EA043318B1 - METHOD AND SYSTEM FOR AUTOMATICALLY GENERATING RECOMMENDATIONS FOR MANAGING THE FINANCIAL MODEL OF USER BEHAVIOR - Google Patents

Description

Field of technology

The present invention relates to the field of computer technology, in particular to automated solutions for managing the user's financial assets.

State of the art

In the field of investment and money management, as a rule, software products are used that are based on predicting indicators of changes in a financial asset, for example, based on the user’s spending or his profile, according to which one or another recommendation for managing his finances can be selected. An example of such a solution is known from patent US 8788388 B2 (American Express Travel Related Services Co Inc, July 22, 2014), which describes a method for analyzing spending indicators for analyzing and developing a strategy for further savings management.

The disadvantage of this approach is that the analysis is performed manually, based on the chronology of financial changes directly generated by the user or company themselves, which ties the forecast to existing transactions. This leads to a limitation of functionality that does not allow the automated generation of financial asset management algorithms themselves, regardless of their type, ensuring the creation of management strategies that can automatically be applied to user profiles.

The closest to the claimed solution is patent EA 041303 (Obolensky et al., 2022.10.05), which discloses a method for generating models for managing the user’s financial assets and strategies for their management based on a retrospective analysis of the behavior of such assets.

In terms of expanding the functionality within the framework of the claimed invention, a subsequent improvement of the known solution is proposed in terms of the possibility of generating recommendations based on models of the user’s financial behavior in relation to his capital and analysis of the movement of his financial flows, which will make it possible to predict optimal management to achieve a given goal.

The essence of the invention

The technical problem to be solved is the implementation of the possibility of automated generation of algorithms for strategies for managing the user's financial flows and the formation of recommendations for their management.

The technical result is to provide automated creation of recommendations for managing the user's financial assets based on an updated financial behavior model, formed on the basis of an analysis of changes in the financial flows of the user's assets.

The claimed technical result is achieved by implementing a method for automatically generating recommendations for managing the financial model of user behavior, which is performed using the processor of a computing device and contains the stages of:

a) obtain data on the user’s financial assets;

b) perform an analysis of financial assets, during which they carry out a retrospective analysis of the movement of financial flows of the user’s assets;

determine the patterns of movement of financial flows over time; determine the amounts of mandatory and additional user expenses;

form a historical financial behavior model (HFM) of the user based on the received data;

c) form a user profile containing at least the cost of the goal and the time range for its achievement;

d) generate algorithms for managing the user’s financial assets based on the data obtained at stage c) and the IFMP, with each algorithm being assigned a weighting coefficient;

e) process the resulting algorithms by analyzing their weighting coefficients according to the following indicators selected from the group:

compound annual growth rate; Sharpe ratio;

Sortino coefficient;

the maximum level of decline in a financial asset;

the average value over a given time range from the obtained maximum levels of decline in a financial asset;

f) perform ranking of control algorithms in terms of identifying the maximum value based on the multiplied indicators of the weighting coefficients;

g) form a recommendatory financial behavior model (RFBM) based on a management algorithm determined during the ranking, containing at least a strategy for managing the financial flows of the user’s assets in given time periods;

h) generate recommendations based on the RFMP, transmitted to the user’s device at specified time intervals;

i) receive data on the implementation or non-compliance of generated recommendations based on an analysis of the movement of financial flows of the user’s assets, and in case of their implementation, use

- 1 043318 take the current RFMP, and if the recommendations are not implemented, then iteratively perform steps a) i) and update the RFMP.

In one of the private embodiments of the method, data on the user’s financial assets comes from the banking system.

In another particular embodiment of the method, the value of the target is updated taking into account changes in its market value when generating recommendations when updating the RFMP.

In another particular embodiment of the method, a request for a loan calculation is generated based on data on the user’s financial assets, the cost of the goal and the time range for achieving it.

In another particular embodiment of the method, when forming the RFMP, data on goods purchased by the user is used, based on an analysis of the movement of financial flows.

In another particular embodiment of the method, when generating recommendations using RFMP, information about analogues of the goods purchased by the user is generated.

In another particular embodiment of the method, information about analogues is transferred to an automated order generation system, which generates an order for analogues, transmitted with recommendations to the user.

In another particular embodiment of the method, recommendations generated on the basis of the RFMP contain at least one of: the amount of savings in a given time range, the use of additional financing through lending, the purchase of securities, or a combination thereof.

In another particular embodiment of the method, the RFMF is adjusted by comparing the amount of incoming financial flows in a given time period, formed within the framework of recommendations, with a given value of the amount of savings.

In another particular embodiment of the method, a ban is established on transactions exceeding the established threshold for changes in financial flows in a given time period.

In another particular embodiment of the method, a ban is set for at least one user payment instrument.

The claimed solution is also implemented using a system for automatically generating recommendations for managing a financial model of user behavior, containing at least one processor and at least one memory storing machine-readable instructions, which, when executed by the processor, perform the above method.

Brief description of the figures

Fig. 1 illustrates a general view of the interaction diagram of the claimed solution.

Fig. 2 illustrates a block diagram of the claimed method.

Fig. Figure 3 illustrates an example of generating recommendations based on an analysis of the user’s financial flows.

Fig. 4 general view of the computing device.

Carrying out the invention

In fig. 1 shows a general diagram (100) of the implementation of the claimed solution, which can be executed on a hardware and software platform (104), for example a server, in terms of processing user information (102) using data exchange through a graphical user interface (GUI) implemented using the device user (103), for example, a smartphone or computer. Information about the user’s financial assets is transmitted to the server (104), which is obtained from various sources (101), such as: banking applications, exchanges, analytical sites, information from the Internet, manual input of information by the user, etc.

A financial asset is understood as a form of ownership that directly or indirectly has a certain material value, which forms the user’s capital (102), for example, shares, mutual funds, bonds, precious metals, bills, cash, cryptocurrency, crypto assets, etc.

The server (104) generates user financial behavior models (FMP) based on the analysis of historical indicators of financial flows associated with the user’s assets, in particular, the analysis is performed on the amount of savings, monthly spending amounts, taking into account mandatory expenses (housing and communal services, cellular communications, products). Based on the fact of the generated FMP, recommendations are generated for managing the user’s assets in terms of achieving the required goals, for example, increasing capital, preserving capital, purchasing goods/services, etc.

In fig. Figure 2 shows a block diagram of the method (200) for automatically generating recommendations for managing the financial model of user behavior. At the first stage (201), information is collected from sources (101), which can be performed by requests via the API, without excluding other sources of information, such as text files, databases, etc. Data collection can be configured depending on the specific implementation of the solution, in particular, one-time setup using data entry (credentials for authorization, access keys, variables for connection frequency, etc.), as well as for periodic collection of information in order to update user profile information.

An example of information obtained from sources (101) is presented in table. 1.

Table 1

Example of information on financial assets

date Fin. asset No. 1 Fin. asset No. 2 01/06/2019 37.17 1491 01/13/2019 37.71 1491.7 01/20/2019 39.1 1492.7 01/27/2019 38.95 1493 02/03/2019 41.85 1493.6 02/10/2019 42.76 1490.6 02/17/2019 42.43 1494 02/24/2019 43.54 1496.8 03.03.2019 43.92 1496.7 03/10/2019 43.87 1519.5 03/17/2019 46.45 1501.7 03/24/2019 47.88 1503.2 03/31/2019 47.91 1505 04/07/2019 49.1 1508.7 04/14/2019 49.65 1510.5 04/21/2019 50.71 1514.8 04/28/2019 51.1 1514.3 05.05.2019 51.07 1519.5 05/12/2019 46.93 1521 05/19/2019 45.88 1519.1 05/26/2019 44.73 1522.7

Data can be saved to a database for subsequent processing. The obtained data from sources (101) is further used at stage (202) for their subsequent analysis, during which a retrospective analysis of the movement of financial flows of the user's assets is performed. The retrospectiveness of the choice of reference point can be chosen arbitrarily using the applied algorithm, for example, taking into account the last calendar year.

Connecting sources (101) of financial data (financial assets) can be entered manually by the user (102) or collected automatically. The necessary information for the subsequent operation of the solution is to obtain financial transactions that are carried out by the user (102) on a daily basis. Financial transactions, based on the nature of their conduct, may belong to one category or another, and the category, to one degree or another, to a certain basket. These entities represent a logical structural combination of the previous one for the purpose of a more objective analysis of the user’s financial transactions.

Based on the data obtained from the user’s financial assets, the patterns of movement of financial flows over time are determined (for example, week, month, year, etc.), while the amounts required (housing and communal services, payment of loans, communication bills, etc.) are taken into account. ) and additional expenses of the user, which can be tracked by the size of the amounts and identifiers of their receipt, which can be automatically collected through the means of payment that the user uses (102). Identification of payment recipients can be carried out using account details, for example, bank details, TIN, OGRN, etc. Additional (optional) user expenses include various movements of funds when purchasing goods/services in a given time period, for example, shopping in stores, buying fuel at a gas station, attending events, ordering goods, etc.

Upon processing of the collected data, a historical financial behavior model (HFM) is generated. At the same time, when forming an IFMP, an automated analysis of incoming data can also be carried out for its integrity (missing dates, anomalous outliers in the data - gap, etc.), and, if available, correct the time series data. Corrections of this kind as a significant gap in the data, called a Gap, are formed according to the following principle: the median value of the distances between close values in the entire time series is calculated and if any of the values is twice as high as the median, the size of this Gap is calculated as the difference neighboring values between which it occurred is subtracted from all values up to the moment when it occurred in time and towards the past, according to the time series. Correction of this kind of omissions in time data is carried out by comparing each value with a calendar time series; if there is a omission, the date of the value is created, and the value itself

- 3 043318 is copied from a neighboring date, earlier in time.

An example of such processing is presented in table. 2.

table 2

Example of preprocessing of financial assets

Fin. Fin. FA analysis Modified FA analysis Modified date asset No. 1 asset No. 2 No. 1 FA#1 No. 2 FA No. 2 01/06/2019 37.17 1491 1.45 37.17 0.05 15.06 01/13/2019 37.71 1491.7 3.69 37.71 0.07 15.07 01/20/2019 39.1 1492.7 -0.38 39.1 0.02 15.08 01/27/2019 38.95 1493 7.45 38.95 0.04 15.08 02/03/2019 41.85 1493.6 2.17 41.85 -0.20 15.09 02/10/2019 42.76 1490.6 -0.77 42.76 0.23 15.06 02/17/2019 42.43 1494 2.62 42.43 0.19 15.09 02/24/2019 43.54 1496.8 0.87 43.54 -0.01 15.12 03.03.2019 43.92 1496.7 -0.11 43.92 1.52 15.12 03/10/2019 43.87 1519.5 5.88 43.87 -1.17 15.35 03/17/2019 46.45 1501.7 3.08 46.45 0.10 15.17 03/24/2019 47.88 1503.2 0.06 47.88 0.12 15.18 03/31/2019 47.91 1505 2.48 47.91 0.25 15.20 04/07/2019 49.1 1508.7 1.12 49.1 0.12 15.24 04/14/2019 49.65 1510.5 2.13 49.65 0.28 15.26 04/21/2019 50.71 1514.8 0.77 50.71 -0.03 15.30 04/28/2019 51.1 1514.3 -0.06 51.1 0.34 15.30 05.05.2019 51.07 1519.5 -8.11 51.07 0.10 15.35 05/12/2019 46.93 1521 -2.24 46.93 -0.12 15.36 05/19/2019 45.88 1519.1 -2.51 45.88 0.24 15.34 05/26/2019 44.73 1522.7 -1.86 44.73 0.25 15.38

After validating the correctness of the data, a correlation analysis is carried out to compare new data with those that have already been obtained from sources (101) at the moment and stored in the database for their similarity.

Next, at step (203), a user profile is created, in which the following basic data is indicated: the cost of the goal and the time range for its achievement. The goal can be different, for example, accumulation of funds, acquisition of real estate, etc. The value of the goal should reflect the monetary equivalent to correlate it with changes in the user's financial flows.

The formation of a user profile at step (203) can be performed using machine learning models and algorithms. This solution can use a model that is trained on user behavior data when purchasing a particular product or service. Data on this kind of user activity can be subject to clustering and calculation of a scoring score, on the basis of which the user is assigned, based on the collection of information at stage (202), to one or another category (group), which will subsequently also be taken into account when forming the IFMP. The classification task within this stage is, as a rule, one-class classification. For this kind of problem, decision trees, random forest algorithm, logistic regression, artificial neural networks and support vector machines (SVMs), for example, such as SVM-Light, LIBSVM, LIBLINEAR and others, are most often used.

Based on the data received at stage (203) and the generated IFMP, algorithms for managing the user’s financial assets are then generated at stage (204). The following approach can be used for this. A pairwise measurement of correlation and data storage in the database is carried out relative to the IFMP data. After making comparisons, based on well-known clustering algorithms, new data is correlated with each of the existing clusters of similar time series. Initially, this calculation is carried out on the entire set. After adding to the cluster closest to the new time series, data on the effectiveness of this series is calculated based on a comparison of the following characteristics:

compound annual growth rate;

Sharpe ratio (https://en.wikipedia.org/wiki/Sharpe_ratio); Sortino ratio (https://en.wikipedia.org/wiki/Sortino_ratio);

the maximum level of decline of a financial asset;

the average value over a given time range from the obtained maximum levels of decline in a financial asset and others.

When forming management algorithms, cash receipts in a given period of time are taken into account, for example, the user’s monthly income, existing encumbrances (for example, loans), as well as related expenses over a time period. During the formation of control algorithms, the best one in the cluster is selected based on simple multiplication of the previously mentioned characteristics and selection of the maximum one based on a weighting coefficient. Data from the financial flows of the corresponding user assets are transferred to a specialized model, for example, built on the basis of machine learning algorithms, which plots the distribution of data from sources (101) according to time patterns (days of the week, days of the month, etc.), according to logical division data from one source to data from another, etc. As a result of the applied algorithmic processing, a decision is made on one or another action with financial assets

- 4 043318 to achieve a goal specified by the user within a given time range.

During the implementation of stage (204), patterns of changes in financial assets over time are identified based on the analysis of historical data (changes in the dynamics of a financial asset over time). A pattern in this case is the repeatability of a sequence of changes in certain values over time, preceded by identical features. For example, one of the algorithms for searching for patterns may be an algorithm for searching for seasonal patterns, when, by analyzing the ordinal number of the day in the year, the ratio of cumulative values is calculated when the time series tended to go in one direction. If patterns have been identified in a time series, which is confirmed by preprocessing performed by module (120), which performs a retrospective calculation of the following indicator on history - the level of dispersion on equal parts of temporary historical data, generated separately for a given time series, which repeat the patterns and form the essence - strategies.

Validation of the identified patterns is carried out, during which the indicators of the financial asset are analyzed, according to which the historical indicators of its behavior are calculated, reflecting the level of deviations of the indicators on equal ranges of temporary historical data. The calculation of deviation is carried out using the calculation of variance and/or standard deviation. Validation should be understood as the calculation of indicators to obtain a relatively stable variance.

Based on the results of data manipulation on each set of characteristics, options are selected based on the selection of the best, which forms the basis for the formation of algorithms for financial asset management strategies based on validated data.

The execution of step (204) can be carried out using the implementation of machine learning algorithms, allowing them to be trained on the data of the behavior of a particular financial asset over time.

Based on the execution of stage (204), the patterns of financial assets are identified, on the basis of which an automated calculation is made and the optimal control algorithm is selected. In table Figure 3 presents an example of data processing to identify patterns on the basis of which the algorithm for strategies for managing financial assets over a time period is calculated.

Table 3

An example of identifying patterns in financial assets

date Week number of the year FA#1 F No. 2 Analysis FA#1 Analysis FA No. 2 FA Strategy Algorithm No. 1 FA strategy algorithm No. 2 01/06/2019 2 37.17 15.06 1.45 0.07 1.45 0.07 01/13/2019 3 37.71 15.07 3.69 0.07 5.138809431 0.132758063 01/20/2019 4 39.1 15.08 -0.38 0.00 4.755177717 0.132758063 01/27/2019 5 38.95 15.08 7.45 0.07 12.20062059 0.19907106 02/03/2019 6 41.85 15.09 2.17 -0.20 14.37505309 0.000263903 02/10/2019 7 42.76 15.06 -0.77 0.20 13.60330379 0.199467091 02/17/2019 8 42.43 15.09 2.62 0.20 16.21937732 0.398274248 02/24/2019 9 43.54 15.12 0.87 0.00 17.092138 0.398274248 03.03.2019 10 43.92 15.12 -0.11 1.52 16.97829465 0.398274248 03/10/2019 eleven 43.87 15.35 5.88 -1.17 22.85930673 0.398274248 03/17/2019 12 46.45 15.17 3.08 0.07 25.93788585 0.464193826 03/24/2019 13 47.88 15.18 0.06 0.13 26.00054249 0.595946131 03/31/2019 14 47.91 15.20 2.48 0.26 28.48436633 0.859104026 04/07/2019 15 49.1 15.24 1.12 0.13 29.60452926 0.990337622 04/14/2019 16 49.65 15.26 2.13 0.26 31.73947387 1.25246082 04/21/2019 17 50.71 15.30 0.77 0.00 32.50855295 1.25246082 04/28/2019 18 51D 15.30 -0.06 0.33 32.44984454 1.579258206 05.05.2019 19 51.07 15.35 -8.11 0.07 24.34332407 1.644404785 05/12/2019 20 46.93 15.36 -2.24 -0.13 22.10594926 1.514196452 05/19/2019 21 45.88 15.34 -2.51 0.26 19.59941046 1.774952645 05/26/2019 22 44.73 15.38 -1.86 0.26 17.74383255 2.035030668

The generated control algorithms are further ranked at stage (205) to select the optimal one to match a specific user profile, taking into account the cost of the goal and the time to achieve it. Within the framework of the proposed solution, automatically, based on the patterns found when analyzing financial flows, it is possible, based on algorithms for combining financial results, to evaluate the effectiveness of a particular management algorithm applied to a selected financial asset, thereby carrying out a comparative analysis and transforming the management algorithm using various risk management algorithms, subsequently forming a recommendation model of financial behavior (RFMP) at stage (206). One of the risk management algorithms can be the calculation of strategy participation coefficients, calculated on the basis of such characteristics as: standard deviation, the total coefficient of cross-correlation with other strategies, the profitability coefficient expressed as the average annual profitability. Management algorithms can be formed using one or more machine learning models, which are trained on changes in financial flows and the effective management of such assets, to quickly respond to possible changes and provide more accurate forecasts for the formation of RFMF.

The ranking of control algorithms at stage (205) can be carried out by random enumeration of weight coefficients, where each coefficient lies in the range from 0 to 1 (in this case, in the control profile, the sum of weight coefficients among strategy algorithms should not exceed 1). Next, the previously specified weighting coefficients are multiplied and the maximum value is selected, which ensures the calculation of the most optimal and effective management algorithms for the corresponding financial assets.

Methods for combining and calculating the weighting coefficients of each time series can be determined by a number of formulas, including equal weighted averages, based on volatility weights (inverse standard deviation), pairwise correlation. After analyzing and selecting optimal sets of strategies for managing the user’s financial assets, the RFMP is formed, on the basis of which its indicators are calculated and stored in the database for further use in terms of generating recommendations at stage (207).

Recommendations at stage (207) are generated regarding the management of financial assets for one or more financial assets of users. For each user profile, a scoring indicator can be additionally calculated, which reflects the preference for managing financial assets, for example, by surveying system users.

Recommendations can also be generated using recommendation algorithms implemented using machine learning (TensofFlow, Apple Core ML), for example, collaborative filtering, content-based, knowledge-based algorithms , hybrid algorithms.

Each user profile may or may not contain already connected financial assets (for example, connection to trust management). Based on the scoring indicator, one or more financial asset management algorithms can be selected and subsequently applied to generate recommendations for a specific user profile.

Next, the generated recommendations are transmitted to the user (102), for example, by displaying them at a given time period (for example, every 5th or 20th of the month) on his device (103), for example, a smartphone. Recommendations can be displayed in a software application, using notifications (PUSH, SMS, etc.), email messages, sending to instant messengers, etc.

Further analysis of following the generated recommendations at stage (207) is performed by the solution using an analysis of the financial flows of the user’s assets (102) according to the generated profile and RFMP. If the change in assets corresponds to the recommendations, then recommendations are subsequently formed for the next time ranges in accordance with the current RFMP (208), but if the recommendations are not followed, then stages (202 - 207) are iteratively repeated with the formation of a new RFMP.

In fig. Figure 3 shows an example of generating recommendations for a user profile (102). When forming a profile, the user (102) specifies the goals (301) for subsequent management in terms of creating recommendations. For example, the cost of a goal is 5 million rubles and the period for achieving it is 3 years. As stated above, the claimed solution, using an automated algorithm, analyzes financial flows and associated user assets (302), as a result of which a retrospective picture is formed with the selection of optimal control algorithms, with the help of which the RFMP and subsequent recommendations are generated (303). Recommendations (3031-3033) are created at specified time intervals for their implementation by the user (102) and can take into account both the user’s own funds (102) and the possibility of attracting borrowed funds, for example, a loan. In this case, the RFMP will take into account the user’s income for optimal management of loan payments.

Data about the target (301) can also be tied to a specific property, for example, an apartment or a car, which also determines the current value of this type of target. The claimed solution also provides constant monitoring of the actual change in the value of the target (301), for example, through integration with external resources that make it possible to obtain dynamic information about jumps in the cost of a particular product, in particular, the real estate market, the automobile market, the value of currency, etc. . If the current value of the goal (301) changes, the RFMP is updated to more accurately generate appropriate recommendations.

Recommendations generated on the basis of the RFMP may contain, for example, the amount of savings in a given time range (for example, the amount that needs to be transferred to the account at the end of the month), the use of additional financing through lending, the purchase of securities, shares, etc. Adjustment of the RFMP can be carried out by comparing the amount of incoming financial flows in a given time period, formed within the framework of recommendations, with a given value of the amount of savings, or based on changes in the value of the goal, etc.

In a particular example of implementation, when generating recommendations using RFMP, data on goods purchased by the user is used, based on an analysis of the movement of financial flows. In this case, an analysis takes place, for example, of the user’s consumer basket, in part of which, by creating a RFMP, analogues/substitutes of goods (for example, less expensive) can be taken into account, information about which can be provided to the user (102). When generating recommendations, a consumer basket can also be formed using integration with relevant applications for ordering and delivering goods. The user (102) can agree on the products offered for direct ordering, as well as make his own changes to its composition.

Also, the claimed invention can track data on the cost of goods in the consumer basket, information about discounts, sets, etc., which is taken into account when forming the RFMP, updating it and generating recommendations. This function can also be tied to the geoposition of the user (102) or the most preferred points of sale of goods or services.

Using the claimed invention, the user (102) also has the ability to set a spending threshold in a time range, for example, the amount of optional spending per month. This can be done using an appropriate software application installed on the user's device (103) connected via API to the server (104). The user (102) can set a ban on transactions using the means of payment available to him (bank card) if the transaction amount is above the established limit. The threshold value can be set for each payment instrument individually.

The application of a financial asset management profile is carried out through its automated management when following the strategy calculated by the system, including automatic sales, purchases of a financial asset, adjustments of orders, etc. The claimed method offers a fully automated solution that allows you to manage financial assets using the dynamic calculation of their management algorithms over time, reducing the need for manual adjustments to the strategy profile. The method can be implemented using a distributed modular system, where each module can perform one or more functions corresponding to the steps of the method.

The claimed invention can be implemented as a software application (client/server application) installed on the user's mobile device, for example, a smartphone. The user creates a profile (102) using the graphical interface of the application, registers in the system and undergoes a survey to calculate his scoring indicator, on the basis of which the system will subsequently calculate the required profile for managing his assets. In this case, the user may not currently own assets to manage them, for which he only needs to create an account to credit funds to it or connect an existing account (settlement or brokerage). The user can also select preferred assets, such as shares or cryptocurrency, as well as the management period (for example, a deposit for 1 or 3 years), based on which the system will automatically calculate the most effective algorithms for managing certain assets.

The claimed invention also allows warning about changes in asset management, and generates automatic changes in assets or their management strategy, notifying the user (102). Users (102) can also change their profiles by changing the type of overall asset management strategy, for example, increasing the management period, changing the asset type to higher/lower risk assets, changing the cost and time frame for achieving the goal, etc.

Users (102) have the opportunity to supplement their profiles with other assets, also connecting them to the system. In this case, the system correlates with existing information obtained from sources (101), calculates algorithms for managing new user assets and selects the most effective management profile, taking into account the existing strategy tied to the current user profile (102). In this case, the user profile (102) may contain several asset management strategies for different asset types, for example, one type of management for shares, another type of management for a foreign currency account, which is used in terms of the formation of RFMP and the subsequent generation of recommendations.

To implement the claimed method (200), one or more machine learning models can be used, with the help of which one or more stages of the method can be performed. At the same time, the principle of training and operation of such models in relation to the functional execution of the corresponding stage of the method (200) should be obvious to a specialist in the field of technology.

In fig. 4 shows an example of a general view of a computing device (400), on the basis of which the claimed system can operate. The device (400) may be part of a computer system, for example, a server, a personal computer, or part of a computing cluster that processes the necessary data to implement the stated technical solution in relation to the method (200).

In the general case, the device (400) contains connected by a common information exchange bus

-

Claims (1)

one or more processors (401), memory means such as RAM (402) and ROM (403), input/output interfaces (404), input/output devices (405), and network communication device (406). The processor (401) (or multiple processors, multi-core processor, etc.) may be selected from a variety of devices commonly used today, for example, from manufacturers such as: Intel™, AMD™, Apple™, Samsung Exynos™, MediaTEK ™, Qualcomm Snapdragon™, etc. The processor or one of the usable processors in the device (400) must also include a graphics processor, such as an NVIDIA or Graphcore GPU, the type of which is also suitable for performing all or part of the method, and can also be used for training and applying machine learning models in various information systems. RAM (402) is a random access memory and is designed to store machine-readable instructions executed by the processor (401) to perform the necessary operations for logical data processing. RAM (402) typically contains executable operating system instructions and associated software components (applications, software modules, etc.). In this case, the available memory capacity of the graphics card or graphics processor can act as RAM (402). ROM (403) is one or more permanent storage devices, such as a hard disk drive (HDD), a solid state drive (SSD), flash memory (EEPROM, NAND, etc.), optical storage media (CD-R) /RW, DVD-R/RW, BlueRay Disc, MD), etc. To organize the operation of device components (400) and organize the operation of external connected devices, various types of I/O interfaces (404) are used. The choice of appropriate interfaces depends on the specific design of the computing device, which can be, but is not limited to: PCI, AGP, PS/2, IrDa, FireWire, LPT, COM, SATA, IDE, Lightning, USB (2.0, 3.0, 3.1, micro, mini, type C), TRS/Audio jack (2.5, 3.5, 6.35), HDMI, DVI, VGA, Display Port, RJ45, RS232, etc. To ensure user interaction with the device (400), various means (405) of I/O information are used, for example, a keyboard, a display (monitor), a touch display, a touch pad, a joystick, a mouse, a light pen, a stylus, a touch pad, a trackball. , speakers, microphone, augmented reality tools, optical sensors, tablet, light indicators, projector, camera, biometric identification tools (retina scanner, fingerprint scanner, voice recognition module), etc. The network communication means (406) provides data transmission via an internal or external computer network, for example, an Intranet, the Internet, a LAN, etc. As one or more means (406), the following may be used, but is not limited to: Ethernet card, GSM modem, GPRS modem, LTE modem, 5G modem, satellite communication module, NFC module, Bluetooth and/or BLE module, Wi-Fi module, etc. The presented materials of the invention disclose preferred examples of implementation of a technical solution and should not be interpreted as limiting other, particular examples of its implementation that do not go beyond the scope of the requested legal protection, which are obvious to specialists in the relevant field of technology. CLAIM 1. A method for automatically generating recommendations for managing the financial model of user behavior, performed using the processor of a computing device and containing stages in which: a) obtain data on the user’s financial assets; b) perform an analysis of financial assets, during which they carry out a retrospective analysis of the movement of financial flows of the user’s assets; determine the patterns of movement of financial flows over time; determine the amount of mandatory and additional user expenses; form a historical financial behavior model (HFM) of the user based on the received data; c) form a user profile containing at least the cost of the goal and the time range for its achievement; d) generate algorithms for managing the user’s financial assets based on the data obtained at stage c) and the IFMP, with each algorithm being assigned a weighting coefficient; e) process the resulting algorithms by analyzing their weighting coefficients according to the following indicators selected from the group: compound annual growth rate; Sharpe ratio; Sortino coefficient; the maximum level of decline of a financial asset; the average value over a given time range from the obtained maximum levels of decline in a financial asset; -