Method and system for matching identity information using probabilistic prediction model

Abstract

One embodiment of the present disclosure provides a user identity matching method. The method comprises the steps of: collecting online behavioral data of a user and analyzing the pattern of said online behavioral data; calculating a matching probability between said user's device data and said online behavioral data based on a pre-established probabilistic prediction model; and linking said user's existing data with additionally collected data of the user based on said matching probability.

Description

Technology Field

[0001] The present disclosure relates to a method and system for matching identity information using a probabilistic prediction model, and more specifically, to a method and system for matching data collected from a device with user data by analyzing patterns in user online behavior data and calculating the matching probability of the same user based on whether device data and user online behavior data match using a pre-established probabilistic prediction model. Background Technology

[0002] As users' behavior shifts from desktops and laptops to mobile devices, it has become crucial in the digital marketing market to collect behavioral data from users accessing from multiple devices and analyze it as the behavior of a single user in order to accurately identify customer interests.

[0003] Traditionally, user behavior data was identified as a single user based on web browser cookies or mobile device app identifiers (AD IDs). However, with the recent introduction of policies to strengthen personal information protection, such as restricting the use of cookies in browsers and limiting app tracking, there is a problem in that it is difficult to identify a single user using existing methods.

[0004] Therefore, there is an increasing need for identity information matching methods by utilizing the company's own customer data or linking customer data collected through user consent. The problem to be solved

[0005] The present disclosure aims to solve the problems of the aforementioned prior art by providing a method and system for matching data collected from a device with user data by analyzing patterns in user online behavior data and calculating the matching probability of the same user based on whether device data and user online behavior data match using a pre-established probabilistic prediction model.

[0006] The technical problems that the present disclosure aims to solve are not limited to the technical problems described above, and other technical problems of the present disclosure may be derived from the following description. means of solving the problem

[0007] As a technical means for solving the aforementioned technical problem, an embodiment according to the first aspect of the present disclosure provides a user identity matching method. The method comprises the steps of: collecting online behavioral data of a user and analyzing the pattern of said online behavioral data; calculating a matching probability between said user's device data and said online behavioral data based on a pre-established probabilistic prediction model; and linking said user's existing data with additionally collected data of the user based on said matching probability.

[0008] Additionally, an embodiment according to a second aspect of the present disclosure provides a user identity matching system. The system comprises a communication module, at least one processor, and a memory electrically connected to the processor and storing at least one code executed by the processor. The memory stores a code that, when executed through the processor, causes the processor to collect online behavior data of a user, analyze patterns of the online behavior data, calculate a matching probability between the user's device data and the online behavior data based on a preset probabilistic prediction model, and link the user's existing data with additionally collected data of the user based on the matching probability. Effects of the invention

[0009] According to the present disclosure, the identity of a user can be interconnected from multiple data by determining that they are the same user even if their email addresses are different.

[0010] In addition, according to the present disclosure, full integration of user profiles can be achieved by interconnecting user identities from multiple data.

[0011] The effects of the present disclosure are not limited to the effects described above and include all effects understood from the following description. Brief explanation of the drawing

[0012] FIGS. 1 and FIGS. 2 are drawings illustrated to explain a user identity matching system according to one embodiment of the present disclosure. Figure 3 is a drawing illustrating the detailed configuration according to an example of the server shown in Figure 1. Figure 4 is a diagram illustrating an example of clustering analysis among data distributions. Figure 5 is a diagram illustrating an example of a diagram linking data pairs based on comprehensive weights. Figure 6 is a diagram illustrating the detailed configuration according to another example of the server shown in Figure 1. FIG. 7 is a flowchart illustrating the sequence of a user identity matching method according to another embodiment of the present disclosure. Figure 8 is a diagram illustrating detailed steps for some steps of the user identity matching method shown in Figure 7. Specific details for implementing the invention

[0013] The present disclosure will be described in detail below with reference to the attached drawings. However, the present disclosure may be implemented in various different forms and is not limited to the embodiments described herein. Furthermore, the attached drawings are intended only to facilitate understanding of the embodiments disclosed herein, and the technical concept disclosed herein is not limited by the attached drawings. All terms used herein, including technical and scientific terms, should be interpreted in the sense generally understood by those skilled in the art to which the present disclosure pertains. Terms defined in advance should be interpreted as having additional meanings consistent with relevant technical literature and the present disclosure, and should not be interpreted in a highly ideal or restrictive sense unless otherwise defined.

[0014] In order to clearly explain the present disclosure in the drawings, parts unrelated to the explanation have been omitted, and the size, form, and shape of each component shown in the drawings may be varied. Throughout the specification, identical or similar parts are denoted by identical or similar reference numerals.

[0015] In the following description, suffixes such as "module" and "part" for components are assigned or used interchangeably solely for the ease of drafting the specification, and do not inherently possess distinct meanings or roles. Furthermore, in describing the embodiments disclosed in this specification, detailed descriptions of related prior art have been omitted where it is determined that such detailed descriptions could obscure the essence of the embodiments disclosed in this specification.

[0016] Throughout the specification, when it is stated that a part is "connected (connected, contacted, or coupled)" to another part, this includes not only cases where they are "directly connected (connected, contacted, or coupled)," but also cases where they are "indirectly connected (connected, contacted, or coupled)" with other members interposed therebetween. Furthermore, when it is stated that a part "includes (provides, or provides)" a certain component, this means that, unless specifically stated otherwise, it does not exclude other components but rather allows for additional "included (provided, or provided)" of other components.

[0017] Terms indicating ordinal numbers, such as first, second, etc., used herein are used solely for the purpose of distinguishing one component from another and do not limit the order or relationship of the components. For example, the first component of the present disclosure may be named the second component, and similarly, the second component may be named the first component. Singular forms used herein should be interpreted to include plural forms unless explicitly to the contrary.

[0018] FIGS. 1 and FIGS. 2 are drawings illustrated to explain a user identity matching system according to one embodiment of the present disclosure.

[0019] Referring to FIGS. 1 and 2, a user identity matching system may include a server (110) and a user terminal (120). More specifically, the user identity matching system may include a server (110) comprising a connection record collection device (240), a connection data processing device (250), a same user inference device (260), and a user identity storage device (270), and a user terminal (120) such as a PC (210), a laptop (220), and a mobile (230).

[0020] The server (110) collects online behavior data of the user and analyzes the patterns of the online behavior data. For example, the online behavior data may include at least one of web log data, app log data, event data, and user log data. For example, the online behavior data may include information such as website page views, scrolls, clicks, the user's IP address, device type, browser, OS, etc., and data regarding at least one of location information, time of use, visited page information, access history, and search history.

[0021] The server (110) calculates the matching probability between the user's device data and online behavior data based on a pre-configured probabilistic prediction model. For example, the device data may include data such as the type of device the user is connected to, OS, browser version, layout engine, and setting language.

[0022] The server (110) links the user's existing data with additionally collected user data based on matching probability.

[0023] The user terminal (120) can be connected to the server (110) via a communication network. The user terminal (120) may refer to any type of handheld wireless communication device, such as a laptop, desktop, laptop equipped with a web browser, a wireless communication device that ensures portability and mobility, or a smartphone, tablet PC, etc.

[0024] The user terminal (120) can transmit the user's data and the user's online behavior data to the server (110).

[0025] FIG. 3 is a diagram illustrating the detailed configuration according to an example of the server shown in FIG. 1. FIG. 4 is a diagram illustrating an example of clustering analysis among data distributions. FIG. 5 is a diagram illustrating an example of a diagram linking data pairs based on comprehensive weights.

[0026] Referring to FIGS. 3 to 5, the server (110) may include a communication module (111), a processor (112), and a memory (113).

[0027] The communication module (111) may include a device comprising hardware and software necessary to transmit and receive signals, such as control signals or data signals, through a wired or wireless connection with another network device.

[0028] The communication module (111) can receive user data and user online behavior data from the user terminal.

[0029] The processor (112) may include various types of devices for controlling and processing data. The processor (112) may refer to a data processing device embedded in hardware having a physically structured circuit to perform a function expressed by code or instructions included in a program.

[0030] In one example, the processor (112) may be implemented in the form of a microprocessor, a central processing unit (CPU), a processor core, a multiprocessor, an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA), etc., but the scope of the invention is not limited thereto.

[0031] The processor (112) performs operations according to the code stored in memory (113).

[0032] The memory (113) can store at least one of the information and data input to the communication module (111), the information and data required for the function performed by the processor (112), and the data generated according to the execution of the processor (112).

[0033] Memory (113) should be interpreted as a general term for non-volatile storage devices that continue to maintain stored information even when power is not supplied, and volatile storage devices that require power to maintain stored information. In addition to volatile storage devices that require power to maintain stored information, memory (113) may include magnetic storage media or flash storage media, but the scope of the present invention is not limited thereto.

[0034] Memory (113) is electrically connected to the processor (112) and stores at least one code executed by the processor (112). Memory (113) stores code that causes the processor (112) to perform the following functions and procedures when executed through the processor (112).

[0035] The memory (113) stores code that causes the user's online behavior data to be collected in real time and the patterns of the online behavior data to be analyzed. For example, the online behavior data includes online behavior data such as user interactions such as page views, clicks, searches, and purchases, and may include system information such as the user's IP address, device type, browser, and OS, as well as data on at least one of location information, time of use, visited page information, access history, and search history.

[0036] In memory (113), code that causes to organize tabular data representing behavior in order to analyze patterns of online behavior data may be stored. For example, each row of the table may be at least one of a user ID, a timestamp, a type of behavior, and a page URL.

[0037] An example of data organization in table format is shown in Table 1 below.

[0038] session user Connecting device IP Connection location Connection time event Access Page element Search terms 123456 789 Windows 10 192.168.1.1 Nonhyeon-dong, Gangnam-gu 2023-05-01T10:15:30Z page_view https: / / example.com / page1 123456 789 Windows 10 192.168.1.1 Nonhyeon-dong, Gangnam-gu 2023-05-01T10:15:30Z click button1 123456 789 Windows 10 192.168.1.1 Nonhyeon-dong, Gangnam-gu 2023-05-01T10:15:30Z search keyword

[0039] Memory (113) may store code that causes the distribution of user behavior to be analyzed based on user online behavior data such as page views, clicks, and searches. Memory (113) may store code that causes the calculation of summary statistics of user behavior indicators such as session duration, page views per session, and time spent on each page.

[0040] Memory (113) may store code that groups user interactions by user ID to analyze behavioral patterns within the same user.

[0041] In memory (113), code may be further stored to cause the user to determine whether the device currently used by the user is the same as the device previously used by the user based on the connection records among online behavior data. For example, memory (113) may determine, based on the stored code, whether at least one of the current user's IP address, device type, OS, browser, and language settings is the same as the previous user's IP address, device type, OS, browser, and language settings.

[0042] The memory (113) stores code that causes the user's device data and online behavior data to be matched based on a preset probabilistic prediction model. For example, the device data may include the user's previous online behavior data, the type of device, and the product name stored on the device the user used to access.

[0043] For example, the memory (113) may store code that converts device data and online behavior data into a dictionary and defines a function to calculate Jaccard similarity, thereby causing the user's online behavior data to be repeated to calculate the matching probability between the user's device data and online behavior data.

[0044] Here, Jaccard similarity is an indicator that measures the similarity between two sets and can be calculated by dividing the intersection by the union. In this case, the two sets may be actions performed by the user and actions associated with the device. The association between user and device data can be evaluated by calculating the Jaccard similarity between the user's online behavior data and the user's device ID. High Jaccard similarity may indicate a high probability that the device and online behavior data are associated with the same user.

[0045] An example of calculating the matching probability between device data and online behavior data is as follows.

[0046] import pandas as pd

[0047] # Device data

[0048] device_data = pd.DataFrame({

[0049] 'user_id': [1, 2, 3],

[0050] 'device_id': ['device1', 'device2', 'device3']

[0051] })

[0052] # Online Behavior Data

[0053] behavior_data = pd.DataFrame({

[0054] 'user_id': [1, 2, 3, 1, 2, 3, 1, 2, 3],

[0055] 'action_type': ['page_view', 'page_view', 'page_view', 'click', 'click', 'search', 'page_view', 'click', 'page_view']

[0056] })

[0057] # Convert device data to a dictionary

[0058] device_dict = device_data.set_index('user_id')['device_id'].to_dict()

[0059] # Define a function to calculate Jaccard similarity

[0060] def jaccard_similarity(device_id, user_behavior):

[0061] device_set = set(device_id)

[0062] behavior_set = set(user_behavior)

[0063] intersection = len(device_set. intersection(behavior_set))

[0064] union = len(device_set.union(behavior_set))

[0065] return intersection / union

[0066] # Calculate Jaccard similarity for each user

[0067] user_similarity = behavior_data.groupby('user_id')['action_type'].apply(lambda x: jaccard_similarity(device_dict[x.name], x)).reset_index()

[0068] print(user_similarity)

[0069] In addition, the pre-established probabilistic prediction model may be an Expectation-Maximization (EM) algorithm based on the Fellegi-Sunter data linkage model. Here, the Fellegi-Sunter data linkage model may be a method that approaches probabilistically by scoring the degree of agreement of the data.

[0070] In memory (113), code that causes the user's existing data and additionally collected user data to be linked based on matching probability may be stored.

[0071] Memory (113) may store code that calculates the combined weight of a probabilistic prediction model based on matching probability and causes data pairs to be determined as linked, potentially linked, and non-linked. For example, data pairs having a combined weight greater than a preset threshold can be classified as linked.

[0072] Here, the composite weight may be the sum of the match weights of the individual items of the data pair being compared. The match weight can be calculated by estimating the m-probability and u-probability and may vary for each item. For example, the m-probability represents the error rate of the data and the u-probability represents the probability of a coincidence; the match weight is defined as the ratio of the m-probability to the u-probability and may be in the form of a logarithm for computational convenience.

[0073] For example, in the case of gender, the error rate is low, so the m-probability can be considered as 0.95, and the u-probability, which is the probability of a coincidence, can be 1 / 2. Calculated using a base-2 logarithm, the match weight when a match occurs is 0.92, and the match weight when a match does not occur is -3.32. In this way, the composite weight can be calculated by summing the match weights for agreement (+) and non-agreement (-) for the items of the data being compared. The numerical value of the composite weight may vary depending on the data being compared. Since the weight has a (+) value when the linked variable fields between two records match and a (-) value when they do not, a higher composite weight may indicate a higher probability of them being the same entity.

[0074] Additionally, the memory (113) may store code that causes a threshold value to be determined for classifying the calculated total weights as linked or non-linked by checking the distribution of the calculated total weights. For example, the threshold value is calculated as an appropriate value by determining the number of records and the frequency of the total weights, and can be adjusted through a verification process.

[0075] The memory (113) may include an initialization process, an expected value calculation process, a maximization process, and an iteration process to calculate the total weight.

[0076] The initialization process may be the process of assigning initial probability values ​​to all record pairs and setting initial parameters of the probability model (e.g., linked probability, unlinked probability).

[0077] The process of calculating the expected value may be a process of calculating the probability that a pair of records represents the same entity using the current parameter estimate for each pair of records. Here, the calculated probability can be interpreted as the posterior probability for each pair of records.

[0078] The maximization process may be a process of updating parameters based on probabilities calculated during the expectation calculation process. Here, parameters are parameters of the probability model and can be classified into linked probabilities, potential linked probabilities, and unlinked probabilities.

[0079] The iterative process may be a process of repeating the expectation calculation and maximization processes until the parameter estimates converge. The iterative process may terminate when a certain criterion is reached (e.g., the change in log-likelihood becoming small).

[0080] Memory (113) may store code that checks the linkage results of data pairs to set a threshold and causes the verification of a probabilistic prediction model. For example, data pairs may be two data sets randomly selected to compare whether they are data of the same user from a data set.

[0081] Memory (113) can store linked data pairs and code that causes them to be linked with additionally collected user data.

[0082] For example, assuming that there are mobile app users 1, 2, 3, ���, n and website visitors A1, B4, C1, D0, ���, NN based on code stored in memory (113), mobile app user 1 can generate and store a diagram linking data pairs based on a comprehensive weight calculated probabilistically for identification variables such as device type, browser type, IP address, and OS, which is 65% with website visitor A1, 77% with website visitor C1, and 21% with website visitor D0.

[0083] Figure 6 is a diagram illustrating the detailed configuration according to another example of the server shown in Figure 1.

[0084] Referring to FIG. 6, the server may include a connection record collection unit (410), a connection data processing unit (420), an identity inference unit (430), an identity linkage unit (440), a prediction model management unit (450), and an identity storage unit (460).

[0085] The connection record collection unit (410) can collect system information such as the user's IP address, device type, browser, and OS, as well as behavioral data of the online user such as location information, usage time, visited page information, and search history in real time.

[0086] The connection data processing unit (420) can distinguish the same device from the system information in the connection records and check the pattern of the behavior data.

[0087] The identity inference unit (430) can infer whether device data and behavior data are the same user by using a probabilistic prediction model. For example, the prediction model may be based on the Fellegi-Sunter data linkage model and may use the Expectation-Maximization (EM) algorithm, and the Fellegi-Sunter model may be a method that approaches probabilistically by scoring the degree of data agreement.

[0088] The identity linkage unit (440) can determine a data pair as linked, potentially linked, or non-linked by calculating a comprehensive weight based on the calculated matching score. For example, a certain cutoff value or threshold value can be set, and a data pair having a comprehensive weight greater than this value can be classified as linked.

[0089] The prediction model management unit (450) checks the linkage results of the identity linkage unit (440) to adjust the threshold and verify the prediction model.

[0090] The prediction model management unit (450) may be a module for finding the optimal threshold value. The prediction model management unit (450) can check the degree of linkage of data pairs while adjusting the threshold value. Additionally, the prediction model management unit (450) may be provided in the form of a dashboard to check the source of the data and the actual data value so as to verify that the linked data pairs are the same user.

[0091] The prediction model management unit (450) can make the prediction model more sophisticated by setting rules for the prediction model, and by setting rules for the combination of data items that measure the sensitivity and accuracy of data items in the previously verified linkage results and compare them.

[0092] The identity storage unit (460) stores the linked data pair determined by the identity linkage unit (440) and can link with additional collected user data.

[0093] The identity storage unit (460) can connect additionally collected user data using a deterministic matching method through linked data pairs determined to be the same user. For example, the deterministic matching method is a method of determining that two records are the same user when the data exactly matches by comparing two records, and the deterministic data linkage matching rate can be increased as the number of linked data pairs increases.

[0094] FIG. 7 is a flowchart illustrating the sequence of a user identity matching method according to another embodiment of the present disclosure.

[0095] The user identity matching method described below may be performed by the user identity matching system or server (110 in FIG. 1) described above with reference to FIG. 1 to 6. Accordingly, the description of the embodiment of the present disclosure described above with reference to FIG. 1 to 6 may be equally applied to the embodiment described below, and any content that overlaps with the description above will be omitted. The steps described below do not necessarily have to be performed in order, the order of the steps can be set in various ways, and the steps may be performed almost simultaneously.

[0096] Referring to FIG. 7, the user identity matching method includes a step of analyzing patterns of online behavior data (S100), a step of calculating the matching probability between device data and online behavior data (S200), and a step of linking the user's existing data with additionally collected user data (S300).

[0097] The online behavioral data pattern analysis step (S100) is a step of collecting online behavioral data of a user and analyzing the patterns of the online behavioral data. For example, the online behavioral data pattern analysis step (S100) may be a step of collecting system information such as the user's IP address, device type, browser, and OS, and user behavioral data such as location information, usage time period, visited page information, and search history, and distinguishing the same device from the collected data and confirming the patterns of the behavioral data.

[0098] The step of calculating the matching probability between device data and online behavior data (S200) is a step of calculating the matching probability between the user's device data and online behavior data based on a pre-established probabilistic prediction model. For example, the step of calculating the matching probability between device data and online behavior data (S200) may be a step of calculating the matching probability of the same user by analyzing the pattern of the behavior data.

[0099] The step of linking the user's existing data with additionally collected user data (S300) is a step of linking the user's existing data with additionally collected user data based on a matching probability. For example, the step of linking the user's existing data with additionally collected user data (S300) may be a step of connecting to a single user based on a matching probability and storing data where the same identity has been confirmed.

[0100] Figure 8 is a diagram illustrating detailed steps for some steps of the user identity matching method shown in Figure 7.

[0101] Referring to FIG. 8, the step of linking the user's existing data with additionally collected user data (S300) may include a step of calculating the comprehensive weights of a probabilistic prediction model (S310) and a step of storing the linked data pairs (S320).

[0102] The step of calculating the comprehensive weight of the probabilistic prediction model (S310) may be a step of calculating the comprehensive weight of the probabilistic prediction model based on matching probabilities and determining a data pair as at least one of linked, potential linked, and non-linked. For example, in the step of calculating the comprehensive weight of the probabilistic prediction model (S310), if the comprehensive weight is greater than a preset threshold, the data pair may be determined as linked.

[0103] The linked data pair storage step (S320) may be a step of storing the linked data pair and linking the user's existing data with additionally collected user data. For example, the linked data pair storage step (S320) may store the data pair determined to be linked and link it with additionally collected user data.

[0104] Those skilled in the art to which this disclosure pertains will understand that, based on the foregoing description, other specific forms can be easily modified without altering the technical spirit or essential features of this disclosure. Therefore, the embodiments described above should be understood as illustrative in all respects and not restrictive. The scope of this disclosure is defined by the claims set forth below, and all modifications or variations derived from the meaning and scope of the claims and equivalents thereof should be interpreted as being included within the scope of this disclosure. The scope of this application is defined by the claims set forth below rather than by the foregoing detailed description, and all modifications or variations derived from the meaning and scope of the claims and equivalents thereof should be interpreted as being included within the scope of this application.

Claims

Claim 1 A method for matching user identity performed by a user identity matching system, comprising: a) collecting online behavioral data of a user and analyzing the pattern of said online behavioral data; b) calculating a matching probability between said user's device data and said online behavioral data based on a pre-established probabilistic prediction model; and c) linking said user's existing data with additionally collected data of said user based on said matching probability, wherein step c) comprises: calculating a comprehensive weight of the probabilistic prediction model based on said matching probability and determining a data pair as at least one of linked, potentially linked, and non-linked; and storing said linked data pair and linking said user's existing data with additionally collected data of said user. Claim 2 A user identity matching method according to claim 1, wherein the online behavior data includes data regarding at least one of system information such as the user’s IP address, device type, browser, OS, location information, usage time zone, visited page information, access history, and search history. Claim 3 A user identity matching method according to claim 1, wherein step a) comprises the steps of converting the user's online behavior data into a table, analyzing the user's behavior distribution based on the online behavior data, and calculating the user's behavior indicator based on the online behavior data. Claim 4 A user identity matching method according to claim 1, wherein step b) includes a step of determining whether the device currently used by the user is the same as the device previously used by the user based on a connection record among the online behavior data. Claim 5 A user identity matching method according to claim 1, wherein the above-mentioned pre-established probabilistic prediction model is an Expectation-Maximization (EM) algorithm based on a pre-established data linkage model. Claim 6 delete Claim 7 A user identity matching system comprising: a communication module; at least one processor; and a memory electrically connected to the processor and storing at least one code executed by the processor, wherein when the memory is executed through the processor, the processor collects online behavior data of a user, analyzes the pattern of the online behavior data, calculates a matching probability between the user's device data and the online behavior data based on a preset probabilistic prediction model, links the user's existing data with additionally collected data of the user based on the matching probability, calculates a composite weight of the probabilistic prediction model based on the matching probability, determines the data pair as linked, potentially linked, and non-linked, stores the linked data pair, and stores a code that causes the user's existing data and additionally collected data of the user to be linked. Claim 8 A user identity matching system according to claim 7, wherein the online behavior data includes system information such as the user’s IP address, device type, browser, OS, etc., location information, usage time zone, visited page information, access history, and search history, and at least one of these. Claim 9 A user identity matching system according to claim 7, wherein the memory stores code that causes the processor to convert the user's online behavior data into a table, analyze the user's behavior distribution based on the online behavior data, and calculate the user's behavior indicator based on the online behavior data. Claim 10 In claim 7, the memory stores code that causes the processor to determine whether the device used by the user is the same as the device previously used by the user based on the connection record among the online behavior data, a user identity matching system. Claim 11 In claim 7, the above-mentioned pre-established probabilistic prediction model is an Expectation-Maximization (EM) algorithm based on a pre-established data linkage model, a user identity matching system. Claim 12 delete

Images