RU2795667C1 - User internet traffic analysis system for user classification coefficient assignment to user device - Google Patents

Abstract

FIELD: computing technology.

SUBSTANCE: user's Internet traffic analysis system for assigning a user classification coefficient to the user device is comprised of a server configured to establish an Internet connection with the user's computing device to analyse the hardware of the user's computing device and assign the user's computing device a coefficient of the user's computing device hardware; establish an Internet connection with the user's computing device to analyse the parameters of the Internet connection, assigning to the user's computing device an Internet connection coefficient of the user's computing device; establish an Internet connection with the user's computing device to analyse the user's behaviour in the Internet session, assigning the user's computing device a user behavioural coefficient; calculate and assign a user classification coefficient to the user device.

EFFECT: increase of accuracy of user classification.

6 cl, 1 dwg

Description

Technical field:

[0001] The invention relates to the field of computer technology, in particular, to a user's Internet traffic monitoring system for assigning a user classification coefficient to a user device.

State of the art:

[0002] Currently, there are many systems for monitoring Internet traffic. One example of such solutions is a method for monitoring Internet traffic by third parties, as described in patent application US 2009/0019148 A1. The known solution discloses a method for monitoring Internet traffic, which includes an analysis by a network service provider of an HTTP transaction involving an Internet user's client. The network service provider responds to the HTTP transaction by sending an intermediate web page containing the monitoring tool to the Internet user's client. After forwarding the intermediate web page to the Internet user's client, the network service provider forwards the web content originally associated with the HTTP transaction to the Internet user's client.

[0003] However, the known solution has disadvantages. The disadvantage of the known solution is that it has a low accuracy of user classification for further provision of advertising content to him. This disadvantage is due to the fact that the well-known classification solution does not provide for assigning a classification coefficient to the user device, and also does not analyze the hardware of the user's computing device, does not take into account the Internet connection parameters for providing media content, and does not analyze the user's behavior as a whole.

Disclosure of the invention:

[0004] The objective of the invention is to eliminate the above disadvantages.

[0005] The technical result is to increase the accuracy of user classification by calculating the user classification coefficient for its further assignment to the user device.

[0006] To achieve a technical result, a system for analyzing user Internet traffic is proposed for assigning a user classification coefficient to a user device, containing a server configured to: establish an Internet connection with at least one user computing device to analyze the hardware of the user computing device with assigning the user computing device a user computing device hardware factor (factor dq1), where the hardware analysis includes identifying the user computing device display matrix, identifying the user computing device video card, identifying the user computing device RAM, identifying the manufacturer of the user computing device, and identifying the secondary characteristics of the user computing device. user devices; establishing an Internet connection with at least one user computing device to analyze the parameters of the Internet connection with assigning to the computing device of the user the coefficient of the Internet connection of the user's computing device (coefficient iq1), where the parameters of the Internet connection include the speed of the Internet connection, the type of connection of the user's computing device to the network Internet, the name of the Internet provider, DNS characteristics, the presence and type of proxy connection, statistics on the activity of using this network segment; establishing an Internet connection with at least one user computing device to analyze user behavior in an Internet session with assigning a user behavioral coefficient (pv1 coefficient) to the user computing device, where user behavior analysis includes identification of Internet sites with the highest user traffic, identification of the amount of Internet connections for a given period of time from the user's device, characteristics of the use of "hot keys", characteristics of the speed and completeness of reading an online page, characteristics of switching between pages, printing speed; and calculating and assigning to the user device a user classification coefficient (coefficient ei1), where calculation of the coefficient ei1 is performed as a function of f from the coefficients dq1, iq1, pv1: ei1= f(dq1, iq1, pv1).

[0007] Additionally, secondary characteristics of the user's computing device include at least one of: the model year of the user's computing device, camera factory specifications, Internet browser version, or combinations thereof.

[0008] Additionally, the type of connection of the user's computing device to the Internet includes at least one of the following connections: Wi-Fi, 3G, 4G, LTE-A, or combinations thereof.

[0009] Additionally, the calculation of the coefficient ei1 as a function of f from the coefficients dq1, iq1, pv1 is performed by machine learning using at least one of the following algorithms: decision trees, GLM/GLZ, gradient boosting.

[0010] Additionally, the server is configured to establish an Internet connection with at least one user computing device to analyze the user computing device software by assigning the user computing device software factor to the user computing device (coefficient pq1).

[0011] Additionally, the server is configured to calculate and assign to the user device a user classification coefficient (coefficient ei1), where the coefficient ei1 is calculated as a function of f from the coefficients dq1, iq1, pv1, pq1: ei1= f(dq1, iq1, pv1, pq1 ).

[0012] Obviously, both the foregoing general description and the following detailed description are for exemplary and illustrative purposes only and are not limitations of the present invention.

Brief description of drawings:

[0013] FIG. 1 is a schematic representation of a user's Internet traffic monitoring system for assigning a user classification factor to a user device, in accordance with an embodiment of the present invention.

Implementation of the invention:

[0014] A schematic representation of the claimed system 100 for monitoring a user's Internet traffic for assigning a user classification factor to a user device is shown in FIG. 1. The system 100 includes a server 101 connected via a network 103 receiving and transmitting data, at least one computing device 102 of the user.

[0015] The user computing device 102 includes at least a transceiver 104, a display matrix 105, a video card 106, and RAM 107.

[0016] In the context of the present solution, the data receiving and transmitting network 103 may be at least one of the following networks: a TCP / IP network, the Internet, Wi-Fi, GPRS, 3G, 4G, 5G, a network based on LTE technologies, or LTE-A, satellite data network and other types.

[0017] The server 101 in the context of the present solution is configured to establish an Internet connection with at least one user computing device to analyze the hardware of the user computing device, assigning to the user computing device a factor of the user computing device hardware (coefficient dq1), for analysis parameters of the Internet connection with the assignment to the user's computing device of the coefficient of the Internet connection of the user's computing device (coefficient iq1), to analyze the user's behavior in the Internet session with the assignment of the user's computing device of the user's behavioral coefficient (coefficient pv1) and the calculation and assignment to the user device of the user classification coefficient (coefficient ei1).

[0018] The hardware analysis includes identifying the display matrix of the user computing device, identifying the video card of the user computing device, identifying the main memory of the user computing device, identifying the manufacturer of the user computing device, and identifying secondary characteristics of the user computing device. Alternatively, the secondary characteristics of the user's computing device may include at least one of the following characteristics: model year of the user's computing device, camera factory specifications, Internet browser version, or combinations thereof.

[0019] The Internet connection parameters include the speed of the Internet connection, the type of connection of the user's computing device to the Internet, the name of the Internet provider.

[0020] Analyzing user behavior includes identifying the Internet sites with the highest user traffic, identifying the number of Internet connections in a given period of time, page reading speed, printing speed, clipboard usage characteristics, and so on.

[0021] The calculation of the coefficient ei1 is performed as a function of f from the coefficients dq1, iq1, pv1: ei1= f(dq1, iq1, pv1). Alternatively, the calculation of the coefficient ei1 as a function of f of the coefficients dq1, iq1, pv1 can be performed by machine learning using at least one of the following algorithms: decision trees, GLM/GLZ, gradient boosting.

[0022] It should be noted that the present solution provides high accuracy of user classification by the calculated user classification coefficient for further provision of advertising content or for evaluating its ability to pay. The solvency assessment can later be used to provide the user with banking products (cash loans, mortgages, etc.). In the context of the present solution, the analysis of Internet connection parameters can be used to evaluate the data transmission channel. The skip channel estimate can be used to determine the amount of media content data that a user's computing device can receive in order to provide advertising content. Also, the analysis of Internet connection parameters can be used to assess the user's income level (high-income subscribers are subscribed to tariff plans with a high level of Internet connection). In particular, for the mentioned analysis, both linear regression, a decision tree, as well as more complex, but in a particular case less stable approaches, such as boosting trees, neural networks, etc., can be used. The following parameters can also be used to analyze the Internet connection: network density in the region (in particular, in the region in which the user's computing device is located), Internet connection speed (calculated using parameters in the browser API), connection method (using parameters in the browser API or by refining the characteristics of network segments using static dictionaries), the time of the Internet session (Internet session) according to the local time of the user, in particular, the start time of the Internet session and secondary characteristics of the Internet connection, for example, the type of Internet connection, the type of entry point to Internet (mobile, Wi-Fi, etc.). Each of the above indicators is taken into account under the model of having a positive value. So, for example, when obtaining (in a particular case, calculating, determining) a high value of the disk system performance, weight is added to the corresponding characteristic, and if it is impossible to determine, then the weight is not added. In a particular case, there are two or more options for defining characteristics, for example, through a value in the browser or through load testing.

[0023] Each of the above indicators can be calculated both by standard methods (GLM / GLZ, decision trees, etc.), and more modern ones (neural networks and non-Gaussian tools, etc.).

[0024] For example, determining a user's estimated income involves dividing disposable income (ei) into deciles ei -> ei1. Next, stable estimates (which in a particular case are rank estimates) are calculated for the quality of the user's computing device (dq1), the quality of the Internet connection (iq1), and a set of predictors of online behavior (pv1). To estimate ei1 as a function of dq1, iq1 and pv1, in particular, ei1= f(dq1, iq1, pv1), various algorithms can be used, for example, decision trees, GLM (generalized linear models), boosting trees ( expandable trees), etc. One sustainable solution is to use GLM/GLZ models in the form ei1=a1*dq1 + a2*iq1 + a3*py1 + a4, where the coefficients a1-a4 are obtained as a result of training the model as part of the machine learning process.

[0025] The value of the network density in the region can be calculated as the ratio of the number of registered ipv4 connections used by user computing devices per thousand inhabitants in the region or by assigning a stable level/weight to the region, which are stable over time and in the particular case do not require adjustments in the range of six to twelve months, since the number of ipv4 is limited and user migration changes by no more than one to three percent per month of the total population of the region (estimates for the dynamics of ipv4 address migration are based on data for three consecutive years for more than 20 countries). This stable level/weight is chosen based on an assessment of the quality of infrastructure, which is highly correlated with local gross regional product (GRP) per capita, which in turn correlates with income in the region. In a particular case, the user's history is tracked, in particular, the user's actions, for example, the history of orders, reservations, etc. If the user has a high income, then he will stay in good hotels with a high-quality Internet connection, where the quality of the Internet connection depends on the cost of the service provided.

[0026] Analysis of the hardware of the user's computing device can be used to identify products from individual manufacturers (eg, NVIDIA, AMD, INTEL, etc.) in order to further provide advertising content from these manufacturers. Also, the analysis of hardware (analysis of the quality of the user's computing device, i.e., analysis of its cost) allows you to assess the level of income of the user for further provision of banking products to the user (cash loans, mortgages, etc.).

[0027] Each of the indicators of the user's computing device, given in the description of this solution, can be calculated both by standard methods (GLM/GLZ, decision trees) and more modern ones (neural networks and non-Gaussian tools).

[0028] The following is an example of comparing the first indicator of the user's income level: The first device (Device 1) is the Apple iPhone 12 device (moreover, the type, model, characteristics of the computing device can be calculated from the information contained in the "useragent" field in the web browser ) has a resolution of 2532x1170 pixels and a device_pixel_ratio equal to three ("3"); The second device (Device 2) is a HUAWEI P40 lite device (the type, model, characteristics of the computing device can be calculated from the information contained in the "useragent" field in the web browser) has a resolution of 2310 × 1080 and a device_pixel_ratio equal to three ("3") .

[0029] In this case, due to the brand (Apple is more expensive than HUAWEI) and the larger screen size (and a number of other characteristics), the first device costs significantly more than the second device.

[0030] An analysis of user behavior in an Internet session provides an assessment of the user's interest in further providing advertising content to the user's computing device or in order to assess his solvency. For example, if the user in the Internet session searched for real estate, then in this case, the user's behavioral coefficient (coefficient pv1) indicates the user's interest in real estate for further provision of advertising content for the purchase of real estate. On the other hand, in this case, the pv1 coefficient estimates the user's income level for further provision of banking products to the user (cash loans, mortgages, etc.).

[0031] After the analysis is performed, the calculation of the user classification coefficient of the user device is performed, which includes dividing the user classification coefficient (ei) into deciles ei -> ei1.

[0032] Next, stable scores (which in a particular case are rank scores) are calculated for the quality of the user's computing device (dq1), the quality of the Internet connection (iq1) and a set of predictors of online behavior (the user's behavioral coefficient pv1). To estimate ei1 as a function of dq1, iq1 and pv1, in particular, ei1= f(dq1, iq1, pv1), various algorithms can be used, for example, decision trees, GLM (generalized linear models), boosting trees ( expandable trees), etc. One sustainable solution is to use GLM/GLZ models in the form ei1=a1*dq1 + a2*iq1 + a3*py1 + a4, where the coefficients a1-a4 are obtained as a result of training the model as part of the machine learning process. Machine learning is used in evaluation, for example, to build the solution architecture in such a way that it learns on its own over time. In the context of this solution, the following machine learning methods can be used: supervised learning, unsupervised learning, reinforcement learning, active learning, partially supervised learning, transductive learning, multitasking learning, multivariate learning, boosting, Bayesian network, etc. .

[0033] It should be noted that the present solution analyzes online data that cannot uniquely identify a user. Those. the data analyzed by this decision are not the following: Last name, first name, patronymic; Year, month, date and place of birth; Passport data or data of any other certifying document; E-mail, name of the account in the social. networks, incl. Skype, WhatsApp, Telegram; Cell phone number; TIN, pension certificate, SNILS, birth certificate; Address of permanent, temporary registration and residence; Family, social, property status; Information about education, profession, income, employment data; Full bank card details full bank card number, bank account details; Detailed payment receipts, including itinerary receipts, order lists, notes, purchases, comments; Information that characterizes the physiological and biological characteristics of a person, on the basis of which it is possible to establish his identity (biometric personal data). For example, a medical card, data from medical tests; Photograph of a person; And any other direct person identifiers or linear/reversible hash functions from these identifiers. In other words, the data analyzed online does not allow the user to be identified.

[0034] While the invention has been shown and described with reference to certain embodiments, those skilled in the art will appreciate that various changes and modifications can be made therein without departing from the actual scope of the invention. Therefore, the described embodiments are intended to cover all such transformations, modifications, and variations that fall within the spirit and scope of the appended claims.

Claims (10)

1. A system for analyzing user Internet traffic for assigning a user classification coefficient to a user device, comprising a server configured to: - establishing an Internet connection with at least one user computing device for analyzing the user computing device hardware with assigning the user computing device hardware coefficient to the user computing device (coefficient dq1), where the hardware analysis includes identifying the display matrix of the user computing device, identifying the video card the user computing device, identifying the user computing device RAM, identifying the manufacturer of the user computing device, and identifying secondary characteristics of the user computing device; - establishing an Internet connection with at least one user computing device to analyze the Internet connection parameters with assigning to the user computing device the coefficient of the Internet connection of the user computing device (coefficient iq1), where the Internet connection parameters include the speed of the Internet connection, the type of connection of the computing the user's devices to the Internet, the name of the Internet provider, DNS characteristics, the presence and type of proxy connection, statistics on the activity of using this network segment; - establishing an Internet connection with at least one user computing device to analyze the user's behavior in an Internet session with assigning a user behavioral coefficient (pv1 coefficient) to the user's computing device, where the analysis of user behavior includes identifying Internet sites with the highest user traffic, identifying the amount of Internet - connections for a given period of time from the user's device, characteristics of the use of "hot keys", characteristics of the speed and completeness of reading an online page, characteristics of switching between pages, printing speed; And - calculating and assigning to the user device a user classification coefficient (coefficient ei1), where the calculation of the coefficient ei1 is performed as a function of f from the coefficients dq1, iq1, pv1: ei1=f(dq1, iq1, pv1). 2. The system of claim. 1, characterized in that the secondary characteristics of the user's computing device include at least one of: the year of manufacture of the model of the user's computing device, the factory characteristics of the camera, the version of the Internet browser, or combinations thereof. 3. The system according to claim 1, characterized in that the type of connection of the user's computing device to the Internet includes at least one of the following connections: Wi-Fi, 3G, 4G, LTE-A, or combinations thereof. 4. The system according to claim 1, characterized in that the calculation of the coefficient ei1 as a function of f from the coefficients dq1, iq1, pv1 is performed by machine learning using at least one of the following algorithms: decision trees, GLM/GLZ, gradient boosting. 5. The system according to claim. 1, characterized in that the server is configured to establish an Internet connection with at least one user computing device for analyzing the software of the user computing device with assigning to the user computing device a coefficient of the software of the user computing device (coefficient pq1 ). 6. The system according to claim 5, characterized in that the server is made with the additional ability to calculate and assign to the user device a user classification coefficient (coefficient ei1), where the calculation of the coefficient ei1 is performed as a function of f from the coefficients dq1, iq1, pv1, pq1: ei1= f(dq1, iq1, pv1, pq1).

Images