Energy Retail Marketing Copywriting Generation Methods and Devices

Abstract

This invention discloses a method and apparatus for generating energy retail marketing copy. The method includes: acquiring user information, information on gas station sites located less than a preset threshold from the user, and information on marketing activities conducted at the sites; inputting the user information, site information, and marketing activity information into a pre-trained marketing activity and marketing copy determination model, and outputting a determination result; and displaying the marketing activity and corresponding marketing copy to the user terminal based on the determination result. This method can improve the efficiency and accuracy of personalized energy retail marketing activity recommendations based on user needs, reduce ineffective activity placement, and thus avoid reducing the exposure of the most attractive marketing activities due to limited exhibition space resources. It also improves user experience and marketing activity conversion rates, reduces marketing costs, adapts to rapid changes in the market environment and user needs, and continuously attracts users.

Description

Technical Field

[0001] This invention relates to the field of energy retail technology, and in particular to a method and apparatus for generating energy retail marketing copy. Background Technology

[0002] This section is intended to provide background or context for embodiments of the present invention. The description herein is not intended to imply that it is prior art simply because it is included in this section.

[0003] In the energy retail technology sector, particularly in the refueling business, flexible and diverse marketing campaigns can help boost sales growth, increase user engagement, and enhance brand influence. However, due to the sheer number and diversity of marketing campaigns, existing marketing campaign recommendation and copywriting generation solutions suffer from the following problems:

[0004] (1) Limited booth space: The limited space in the homepage area of ​​the APP (Application), the mini-program operation booth, the pop-up ad, etc., makes it impossible to display all marketing activities at the same time, thus limiting the exposure of the activities.

[0005] (2) Users’ attention is scattered: Users are exposed to a huge amount of information on the Internet every day and their attention is scattered. They may feel stressed due to receiving too much information about activities, which may lead to a decrease in their interest in marketing activities or even resistance.

[0006] (3) Difficulty in meeting personalized needs: Different users have different interests and needs for activities. Large-scale, undifferentiated operation activities and undifferentiated marketing copy generation cannot accurately identify and match user needs, making it difficult to achieve personalized recommendations for each user, resulting in unsatisfactory conversion results.

[0007] (4) Efficiency issues: The sheer number and diversity of marketing campaigns make it inefficient to filter out campaigns that users are interested in from a large number of campaigns.

[0008] (5) Accuracy issues: Rapid changes in the market environment and user needs may cause some activities to become outdated quickly. It is difficult to keep up with market changes and the accuracy of filtering out marketing activities that users are interested in from a large number of marketing activities is low. Summary of the Invention

[0009] This invention provides a method for generating energy retail marketing copy, which improves the efficiency and accuracy of personalized energy retail marketing campaign recommendations based on user needs, reduces ineffective campaign placements, avoids reduced exposure of the most attractive marketing campaigns due to limited booth resources, enhances user experience and marketing campaign conversion rates, reduces marketing costs, adapts to rapid changes in the market environment and user needs, and continuously attracts users. The method includes:

[0010] Obtain user information, gas station information within a preset threshold distance from the user, and information on marketing activities conducted by the gas station; user information includes user tags;

[0011] User information, site information, and marketing campaign information are input into a pre-trained marketing campaign and marketing copy determination model, and the determination result is output. The determination result is the most attractive marketing campaign and corresponding marketing copy for the site for the user. The marketing campaign and marketing copy determination model is trained on a large language model based on historical user information, historical site information, historical marketing campaign information, and a pre-established energy retail marketing terminology database. The energy retail marketing terminology database includes one or any combination of energy retail terms, marketing domain terms, user tag names, and historical marketing campaign terms.

[0012] Based on the determined results, the marketing campaign and corresponding marketing copy are displayed to the user terminal.

[0013] This invention also provides an energy retail marketing copy generation device to improve the efficiency and accuracy of personalized energy retail marketing campaign recommendations based on user needs, reduce ineffective campaign placements, and thus avoid reducing the exposure of the most attractive marketing campaigns due to limited booth resources. This enhances user experience and marketing campaign conversion rates, reduces marketing costs, adapts to rapid changes in the market environment and user needs, and continuously attracts users. The device includes:

[0014] The acquisition module is used to acquire user information, gas station information that is less than a preset threshold away from the user, and information on marketing activities carried out by the stations; user information includes user tags;

[0015] The output module is used to input user information, site information, and marketing campaign information into a pre-trained marketing campaign and marketing copy determination model, and output the determination result. The determination result is the most attractive marketing campaign and corresponding marketing copy for the user on the site. The marketing campaign and marketing copy determination model is trained on a large language model based on historical user information, historical site information, historical marketing campaign information, and a pre-established energy retail marketing terminology database. The energy retail marketing terminology database includes one or any combination of energy retail terms, marketing domain terms, user tag names, and historical marketing campaign terms.

[0016] The display module is used to show marketing campaigns and corresponding marketing copy to user terminals based on the determined results.

[0017] Compared with existing technologies for marketing campaign recommendation and marketing copy generation, this invention obtains user information, information on gas station sites within a preset threshold distance of the user, and information on marketing activities conducted at those sites. User information includes user tags. The user information, site information, and marketing activity information are input into a pre-trained marketing campaign and marketing copy determination model, which outputs a determination result. The determination result is the most attractive marketing campaign and corresponding marketing copy for the user. The marketing campaign and marketing copy determination model is based on historical user information, historical site information, historical marketing activity information, and a pre-established energy retail... The marketing terminology database is generated by training a large language model. The energy retail marketing terminology database includes one or any combination of energy retail terms, marketing field terms, user tag names, and terms from historical marketing campaigns. Based on the determined results, marketing campaigns and corresponding marketing copy are displayed to user terminals. This improves the efficiency and accuracy of personalized energy retail marketing campaign recommendations tailored to user needs, reduces ineffective campaign placements, and avoids reducing the exposure of the most attractive marketing campaigns due to limited booth resources. It enhances user experience and marketing campaign conversion rates, reduces marketing costs, adapts to rapid changes in the market environment and user needs, and continuously attracts users. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. In the drawings:

[0019] Figure 1 This is a flowchart illustrating a method for generating energy retail marketing copy in an embodiment of the present invention;

[0020] Figure 2 A flowchart illustrating a specific example of an energy retail marketing copy generation method provided in this embodiment of the invention;

[0021] Figure 3 This is a schematic diagram of an energy retail marketing copy generation device provided in an embodiment of the present invention;

[0022] Figure 4 This is a schematic diagram illustrating a specific example of an energy retail marketing copy generation device provided in an embodiment of the present invention;

[0023] Figure 5 This is a schematic diagram of a computer device provided in an embodiment of the present invention. Detailed Implementation

[0024] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the embodiments of the present invention will be further described in detail below with reference to the accompanying drawings. Here, the illustrative embodiments of the present invention and their descriptions are used to explain the present invention, but are not intended to limit the present invention.

[0025] The acquisition, storage, use, and processing of data in this application comply with relevant laws and regulations.

[0026] In this document, the term "and / or" merely describes a relationship, indicating that three relationships can exist. For example, A and / or B can represent three cases: A alone, A and B simultaneously, and B alone. Furthermore, the term "at least one" in this document means any combination of at least two of any one or more elements. For example, including at least one of A, B, and C can mean including any one or more elements selected from the set consisting of A, B, and C.

[0027] In the description of this specification, the terms "comprising," "including," "having," and "containing" are open-ended terms, meaning that they include but are not limited to. The terms "an embodiment," "a specific embodiment," "some embodiments," and "for example," etc., refer to specific features, structures, or characteristics described in connection with that embodiment or example that are included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, or characteristics described can be combined in any suitable manner in one or more embodiments or examples. The order of steps involved in the various embodiments is used to illustrate the implementation of this application, and the order of steps is not limited and can be adjusted appropriately as needed.

[0028] To address the problems of existing technologies, this invention provides a method and apparatus for generating marketing copy for energy retail. This method and apparatus, based on site marketing activity data and combined with user basic information, user tags, and historical consumption behavior data, determines user activity preferences. Based on these preferences, it selects the most attractive activity from all available activities and recommends it to the user. After determining the recommended activity, the method, combined with a customized copy generation function, enables AI (Artificial Intelligence) empowerment across the entire process of activity selection and copy creation, thereby improving the conversion rate of marketing activities.

[0029] Figure 1 This is a flowchart of a method for generating energy retail marketing copy provided in an embodiment of the present invention, such as... Figure 1 As shown, the method may include:

[0030] Step 101: Obtain user information, gas station information at locations less than a preset threshold from the user, and information on marketing activities conducted at these gas stations; user information includes user tags;

[0031] Step 102: Input user information, site information, and marketing campaign information into a pre-trained marketing campaign and marketing copy determination model, and output the determination result; the determination result is the most attractive marketing campaign and corresponding marketing copy for the user on the site; the marketing campaign and marketing copy determination model is trained on a large language model based on historical user information, historical site information, historical marketing campaign information, and a pre-established energy retail marketing terminology database; the energy retail marketing terminology database includes one or any combination of: energy retail terms, marketing field terms, user tag names, and historical marketing campaign terms;

[0032] Step 103: Based on the determined results, display the marketing campaign and corresponding marketing copy to the user terminal.

[0033] In one embodiment, user information may include user tags and basic information.

[0034] In one embodiment, user tags may include one or any combination of the user's oil consumption information, membership information, user attribute information, card binding information, non-oil consumption information, recharge consumption information, and discount habit information.

[0035] The fuel consumption information may include: annual consumption frequency, annual fuel transaction amount, annual fuel transaction amount level, annual gasoline transaction amount, first transaction time of the year, most recent fuel transaction date of the year, most recent gasoline transaction date of the year, most recent diesel transaction date of the year, most recent refueling type of the year, most recent transaction time interval of the year, most recent fuel grade of the year, most recent refueling city of the year, most recent refueling station of the year, date of highest refueling frequency of the year, minimum single refueling volume of the year, maximum single refueling volume of the year, most recent payment method of the year, most frequently refueling city in the past six months, most frequently refueling station in the past six months, fuel grade with the highest consumption frequency in the past six months, most commonly used payment method in the past six months, average monthly consumption frequency of the quarter, average monthly consumption amount of the quarter, average monthly consumption liters of the quarter, and average purchase interval of the quarter, or any combination thereof; member information includes: member registration channel, member registration time, member registration province, member registration city, member gender, member age, member birth month, member level, number of linked cards, and most frequently used cities, or any combination thereof.

[0036] Oil consumption information may also include: number of consumers, annual single transaction volume (liters), annual single transaction amount, fuel consumption volume (liters) in the last 30 days, and fuel consumption amount in the last 30 days.

[0037] User attribute information may include: user lifecycle and / or user RFM (Recency-Frequency-Monetary, customer relationship management model) classification, and may also include the gasoline or diesel attribute of the user's refueling, the user's refueling rate, etc.

[0038] Card binding information includes: card activation date, card holding period, activation city, activation location, card balance, card points, and card level, or any combination thereof. Card balance includes reserve funds.

[0039] Non-oil consumption information includes: annual non-oil transaction amount, annual non-oil transaction amount level, annual single non-oil consumption amount, annual most recent non-oil transaction date, annual most frequently purchased non-oil products at gas stations, annual non-oil discount rate level, percentage of non-oil product purchases made alone within six months, quarterly average monthly non-oil transaction amount, and quarterly average monthly non-oil transaction frequency, or any combination thereof. It may also include the percentage of combined purchases of oil and non-oil products within six months.

[0040] Recharge and consumption information includes: the first recharge amount of the year, the most recent recharge amount of the year, the recharge cycle within six months, the ratio of online to offline recharges within six months, the member account points balance, the frequency of member account points consumption within six months, and the total amount of member account points consumption within six months, or any combination thereof, and also includes the time interval of the most recent recharge within the year.

[0041] Promotional information includes: the quarterly card discount amount and / or the quarterly card discount percentage.

[0042] In one embodiment, obtaining user information, gas station information at locations less than a preset threshold, and marketing activity information of the gas stations may include: determining the user's location, identifying gas station locations at locations less than a preset threshold based on the user's location, and obtaining the gas station information; the gas station information includes: station name, station address, business hours, contact number, and station distance.

[0043] In this embodiment of the invention, LBS (Location Based Services) technology can be used to determine the gas station locations near the user. The specific implementation process of determining the gas station locations near the user based on LBS technology can be as follows:

[0044] (1) Establish a gas station location information database, which may include: gas station name, address, geographical coordinates (latitude and longitude), etc.

[0045] (2) Obtaining user location information: Use one or a combination of various methods such as base stations, Wi-Fi (wireless fidelity) hotspots, Bluetooth beacons, and GPS (Global Positioning System) technology to obtain the user's geographical location coordinates;

[0046] (3) Distance calculation and gas station recommendation: Combining the geographical coordinates of the user and the gas station, a distance-based search algorithm is used to find the gas stations around the user.

[0047] In this embodiment, obtaining user information, gas station information at locations less than a preset threshold, and marketing activity information conducted by these gas stations can include: after determining the gas station information at locations less than the preset threshold, sending a marketing activity information acquisition request to the gas station; receiving marketing activity information returned by the gas station after verifying the request; or receiving marketing activity information sent by the gas station after detecting user actions through the fuel pump or gas station convenience store; the user actions include: user entering the station, user activating a fuel card, user recharging a fuel card, user consuming fuel, user consuming non-fuel products, and user redeeming electronic coupons; the marketing activity categories included in the marketing activity information include: single-item promotion, timed coupon distribution, points distribution, payment discounts, combination promotions, and gift promotions, or any combination thereof; one marketing activity category corresponds to one or more discount forms; the discount forms include threshold-based discounts, discounted unit prices, unit price deductions, unit price discounts, discounted total prices, total price deductions, and total price discounts, or any combination thereof.

[0048] Real-time data collection can be conducted through multiple touchpoints, including apps (the apps used after copy generation or apps that cooperate with the app, such as bank apps), mini-programs, gas station sites, fuel pumps, and dedicated convenience stores. This includes: acquiring information on user actions such as entering the station, activating fuel cards, recharging fuel cards, fuel consumption, non-fuel consumption, and redeeming e-coupons through gas station sites, fuel pumps, and gas station convenience stores, as well as information on marketing activities currently underway. Marketing activities include 15 categories such as timed coupon distribution, points distribution, payment discounts, bundled promotions, and gift promotions, and the current discount format for each activity can be obtained. Each activity category supports different discount formats. For example, the single-item promotion category can set a minimum purchase threshold (e.g., a minimum quantity or a minimum amount), and can also set six formats: discounted unit price, unit price deduction, unit price discount, discounted total price, total price deduction, and total price discount.

[0049] To improve the accuracy of energy retail marketing campaign identification and copywriting generation, in one embodiment, after obtaining user information, gas station information at locations less than a preset threshold, and marketing campaign information at these locations, the process may further include: analyzing the fields in the user information, gas station information at locations less than a preset threshold, and marketing campaign information to identify missing and outliers; filling in missing values ​​with constant or statistical values; and removing or replacing outliers with constant or statistical values.

[0050] Observe the fields in the collected data to identify missing and outlier values, and remove fields irrelevant to the model input. For missing and outlier values ​​in the model input fields, handle them using methods such as constant imputation and statistical value imputation.

[0051] Figure 2 A flowchart illustrating a specific example of an energy retail marketing copy generation method provided in this embodiment of the invention, such as... Figure 2 As shown, in one embodiment, the energy retail marketing copy generation method may further include:

[0052] Step 201: Obtain historical user information, historical site information, and historical marketing campaign information;

[0053] Step 202: Establish an energy retail marketing terminology database based on historical user information, historical site information, historical marketing campaign information, and a set of professional terms in the petroleum industry;

[0054] Step 203: Determine the historical user information, historical site information, historical marketing activity information, and energy retail marketing terminology database with a preset quantity and preset ratio as the training set, and pre-train the initial large language model to obtain the pre-trained large language model; the initial large language model is a large language model with preset parameters.

[0055] Step 204: Validate the pre-trained large model and fine-tune the proportions of the training set based on the validation results;

[0056] Step 205: Train the pre-trained large language model using the fine-tuned training set to generate a marketing campaign and marketing copy determination model. The initial large language model, the pre-trained large language model, and the marketing campaign and marketing copy determination model are all used for: de-identifying the input data (identity document); concatenating the de-identified input data before inputting it into the input layer; and using prompts to describe the target task. The target task is: based on historical user information, historical site information, historical marketing campaign information, and an energy retail marketing terminology database, recommend the most attractive marketing campaigns and corresponding marketing copy conducted by the site to the user.

[0057] In this embodiment, an energy retail marketing terminology database (a professional terminology database for the marketing field and the petroleum industry) was pre-built, including energy industry terms such as "refueling," "oil coupons," "non-oil coupons," "premium cards," "98#," and "95#"; marketing field terms such as "discount coupons," "single-item promotions," and "annual non-oil discount rate levels" and "annual non-oil single transaction amount"; and historical event terms such as "Extraordinary 10 Discounts" and "Auspicious Rabbit Brings Blessings, Heading for Oil Gifts." Personalized recommendations for marketing activities and customized creation of marketing copy were achieved within the specific operational scenarios of the energy industry.

[0058] In one embodiment, the training set can be updated in real time, including an energy retail marketing terminology database, to avoid the selected activities becoming outdated due to rapid changes in the market environment and user needs, which would prevent the accurate selection of marketing activities that users are interested in from a large number of marketing activities.

[0059] In this embodiment, the initial large language model is trained based on the preprocessed data. The training process can be divided into: initial large language model pre-training + fine-tuning + prompt (contingency words). This learning paradigm has significant advantages and can help train an excellent model.

[0060] Large language models possess powerful text representation and understanding capabilities, and recommendation scenarios involve many features, such as product titles and various text descriptions. However, the problem of new items in recommendation systems is particularly challenging because traditional recommendation systems are primarily ID-based. Each new item must have an ID, and a representation of the item is learned based on this ID, ultimately leading to recommendations. However, if a general model existed that could understand language exceptionally well, it could directly describe items using language, removing the ID and obtaining the text representation of the item description as the item's representation. Specifically, for an item sequence, each item contains not only an ID but may also contain some category features. The model's function is to construct an item sentence, concatenating all item-related attributes, such as title, brand, and price, into a long sentence.

[0061] The core idea of ​​prompt learning is to use a prompt to describe each task. Taking sentiment analysis as an example, traditionally, this type of task involved classifying the input text to predict its sentiment (positive or negative), a more discriminative approach. Now, given an input, a prompt describes the sentiment analysis task, and the model decodes and generates a result. This generated result is then used to determine the sentiment's positive or negative tendency. In summary, the core idea of ​​prompt learning is to construct a prompt to describe the task and use a generative model to generate the task result.

[0062] This invention utilizes a large language model for data de-identification: all item representations are converted into text, and the powerful text understanding capabilities of the language model are used to comprehend and represent the items. In this embodiment, for example, member tags contain a wealth of attribute information such as the current member's gender, age, birth month, annual consumption frequency, fuel expenditure in the last 30 days, and annual non-fuel transaction amount. During the model de-identification process, all the current member's information is concatenated to construct an item sentence, forming a long sentence, which is then input into the large model for learning. Similarly, the marketing campaign information of the current site is constructed by concatenating attributes such as the start and end times of the marketing campaign, the target audience, and discount rules, which serve as input information for the large model's learning. The large model then uses the learned user and marketing campaign representations to recommend the most attractive marketing campaigns for the current user.

[0063] Adjust the parameters of the pre-trained large language model: clarify that the large model needs to complete the core task of recommending the most attractive marketing activities of the current site based on member tags and historical site consumption behavior data, refine the instructions, optimize the language, and avoid ambiguity.

[0064] Construct a prompt to describe the task, and use a generative model to generate the task results. After the model generates the output, monitor and evaluate it to ensure that the output meets expectations, and adjust the feature selection, item sentence, or prompt construction as needed.

[0065] Compared to general large models, the large language model provided in this embodiment of the invention has the advantages of being more professional, customizable to specific needs, and having better data security and privacy protection. Its application scenarios can mainly include: oil consumption, non-oil consumption, and non-oil promotions for oil (offline purchases or online orders of non-oil products with oil coupons).

[0066] Compared to traditional models, the advantages of the large language model provided in this embodiment of the invention are specifically as follows:

[0067] ① Powerful modeling capabilities: The large language model has strong capabilities, thanks to the model structure such as Transformer used by the large language model, which itself has very powerful modeling capabilities.

[0068] ② Excellent learning paradigm: The learning paradigm of large language model pre-training + fine-tuning + prompt also has great advantages and can help train excellent models.

[0069] ③ Use large models to understand and represent items (de-ID): Large language models have powerful text representation and understanding capabilities.

[0070] ④ Establish a large-scale recommendation model paradigm: 1) Use a unified text representation to solve the problem of ID dependence, which can ignore cross-domain and interaction issues, and also solve many long-tail problems well; 2) Use a unified prompt task to achieve open-ending tasks, use a trained language model to achieve cross-domain tasks, and finally obtain a basic model with open-ending tasks and domains.

[0071] and:

[0072] ①Regarding representation:

[0073] The core idea is to remove the ID, converting all item representations into text, and using the powerful text understanding capabilities of language models to understand and represent the items.

[0074] ② Regarding Prompt learning:

[0075] The advantage of prompt learning lies in its ability to elevate learning from the sample level to the task level. Traditionally, supervised, unsupervised, and contrastive learning all operate at the sample level. With prompt learning, in addition to samples, there's an extra input—the prompt—describing the task. This transforms model learning into task-level learning, a higher level of abstraction compared to sample-level learning, making this paradigm more advantageous. Compared to traditional discriminative methods, constructing prompts for task description and using generative models to generate task results achieves this leap from sample-level to task-level learning.

[0076] In one embodiment, displaying marketing activities and corresponding marketing copy to a user terminal based on the determined result may include: displaying marketing activities and corresponding marketing copy to a user terminal when actions such as user registration, login, real-name authentication, page browsing, button clicking, following a public account, or scanning a QR code to join a group are detected based on the determined result.

[0077] In this embodiment, marketing campaigns and copywriting can be implemented when user registration, login, real-name authentication, page browsing, button clicking, following official accounts, scanning QR codes to join groups, and other similar actions are detected.

[0078] This invention also proposes an energy retail marketing copy generation device, the principle of which is similar to the energy retail marketing copy generation method, and will not be described in detail here.

[0079] Figure 3 This is a schematic diagram of an energy retail marketing copy generation device provided in an embodiment of the present invention, such as... Figure 3 As shown, the energy retail marketing copy generation device may include:

[0080] The acquisition module 301 is used to acquire user information, gas station information that is less than a preset threshold away from the user, and information on marketing activities carried out by the stations; the user information includes user tags;

[0081] Output module 302 is used to input user information, site information, and marketing activity information into a pre-trained marketing activity and marketing copy determination model, and output the determination result; the determination result is the most attractive marketing activity and corresponding marketing copy for the user on the site; the marketing activity and marketing copy determination model is trained on a large language model based on historical user information, historical site information, historical marketing activity information, and a pre-established energy retail marketing terminology database; the energy retail marketing terminology database includes one or any combination of: energy retail terms, marketing domain terms, user tag names, and historical marketing activity terms;

[0082] The display module 303 is used to display marketing activities and corresponding marketing copy to the user terminal based on the determined results.

[0083] In one embodiment, user tags include one or any combination of user's oil consumption information, membership information, user attribute information, card binding information, non-oil consumption information, recharge consumption information, and discount habit information;

[0084] Oil consumption information includes: annual consumption frequency, annual oil transaction amount, annual oil transaction amount level, annual gasoline transaction amount, first transaction time of the year, most recent oil transaction date of the year, most recent gasoline transaction date of the year, most recent diesel transaction date of the year, most recent refueling type of the year, most recent transaction time interval of the year, most recent refueling grade of the year, most recent refueling city of the year, most recent refueling station of the year, date of highest refueling frequency of the year, minimum single refueling volume of the year, maximum single refueling volume of the year, most recent payment method of the year, most frequently refueling city in the past six months, most frequently refueling station in the past six months, most frequently consumed oil grade of the past six months, most commonly used payment method in the past six months, average monthly consumption frequency of the quarter, average monthly consumption amount of the quarter, average monthly consumption liters of the quarter, and average purchase interval of the quarter, or any combination thereof.

[0085] Member information includes: member registration channel, member registration time, member registration province, member registration city, member gender, member age, member birth month, member level, number of bank cards linked, and one or any combination of frequently used cities;

[0086] User attribute information includes: user lifecycle and / or user customer relationship management RFM model classification;

[0087] Card binding information includes: card activation time, card holding period, card activation city, card activation location, card balance, card points, card level, or any combination thereof;

[0088] Non-oil consumption information includes: annual non-oil transaction amount, annual non-oil transaction amount level, annual single non-oil consumption amount, annual most recent non-oil transaction date, annual most frequently purchased non-oil products at gas stations, annual non-oil discount rate level, percentage of non-oil product purchases made alone within the past six months, quarterly average monthly non-oil transaction amount, and quarterly average monthly non-oil transaction frequency, or any combination thereof.

[0089] Recharge and consumption information includes: the first recharge amount of the year, the most recent recharge amount of the year, the recharge cycle within six months, the ratio of online to offline recharges within six months, the member account points balance, the frequency of member account points consumption within six months, and the total amount of member account points consumption within six months, or any combination thereof.

[0090] Promotional information includes: the quarterly card discount amount and / or the quarterly card discount percentage.

[0091] In one embodiment, the acquisition module 301 is specifically used for:

[0092] Determine the user's location and identify gas station locations that are less than a preset threshold away from the user;

[0093] Obtain the site information of the gas station; the site information includes: site name, site address, business hours, contact number, and site distance.

[0094] In this embodiment, the acquisition module 301 is specifically used for:

[0095] After identifying gas station locations that are less than a preset threshold away from the user, a marketing activity information request is sent to the gas station location.

[0096] Receive marketing activity information returned by the gas station after its request for marketing activity information has been verified;

[0097] or,

[0098] The system receives marketing activity information sent by gas stations after detecting user actions via fuel pumps or gas station convenience stores. These user actions include: user entering the station, user activating a fuel card, user topping up a fuel card, user consuming fuel, user consuming non-fuel products, and user redeeming electronic coupons. The marketing activity categories included in the information include: single-item promotions, timed coupon distribution, points distribution, payment discounts, bundled promotions, and gift promotions, or any combination thereof. Each marketing activity category corresponds to one or more discount formats. These discount formats include: minimum purchase threshold discounts, discounted unit price, unit price deduction, unit price discount, discounted total price, total price deduction, and total price discount, or any combination thereof.

[0099] In one embodiment, the energy retail marketing copy generation device further includes: a preprocessing module, used for:

[0100] Analyze fields in user information, gas station information at locations less than a preset threshold, and marketing activities conducted by these gas stations to identify missing and outlier values.

[0101] Missing values ​​are filled by using constant values ​​or statistical values.

[0102] Outliers can be removed or replaced with constant or statistical values.

[0103] Figure 4 This is a schematic diagram illustrating a specific example of an energy retail marketing copy generation device provided in an embodiment of the present invention, such as... Figure 4 As shown, in one embodiment, the energy retail marketing copy generation device further includes: a training module 401, used for:

[0104] Obtain historical user information, historical site information, and historical marketing campaign information;

[0105] An energy retail marketing terminology database will be established based on historical user information, historical site information, historical marketing campaign information, and a set of professional terms in the petroleum industry.

[0106] The training set is determined by setting a preset quantity and preset ratio of historical user information, historical site information, historical marketing activity information, and energy retail marketing terminology database. The initial large language model is pre-trained to obtain a pre-trained large language model. The initial large language model is a large language model with preset parameters.

[0107] Validate the pre-trained large model and fine-tune the allocation of the training set based on the validation results;

[0108] The pre-trained large language model is trained using a finely tuned training set to generate a marketing campaign and marketing copy determination model. The initial large language model, the pre-trained large language model, and the marketing campaign and marketing copy determination model are all used for: de-identifying and ID-izing the input data; concatenating the de-identified input data before inputting it into the input layer; and using prompt words to describe the target task. The target task is: based on historical user information, historical site information, historical marketing campaign information, and an energy retail marketing terminology database, to recommend the most attractive marketing campaigns and corresponding marketing copy for users.

[0109] In one embodiment, the display module 303 is used for:

[0110] Based on the determined results, when user actions such as registration, login, real-name authentication, page browsing, button clicking, following official accounts, and scanning QR codes to join groups are detected, marketing activities and corresponding marketing copy will be displayed to the user's terminal.

[0111] Compared with existing technologies for marketing campaign recommendation and marketing copy generation, this invention obtains user information, information on gas station sites within a preset threshold distance of the user, and information on marketing activities conducted at those sites. User information includes user tags. The user information, site information, and marketing activity information are input into a pre-trained marketing campaign and marketing copy determination model, which outputs a determination result. The determination result is the most attractive marketing campaign and corresponding marketing copy for the user. The marketing campaign and marketing copy determination model is based on historical user information, historical site information, historical marketing activity information, and a pre-established energy retail... The marketing terminology database is generated by training a large language model. The energy retail marketing terminology database includes one or any combination of energy retail terms, marketing field terms, user tag names, and terms from historical marketing campaigns. Based on the determined results, marketing campaigns and corresponding marketing copy are displayed to user terminals. This improves the efficiency and accuracy of personalized energy retail marketing campaign recommendations tailored to user needs, reduces ineffective campaign placements, and avoids reducing the exposure of the most attractive marketing campaigns due to limited booth resources. It enhances user experience and marketing campaign conversion rates, reduces marketing costs, adapts to rapid changes in the market environment and user needs, and continuously attracts users.

[0112] In summary, this invention provides a method and apparatus for personalized recommendation and customized copywriting generation for marketing campaigns based on site location. Compared with traditional marketing campaign recommendation methods, this invention has the following advantages:

[0113] (1) Avoid booth resource limitations: Personalize marketing activities for different users and do not display marketing activities that users are not interested in, so as to avoid reducing the exposure of the most attractive marketing activities due to limited booth resources.

[0114] (2) Overcome the problem of user attention being distracted: Do not provide too much activity information, only recommend marketing activities that users are interested in, so as to avoid users’ interest declining or even developing resistance.

[0115] (3) Meeting personalized needs: Taking into account the different interests and needs of different users, avoiding large-scale, undifferentiated operational activities, accurately identifying and matching user needs, and achieving personalized recommendations for each user to improve conversion rates.

[0116] (4) High efficiency: It can efficiently filter out marketing activities that users are interested in from a large number and diverse range of marketing activities.

[0117] (5) High accuracy: By updating the training set in real time, including the energy retail marketing terminology database, it can adapt to the rapid changes in the market environment and user needs, avoid marketing activities from becoming outdated, and accurately select marketing activities that users are interested in from a large number of marketing activities.

[0118] This invention also provides a computer device. Figure 5 This is a schematic diagram of a computer device in an embodiment of the present invention. The computer device 500 includes a memory 510, a processor 520, and a computer program 530 stored in the memory 510 and executable on the processor 520. When the processor 520 executes the computer program 530, it implements the above-mentioned energy retail marketing copy generation method.

[0119] This invention also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described energy retail marketing copy generation method.

[0120] This invention also provides a computer program product, which includes a computer program that, when executed by a processor, implements the above-described energy retail marketing copy generation method.

[0121] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0122] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0123] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0124] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0125] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for generating energy retail marketing copy, characterized in that, include: Obtain user information, gas station information that is less than a preset threshold away from users, and information on marketing activities carried out by the stations; User information includes user tags; Input user information, site information, and marketing campaign information into a pre-trained marketing campaign and marketing copy determination model, and output the determination result; The results are determined as the most attractive marketing campaigns and corresponding marketing copy conducted by the site for the user; the marketing campaign and marketing copy determination model is obtained by training a large language model based on historical user information, historical site information, historical marketing campaign information and a pre-established energy retail marketing terminology database. The energy retail marketing terminology database includes: energy retail terms, marketing field terms, user tag names, and terms from historical marketing campaigns, or any combination thereof. Based on the determined results, the marketing campaign and corresponding marketing copy are displayed to the user terminal.

2. The method as described in claim 1, characterized in that, The user tags include one or any combination of the user's oil consumption information, membership information, user attribute information, card binding information, non-oil consumption information, recharge consumption information, and discount habit information; Oil consumption information includes: annual consumption frequency, annual oil transaction amount, annual oil transaction amount level, annual gasoline transaction amount, first transaction time of the year, most recent oil transaction date of the year, most recent gasoline transaction date of the year, most recent diesel transaction date of the year, most recent refueling type of the year, most recent transaction time interval of the year, most recent refueling grade of the year, most recent refueling city of the year, most recent refueling station of the year, date of highest refueling frequency of the year, minimum single refueling volume of the year, maximum single refueling volume of the year, most recent payment method of the year, most frequently refueling city in the past six months, most frequently refueling station in the past six months, most frequently consumed oil grade of the past six months, most commonly used payment method in the past six months, average monthly consumption frequency of the quarter, average monthly consumption amount of the quarter, average monthly consumption liters of the quarter, and average purchase interval of the quarter, or any combination thereof. Member information includes: member registration channel, member registration time, member registration province, member registration city, member gender, member age, member birth month, member level, number of bank cards linked, and one or any combination of frequently used cities; User attribute information includes: user lifecycle and / or user customer relationship management RFM model classification; Card binding information includes: card activation time, card holding period, card activation city, card activation location, card balance, card points, card level, or any combination thereof; Non-oil consumption information includes: annual non-oil transaction amount, annual non-oil transaction amount level, annual single non-oil consumption amount, annual most recent non-oil transaction date, annual most frequently purchased non-oil products at gas stations, annual non-oil discount rate level, percentage of non-oil product purchases made alone within the past six months, quarterly average monthly non-oil transaction amount, and quarterly average monthly non-oil transaction frequency, or any combination thereof. Recharge and consumption information includes: the first recharge amount of the year, the most recent recharge amount of the year, the recharge cycle within six months, the ratio of online to offline recharges within six months, the member account points balance, the frequency of member account points consumption within six months, and the total amount of member account points consumption within six months, or any combination thereof. Promotional information includes: the quarterly card discount amount and / or the quarterly card discount percentage.

3. The method as described in claim 1, characterized in that, Obtain user information, site information of gas stations located less than a preset threshold from users, and information on marketing activities conducted at these stations, including: Determine the user's location and identify gas station locations that are less than a preset threshold away from the user; Obtain the site information of the gas station; the site information includes: site name, site address, business hours, contact number, and site distance.

4. The method as described in claim 3, characterized in that, Obtain user information, site information of gas stations located less than a preset threshold from users, and information on marketing activities conducted at these stations, including: After identifying gas station locations that are less than a preset threshold away from the user, a marketing activity information request is sent to the gas station location. Receive marketing activity information returned by the gas station after its request for marketing activity information has been verified; or, The system receives marketing activity information sent by gas stations after detecting user actions via fuel pumps or gas station convenience stores. These user actions include: user entering the station, user activating a fuel card, user topping up a fuel card, user consuming fuel, user consuming non-fuel products, and user redeeming electronic coupons. The marketing activity categories included in the information include: single-item promotions, timed coupon distribution, points distribution, payment discounts, bundled promotions, and gift promotions, or any combination thereof. Each marketing activity category corresponds to one or more discount formats. These discount formats include: minimum purchase threshold discounts, discounted unit price, unit price deduction, unit price discount, discounted total price, total price deduction, and total price discount, or any combination thereof.

5. The method as described in claim 1, characterized in that, After obtaining user information, information on gas station sites located less than a preset threshold from the user, and information on marketing activities conducted at those sites, the following is also included: Analyze fields in user information, gas station information at locations less than a preset threshold, and marketing activities conducted by these gas stations to identify missing and outlier values. Missing values ​​are filled by using constant values ​​or statistical values. Outliers can be removed or replaced with constant or statistical values.

6. The method as described in claim 1, characterized in that, Also includes: Obtain historical user information, historical site information, and historical marketing campaign information; An energy retail marketing terminology database will be established based on historical user information, historical site information, historical marketing campaign information, and a set of professional terms in the petroleum industry. The training set is determined by setting a preset quantity and preset ratio of historical user information, historical site information, historical marketing activity information, and energy retail marketing terminology database. The initial large language model is pre-trained to obtain a pre-trained large language model. The initial large language model is a large language model with preset parameters. Validate the pre-trained large model and fine-tune the allocation of the training set based on the validation results; The pre-trained large language model is trained using a finely tuned training set to generate a marketing campaign and marketing copy determination model. The initial large language model, the pre-trained large language model, and the marketing campaign and marketing copy determination model are all used for: de-identifying and ID-izing the input data; concatenating the de-identified input data before inputting it into the input layer; and using prompt words to describe the target task. The target task is: based on historical user information, historical site information, historical marketing campaign information, and an energy retail marketing terminology database, to recommend the most attractive marketing campaigns and corresponding marketing copy for users.

7. The method as described in claim 1, characterized in that, Based on the determined results, the marketing campaign and corresponding marketing copy are displayed to the user terminal, including: Based on the determined results, when user actions such as registration, login, real-name authentication, page browsing, button clicking, following official accounts, and scanning QR codes to join groups are detected, marketing activities and corresponding marketing copy will be displayed to the user's terminal.

8. An energy retail marketing copy generation device, characterized in that, include: The acquisition module is used to acquire user information, gas station information that is less than a preset threshold away from the user, and information on marketing activities carried out by the stations. User information includes user tags; The output module is used to input user information, site information, and marketing campaign information into a pre-trained marketing campaign and marketing copy determination model, and output the determination result. The results are determined as the most attractive marketing campaigns and corresponding marketing copy conducted by the site for the user; the marketing campaign and marketing copy determination model is obtained by training a large language model based on historical user information, historical site information, historical marketing campaign information and a pre-established energy retail marketing terminology database. The energy retail marketing terminology database includes: energy retail terms, marketing field terms, user tag names, and terms from historical marketing campaigns, or any combination thereof. The display module is used to show marketing campaigns and corresponding marketing copy to user terminals based on the determined results.

9. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the method of any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the method of any one of claims 1 to 7.

11. A computer program product, characterized in that, The computer program product includes a computer program that, when executed by a processor, implements the method of any one of claims 1 to 7.

Images