User level calculation method and device for gas station and medium

Abstract

The embodiment of the invention discloses a user level calculation method and device for a gas station, and a medium, belongs to the technical field of gas station data processing, and solves the problem of loss of a large number of customers caused by low preferential strategy pushing accuracy of an existing gas station. Determining a user member type through a preset member calculation model and refueling and recharging information; matching the license plate information with a historical refueling database, and determining a refueling track corresponding to the refueling user; constructing a refueling state transition matrix based on the refueling trajectory to obtain predicted refueling information corresponding to the refueling user; based on the predicted refueling information, the user member type and the refueling track are sent to the predicted refueling station within the predicted refueling time period; under the condition that the refueling user arrives at the predicted gas station, constructing a user refueling information portrait; and determining the current member level of the user based on the member type and the user refueling information portrait, and pushing a corresponding refueling discount strategy to the refueling user.

Description

Technical Field

[0001] This application relates to the technical field of gas station data processing, and particularly to a user level calculation method, device, and medium for gas stations. Background Art

[0002] With the continuous increase in the global automobile ownership, the gas station industry is facing increasingly fierce market competition. In the past, the gas station operation mode was relatively traditional, and the application of customer relationship management and marketing means was relatively single, mainly focusing on the supply of oil products, and lacking sufficient attention to the precise marketing of user consumption behaviors.

[0003] In recent years, with the rapid development of information technology, the gas station industry has begun to actively introduce digital management methods. In terms of user data collection, most gas stations have established basic user databases that can store users' refueling and recharge information, including recharge balances and coupon data, etc.

[0004] However, the existing gas station preferential policy push technology mainly relies on simple statistical analysis. For example, based on the user's historical average refueling amount and refueling frequency to roughly infer the preferential types that the user may be interested in, resulting in the disconnection between the preferential policy push and the actual needs of users. For example, high-value preferential strategies are pushed to users who are not price-sensitive and have low consumption capabilities, while users with high consumption potential do not receive matching preferences. This not only causes waste of resources, but also reduces users' satisfaction and loyalty to gas stations, leading to a large number of customer losses. Summary of the Invention

[0005] Embodiments of this application provide a user level calculation method, device, and medium for gas stations to solve the following technical problems: The accuracy of the existing gas station preferential strategy push technology is relatively low, which not only causes waste of resources, but also reduces users' satisfaction and loyalty to gas stations, leading to a large number of customer losses.

[0006] Embodiments of this application adopt the following technical solutions:

[0007] An embodiment of the present application provides a method for calculating user levels for gas stations. The method includes: based on the gas station user database, determining the refueling recharge information corresponding to the refueling user, and determining the user membership type through a preset membership calculation model and the refueling recharge information; wherein, the refueling recharge information includes the recharge balance and coupon data; determining the license plate information corresponding to the refueling user in the gas station database, matching the license plate information with the historical refueling database, and determining the refueling trajectory corresponding to the refueling user; based on the time series data in the refueling trajectory, constructing a refueling state transition matrix to perform refueling probability distribution processing through the refueling state transition matrix, and obtaining the predicted refueling information corresponding to the refueling user; based on the predicted refueling information, sending the user membership type and the refueling trajectory to the predicted refueling station during the predicted refueling time period; in the case where the refueling user arrives at the predicted refueling station, combining the refueling trajectory and the predicted refueling information to construct a user refueling information portrait; based on the membership type and the user refueling information portrait, determining the current user membership level, and pushing corresponding refueling preferential policies to the refueling user based on the current membership level.

[0008] In an embodiment of the present application, by analyzing the recharge balance and coupon data of users and using a preset membership calculation model, the user membership type can be carefully distinguished. A judgment matrix is constructed, incorporating the membership type and the user refueling information portrait, making the determination of the membership level more scientific and comprehensive. Information such as the gas station location, refueling frequency, amount, and time in the refueling trajectory, combined with the predicted refueling information, can vividly present the user's refueling habits and potential needs. Constructing a refueling state transition matrix based on the time series data in the refueling trajectory can deeply explore the rules of the user's refueling behavior. By analyzing data such as historical refueling time and frequency, the future refueling probability distribution of the user is predicted, and then accurate predicted refueling information is obtained. During the predicted refueling time period, the user membership type and the refueling trajectory are sent to the predicted refueling station to achieve service preposition. Based on the accurate membership level and the comprehensive user refueling information portrait, the gas station can formulate highly personalized refueling preferential policies, avoid blind pushing in the past, and ensure that the preferential policies accurately reach the target users.

[0009] In one implementation manner of the present application, before determining the user membership type through the preset membership calculation model and the refueling recharge information, the method further includes: constructing a coupon type indication function based on the refueling coupon type and a first preset condition value; constructing a coupon quantity indication function based on the coupon quantity corresponding to the user and a second preset condition value; constructing a balance indication function based on the recharge balance corresponding to the user and a third preset condition value; constructing a preset membership calculation model through the coupon type indication function, the coupon quantity indication function, and the balance indication function:

[0010] Result=I1·(I2·SVIP 绿 +(1-I2)·I3·SVIP黄 +(1 - I2)·(1 - I3)·I4·VIP 绿 +(1 - I2)·(1 - I3)·(1 - I4)·VIP 黄 )+(1 - I1)·V 会

[0011] Wherein, Result is the user membership type; I1 is the coupon type indication function; I2 is the first coupon quantity indication function; I3 is the second coupon quantity indication function; I4 is the balance indication function; SVIP 绿 indicates that the membership level is green SVIP; SVIP 黄 indicates that the membership level is yellow SVIP; VIP 黄 indicates that the membership level is yellow VIP; VIP 绿 indicates that the membership level is green VIP; V 会 indicates that the user is a new user or a recalled user.

[0012] In an implementation manner of the present application, the license plate information is matched with the historical refueling database to determine the refueling trajectory corresponding to the refueling user, which specifically includes: querying in the historical refueling database to determine the refueling data corresponding to the refueling user within a preset time period; obtaining the gas station locations in the refueling data, and constructing a refueling station map corresponding to the refueling user based on the gas station locations; connecting the gas stations in the refueling station map in the order of the refueling time to construct a refueling path, and performing a first assignment on each gas station in the path based on the number of connection lines; determining the refueling time and refueling amount corresponding to each gas station respectively, and performing a second assignment on each gas station based on the refueling time and refueling amount; screening the gas stations in the refueling path based on the first assignment and the second assignment, so as to determine the refueling trajectory corresponding to the refueling user through the screened gas stations and the refueling data of the screened gas stations within the preset time period.

[0013] In an implementation manner of the present application, based on the time series data in the refueling trajectory, a refueling state transition matrix is constructed to perform refueling probability distribution processing through the refueling state transition matrix, and obtain the predicted refueling information corresponding to the refueling user, which specifically includes: clustering the time series data into multiple groups according to the density distribution corresponding to the time series data, and performing regular refueling data identification and irregular refueling data identification on each group according to the periodic data in each group; constructing a state transition matrix based on the data corresponding to the regular refueling data identification; where the rows in the state transition matrix correspond to refueling time states, the columns correspond to refueling station states, and the elements in the state transition matrix represent the number of times transferred from the first time-station state to the second time-station state; determining the state transition probability matrix corresponding to the state transition matrix based on the ratio between each element in each row of the state transition matrix and the sum of all elements in the current row; determining the initial element in the state transition probability matrix according to the most recent refueling time state and the most recent refueling station state corresponding to the refueling user, and determining the probability distribution corresponding to the initial element; determining the predicted refueling time and the predicted refueling station corresponding to the refueling user through the random number generator and the probability distribution corresponding to the initial element.

[0014] In an implementation manner of the present application, based on the predicted refueling information, the user membership type and the refueling trajectory are sent to the predicted refueling station within the predicted refueling time period, which specifically includes: determining the historical refueling behavior data based on the historical refueling information of the refueling user at the predicted refueling station; determining the historical information sending data based on the historical information sending time of the refueling user corresponding to the predicted refueling station and the correct rate feedback information corresponding to the historical information sending time; constructing a time prediction feature based on the historical refueling behavior data and the historical information sending data; inputting the time prediction feature into a preset decision tree model to output the information gain corresponding to different time prediction features through the preset decision tree model, and determining the reference refueling time according to the information gain; expanding the reference refueling time into a predicted refueling time period based on a preset time error region; binding the user membership type corresponding to the refueling user and the refueling trajectory, and pushing the bound information to the predicted refueling station within the reference refueling time period.

[0015] In an implementation manner of the present application, after combining the refueling trajectory and the predicted refueling information to construct a user refueling information portrait, the method further includes: dividing the user refueling information portrait into multiple first groups based on the refueling frequency corresponding to each refueling station, and arranging the first groups in chronological order; dividing the sequential refueling data corresponding to the user refueling information portrait in each first group into multiple refueling subsequences according to a preset time period section, and numbering the subsequences; in the first group, dividing the multiple refueling subsequences into multiple second groups based on chronological order; based on the subsequence numbers, making multiple rounds of random selections in each second group to determine a first reference portrait; making a random selection of the first reference portraits corresponding to the multiple first groups respectively to determine a second reference portrait; after multiple rounds of selections, combining each selected second reference portrait with the non-sequential refueling data in the user refueling information portrait to obtain multiple rounds of reference third portraits, so as to determine the current membership level of the user based on the reference third portraits.

[0016] In an implementation manner of the present application, determining the current membership level of a refueling user based on the membership type and the user refueling information portrait specifically includes: determining user level indicators sorted by importance based on the reference third portrait; wherein, the user level indicators at least include one of refueling frequency, average refueling amount, continuous refueling duration, peak period refueling proportion, and oil product diversity selection; constructing a user level judgment matrix through the user level indicators and a preset fuzzy scale; in the user level judgment matrix, determining the first weight corresponding to each user level indicator respectively through a weighted average type fuzzy composition operator; constructing a weight optimization function based on the first weight and the preset fuzzy scale, so as to obtain the second weight corresponding to each user level indicator respectively based on the weight optimization function; obtaining a reference level based on the user level indicators and the second weight; dividing the membership type corresponding to the refueling user based on the reference level to determine the current membership level of the user.

[0017] In an implementation manner of the present application, pushing corresponding refueling preferential policies to refueling users based on the current membership level specifically includes: determining multiple corresponding refueling preferential policies in the preferential information library based on the current membership level; inputting the current membership level, the basic information corresponding to the refueling user, and the refueling trajectory into a preset preferential acceptance model, so as to output the acceptance probability of the refueling user for different refueling preferential policies through the preferential acceptance model; sorting the multiple refueling preferential policies based on the acceptance probability, so as to sequentially send the preferential policies to the refueling user based on the sorting order.

[0018] An embodiment of the present application provides a user level calculation device for a gas station, including: at least one processor; and a memory communicatively connected to the at least one processor; wherein, the memory stores instructions executable by the at least one processor, and when the instructions are executed by the at least one processor, the at least one processor is enabled to: based on the gas station user database, determine the refueling recharge information corresponding to the refueling user, and determine the user membership type through a preset membership calculation model and the refueling recharge information; wherein, the refueling recharge information includes the recharge balance and coupon data; determine the license plate information corresponding to the refueling user in the gas station database, match the license plate information with the historical refueling database, and determine the refueling trajectory corresponding to the refueling user; based on the time series data in the refueling trajectory, construct a refueling state transition matrix to perform refueling probability distribution processing through the refueling state transition matrix, and obtain the predicted refueling information corresponding to the refueling user; based on the predicted refueling information, send the user membership type and the refueling trajectory to the predicted refueling station during the predicted refueling time period; in the case where the refueling user arrives at the predicted refueling station, combine the refueling trajectory and the predicted refueling information to construct a user refueling information portrait; based on the membership type and the user refueling information portrait, determine the current user membership level, and push corresponding refueling preferential policies to the refueling user based on the current membership level.

[0019] A non-volatile computer storage medium provided by an embodiment of the present application stores computer-executable instructions, and the computer-executable instructions are set to: based on the gas station user database, determine the refueling recharge information corresponding to the refueling user, and determine the user membership type through a preset membership calculation model and the refueling recharge information; wherein, the refueling recharge information includes the recharge balance and coupon data; determine the license plate information corresponding to the refueling user in the gas station database, match the license plate information with the historical refueling database, and determine the refueling trajectory corresponding to the refueling user; based on the time series data in the refueling trajectory, construct a refueling state transition matrix to perform refueling probability distribution processing through the refueling state transition matrix, and obtain the predicted refueling information corresponding to the refueling user; based on the predicted refueling information, send the user membership type and the refueling trajectory to the predicted refueling station during the predicted refueling time period; in the case where the refueling user arrives at the predicted refueling station, combine the refueling trajectory and the predicted refueling information to construct a user refueling information portrait; based on the membership type and the user refueling information portrait, determine the current user membership level, and push corresponding refueling preferential policies to the refueling user based on the current membership level.

[0020] The above at least one technical solution adopted in the embodiments of the present application can achieve the following beneficial effects: By analyzing the user's recharge balance and coupon data and using a preset membership calculation model, the embodiments of the present application can carefully distinguish the user membership types. A judgment matrix is constructed and integrated with the membership type and the user's refueling information portrait to make the membership level determination more scientific and comprehensive. Information such as the gas station location, refueling frequency, amount, and time in the refueling trajectory, combined with the predicted refueling information, can vividly present the user's refueling habits and potential needs. Based on the time series data in the refueling trajectory, a refueling state transition matrix is constructed to deeply mine the user's refueling behavior patterns. By analyzing data such as historical refueling time and frequency, the future refueling probability distribution of the user is predicted, and then accurate predicted refueling information is obtained. During the predicted refueling time period, the user membership type and the refueling trajectory are sent to the predicted refueling station to achieve service preposition. Based on the accurate membership level and the comprehensive user refueling information portrait, the gas station can formulate highly personalized refueling preferential strategies, avoid blind pushing in the past, and ensure that the preferential policies accurately reach the target users. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments recorded in the present application. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings. In the drawings:

[0022] Figure 1 It is a flowchart of a method for calculating user levels for a gas station provided by an embodiment of the present application;

[0023] Figure 2 It is a schematic structural diagram of a device for calculating user levels for a gas station provided by an embodiment of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0024] The embodiments of the present application provide a method, a device, and a medium for calculating user levels for a gas station.

[0025] In order to enable those skilled in the art to better understand the technical solutions in the present application, the following will clearly and completely describe the technical solutions in the embodiments of the present application with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments of this specification, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present application.

[0026] The technical solutions proposed in the embodiments of the present application will be described in detail below with reference to the drawings.

[0027] Figure 1 The following is a flowchart of a user level calculation method for a gas station provided by an embodiment of the present application. As Figure 1 shown, the user level calculation method for a gas station includes the following steps:

[0028] S101. Based on the gas station user database, determine the refueling recharge information corresponding to the refueling user, and determine the user membership type through a preset membership calculation model and the refueling recharge information.

[0029] In an embodiment of the present application, first extract data from the gas station user database to determine the refueling recharge information of the current refueling user. Among them, the refueling recharge information includes the recharge balance and coupon data. For example, according to the name and license plate information of the current refueling user, data can be extracted from the gas station user database to obtain data such as the refueling recharge balance, refueling times, and the remaining coupon usage situation of the user.

[0030] In an embodiment of the present application, based on the refueling coupon type and the first preset condition value, construct a coupon type indication function. Based on the number of coupons corresponding to the user and the second preset condition value, construct a coupon quantity indication function. Based on the recharge balance corresponding to the user and the third preset condition value, construct a balance indication function.

[0031] Specifically, based on the refueling coupon type and the first preset condition value, construct a coupon type indication function:

[0032]

[0033] Based on the number of coupons corresponding to the user and the second preset condition value, construct a coupon quantity indication function:

[0034]

[0035] Based on the recharge balance corresponding to the user and the third preset condition value, construct a balance indication function:

[0036]

[0037] Among them, I1 is the coupon type indication function; I2 is the first coupon quantity indication function; I3 is the second coupon quantity indication function; I4 is the balance indication function; T 券 represents the coupon type; q 新 represents the new user coupon; q 召 : represents the recalled coupon; y 余 represents the balance on the account; X represents the number of coupons.

[0038] Further, a preset membership calculation model is constructed through a coupon type indication function, a coupon quantity indication function, and a balance indication function:

[0039] Result=I1·(I2·SVIP green+(1-I2)·I3·SVIP yellow+(1-I2)·(1-I3)·I4·VIP green+

[0040] (1-I2)·(1-I3)·(1-I4)·VIP 黄 )+(1-I1)·V 会 ;

[0041] Wherein, Result is the user membership type; I1 is the coupon type indication function; I2 is the first coupon quantity indication function; I3 is the second coupon quantity indication function; I4 is the balance indication function; SVIP 绿 indicates that the membership level is green SVIP; SVIP 黄 indicates that the membership level is yellow SVIP; VIP 黄 indicates that the membership level is yellow VIP; VIP 绿 indicates that the membership level is green VIP; V 会 indicates that the user is a new user or a recalled user.

[0042] S102. Determine the license plate information corresponding to the refueling user in the gas station database, match the license plate information with the historical refueling database, and determine the refueling trajectory corresponding to the refueling user.

[0043] In an embodiment of the present application, query in the historical refueling database to determine the refueling data corresponding to the refueling user within a preset time period. Obtain the gas station locations in the refueling data, and construct a refueling station map corresponding to the refueling user based on the gas station locations. Connect the gas stations in the refueling station map in the order of refueling time to construct a refueling path, and perform a first assignment to each gas station in the path based on the number of connection lines. Determine the refueling time and refueling amount corresponding to each gas station respectively, and perform a second assignment to each gas station based on the refueling time and refueling amount. Based on the first assignment and the second assignment, screen the gas stations in the refueling path, so as to determine the refueling trajectory corresponding to the refueling user through the screened gas stations and the refueling data of the screened gas stations within the preset time period. The refueling trajectory includes gas station location, refueling frequency, refueling amount, and refueling time.

[0044] Specifically, the preset time period in the embodiments of the present application can be one month, one quarter, or half a year, etc. In the historical refueling database, according to the user's identification, such as mobile phone number, membership ID, etc., query to find all the refueling records of the user in the past preset time period, and these records are the refueling data. From the refueling data queried above, extract the location information of the gas station corresponding to each refueling record. This location information can be a specific address or represented by geographical coordinates such as longitude and latitude. Mark different gas station locations on the map to construct a refueling station map corresponding to the refueling users. Further, in the order of refueling time, connect the gas stations in the refueling station map with lines in sequence.

[0045] For example, assume that the refueling user refuels at gas station A for the first time, at gas station B for the second time, at gas station C for the third time, returns to gas station A for the fourth time, and at gas station D for the fifth time. Then starting from the first gas station A, draw a line to connect to gas station B, then connect from gas station B to gas station C, then connect from gas station C to gas station A where the second refueling occurs, and finally connect from gas station A to gas station D to form a complete refueling path. At the same time, perform a first assignment to each gas station in the path based on the number of connecting lines. For example, gas station A appears twice and there are three lines connecting it, so its first assignment can be 3; gas station B appears only once and there is one line connecting it, so its first assignment is 1; gas station C is the same, and its first assignment is also 1; the first assignment of gas station D is also 1.

[0046] Further, determine the refueling time and refueling amount corresponding to each gas station respectively. For example, assume that the first refueling time of the refueling user at gas station A is January 10th and the refueling amount is 200 yuan, and calculate comprehensively based on the height of the refueling amount and the distance from the current time. Assume the second assignment = refueling amount × (1 / (current time - refueling time)). Taking the first refueling at gas station A as an example, calculate the second assignment as 200 × (1 / (the number of days converted from the current time - January 10th)). Each gas station will obtain a second assignment based on its refueling time and amount, and this assignment can reflect the importance of the gas station at the level of the user's actual consumption. Gas stations with a high refueling amount and a relatively short distance from the current time have a relatively high second assignment.

[0047] Further, based on the preset weight values, weighted calculations are performed on the first assignment and the second assignment corresponding to each gas station to determine the weighted values corresponding to each gas station. The weighted values are compared with the preset weighted threshold to delete the gas stations with weighted values lower than the threshold, thereby obtaining the filtered gas stations. Based on the filtered gas stations and their refueling data within a preset time period: gas station location, refueling time, and refueling amount, the refueling trajectory corresponding to the user is determined. The refueling trajectory in the embodiments of the present application not only clearly presents the locations of the gas stations that the user often visits, but also reflects the refueling frequency through the refueling time, and reflects the consumption situation of the user at different gas stations through the refueling amount.

[0048] S103. Based on the time series data in the refueling trajectory, construct a refueling state transition matrix to perform refueling probability distribution processing through the refueling state transition matrix, and obtain the predicted refueling information corresponding to the refueling user.

[0049] In an embodiment of the present application, according to the density distribution corresponding to the time series data, the time series data is clustered into multiple groups, and based on the periodic data in each group, regular refueling data identification and irregular refueling data identification are performed on each group. Based on the data corresponding to the regular refueling data identification, a state transition matrix is constructed; where the rows in the state transition matrix correspond to refueling time states, the columns correspond to refueling station states, and the elements in the state transition matrix represent the number of times of transitioning from the first time-station state to the second time-station state. Based on the ratio between each element in each row of the state transition matrix and the sum of all elements in the current row, the state transition probability matrix corresponding to the state transition matrix is determined. According to the most recent refueling time state and the most recent refueling station state corresponding to the refueling user, an initial element is determined in the state transition probability matrix, and the probability distribution corresponding to the initial element is determined. Through the random number generator and the probability distribution corresponding to the initial element, the predicted refueling time and the predicted refueling station corresponding to the refueling user are determined.

[0050] Specifically, time series data is determined in the refueling trajectory, such as refueling time and corresponding refueling station and other information. According to the density distribution corresponding to the time series data, a clustering algorithm is used to cluster these data into multiple groups, where the density distribution reflects the degree of aggregation of the data in the time and gas station dimensions. For example, within certain time periods, the user refuels frequently at specific several gas stations, and these data points will show obvious aggregation areas on the density distribution graph, and the clustering algorithm will divide the data in these aggregation areas into one group, thereby classifying the data with similar refueling patterns into one category.

[0051] Furthermore, for each clustering group, determine whether the user refuels at specific time intervals or according to certain patterns, such as refueling at a specific gas station every Monday, or rotating among several fixed gas stations at the beginning of each month. For data groups with periodicity, identify the regular refueling data; while for those data groups without obvious periodicity, mark them as irregular refueling data. For example, after analysis, it is found that in a certain data group, the user always refuels at the same gas station about every 7 days, and this data group is identified as regular refueling data; while in another data group, the refueling time and station are completely irregular and are marked as irregular refueling data.

[0052] Furthermore, based on the data corresponding to the regular refueling data identification, construct a state transition matrix. Among them, the rows in the state transition matrix correspond to the refueling time states, for example, they can be divided according to different time periods, such as morning, afternoon, evening, or different time intervals from Monday to Sunday, etc.; the columns correspond to the refueling station states, that is, different gas stations. The elements in the state transition matrix represent the number of times of transitioning from the first time-station state to the second time-station state. For example, within a certain time period, the number of times the user refuels at gas station A in the morning and then transfers to gas station B in the afternoon is 5 times. Then, in the state transition matrix, the element value corresponding to the row of morning-gas station A and the column of afternoon-gas station B is 5. By counting the number of transitions between different time-station states in a large amount of regular refueling data, a complete state transition matrix is constructed.

[0053] Furthermore, for the constructed state transition matrix, calculate the ratio of each element in each row to the sum of all elements in the current row. For example, the sum of all elements in a certain row is 50, and one of the element values is 10, then the ratio corresponding to this element is 10÷50 = 0.2. Calculate the ratio of each element in the state transition matrix according to the above method to obtain a new matrix, and this matrix is the state transition probability matrix. In the state transition probability matrix, the value of each element represents the probability of transitioning from one time-station state to another time-station state.

[0054] Further, obtain the status of the most recent refueling time and the status of the most recent refueling station corresponding to the refueling user. For example, the user's most recent refueling was on Tuesday afternoon at Gas Station C. In the state transition probability matrix, find the element corresponding to the state of Tuesday afternoon - Gas Station C. This element is the initial element. At the same time, determine all the elements in the row where the initial element is located. These elements together constitute the probability distribution corresponding to the initial element. For example, the other elements in the row where the initial element is located respectively represent the probabilities of transitioning from Tuesday afternoon - Gas Station C to other different time - station states. The set of these probability values is the probability distribution corresponding to the initial element. Use a random number generator to generate a random number between 0 and 1. For example, the generated random number is 0.45. According to the probability distribution corresponding to the initial element, accumulate the probability values in the probability distribution in sequence. For example, the probability distribution in the row where the initial element is located is [0.1, 0.3, 0.2, 0.4], and after accumulation, it becomes [0.1, 0.4, 0.6, 1]. Find the accumulated interval into which the random number falls. For example, 0.45 falls into the interval [0.4, 0.6]. The time - station state corresponding to this interval is the predicted refueling time and station. Assume that the interval [0.4, 0.6] corresponds to Wednesday morning - Gas Station D. Then it is predicted that the user's next refueling time may be Wednesday morning and the refueling station may be Gas Station D.

[0055] It should be noted that the number of predicted gas stations in the embodiments of the present application can be one or more, and the predicted refueling time can also be one or more. Information is sent to one or more gas stations within the predicted time period. At the gas station where the user arrives, a subsequent refueling portrait of the user is generated to determine an accurate preferential strategy, and the preferential strategy is pushed to the refueling user through the staff at the gas station where the user arrives.

[0056] S104. Based on the predicted refueling information, send the user's membership type and refueling trajectory to the predicted refueling station within the predicted refueling time period.

[0057] In an embodiment of the present application, based on the historical refueling information of the refueling user at the predicted refueling station, historical refueling behavior data is determined. Based on the historical information sending time of the refueling user corresponding to the predicted refueling station and the correct rate feedback information corresponding to the historical information sending time, historical information sending data is determined. Based on the historical refueling behavior data and the historical information sending data, a time prediction feature is constructed. The time prediction feature is input into a preset decision tree model to output the information gain corresponding to different time prediction features through the preset decision tree model, and the reference refueling time is determined according to the information gain. Based on the preset time error region, the reference refueling time is expanded into a predicted refueling time period. The user's membership type corresponding to the refueling user is bound to the refueling trajectory, and the bound information is pushed to the predicted refueling station within the reference refueling time period.

[0058] Specifically, after determining the predicted refueling station for the refueling user, search the database for all historical refueling records of this user at this specific station. These records constitute historical refueling behavior data, including detailed information such as the time of each refueling, the refueling volume, the refueling amount, and the type of fuel product. At the same time, search for historical information sending records related to this user at the predicted refueling station, including the time of each information sent to the user, such as the time of preferential push, service reminder, etc., that is, the historical information sending time. For each information sending, there is corresponding accuracy feedback information. Among them, the accuracy feedback information in the embodiments of the present application is the reaction of the user to the received information. For example, whether the user has refueled due to the received information, or whether the user has performed interactive behaviors such as clicking to view the information. If the user has refueled due to the received information, then it can be considered that this information sending is correct and effective, and the feedback information may be marked as "correct"; if the user has no reaction, the feedback information may be marked as "wrong". By collecting these historical information sending times and the corresponding accuracy feedback information, historical information sending data is formed.

[0059] Furthermore, integrate and process the obtained historical refueling behavior data and historical information sending data to construct time prediction features. Among them, the time prediction features in the embodiments of the present application include various time-related information. For example, calculate the average value, median value of the time interval between each refueling of the user at this station, and the standard deviation of the time interval to reflect the central tendency and dispersion degree of the user's refueling time interval; analyze the changes in the refueling frequency of the user in different seasons, on different weekdays, and in different time periods, as well as the time difference distribution between the information sending time and the user's refueling time. Input the constructed time prediction features into the pre-set decision tree model. Among them, the decision tree model is a classification and prediction model based on a tree structure. The pre-set decision tree model in the embodiments of the present application can analyze the input time prediction features.

[0060] Specifically, calculate the information gain corresponding to different time prediction features through the decision tree model. For example, through calculation, it is found that the information gain of the feature "average value of the time interval between refuelings of the user at this station" is relatively high, that is, this feature plays an important role in predicting the user's next refueling time. According to the information gain sizes of each time prediction feature, the decision tree model will select the feature or feature combination with the largest information gain to determine the reference refueling time. For example, through analysis, the decision tree model finds that by combining the average value of the past refueling time interval of the user at this station and the refueling frequency on weekends, the user's next refueling time can be predicted most accurately, thereby obtaining a reference refueling time, such as next Saturday afternoon.

[0061] Furthermore, considering that there may be certain fluctuations in the refueling time of users in actual situations, it is necessary to expand the reference refueling time into a time period, that is, the predicted refueling time period. Specifically, a time fluctuation range is obtained through historical data statistics, and the predicted time point is expanded according to the time fluctuation range. For example, the preset time error range is set to 3 hours before and after. If the reference refueling time is 3 pm next Saturday, then the predicted refueling time period is from 12 noon to 6 pm next Saturday. Such a setting can improve the accuracy and practicality of the prediction and better cover the time range when users may actually refuel.

[0062] Furthermore, bind the user membership type corresponding to the refueling user to the previously determined refueling trajectory. User membership types, such as ordinary members, VIP members, SVIP members, etc., reflect the user's level and the rights and interests enjoyed. The refueling trajectory contains the refueling behavior information of the user at various gas stations in the past, such as gas station location, refueling frequency, refueling amount, refueling time, etc. After binding these two, a data packet containing comprehensive user information is formed. During the determined reference refueling time period, the bound information data packet is pushed to the predicted refueling station. The staff at the predicted refueling station can understand the relevant information of the user during the predicted arrival time period of the user and provide more personalized services for the user.

[0063] S105. In the case where the refueling user arrives at the predicted refueling station, combine the refueling trajectory with the predicted refueling information to construct a user refueling information portrait.

[0064] In an embodiment of the present application, when the user arrives at the predicted refueling station, integrate the previously collected refueling trajectory data with the predicted refueling information. For example, summarize the information such as the refueling location, frequency, amount, and time of the user at various gas stations in the past, and the information such as the predicted refueling time, station, and predicted amount of this time to construct a user refueling information portrait.

[0065] When the user arrives at the predicted refueling station, integrate the previously collected refueling trajectory data with the predicted refueling information. For example, summarize the information such as the refueling location, frequency, amount, and time of the user at various gas stations in the past, and the information such as the predicted refueling time, station, and predicted amount of this time together.

[0066] In one embodiment of the present application, based on the refueling frequencies corresponding to each gas station, the user refueling information portrait is divided into multiple first groups, and the first groups are arranged in chronological order. According to a preset time period, the time-series refueling data corresponding to the user refueling information portrait in each first group is divided into multiple refueling subsequences, and subsequence numbers are assigned. In the first group, based on chronological order, the multiple refueling subsequences are divided into multiple second groups. Based on the subsequence numbers, multiple rounds of random selection are performed in each second group to determine a first reference portrait. Random selection is performed on the first reference portraits corresponding to the multiple first groups respectively to determine a second reference portrait. After multiple rounds of selection, the second reference portraits selected in each round are respectively combined with the non-time-series refueling data in the user refueling information portrait to obtain multiple rounds of reference third portraits, so as to determine the user's current membership level based on the reference third portraits.

[0067] Specifically, the refueling frequency data corresponding to each gas station in the user refueling information portrait is obtained. Based on these refueling frequency data, the user refueling information portrait is divided into multiple first groups. The division method is to group according to the high and low refueling frequencies. For example, the portrait information corresponding to the gas stations with high refueling frequencies is divided into one group, those with medium refueling frequencies into another group, and those with low refueling frequencies into another group. Finally, the first groups are arranged in chronological order. For example, if the first refueling record time of the high-frequency group is the earliest, the medium-frequency group is the second, and the low-frequency group is the latest, then the arrangement order is the high-frequency group first, followed by the medium-frequency group, and finally the low-frequency group.

[0068] Furthermore, for the user refueling information portrait in each first group, its time-series refueling data is divided according to a preset time period. The preset time period in the embodiment of the present application can be set according to the actual analysis needs of the user, for example, set to one month or half a month. Taking a preset time period of one month as an example, assuming the refueling records of a user in a certain first group in the past 6 months, the data for these 6 months is divided into 6 refueling subsequences according to each month as a time period. Each refueling subsequence contains information such as the refueling time and refueling amount of the user at the relevant gas station within that month. After the division is completed, each refueling subsequence is assigned a subsequence number.

[0069] Further, in each first group, based on the chronological order, multiple refueling subsequences are divided into multiple second groups. Specifically, the refueling subsequences for several consecutive months are divided into one group. Suppose there are 6 refueling subsequences in a certain first group, and they are divided into groups according to every two months. Then 3 second groups can be obtained. The first group contains the refueling subsequences labeled 1 and 2, the second group contains the refueling subsequences labeled 3 and 4, and the third group contains the refueling subsequences labeled 5 and 6. In each second group, multiple rounds of random selection are performed based on the subsequence labels. For example, in a second group containing refueling subsequences labeled 1 and 2, two rounds of random selection are performed. In the first round, the refueling subsequence labeled 1 may be randomly selected, and in the second round, the refueling subsequence labeled 2 may be randomly selected. After multiple rounds of random selection, the user refueling information portrait parts corresponding to the refueling subsequences selected each time are determined as the first reference portraits.

[0070] Further, random selection is performed on the first reference portraits corresponding to multiple first groups respectively. For example, there are three first groups, and multiple first reference portraits are determined for each first group through the previous steps. One portrait is randomly selected from the first reference portrait set of the first first group, one portrait is randomly selected from the first reference portrait set of the second first group, and one portrait is randomly selected from the first reference portrait set of the third first group in the same way. These three first reference portraits randomly selected from different first groups are combined to determine the second reference portrait.

[0071] Further, after multiple rounds of selection, the second reference portraits selected in each round are respectively combined with the non-temporal refueling data in the user refueling information portrait to construct the third reference portrait. Among them, the non-temporal refueling data in the embodiments of the present application refers to the user refueling-related information that does not depend on the time order, such as the user's membership type, the user's basic information: age, gender, etc. Based on the reference third portrait, the user's current membership level is determined.

[0072] S106. Based on the membership type and the user refueling information portrait, determine the user's current membership level, and push the corresponding refueling preferential strategy to the refueling user based on the current membership level.

[0073] In one embodiment of the present application, based on a reference third image, user level indicators sorted by importance are determined; wherein, the user level indicators at least include one of refueling frequency, average refueling amount, continuous refueling duration, proportion of refueling during peak periods, and selection of fuel variety diversity. A user level judgment matrix is constructed through the user level indicators and a preset fuzzy scale. In the user level judgment matrix, the first weight corresponding to each user level indicator is determined through a weighted average type fuzzy composition operator. Based on the first weight and the preset fuzzy scale, a weight optimization function is constructed to obtain the second weight corresponding to each user level indicator based on the weight optimization function. Based on the user level indicators and the second weight, a reference level is obtained. The membership type corresponding to the refueling user is classified based on the reference level to determine the current membership level of the user.

[0074] Specifically, by determining the refueling records of a user within a period of time, counting the number of refueling times at each gas station, the refueling frequency is obtained; the sum of the refueling amounts each time is calculated and divided by the number of refueling times to obtain the average refueling amount; the continuous refueling duration is determined from the time interval from the first refueling to the most recent refueling; the proportion of the number of refueling times during the peak period of the gas station to the total number of refueling times is counted to obtain the proportion of refueling during peak periods; the number of different fuel types used by the user is counted and divided by the total number of refueling times to obtain the fuel variety diversity selection index. Then, based on these indicators, the importance of user level evaluation is sorted.

[0075] Furthermore, for every two user level indicators, a value is determined based on their relative importance on the basis of the fuzzy scale. Among them, the preset fuzzy scale in the embodiment of the present application is a scale preset for measuring the relative importance between indicators, such as the 1-9 scale method, where 1 indicates that the two indicators are equally important, and 9 indicates that one indicator is extremely more important than the other. For example, assuming the two indicators of refueling frequency and average refueling amount, if it is considered that the refueling frequency is slightly more important than the average refueling amount, according to the 1-9 scale method, the relationship between them may be assigned a value of 3. By making such comparisons and assignments for all pairs of indicators, a user level judgment matrix is constructed. Among them, the rows and columns of the user level judgment matrix correspond to each user level indicator, and the elements in the matrix represent the relative importance between the corresponding indicators.

[0076] Further, in the constructed user level judgment matrix, a weighted average type fuzzy composition operator is used to determine the first weight corresponding to each user level index by summing the rows or columns of the matrix. For example, for a judgment matrix containing three indicators: refueling frequency, average refueling amount, and continuous refueling duration, after calculation using the weighted average type fuzzy composition operator, the first weight of the refueling frequency is 0.3, the first weight of the average refueling amount is 0.4, and the first weight of the continuous refueling duration is 0.3. These weights reflect the relative importance of each indicator in the user level evaluation under the current judgment matrix.

[0077] Further, based on the previously obtained first weights and the preset fuzzy scale, a weight optimization function is constructed, and by solving the weight optimization function, the second weights corresponding to each user level index are obtained.

[0078] Among them, the following weight optimization function is constructed:

[0079]

[0080] where w1, w2, … w n are the weights corresponding to each user level index; represents the mean value of the i-th user level index; σi represents the standard deviation of the i-th index; if the refueling frequency of the user fluctuates less within a certain period of time, it indicates that their consumption behavior is relatively stable, and the σ i value is small; conversely, the σ i value is large; a ij is the relative importance of indicator i relative to indicator j in the judgment matrix constructed based on the preset fuzzy scale; γ is a balance coefficient used to adjust the relative importance of the judgment matrix consistency error and the consumption behavior stability in the weight optimization.

[0081] After determining the user level indicators and the corresponding second weights, multiply the value of each user level indicator by its corresponding second weight, and then sum these products. Based on the obtained reference level, the membership types corresponding to the refueling users are classified. For example, users with a reference level above 80 points are classified as high-level members, those with a reference level between 60 - 80 points are classified as intermediate-level members, and those with a reference level below 60 points are classified as ordinary members. According to the user's reference level, they are classified into the corresponding membership levels, thereby determining the user's current membership level.

[0082] In one embodiment of the present application, based on the current membership level, a corresponding plurality of fueling discount strategies are determined in the discount information database. The current membership level, the basic information corresponding to the fueling user, and the fueling trajectory are input into a preset discount acceptance model, so as to output the acceptance probabilities of the fueling user for different fueling discount strategies through the discount acceptance model. The plurality of fueling discount strategies are sorted based on the acceptance probabilities, and the discount strategies are sequentially sent to the fueling user based on the sorting order.

[0083] Specifically, a comprehensive discount information database is pre-constructed in the gas station in the embodiment of the present application, which covers various types of fueling discount strategies. These strategies are classified and stored according to the membership level. For example, different levels such as ordinary members, VIP senior members, and SVIP senior members correspond to different discount sets. Each discount strategy includes detailed information such as the discount type, the discount intensity, and the applicable conditions. When it is necessary to push a discount strategy to a certain fueling user, the system first identifies the current membership level of the user. For example, if the user is a VIP member, the system will locate the discount strategy set corresponding to the VIP member in the discount information database. From this set, the system filters out all fueling discount strategies that meet the membership level to form a preliminary discount strategy list for this user.

[0084] Furthermore, a preset discount acceptance model is provided in the embodiment of the present application. The training process of this discount acceptance model is as follows: The historical membership level, the basic information corresponding to the historical fueling user, and the historical fueling trajectory are used as input samples, and the user acceptance probability value corresponding to the input sample is used as the output sample to train the preset neural network model to obtain this discount acceptance model. For the fueling user who currently needs to push a discount strategy, the current membership level, the basic information, and the fueling trajectory are sorted into the input format required by the model, and then the processed information is input into the preset discount acceptance model. The model calculates the acceptance probabilities of the user for different fueling discount strategies respectively according to the mapping relationship it has learned. The plurality of fueling discount strategies are sorted according to the acceptance probabilities output by the discount acceptance model. The discount strategy with a higher acceptance probability is ranked higher in the sorting. According to the sorting order, the fueling discount strategies are sequentially sent to the fueling user. By means of orderly pushing, the discount strategies that the user is most likely to accept are preferentially shown to the user, improving the conversion rate of the discount strategy and the user participation, and at the same time avoiding disturbing the user by pushing too many discount messages to the user at one time.

[0085] Figure 2 It is a schematic structural diagram of a user level calculation device for a gas station provided by an embodiment of the present application. As Figure 2As shown, the user level calculation device for a gas station includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein, the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to: determine the refueling recharge information corresponding to the refueling user based on the gas station user database, and determine the user membership type through the preset membership calculation model and the refueling recharge information; wherein, the refueling recharge information includes the recharge balance and coupon data; determine the license plate information corresponding to the refueling user in the gas station database, match the license plate information with the historical refueling database, and determine the refueling trajectory corresponding to the refueling user; construct a refueling state transition matrix based on the time series data in the refueling trajectory, so as to perform refueling probability distribution processing through the refueling state transition matrix to obtain the predicted refueling information corresponding to the refueling user; based on the predicted refueling information, send the user membership type and the refueling trajectory to the predicted refueling station during the predicted refueling time period; in the case that the refueling user arrives at the predicted refueling station, combine the refueling trajectory and the predicted refueling information to construct a user refueling information portrait; determine the current user membership level based on the membership type and the user refueling information portrait, and push the corresponding refueling preferential strategy to the refueling user based on the current membership level.

[0086] A non-volatile computer storage medium provided by an embodiment of the present application stores computer-executable instructions, and the computer-executable instructions are set to: determine the refueling recharge information corresponding to the refueling user based on the gas station user database, and determine the user membership type through the preset membership calculation model and the refueling recharge information; wherein, the refueling recharge information includes the recharge balance and coupon data; determine the license plate information corresponding to the refueling user in the gas station database, match the license plate information with the historical refueling database, and determine the refueling trajectory corresponding to the refueling user; construct a refueling state transition matrix based on the time series data in the refueling trajectory, so as to perform refueling probability distribution processing through the refueling state transition matrix to obtain the predicted refueling information corresponding to the refueling user; based on the predicted refueling information, send the user membership type and the refueling trajectory to the predicted refueling station during the predicted refueling time period; in the case that the refueling user arrives at the predicted refueling station, combine the refueling trajectory and the predicted refueling information to construct a user refueling information portrait; determine the current user membership level based on the membership type and the user refueling information portrait, and push the corresponding refueling preferential strategy to the refueling user based on the current membership level.

[0087] The various embodiments in the present application are described in a progressive manner. For the same or similar parts among the various embodiments, reference can be made to each other. Each embodiment focuses on the differences from other embodiments. In particular, for the embodiments of the device, equipment, and non-volatile computer storage medium, since they are basically similar to the method embodiments, the description is relatively simple. For the relevant parts, reference can be made to the corresponding descriptions in the method embodiments.

[0088] The above description is only for the embodiments of the present application and is not intended to limit the present application. For those skilled in the art, various modifications and changes can be made to the embodiments of the present application. These modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the various embodiments of the present application.

Claims

1. A method for calculating user levels for gas stations, characterized in that, The method includes: Based on the gas station user database, determine the refueling recharge information corresponding to the refueling user, and determine the user membership type through a preset membership calculation model and the refueling recharge information; wherein, the refueling recharge information includes the recharge balance and coupon data. Determine the license plate information corresponding to the refueling user in the gas station database, match the license plate information with the historical refueling database, and determine the refueling trajectory corresponding to the refueling user. Based on the time series data in the refueling trajectory, construct a refueling state transition matrix to perform refueling probability distribution processing through the refueling state transition matrix, and obtain the predicted refueling information corresponding to the refueling user. Based on the predicted refueling information, send the user membership type and the refueling trajectory to the predicted refueling station during the predicted refueling time period. In the case where the refueling user arrives at the predicted refueling station, combine the refueling trajectory and the predicted refueling information to construct a user refueling information profile. Based on the membership type and the user refueling information profile, determine the current user membership level, and push corresponding refueling preferential policies to the refueling user based on the current membership level.

2. The user level calculation method for a gas station according to claim 1, wherein Before determining the user membership type through the preset membership calculation model and the refueling recharge information, the method further includes: Construct a coupon type indicator function based on the refueling coupon type and the first preset condition value. Construct a coupon quantity indicator function based on the coupon quantity corresponding to the user and the second preset condition value. Construct a balance indicator function based on the recharge balance corresponding to the user and the third preset condition value. Construct the preset membership calculation model through the coupon type indicator function, the coupon quantity indicator function, and the balance indicator function: Result = I1·(I2·SVIP Green+(1 - I2)·I3·SVIP Yellow+(1 - I2)·(1 - I3)·I4·VIP Green+ (1 - I2)·(1 - I3)·(1 - I4)·VIP 黄 ) + (1 - I1)·V 会 ; Among them, Result is the user's membership type; I1 is the coupon type indication function; I2 is the first coupon quantity indication function; I3 is the second coupon quantity indication function; I4 is the balance indication function; SVIP 绿 indicates that the membership level is green SVIP; SVIP 黄 indicates that the membership level is yellow SVIP; VIP 黄 indicates that the membership level is yellow VIP; VIP 绿 indicates that the membership level is green VIP; V 会 indicates that the user is a new member or a recalled member.

3. The user level calculation method for a gas station according to claim 1, characterized in that, The step of matching the license plate information with the historical refueling database to determine the refueling trajectory corresponding to the refueling user specifically includes: Query in the historical refueling database to determine the refueling data corresponding to the refueling user within a preset time period. Obtain the gas station locations in the refueling data, and construct a refueling station map corresponding to the refueling user based on the gas station locations. Connect the gas stations in the refueling station map based on the chronological order of refueling times to construct a refueling path, and perform a first assignment to each gas station in the path based on the number of connection lines. Determine the refueling time and refueling amount corresponding to each gas station, and perform a second assignment to each gas station based on the refueling time and the refueling amount. Based on the first assignment and the second assignment, screen the gas stations in the refueling path, and determine the refueling trajectory corresponding to the refueling user through the screened gas stations and the refueling data of the screened gas stations within the preset time period.

4. A method for calculating user levels for a gas station according to claim 1, wherein Based on the time series data in the refueling trajectory, a refueling state transition matrix is constructed to perform refueling probability distribution processing through the refueling state transition matrix to obtain the predicted refueling information corresponding to the refueling user, specifically including: According to the density distribution corresponding to the time series data, the time series data is clustered into multiple groups, and based on the periodic data in each group, each group is marked with regular refueling data and irregular refueling data; Based on the data corresponding to the regular refueling data mark, a state transition matrix is constructed; wherein, the rows in the state transition matrix correspond to refueling time states, the columns correspond to refueling station states, and the elements in the state transition matrix represent the number of times transferred from the first time-station state to the second time-station state; Based on the ratio between each element in each row of the state transition matrix and the sum of all elements in the current row, the state transition probability matrix corresponding to the state transition matrix is determined; According to the most recent refueling time state and the most recent refueling station state corresponding to the refueling user, an initial element is determined in the state transition probability matrix, and the probability distribution corresponding to the initial element is determined; Through the random number generator and the probability distribution corresponding to the initial element, the predicted refueling time and the predicted refueling station corresponding to the refueling user are determined.

5. A method for calculating user levels for a gas station according to claim 1, characterized in that, Based on the predicted refueling information, during the predicted refueling time period, the user membership type and the refueling trajectory are sent to the predicted refueling station, specifically including: Based on the historical refueling information of the refueling user at the predicted refueling station, the historical refueling behavior data is determined; Based on the historical information sending time of the refueling user corresponding to the predicted refueling station and the correct rate feedback information corresponding to the historical information sending time, the historical information sending data is determined; Based on the historical refueling behavior data and the historical information sending data, a time prediction feature is constructed; The time prediction feature is input into a preset decision tree model, so that the information gain corresponding to different time prediction features is output through the preset decision tree model, and the reference refueling time is determined according to the information gain; Based on a preset time error region, the reference refueling time is expanded into a predicted refueling time period; The user membership type corresponding to the refueling user is bound to the refueling trajectory, and the bound information is pushed to the predicted refueling station during the reference refueling time period.

6. The user level calculation method for a gas station according to claim 1, characterized in that, After combining the refueling trajectory and the predicted refueling information to construct a user refueling information portrait, the method further includes: Based on the refueling frequency corresponding to each refueling station, the user refueling information portrait is divided into multiple first groups, and each of the first groups is arranged in chronological order; According to a preset time period, the sequential refueling data corresponding to the user refueling information portrait in each of the first groups is divided into multiple refueling subsequences, and subsequence numbers are assigned; In the first group, based on chronological order, the multiple refueling subsequences are divided into multiple second groups; Based on the subsequence labels, perform multiple rounds of random selection in each of the second groups to determine a first reference image; Randomly select the first reference images corresponding to the multiple first groups respectively to determine a second reference image; After multiple rounds of selection, combine the second reference images selected in each round with the non-sequential refueling data in the user refueling information image respectively to obtain multiple rounds of reference third images, so as to determine the current membership level of the user based on the reference third images.

7. A user level calculation method for a gas station according to claim 6, characterized in that, The determining of the current membership level of the user based on the membership type and the user refueling information image specifically includes: Based on the reference third images, determine user level indicators sorted by importance; wherein, the user level indicators at least include one of refueling frequency, average refueling amount, continuous refueling duration, proportion of refueling during peak periods, and selection of fuel variety; Construct a user level judgment matrix through the user level indicators and a preset fuzzy scale; In the user level judgment matrix, determine the first weights corresponding to the respective user level indicators through a weighted average type fuzzy composition operator; Based on the first weights and the preset fuzzy scale, construct a weight optimization function to obtain the second weights corresponding to the respective user level indicators based on the weight optimization function; Based on the user level indicators and the second weights, obtain a reference level; Based on the reference level, classify the membership type corresponding to the refueling user to determine the current membership level of the user.

8. A method for calculating user levels for a gas station according to claim 1, characterized in that, The pushing of corresponding refueling preferential policies to the refueling user based on the current membership level specifically includes: Based on the current membership level, determine a corresponding plurality of refueling preferential policies in the preferential information database; Input the current membership level, the basic information corresponding to the refueling user, and the refueling trajectory into a preset preferential acceptance model, so as to output the acceptance probabilities of the refueling user for different refueling preferential policies through the preferential acceptance model; Sort the multiple refueling preferential policies based on the acceptance probabilities, and sequentially send the preferential policies to the refueling user based on the sorting order.

9. A user level calculation device for a gas station, characterized in that, The device includes a memory for storing computer program instructions and a processor for executing the program instructions, wherein when the computer program instructions are executed by the processor, the device is triggered to execute the method according to any one of claims 1-8.

10. A non-volatile computer storage medium stores computer-executable instructions, characterized in that, The computer-executable instructions can execute the method according to any one of claims 1-8.

Images