical consumption information threshold, the first group of people to be processed is determined as the target first group of people, wherein the second historical consumption information threshold is greater than the first historical consumption information threshold. Obtain the first group feature information corresponding to the target first group group.

5. The method as described in claim 4, characterized in that, Obtaining the first group feature information corresponding to the target first group group includes: Determine the first feature information to be processed corresponding to each first user in the target first group packet; By integrating the first feature information to be processed, the first group feature information corresponding to the target first group group is generated.

6. The method as described in claim 1, characterized in that, Based on the population package attribute information and historical consumption information of each first population package in the first population package set, first feature information corresponding to each first population package is generated, including: Based on the population attribute information of each first population packet in the first population packet set, the static feature information of the first population packet is determined; Based on the historical consumption information of each first group of people in the first group of people set, the temporal characteristic information of the first group of people is determined. Based on the static feature information and the temporal feature information, first feature information corresponding to each first group packet is generated.

7. The method as described in claim 6, characterized in that, Based on the first feature information, predict the predicted consumption information corresponding to the first group of people, including: Determine the first feature information of the first group of people to be predicted, wherein the first group of people to be predicted is any one of the first group of people sets; The temporal feature information in the first feature information is input into the temporal feature decomposition model to obtain the temporal decomposition features output by the temporal feature decomposition model; The time-series decomposition features and the static feature information in the first feature information are input into the target consumption prediction model to obtain the predicted consumption information of the first group of people to be predicted output by the target consumption prediction model.

8. The method as described in claim 7, characterized in that, The temporal feature information in the first feature information is input into the temporal feature decomposition model to obtain the temporal decomposition features output by the temporal feature decomposition model, including: The temporal feature information in the first feature information is input into the temporal feature decomposition model; Based on the time series feature decomposition model, trend terms, periodic terms, holiday terms, autoregressive terms, and custom regression terms are obtained. The trend term, periodic term, holiday term, autoregressive term, and custom regression term are identified as the time series decomposition features output by the time series feature decomposition model.

9. The method as described in claim 1, characterized in that, The predicted consumption information for each second group is determined based on the predicted consumption information corresponding to each associated first group, including: Determine the second group packet to be processed, wherein the second group packet to be processed is any one of the second group packets set; The predicted consumption information corresponding to each associated first group of people is weighted and summed to obtain the predicted consumption information of the second group of people to be processed.

10. The method as described in claim 1, characterized in that, Obtain the first group packet set and the second group packet set, including: Obtain a set of audience segments, and obtain the creation time corresponding to each audience segment in the set of audience segments; Determine the target audience group, wherein the target audience group is any one of the audience group sets; If the creation time of the pending audience package is greater than the creation time threshold, and it is determined that the historical consumption information of the pending audience package is greater than the first historical consumption information threshold, the pending audience package is added to the first audience package set. If the creation time of the pending audience package is less than the creation time threshold, and the historical consumption information of the pending audience package is less than the first historical consumption information threshold, the pending audience package is added to the second audience package set.

11. The method as described in claim 1, characterized in that, Based on the predicted consumption information corresponding to each first and second population group, target population groups are selected from the first and second population group sets, including: Obtain the predicted consumption information corresponding to each first group and each second group, sort the predicted consumption information, and obtain the sorting results; Based on preset filtering conditions and the sorting results, at least one target predicted consumption information is obtained through filtering. Based on the predicted consumption information for each target, the target audience package is determined.

12. A data processing apparatus, characterized in that, include: The acquisition unit is configured to acquire a first group packet set and a second group packet set, wherein the historical consumption information of the first group packet in the first group packet set is greater than a first historical consumption information threshold, and the historical consumption information of the second group packet in the second group packet set is less than the first historical consumption information threshold. The processing unit is configured to generate first feature information corresponding to each first group of people based on the group of people attribute information and the group of people historical consumption information of each first group of people in the first group of people set, and to predict the predicted consumption information corresponding to the first group of people based on the first feature information. The prediction unit is configured to determine at least one associated first population group corresponding to each second population group in the second population group set, and to determine the prediction consumption information of each second population group based on the prediction consumption information corresponding to each associated first population group. The filtering unit is configured to filter target population packages from the first population package set and the second population package set based on the predicted consumption information corresponding to each first population package and each second population package.

13. A computing device, characterized in that, include: Memory and processor; The memory is used to store computer programs or instructions, and the processor is used to execute the computer programs or instructions, which, when executed by the processor, implement the steps of the method according to any one of claims 1 to 11.

14. A computer-readable storage medium storing a computer program or instructions, characterized in that, When the computer program or instructions are executed by a processor, they implement the steps of the method according to any one of claims 1 to 11.

15. A computer program product, comprising a computer program or instructions, characterized in that, When the computer program or instructions are executed by a processor, they implement the steps of the method according to any one of claims 1 to 11.

Images