Data analysis methods, apparatus, equipment and storage media

Abstract

This disclosure relates to a data analysis method, apparatus, device, and storage medium. The method includes: acquiring conversion data generated upon completion of a preset operation corresponding to promotional information; parsing the conversion data and determining multiple candidate trigger data matching the conversion data based on the parsing results; wherein the trigger data is data obtained when promotional information is triggered on different information publishing platforms; determining the matching confidence level between each candidate trigger data and the conversion data; and determining the target trigger data corresponding to the conversion data based on the matching confidence level corresponding to each candidate trigger data. Through the above technical solution, the matching confidence level between each candidate trigger data and the conversion data is determined without introducing new data, thereby identifying a target trigger data that best matches the conversion data, improving the accuracy and recall rate of the effect analysis of promotional information conversion on various information publishing platforms.

Description

Technical Field

[0001] This disclosure relates to the field of computer technology, and more particularly to a data analysis method, apparatus, device, and storage medium. Background Technology

[0002] With the development of internet technology, content delivery platforms are becoming increasingly diversified. To analyze the performance of each platform, it is necessary to analyze the conversion rate brought by the platform's traffic, that is, to determine which delivery platform the conversion effect comes from.

[0003] However, existing conversion analysis methods typically produce low-precision results, or even fail to yield any results at all. Summary of the Invention

[0004] To solve the above-mentioned technical problems, or at least partially solve them, this disclosure provides a data analysis method, apparatus, device, and storage medium.

[0005] This disclosure provides a data analysis method, the method comprising:

[0006] Retrieve conversion data generated when the preset operation corresponding to the promotion information is completed;

[0007] The conversion data is analyzed, and multiple candidate trigger data matching the conversion data are determined based on the analysis results; wherein, the candidate trigger data is data obtained when the promotional information is triggered on different information publishing platforms;

[0008] Determine the matching confidence level between each candidate trigger data and the conversion data;

[0009] Based on the matching confidence level corresponding to each of the candidate trigger data, the target trigger data corresponding to the conversion data is determined.

[0010] This disclosure also provides a data analysis apparatus, the apparatus comprising:

[0011] The conversion data acquisition module is used to acquire conversion data generated when a preset operation corresponding to the promotion information is completed;

[0012] The candidate trigger data determination module is used to parse the conversion data and determine multiple candidate trigger data that match the conversion data based on the parsing results; wherein, the candidate trigger data is data obtained when the promotional information is triggered on different information publishing platforms;

[0013] The matching confidence determination module is used to determine the matching confidence of each candidate trigger data with the conversion data;

[0014] The target trigger data determination module is used to determine the target trigger data corresponding to the conversion data based on the matching confidence level corresponding to each candidate trigger data.

[0015] This disclosure also provides an electronic device, which includes:

[0016] Processor and memory;

[0017] The processor executes the steps of the data analysis method described in any embodiment of the present invention by calling the program or instructions stored in the memory.

[0018] This disclosure also provides a computer-readable storage medium storing a program or instructions that cause a computer to perform the steps of the data analysis method described in any embodiment of the present invention.

[0019] The data analysis scheme provided in this disclosure acquires conversion data generated when a preset operation corresponding to promotional information is completed; parses the conversion data, and determines multiple candidate trigger data that match the conversion data based on the parsing results; wherein, the candidate trigger data are data obtained when promotional information is triggered on different information publishing platforms; determines the matching confidence level between each candidate trigger data and the conversion data; and determines the target trigger data corresponding to the conversion data based on the matching confidence level corresponding to each candidate trigger data. This achieves the goal of selecting the target trigger data that best matches the conversion data by analyzing the matching confidence level between each candidate trigger data and the conversion data, thereby improving the accuracy of conversion analysis of promotional information. It also allows for flexible adjustment of the confidence level threshold for candidate trigger data selection, thereby improving the recall rate of conversion analysis of promotional information. Attached Figure Description

[0020] The accompanying drawings, which are incorporated in and form a part of this specification, illustrate embodiments consistent with this disclosure and, together with the description, serve to explain the principles of this disclosure.

[0021] To more clearly illustrate the technical solutions in the embodiments of this disclosure or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0022] Figure 1 A flowchart illustrating a data analysis method provided in an embodiment of this disclosure;

[0023] Figure 2 A flowchart illustrating yet another data analysis method provided in this disclosure embodiment;

[0024] Figure 3 A flowchart illustrating yet another data analysis method provided in this disclosure embodiment;

[0025] Figure 4 This is a schematic diagram of the structure of a data analysis device provided in an embodiment of the present disclosure;

[0026] Figure 5 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this disclosure. Detailed Implementation

[0027] To better understand the above-mentioned objectives, features, and advantages of this disclosure, the solutions disclosed herein will be described in further detail below. It should be noted that, unless otherwise specified, the embodiments and features described herein can be combined with each other.

[0028] Numerous specific details are set forth in the following description in order to provide a full understanding of this disclosure, but this disclosure may also be implemented in other ways different from those described herein; obviously, the embodiments in the specification are only some, and not all, of the embodiments of this disclosure.

[0029] The data analysis method provided in this disclosure is mainly applicable to the analysis of the conversion effect of promotional information on different distribution channels (also known as information publishing platforms) (also known as channel attribution). It is particularly suitable for channel attribution scenarios where unique identifiers such as device numbers or channel numbers cannot be obtained from the trigger data generated when promotional information is triggered, such as short-link distribution scenarios. The data analysis method provided in this disclosure can be executed by a data analysis device, which can be implemented in software and / or hardware. This device can be integrated into an electronic device with a certain computing power, such as a mobile phone, tablet computer, laptop computer, desktop computer, or server.

[0030] Figure 1 This is a flowchart of a data analysis method provided in an embodiment of this disclosure. See also... Figure 1 The data analysis method specifically includes:

[0031] S110. Obtain the conversion data generated when the preset operation corresponding to the promotion information is completed.

[0032] Promotional information refers to information published and promoted through various information publishing platforms, such as applications (apps) or e-commerce items. Preset actions refer to the actions corresponding to the conversion of promotional information, such as cold-starting and registering an application, upgrading the application's usage level, browsing, adding to cart, and / or purchasing items on an e-commerce platform, or clicking and redirecting to the promotional webpage. Conversion data refers to the data generated by the device when promotional information is converted, including the user-authorized IP address (i.e., the conversion IP address) and the user-authorized device attribute information (such as device ID, device operating system information, device brand, and model).

[0033] Specifically, to analyze the information conversion effect of different information publishing platforms for promotional information, the first step is to obtain the data when the promotional information is converted, that is, to obtain the conversion data generated when the preset operation corresponding to the promotional information is completed.

[0034] S120. Analyze the transformation data and determine multiple candidate trigger data that match the transformation data based on the analysis results.

[0035] The trigger data refers to the data obtained when promotional information is triggered on different information publishing platforms. For example, it could be data generated when promotional information is triggered by clicking, voice control, eye tracking, or other methods on any information publishing platform. The trigger data may include the user-authorized IP address of the corresponding device when the promotional information is triggered (i.e., the trigger IP address) and the user-authorized user agent string (User-Agent, UA). The user-authorized device attribute information can be parsed from the user-authorized UA string, but this user-authorized device attribute information may not include the device ID.

[0036] Specifically, the conversion data of the promotional information is parsed to obtain the parsing results. These results are then roughly matched with all corresponding trigger data to obtain multiple trigger data that match the parsing results, serving as candidate trigger data. In some embodiments, conversion analysis of the advertising platform based on user-authorized IP addresses and user-authorized UA strings can be used to roughly match the conversion data and each trigger data to obtain candidate trigger data. In some embodiments, the candidate trigger data corresponding to the promotional information can be directly obtained from the conversion analysis results based on user-authorized IP addresses and user-authorized UA strings.

[0037] S130. Determine the matching confidence level between each candidate trigger data and the conversion data.

[0038] In this context, the matching confidence level refers to the reliability of a correct match between two data points. The higher the matching confidence level, the greater the likelihood that the transformed data will be derived from the candidate trigger data corresponding to that matching confidence level.

[0039] When the device ID or channel ID cannot be obtained from the trigger data, two approaches are used in related technologies to analyze the conversion of promotional information across various information publishing platforms. One approach is based on the user-authorized IP address and user-authorized UA string for conversion analysis. However, this approach yields low accuracy results, and may even fail to obtain the target trigger data. Furthermore, the low accuracy of the target trigger data obtained by this approach makes it susceptible to data manipulation. The other approach involves adding uniquely identifying information to both the trigger data and the conversion data to achieve precise matching. However, this approach introduces new data and has poor compatibility with existing operational services.

[0040] In view of the above, this embodiment of the disclosure, without introducing new data, utilizes the obtained user-authorized IP address and user-authorized UA string to calculate the matching confidence between the conversion data and each candidate trigger data. For example, the role and influence trend of each attribute dimension in the user-authorized device attribute information obtained from the user-authorized UA string can be analyzed in attribution matching, and a probabilistic model for confidence calculation can be constructed based on Bayesian principles. Then, the constructed probabilistic model is used to calculate the matching confidence between the conversion data and each candidate trigger data.

[0041] S140. Based on the matching confidence level of each candidate trigger data, determine the target trigger data corresponding to the conversion data.

[0042] Specifically, based on the matching confidence level corresponding to each candidate trigger data, a final attribution result, namely the target trigger data, is determined from all candidate trigger data for the conversion data.

[0043] In some embodiments, the process of determining the target trigger data can be as follows: Candidate trigger data is filtered by setting a threshold for matching confidence, thereby determining the target trigger data. For example, candidate trigger data exceeding the threshold is determined as the target trigger data. If there are multiple candidate trigger data exceeding the threshold, further filtering can be performed. For example, one candidate trigger data can be randomly selected from the candidate trigger data exceeding the threshold as the target trigger data; or, from the matching confidence scores of the candidate trigger data exceeding the threshold, the matching confidence score corresponding to the highest or median, etc., can be selected, and the candidate trigger data corresponding to the selected matching confidence score can be used as the target trigger data. In this embodiment, the threshold for matching confidence can be flexibly adjusted according to business needs. For example, if the business needs focus on the recall rate of the target trigger data, then the threshold can be appropriately reduced if the matching confidence scores of the candidate trigger data corresponding to the conversion data are generally low, ensuring that more conversion data can match the target trigger data. Conversely, if the business needs focus on the accuracy of the target trigger data, the threshold can be appropriately increased.

[0044] In some embodiments, the process of determining the target trigger data can be as follows: the candidate trigger data corresponding to the maximum matching confidence score is determined as the target trigger data corresponding to the conversion data. In this embodiment, in order to balance accuracy and recall, the candidate trigger data corresponding to the highest matching confidence score (i.e., the maximum matching confidence score) among all matching confidence scores is determined as the target trigger data.

[0045] In some embodiments, the process of determining the target trigger data can be as follows: if there are multiple maximum matching confidence levels among the matching confidence levels, then from the candidate trigger data corresponding to each maximum matching confidence level, the candidate trigger data whose data generation time is closest to the data generation time of the conversion data is selected, and the selected candidate trigger data is determined as the target trigger data corresponding to the conversion data. In this embodiment, if there are multiple maximum matching confidence levels, then the candidate trigger data corresponding to these multiple maximum matching confidence levels needs to be screened again. Considering that users are more likely to convert immediately after triggering promotional information, the candidate trigger data is screened again based on the proximity of the data generation time of the candidate trigger data to the data generation time of the conversion data, so as to further improve the accuracy of data analysis. In specific implementation, the candidate trigger data whose data generation time is closest to the data generation time of the conversion data among the candidate trigger data corresponding to the maximum matching confidence level is selected, and this is taken as the target trigger data of the conversion data.

[0046] The technical solution described in this embodiment acquires conversion data generated upon completion of a preset operation corresponding to promotional information; parses the conversion data and determines multiple candidate trigger data matching the conversion data based on the parsing results; wherein the candidate trigger data are data obtained when promotional information is triggered on different information publishing platforms; determines the matching confidence level between each candidate trigger data and the conversion data; and determines the target trigger data corresponding to the conversion data based on the matching confidence level corresponding to each candidate trigger data. This achieves improved accuracy in conversion analysis of promotional information by analyzing the matching confidence level between each candidate trigger data and the conversion data, and then selecting the target trigger data that best matches the conversion data based on the matching confidence level. It also improves the recall rate of conversion analysis of promotional information by flexibly adjusting the confidence level threshold for candidate trigger data selection.

[0047] Figure 2 This is a flowchart of another data analysis method provided in this disclosure. It further optimizes the step of "determining multiple candidate trigger data for promotional information matching conversion data based on the analysis results." Furthermore, it can optimize the step of "determining the matching confidence level between each candidate trigger data and conversion data." Based on the above, it can further add steps related to trigger data filtering. Explanations of terms that are the same as or corresponding to those in the above embodiments are not repeated here. See [link to documentation]. Figure 2 The data analysis method includes:

[0048] S201. Based on the parsing results, determine the user-authorized conversion IP address and user-authorized conversion device attribute information corresponding to the conversion data.

[0049] The conversion device attribute information refers to the device information of the device used by the user when performing the preset operation. In some embodiments, the conversion device attribute information includes the device brand (such as the device manufacturer), device model, operating system type (such as Android or iOS), and system version number.

[0050] Specifically, ETL attribute calculations are performed on the acquired conversion data. For example, the conversion data is extracted and normalized according to IP address, device brand, device model, operating system type, and system version number to obtain the user-authorized conversion IP address and user-authorized conversion device attribute information contained in the conversion data, which are used for the subsequent initial matching process.

[0051] S202. Obtain the trigger data generated when the promotional information in each information publishing platform is triggered, and parse each trigger data to obtain the user-authorized trigger device attribute information and user-authorized trigger IP address corresponding to each trigger data.

[0052] The trigger device attribute information refers to the device information of the device used when the user executes the trigger action. The information contained in the trigger device attribute information depends on the device information that can be obtained during the trigger data acquisition process. In some embodiments, the trigger device attribute information includes attribute information on at least one of the following attribute dimensions: device brand, device model, operating system type, and system version number.

[0053] Specifically, within the statistical attribution time window, trigger data generated when the promotional information was triggered is obtained from various information publishing platforms. Then, ETL attribute calculations are performed on each trigger data point. For example, the user-authorized trigger IP address is extracted from the trigger data, and the user-authorized UA string is parsed to obtain device attributes with substantial data in four attribute dimensions: device brand, device model, operating system type, and system version number. These attributes are then normalized to obtain the user-authorized trigger IP address and user-authorized trigger device attribute information contained in each trigger data point, which is used for the subsequent initial matching process.

[0054] It should be noted that the obtained trigger data is stored in Redis as a HashMap so that all possible matching trigger data can be obtained in O(1) time complexity, further improving attribution efficiency.

[0055] S203. Determine whether the triggering IP address authorized by the user belongs to an abnormal IP category or an excessively large public IP category.

[0056] Among them, the "excessively large public IP category" refers to IP categories where the number of users sharing the IP address exceeds a set threshold. This set threshold can be configured according to actual business needs.

[0057] Specifically, considering that too many users have the same IP address under the "excessively large public IP" category, which can easily lead to attribution errors, this embodiment removes trigger data belonging to the "excessively large public IP" category to further improve attribution accuracy. Additionally, considering that abnormal IPs have no practical attribution significance, trigger data belonging to the "abnormal IP" category is also removed. In specific implementation, the IP category of each user's authorized trigger IP address is identified to determine whether it belongs to the "abnormal IP" category or the "excessively large public IP" category. If it does not belong to either, step S205 is executed; if it belongs to one of the categories, step S204 is executed.

[0058] In some embodiments, before determining whether the triggering IP address authorized by the user belongs to the abnormal IP category or the excessively large public IP category, the data analysis method further includes: collecting multiple user-authorized IP addresses and statistically classifying the multiple user-authorized IP addresses to obtain abnormal IP categories and excessively large public IP categories. The abnormal IP category contains each abnormal IP address, and the excessively large public IP category contains each excessively large public IP address. In this embodiment, multiple user-authorized IP addresses are collected in advance. For example, based on a big data processing platform, Hive is used to analyze log data stored in Hadoop on a daily basis to obtain user-authorized IP addresses from the entire network. Then, IP field classification operations are performed on these user-authorized IP addresses. All abnormal IP addresses are classified into the abnormal IP category, and all excessively large public IP addresses are classified into the excessively large public IP category.

[0059] Accordingly, in this embodiment, S203 can be implemented as follows: matching the user-authorized trigger IP address with each abnormal IP address and each excessively large public IP address; if a match is successful with an abnormal IP address, it is determined that the user-authorized trigger IP address belongs to the abnormal IP category; if a match is successful with an excessively large public IP address, it is determined that the user-authorized trigger IP address belongs to the excessively large public IP category. If no match is found, it indicates that the user-authorized trigger IP address does not belong to either the abnormal IP category or the excessively large public IP category. This improves the identification efficiency of the user-authorized trigger IP address, thereby improving the efficiency of trigger data filtering.

[0060] S204. If so, then remove the trigger data corresponding to the trigger IP address authorized by the corresponding user.

[0061] Specifically, trigger data whose user-authorized trigger IP addresses belong to abnormal IP categories or excessively large public IP categories are removed from all collected trigger data, and the remaining trigger data participates in the subsequent initial matching operation.

[0062] S205. Based on the user-authorized conversion IP address, the user-authorized trigger IP address, the user-authorized conversion device attribute information, and the user-authorized trigger device attribute information, match the conversion data and each trigger data to obtain each candidate trigger data that matches the conversion data.

[0063] Specifically, for conversion data and any trigger data, the system matches the user-authorized conversion IP address with the user-authorized trigger IP address, and simultaneously matches the user-authorized conversion device attribute information with the user-authorized trigger device attribute information. The degree of matching between these two types of data determines whether the trigger data and the conversion data meet the initial matching requirements. If the requirements are met, the trigger data is considered a candidate trigger data.

[0064] It should be understood that there is no information conflict between the candidate trigger data and the conversion data in the initial matching. For example, the candidate trigger data may not contain attribute information for certain attribute dimensions, but the attribute information for other attribute dimensions it contains is consistent with the attribute information for the corresponding attribute dimensions in the conversion data.

[0065] S206. Parse each candidate trigger data to obtain the trigger device attribute information and the user-authorized trigger IP address corresponding to each candidate trigger data.

[0066] The device attribute information includes attribute information from multiple attribute dimensions.

[0067] Specifically, through the operation of S220, the user-authorized trigger device attribute information and the user-authorized trigger IP address corresponding to each candidate trigger data can be obtained.

[0068] S207. Based on the user-authorized conversion device attribute information and the user-authorized trigger device attribute information in the parsing results of the conversion data, determine at least one attribute combination with the same attribute information on the same attribute dimension.

[0069] Among them, attribute combination refers to the combination of attributes in at least two dimensions.

[0070] Specifically, if a triggering data point is the result of a conversion analysis of conversion data, then the user-authorized IP address and user-authorized device attribute information between the triggering data and the conversion data should have a high degree of consistency. Based on this, in this embodiment of the disclosure, the matching confidence level is calculated according to the attribute combination of IP address and device attribute information.

[0071] Based on the above explanation, the attribute information of the user-authorized conversion device and the attribute information of the user-authorized triggering device are both attribute information in four dimensions: device brand, device model, operating system type, and system version number. Therefore, in this operation, the attributes with the same attribute value in the same attribute dimension of the user-authorized conversion device attribute information and the attribute information of each user-authorized triggering device are first combined to obtain the attribute combinations.

[0072] For example, suppose we use 1 to represent that there is an attribute value in a certain attribute dimension and the attribute values ​​are consistent, and 0 to represent that there is no attribute value in a certain attribute dimension. The digital representation of the user-authorized conversion device attribute information in the conversion data is (1,1,0,1). Under the same IP address, there are 5 candidate trigger data, and their user-authorized trigger device attribute information is represented as (1,1,0,1), (1,1,0,1), (1,0,0,1), (1,0,1,1), and (0,1,1,1), respectively. Then, the determined attribute combinations are (1,1,0,1), (1,0,0,1), and (0,1,0,1).

[0073] S208. Determine the probability distribution value for each attribute combination.

[0074] The probability distribution value refers to the rarity of an attribute combination within a certain statistical range. Within the statistical range, the fewer times a certain attribute combination occurs, the rarer its existence, and the smaller its probability distribution value.

[0075] Specifically, the rarer an attribute combination is within the attribution statistics range, the greater the likelihood that candidate trigger data containing that attribute combination will be identified as target trigger data containing the conversion data of that attribute combination. Based on this, in this embodiment, the distribution probability value of attribute combinations is used as the basic data to calculate the matching confidence. The distribution probability value of each attribute combination within the attribution statistics range can be determined by acquiring data in real time and performing statistical calculations; alternatively, it can be obtained by pre-calculating and storing each distribution probability value, and here, by directly querying the stored data based on the attribute combination.

[0076] S209. For any user-authorized trigger IP address and any attribute combination contained in each candidate trigger data, determine the matching confidence level between the candidate trigger data and the conversion data corresponding to the attribute combination under the user-authorized trigger IP address based on the probability distribution value of the attribute combination.

[0077] Specifically, for any user-authorized trigger IP address and any attribute combination, using the probability distribution value of the attribute combination as the basic data, and based on the Bayesian principle and the idea that the rarer the existence of the attribute combination, the greater the matching confidence, the matching confidence between the candidate trigger data and the conversion data corresponding to the attribute combination under the user-authorized trigger IP address is calculated.

[0078] S210. Based on the matching confidence level of each candidate trigger data, determine the target trigger data corresponding to the conversion data.

[0079] The technical solution described in this disclosure determines whether the user-authorized trigger IP address in each trigger data belongs to an abnormal IP category or an excessively large public IP category; if so, the trigger data corresponding to the user-authorized trigger IP address is removed. This achieves preliminary filtering of trigger data based on IP address, reduces the amount of data in the attribution process, and avoids the participation of trigger data with large attribution errors, thereby further improving the accuracy of subsequent attribution. It determines the user-authorized conversion IP address and user-authorized conversion device attribute information corresponding to the conversion data based on the parsing results; obtains each trigger data generated when promotional information in each information publishing platform is triggered, and parses each trigger data to obtain the user-authorized trigger device attribute information and user-authorized trigger IP address corresponding to each trigger data; based on the user-authorized conversion IP address, each user-authorized trigger IP address, user-authorized conversion device attribute information, and each user-authorized trigger device attribute information, the conversion data and each trigger data are matched to obtain each candidate trigger data that matches the conversion data. This achieves preliminary screening of each trigger data, providing a data foundation for subsequent accurate attribution. By analyzing the attribute information of user-authorized conversion devices and trigger devices from the parsing results, at least one attribute combination with identical attribute information across the same attribute dimension is identified. The probability distribution value of each attribute combination is determined. For any user-authorized trigger IP address and any attribute combination contained in each candidate trigger data set, the matching confidence level between the candidate trigger data and the conversion data corresponding to the attribute combination under that user-authorized trigger IP address is determined based on the probability distribution value of the attribute combination. This approach uses the user-authorized trigger IP address and attribute combination as the unit of confidence calculation, and calculates the matching confidence level based on the probability distribution value representing the rarity of the attribute combination, improving the accuracy of the matching confidence level and thus further enhancing the attribution accuracy.

[0080] Figure 3 This is a flowchart of another data analysis method provided in this disclosure. It further optimizes the step of "determining the probability distribution value of each attribute combination". Based on this, it can be further optimized to "determine the matching confidence level between the candidate trigger data and the conversion data corresponding to any user-authorized trigger IP address and any attribute combination contained in each candidate trigger data, based on the probability distribution value of the attribute combination". Explanations of terms that are the same as or corresponding to those in the above embodiments are not repeated here. See [link to documentation] Figure 3 The data analysis method includes:

[0081] S310. Obtain the conversion data generated when the preset operation corresponding to the promotion information is completed.

[0082] S320. Analyze the transformation data and determine multiple candidate trigger data that match the transformation data based on the analysis results.

[0083] S330. Parse each candidate trigger data to obtain the user-authorized trigger device attribute information and the user-authorized trigger IP address corresponding to each candidate trigger data.

[0084] S340. Based on the user-authorized conversion device attribute information and the user-authorized trigger device attribute information in the parsing results, determine at least one attribute combination with the same attribute information on the same attribute dimension.

[0085] S350. Based on the attribute combination, query the attribute probability dictionary to determine the distribution probability value of the attribute combination.

[0086] The attribute probability dictionary stores the probability distribution values ​​of each attribute combination. To improve query efficiency, it is stored in a dictionary format with high read and write efficiency.

[0087] Specifically, for each identified attribute combination, the attribute probability dictionary is queried using that combination as an index, and the query result is the probability distribution value of the corresponding attribute combination.

[0088] In some embodiments, the attribute probability dictionary is pre-obtained by statistically analyzing device attribute information authorized by multiple users. The pre-establishment process involves: for any attribute combination, determining the number of attributes in the device attribute information authorized by multiple users that match that attribute combination, and determining the ratio of this number of attributes to the total number of device attribute information authorized by multiple users as the probability distribution value of that attribute combination. Specifically, each attribute combination Attr is determined using the pre-collected device attribute information authorized by multiple users, and the number of times N of each attribute combination Attr occurs is counted. Attr The probability distribution of the corresponding attribute combination is determined by the ratio of the number of occurrences of all attribute combinations, N, and then the ratio of the two is called the probability distribution value of the corresponding attribute combination, i.e., P(Attr) = N. Attr / N.

[0089] S360. For any user-authorized trigger IP address and any attribute combination, determine the number of candidate trigger data corresponding to the user-authorized trigger IP address, and based on the distribution probability value of the attribute combination, determine the initial probability value of matching different combinations of candidate trigger data with conversion data.

[0090] The number of combinations ranges from [1, number of data], and the number of combinations is a natural number.

[0091] Specifically, for any user-authorized trigger IP address and any attribute combination, candidate trigger data containing the user-authorized trigger IP address and attribute combination are determined from all candidate trigger data, and the number of these determined candidate trigger data (i.e., the number of data) n is counted. Then, according to formula (1), based on the probability distribution value P(Attr) of the attribute combination as the basic probability data, the probability that i candidate trigger data among the n candidate trigger data can match the conversion data is determined respectively:

[0092]

[0093] Where i ranges from [1, n], representing the number of different combinations from 1 to n; P(Attr, ip, i) represents the initial probability value when the number of combinations is i under the trigger IP address and attribute combination Attr authorized by the user.

[0094] According to formula (1), the initial probability values ​​of the trigger IP address authorized by the user and the number of combinations of the attribute can be calculated from 1 to n: P(Attr,ip,1), P(Attr,ip,2), ..., P(Attr,ip,n).

[0095] S370. For any user-authorized trigger IP address and any attribute combination, based on the number of data and initial probability value corresponding to the user-authorized trigger IP address and the attribute combination, determine the matching confidence of the candidate trigger data and conversion data corresponding to the attribute combination under the user-authorized trigger IP address.

[0096] Specifically, for any user-authorized trigger IP address and any attribute combination Attr, according to formula (2), the matching confidence P(Attr,ip) between the candidate trigger data and conversion data corresponding to the user-authorized trigger IP address and the attribute combination Attr is calculated based on the initial probability value of all combinations under the user-authorized trigger IP address and the attribute combination Attr:

[0097]

[0098] As can be seen from formula (2), if there are multiple candidate trigger data corresponding to the trigger IP address authorized by the user and the attribute combination Attr, then the matching confidence of these multiple candidate trigger data is the same.

[0099] Taking the user-authorized conversion device attribute information as (1,1,0,1) and the triggering IP address authorized by the same user, the triggering device attribute information of the 5 users authorized are (1,1,0,1), (1,1,0,1), (1,0,0,1), (1,0,1,1) and (0,1,1,1), respectively. The resulting attribute combinations are (1,1,0,1), (1,0,0,1) and (0,1,0,1). The matching confidence of the 5 candidate triggering data are P1, P1, P2, P2 and P3, respectively. Under the premise that the initial probability values ​​are different, P1, P2 and P3 are not equal.

[0100] S380. Based on the matching confidence level of each candidate trigger data, determine the target trigger data corresponding to the conversion data.

[0101] The technical solution described in this disclosure improves the efficiency of determining the distribution probability value by querying an attribute probability dictionary based on attribute combinations, thereby further improving channel attribution efficiency. By determining the number of candidate trigger data corresponding to any user-authorized trigger IP address and any attribute combination, and based on the distribution probability value of the attribute combinations, determining the initial probability value for matching different combinations of candidate trigger data with conversion data; and based on the number of data and the initial probability value corresponding to the user-authorized trigger IP address and attribute combination, determining the matching confidence level between candidate trigger data and conversion data corresponding to the attribute combination under the user-authorized trigger IP address. This achieves the calculation of matching confidence level under the same user-authorized trigger IP address and the same attribute combination, further improving the reliability of matching confidence level, thereby further improving the stability and reliability of conversion analysis results under short-link advertising and other promotional information delivery platforms, and thus increasing the number of promotional information delivery platforms.

[0102] Figure 4 This is a schematic diagram of a data analysis device provided in an embodiment of the present disclosure. The device can be implemented by software and / or hardware, and is generally integrated into an electronic device. It can execute data analysis methods to match more accurate target trigger data for transformed data. Figure 4 As shown, the device includes:

[0103] The conversion data acquisition module 410 is used to acquire conversion data generated when a preset operation corresponding to the promotion information is completed.

[0104] The candidate trigger data determination module 420 is used to parse the conversion data and determine multiple candidate trigger data that match the conversion data based on the parsing results; wherein, the candidate trigger data is the data obtained when the promotional information is triggered on different information publishing platforms;

[0105] The matching confidence determination module 430 is used to determine the matching confidence of each candidate trigger data and the conversion data;

[0106] The target trigger data determination module 440 is used to determine the target trigger data corresponding to the conversion data based on the matching confidence level corresponding to each candidate trigger data.

[0107] In some embodiments, the matching confidence determination module 430 includes:

[0108] The candidate trigger data parsing submodule is used to parse each candidate trigger data to obtain the user-authorized trigger device attribute information and the user-authorized trigger IP address corresponding to each candidate trigger data; among which, the device attribute information includes attribute information of multiple attribute dimensions;

[0109] The attribute combination determination submodule is used to determine at least one attribute combination with the same attribute information on the same attribute dimension based on the user-authorized conversion device attribute information and the user-authorized trigger device attribute information in the parsing results.

[0110] The probability distribution value determination submodule is used to determine the probability distribution value for each attribute combination;

[0111] The matching confidence determination submodule is used to determine the matching confidence between candidate trigger data and conversion data for any user-authorized trigger IP address and any attribute combination contained in each candidate trigger data, based on the distribution probability value of the attribute combination.

[0112] In some embodiments, the user-authorized conversion device attribute information includes device brand, device model, operating system type, and system version number; the user-authorized triggering device attribute information includes at least one of device brand, device model, operating system type, and system version number.

[0113] In some embodiments, the probability distribution value determination submodule is specifically used for:

[0114] Based on the attribute combination, the attribute probability dictionary is queried to determine the distribution probability value of the attribute combination; the attribute probability dictionary is obtained in advance by statistically analyzing the device attribute information authorized by multiple users.

[0115] In some embodiments, the apparatus further includes an attribute probability dictionary building module for pre-building an attribute probability dictionary in the following manner:

[0116] For any attribute combination, determine the number of attributes that match the attribute combination among the device attribute information authorized by multiple users, and determine the ratio of the number of attributes to the total number of device attribute information authorized by multiple users as the probability distribution value of the attribute combination.

[0117] In some embodiments, the matching confidence determination submodule is specifically used for:

[0118] For any user-authorized trigger IP address and any attribute combination, determine the number of candidate trigger data corresponding to the user-authorized trigger IP address, and based on the distribution probability value of the attribute combination, determine the initial probability value of matching candidate trigger data with conversion data for different combinations; wherein, the value range of the number of combinations is [1, number of data], and the number of combinations is a natural number;

[0119] For any user-authorized trigger IP address and any attribute combination, based on the number of data and initial probability values ​​corresponding to the user-authorized trigger IP address and attribute combination, determine the matching confidence of candidate trigger data and conversion data corresponding to the attribute combination under the user-authorized trigger IP address.

[0120] In some embodiments, the target trigger data determination module 440 is specifically used for:

[0121] The candidate trigger data corresponding to the highest matching confidence is determined as the target trigger data corresponding to the conversion data.

[0122] In some embodiments, the target trigger data determination module 440 is specifically used for:

[0123] If there are multiple maximum matching confidence levels among the matching confidence levels, then from the candidate trigger data corresponding to each maximum matching confidence level, determine the candidate trigger data whose data generation time is closest to the data generation time of the conversion data, and determine the selected candidate trigger data as the target trigger data corresponding to the conversion data.

[0124] In some embodiments, the candidate trigger data determination module 420 includes:

[0125] The conversion data parsing submodule is used to determine the user-authorized conversion IP address and user-authorized conversion device attribute information corresponding to the conversion data based on the parsing results.

[0126] The trigger data parsing submodule is used to obtain the trigger data generated when the promotion information in each information publishing platform is triggered, and to parse each trigger data to obtain the user-authorized trigger device attribute information and the user-authorized trigger IP address corresponding to each trigger data.

[0127] The candidate trigger data acquisition submodule is used to match conversion data and each trigger data based on the user-authorized conversion IP address, the user-authorized trigger IP address, the user-authorized conversion device attribute information, and the user-authorized trigger device attribute information to obtain each candidate trigger data that matches the conversion data.

[0128] In some embodiments, the candidate trigger data determination module 420 further includes a trigger data filtering submodule, used for:

[0129] After parsing each trigger data and obtaining the user-authorized trigger device attribute information and user-authorized trigger IP address corresponding to each trigger data, it is determined whether the user-authorized trigger IP address belongs to an abnormal IP category or an excessively large public IP category. The excessively large public IP category is an IP category in which the number of people sharing the IP address exceeds a set threshold.

[0130] If so, then remove the trigger data corresponding to the trigger IP address authorized by the user.

[0131] In some embodiments, the apparatus further includes an IP address classification module, for:

[0132] Before determining whether the triggering IP address authorized by the user belongs to the abnormal IP category or the excessive public IP category, multiple user-authorized IP addresses are collected, and statistical classification is performed on these multiple user-authorized IP addresses to obtain the abnormal IP category and the excessive public IP category. The abnormal IP category contains each abnormal IP address, and the excessive public IP category contains each excessive public IP address.

[0133] Accordingly, the trigger data filtering submodule is specifically used for:

[0134] The user-authorized trigger IP address is matched with each abnormal IP address and each excessively large public IP address;

[0135] If a match is found with an abnormal IP address, it is determined that the triggering IP address authorized by the user belongs to the abnormal IP category;

[0136] If a match is found with an excessively large public IP address, it is determined that the triggering IP address authorized by the user belongs to the excessively large public IP category.

[0137] The data analysis device provided in this disclosure improves the accuracy of conversion analysis of promotional information by analyzing the matching confidence level between each candidate trigger data and conversion data, and then selecting the target trigger data that best matches the conversion data based on the matching confidence level. It can also improve the recall rate of conversion analysis of promotional information by flexibly adjusting the confidence level threshold of candidate trigger data screening.

[0138] The data analysis apparatus provided in this disclosure can execute the data analysis method provided in any embodiment of this disclosure, and has the corresponding functional modules and beneficial effects for executing the method.

[0139] It is worth noting that in the embodiments of the above data analysis device, the various modules and sub-modules are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be achieved; in addition, the specific names of each functional module and functional sub-module are only for easy differentiation and are not used to limit the protection scope of this disclosure.

[0140] Figure 5 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this disclosure. Figure 5 As shown, the electronic device 500 includes one or more processors 501 and memory 502.

[0141] The processor 501 may be a central processing unit (CPU) or other form of processing unit with data processing capabilities and / or instruction execution capabilities, and may control other components in the electronic device 500 to perform desired functions.

[0142] The memory 502 may include one or more computer program products, which may include various forms of computer-readable storage media, such as volatile memory and / or non-volatile memory. Volatile memory may include, for example, random access memory (RAM) and / or cache memory. Non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc. One or more computer program instructions may be stored on the computer-readable storage medium, and the processor 501 may execute the program instructions to implement the data analysis method and / or other desired functions described in the embodiments of this disclosure. Various contents, such as attribute probability dictionaries, IP categories and their subordinate IP address fields, may also be stored in the computer-readable storage medium.

[0143] In one example, the electronic device 500 may further include an input device 503 and an output device 504, these components being interconnected via a bus system and / or other forms of connection mechanisms (not shown). The input device 503 may include, for example, a keyboard, a mouse, etc. The output device 504 may output various information to the outside, including target trigger data corresponding to converted data and its matching confidence level, etc. The output device 504 may include, for example, a display, a speaker, a printer, and a communication network and its connected remote output devices, etc.

[0144] Of course, to simplify, Figure 5 Only some of the components of the electronic device 500 relevant to this disclosure are shown, omitting components such as buses, input / output interfaces, etc. In addition, the electronic device 500 may include any other suitable components depending on the specific application.

[0145] In addition to the methods and apparatus described above, embodiments of this disclosure may also be computer program products, which include computer program instructions that, when executed by a processor, cause the processor to perform the steps of the data analysis method provided in any embodiment of this disclosure.

[0146] Computer program products can be written in any combination of one or more programming languages ​​to perform the operations of embodiments of this disclosure. The programming languages ​​include object-oriented programming languages ​​such as Java and C++, as well as conventional procedural programming languages ​​such as C or similar languages. The program code can be executed entirely on a user's computing device, partially on a user's computing device, as a standalone software package, partially on a user's computing device and partially on a remote computing device, or entirely on a remote computing device or server.

[0147] Furthermore, embodiments of this disclosure may also be computer-readable storage media storing computer program instructions that, when executed by a processor, cause the processor to perform the steps of the data analysis method provided in any embodiment of this disclosure.

[0148] The aforementioned computer-readable storage medium may be any combination of one or more readable media. A readable medium may be a readable signal medium or a readable storage medium. A readable storage medium may, for example, include, but is not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatuses, or devices, or any combination thereof. More specific examples of readable storage media (a non-exhaustive list) include: electrical connections having one or more wires, portable disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0149] It should be noted that the terminology used in this disclosure is for the purpose of describing specific embodiments only and is not intended to limit the scope of this application. As shown in this disclosure and the claims, unless the context clearly indicates otherwise, the words "a," "an," "an," and / or "the" are not specifically singular and may include plural forms. The term "and / or" includes any one and all combinations of one or more of the associated listed items. The terms "comprising," "including," or any other variations thereof are intended to cover a non-exclusive inclusion, such that a process, method, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, or apparatus. Without further limitations, an element defined by the phrase "comprising an..." does not exclude the presence of other identical elements in the process, method, or apparatus that includes said element.

[0150] The above description is merely a specific embodiment of this disclosure, enabling those skilled in the art to understand or implement it. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this disclosure. Therefore, this disclosure is not to be limited to the embodiments described herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A data analysis method, characterized in that, include: Retrieve conversion data generated when the preset operation corresponding to the promotion information is completed; The conversion data is analyzed, and multiple candidate trigger data that match the conversion data are determined based on the analysis results; wherein, the trigger data is the data obtained when the promotional information is triggered on different information publishing platforms; Determine the matching confidence level between each candidate trigger data and the conversion data; wherein, the matching confidence level is determined based on the attribute combination of the trigger IP address and device attribute information corresponding to each candidate trigger data, and the attribute combination is at least one attribute combination with the same attribute information on the same attribute dimension determined based on the conversion device attribute information in the parsing result and the trigger device attribute information corresponding to each candidate trigger data; Based on the matching confidence level corresponding to each of the candidate trigger data, the target trigger data corresponding to the conversion data is determined.

2. The method according to claim 1, characterized in that, Determining the matching confidence level between each candidate trigger data and the conversion data includes: Parse each of the candidate trigger data to obtain the trigger device attribute information and trigger IP address corresponding to each candidate trigger data; wherein, the device attribute information includes attribute information of multiple attribute dimensions; Based on the conversion device attribute information and each triggering device attribute information in the parsing results, at least one attribute combination with the same attribute information on the same attribute dimension is determined. Determine the probability distribution value for each of the attribute combinations; For any triggering IP address and any attribute combination contained in each of the candidate triggering data, the matching confidence level between the candidate triggering data and the conversion data corresponding to the attribute combination under the triggering IP address is determined based on the distribution probability value of the attribute combination.

3. The method according to claim 2, characterized in that, Determining the probability distribution value of the attribute combination includes: Based on the attribute combination, the attribute probability dictionary is queried to determine the distribution probability value of the attribute combination; wherein, the attribute probability dictionary is obtained in advance by statistically analyzing multiple device attribute information.

4. The method according to claim 3, characterized in that, The attribute probability dictionary is pre-built in the following manner: For any of the attribute combinations, determine the number of attributes in the plurality of device attribute information that are consistent with the attribute combination, and determine the ratio of the number of attributes to the total number of the plurality of device attribute information as the probability distribution value of the attribute combination.

5. The method according to claim 2, characterized in that, The step of determining the matching confidence level between the candidate trigger data and the conversion data corresponding to the attribute combination under the trigger IP address, based on the probability distribution value of the attribute combination, for any trigger IP address and any attribute combination contained in each of the candidate trigger data, includes: For any of the aforementioned triggering IP addresses and any of the aforementioned attribute combinations, determine the number of candidate triggering data corresponding to the triggering IP address, and based on the distribution probability value of the attribute combination, determine the initial probability value for matching different combinations of candidate triggering data with the converted data; wherein, the value range of the number of combinations is [1, the number of data], and the number of combinations is a natural number; For any of the aforementioned triggering IP addresses and any of the aforementioned attribute combinations, based on the number of data corresponding to the triggering IP address and the attribute combination and the initial probability value, the matching confidence level between the candidate triggering data and the conversion data corresponding to the attribute combination under the triggering IP address is determined.

6. The method according to claim 1, characterized in that, The step of determining the target trigger data corresponding to the conversion data based on the matching confidence level corresponding to each of the candidate trigger data includes: The candidate trigger data corresponding to the maximum matching confidence is determined as the target trigger data corresponding to the conversion data.

7. The method according to claim 1, characterized in that, The step of determining the target trigger data corresponding to the conversion data based on the matching confidence level corresponding to each of the candidate trigger data includes: If there are multiple maximum matching confidence scores among the matching confidence scores, then from the candidate trigger data corresponding to each maximum matching confidence score, the candidate trigger data whose data generation time is closest to the data generation time of the conversion data is selected, and the selected candidate trigger data is determined as the target trigger data corresponding to the conversion data.

8. The method according to claim 1, characterized in that, The multiple candidate trigger data for the promotional information that matches the conversion data, determined based on the analysis results, include: Based on the parsing results, the conversion IP address and conversion device attribute information corresponding to the conversion data are determined; The system obtains the trigger data generated when the promotional information in each of the information publishing platforms is triggered, and parses each trigger data to obtain the trigger device attribute information and trigger IP address corresponding to each trigger data. Based on the conversion IP address, each of the trigger IP addresses, the conversion device attribute information, and each of the trigger device attribute information, the conversion data and each of the trigger data are matched to obtain each of the candidate trigger data that matches the conversion data.

9. The method according to claim 8, characterized in that, After parsing each of the trigger data to obtain the trigger device attribute information and trigger IP address corresponding to each trigger data, the method further includes: Determine whether the triggering IP address belongs to an abnormal IP category or an excessively large public IP category, wherein the excessively large public IP category is an IP category in which the number of people sharing the IP address exceeds a set threshold; If so, then the trigger data corresponding to the triggering IP address will be removed.

10. The method according to claim 9, characterized in that, Before determining whether the triggering IP address belongs to an abnormal IP category or an excessively large public IP category, the method further includes: Collect multiple IP addresses and statistically classify them to obtain the abnormal IP category and the excessive public IP category. The abnormal IP category contains each abnormal IP address, and the excessive public IP category contains each excessive public IP address. The determination of whether the triggering IP address belongs to an abnormal IP category or an excessively large public IP category includes: The triggering IP address is matched with each of the abnormal IP addresses and each of the excessively large public IP addresses; If a match is found with the abnormal IP address, then the triggering IP address is determined to belong to the abnormal IP category; If a match is found with the excessively large public IP address, then the triggering IP address is determined to belong to the excessively large public IP category.

11. A data analysis device, characterized in that, include: The conversion data acquisition module is used to acquire conversion data generated when a preset operation corresponding to the promotion information is completed; The candidate trigger data determination module is used to parse the conversion data and determine multiple candidate trigger data that match the conversion data based on the parsing results; wherein, the candidate trigger data is data obtained when the promotional information is triggered on different information publishing platforms; The matching confidence determination module is used to determine the matching confidence of each candidate trigger data and the conversion data; wherein, the matching confidence is determined based on the attribute combination of the trigger IP address and device attribute information corresponding to each candidate trigger data, and the attribute combination is at least one attribute combination with the same attribute information on the same attribute dimension determined based on the conversion device attribute information in the parsing result and the trigger device attribute information corresponding to each candidate trigger data; The target trigger data determination module is used to determine the target trigger data corresponding to the conversion data based on the matching confidence level corresponding to each candidate trigger data.

12. An electronic device, characterized in that, The electronic device includes: Processor and memory; The processor executes the steps of the method as described in any one of claims 1 to 10 by invoking programs or instructions stored in the memory.

13. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a program or instructions that cause a computer to perform the steps of the method as described in any one of claims 1 to 10.

Images