A refueling recommendation method, device and computer readable medium

Abstract

This application provides a refueling recommendation method, device, and computer-readable medium. After obtaining a user's refueling demand information, the method filters candidate gas stations based on this information, then calculates fuel price predictions, coupon recommendations, service quality information, and distance information. It further obtains a comprehensive recommendation score for each candidate gas station and finally determines the recommended gas station based on this comprehensive score. This method does not rely on simple user preferences but instead calculates multiple influencing factors, including fuel price predictions, coupon recommendations, and service quality information, to achieve more personalized refueling service recommendations. This results in recommendations with better convenience, economy, and service quality, effectively improving recommendation accuracy and better meeting users' actual refueling needs.

Description

Technical Field

[0001] This application relates to the field of information technology, and in particular to a refueling recommendation method, device, and computer-readable medium. Background Technology

[0002] With the rapid growth of car ownership and the widespread adoption of mobile internet, car owners' demands for refueling services are becoming increasingly diversified. Simultaneously, internet platform companies operating refueling services have emerged in the market. These companies can access data from refueling service providers to recommend suitable gas stations to users. However, current refueling recommendation schemes are generally based on simple user preferences, such as distance and price preferences, recommending gas stations that meet those preferences. Such recommendation schemes often only provide basic recommendations and fail to meet users' comprehensive requirements for convenience, affordability, and service quality in refueling services.

[0003] In addition, the existing refueling recommendation scheme has the following problems: The heterogeneity of multi-source data: Different refueling service providers have different data formats, interface standards, and update frequencies. When the platform obtains data from these different data sources, the data quality varies, which affects the accuracy of subsequent recommendation processing results.

[0004] Insufficient price prediction capability: Existing technology mainly relies on simple comparisons based on historical price data, which cannot predict future price trends, and users cannot plan the best time to refuel in advance.

[0005] Low coupon matching efficiency: Existing technology lacks a suitable fuel coupon distribution mechanism, resulting in low coupon usage and making it difficult for users to obtain the maximum discount.

[0006] Service quality assessment is lagging: Existing technologies mainly rely on user feedback after the fact, lacking a real-time service quality monitoring and assessment mechanism, making it difficult to assess the service quality of gas stations in a timely and accurate manner. Summary of the Invention

[0007] One object of this application is to provide a refueling recommendation method, device, and computer-readable medium to solve the problems existing in the prior art.

[0008] To achieve the above objectives, embodiments of this application provide a refueling recommendation method, the method comprising: Obtain users' refueling needs information; Based on the refueling demand information, filter candidate gas stations; Calculate fuel price forecasts, coupon recommendations, service quality, and distance information for candidate gas stations; Based on the fuel price forecast information, coupon recommendation information, service quality information, and distance information, calculate the comprehensive recommendation score for the candidate gas stations; Based on the comprehensive recommendation score, recommended gas stations are determined from the candidate gas stations, and the recommended gas stations are pushed to the user.

[0009] Furthermore, the calculation of fuel price forecast information for candidate gas stations includes: Obtain temporal, regional, supply and demand, market, and historical fuel price characteristics of candidate gas stations to construct multidimensional feature data; Construct a deep learning model based on LSTM or Transformer architecture, and train the deep learning model using the multidimensional feature data to obtain a price prediction model; The price prediction model is used to obtain fuel price prediction information for candidate gas stations.

[0010] Furthermore, the method also includes: Predict users' expected refueling time based on user behavior information; Calculate coupon recommendation information and issue coupons to the target users based on the coupon recommendation information, wherein the target users are users whose expected refueling time is within a preset time range.

[0011] Furthermore, coupon recommendation information is calculated, including: Calculate the matching degree between coupons and users based on user preference information and coupon conditions; Calculate the commercial value of the coupon based on the expected transaction amount after using the coupon and the platform's revenue rate; Calculate the probability of using the coupon based on its relevant usage characteristics; Based on the coupon's matching degree, commercial value, probability of use, and timeliness factor, coupon recommendation information is calculated, wherein the timeliness factor is positively correlated with the coupon's validity period.

[0012] Furthermore, the method also includes: When obtaining users' refueling needs, based on the matching degree between coupons and users, a prompt message about the target coupon is pushed to the user, wherein the target coupon is a coupon whose matching degree meets the push conditions.

[0013] Furthermore, the service quality information of the candidate gas stations is calculated, including: Obtain service efficiency indicators, service attitude indicators, facility quality indicators, and price accuracy indicators for candidate gas stations; Calculate the standardized scores corresponding to the service efficiency index, service attitude index, facility quality index, and price accuracy index respectively; Based on the standardized score, the service quality information of the candidate gas stations is calculated.

[0014] Further, based on the standardized score, the service quality information of the candidate gas stations is calculated, including: Based on the standardized scores, dimensional weights, and credibility weights corresponding to service efficiency indicators, service attitude indicators, facility quality indicators, and price accuracy indicators, a weighted calculation is performed to obtain the service quality information of the candidate gas stations. The credibility weight is positively correlated with the reliability of the data source for the corresponding indicator.

[0015] Further, based on the comprehensive recommendation score, recommended gas stations are determined from the candidate gas stations, and the recommended gas stations are pushed to the user, including: If the service quality information of a candidate gas station meets the recommendation weight adjustment conditions, the overall recommendation score of the candidate gas station will be adjusted. Based on the adjusted comprehensive recommendation score, recommended gas stations are determined from the candidate gas stations and pushed to the user.

[0016] Furthermore, if the service quality information of the candidate gas station meets the recommendation weight adjustment conditions, the comprehensive recommendation score of the candidate gas station is adjusted, including: The recommendation weight coefficient is determined based on the numerical range of the service quality information of the candidate gas stations; The adjusted comprehensive recommendation score is obtained by multiplying the comprehensive recommendation score by the recommendation weight coefficient.

[0017] Furthermore, based on the fuel price forecast information, coupon recommendation information, service quality information, and distance information, a comprehensive recommendation score for the candidate gas stations is calculated, including: Price weight, discount weight, service weight, and distance weight are determined based on user preference information; Based on the fuel price forecast information, coupon recommendation information, service quality information, and distance information, and combined with the corresponding price weight, discount weight, service weight, and distance weight, a weighted calculation is performed to obtain the comprehensive recommendation score of the candidate gas stations.

[0018] Furthermore, the method also includes: Retrieve data to be processed from multiple data sources for the same data object; Based on the mapping rules, the data to be processed from different data sources is converted into the same data format; Calculate the credibility coefficient of each data source; Based on the credibility coefficient and weight of each data source, the data to be processed for each identical data object is fused to obtain fused data.

[0019] Furthermore, the credibility coefficients of each data source are calculated, including: Obtain data quality scores, supplier historical performance, data update frequency, and reporting success rate for each data source; The credibility coefficient of each data source is calculated based on the data quality score, the supplier's historical performance, the data update frequency, and the reporting normality rate.

[0020] Furthermore, obtain the data quality scores for each data source: Based on the data integrity, timeliness, accuracy, and consistency of the data to be processed, a weighted calculation is performed to obtain the data quality score for each data source.

[0021] Furthermore, obtaining users' refueling needs information includes at least one of the following: Based on the user's input of refueling-related actions, obtain the user's refueling demand information; The best time to refuel is predicted based on fuel price forecasts, and the user's refueling demand information is generated based on the best time to refuel. Predict the user's expected refueling time based on user behavior information, and generate the user's refueling demand information based on the expected refueling time.

[0022] Furthermore, after determining recommended gas stations from the candidate gas stations based on the comprehensive recommendation score and pushing the recommended gas stations to the user, the process also includes: Obtain the user's refueling order information based on the recommended gas stations; Based on the refueling order information, a reservation order is generated, and the recommended gas station is assigned to the fuel nozzle position for the reservation time period corresponding to the reservation order. Send order confirmation and navigation information to users.

[0023] Some embodiments of this application also provide a refueling recommendation device, wherein the device includes a memory for storing computer program instructions and a processor for executing the computer program instructions, wherein when the computer program instructions are executed by the processor, the device is triggered to execute the aforementioned refueling recommendation method.

[0024] Other embodiments of this application also provide a computer-readable medium having computer program instructions stored thereon, which can be executed by a processor to implement the refueling recommendation method.

[0025] Compared to existing technologies, this application provides a refueling recommendation scheme. After obtaining a user's refueling needs information, the scheme filters candidate gas stations based on these needs. It then calculates the fuel price forecast, coupon recommendations, service quality, and distance information of each candidate gas station. Based on these factors, a comprehensive recommendation score is calculated for each candidate gas station. Finally, a recommended gas station is selected from the candidate stations based on this comprehensive recommendation score and pushed to the user. This refueling recommendation scheme no longer relies on simple user preferences but instead calculates and integrates multiple influencing factors, such as fuel price forecasts, coupon recommendations, and service quality information, to achieve more personalized refueling service recommendations. This results in recommendations with better convenience, economy, and service quality, effectively improving recommendation accuracy and better meeting the user's actual refueling needs.

[0026] In another embodiment of this application, a refueling recommendation scheme can acquire time-related information, regional information, supply and demand information, market information, and historical oil price information of candidate gas stations to construct multi-dimensional feature data. Then, a deep learning model based on LSTM or Transformer architecture is constructed, and the deep learning model is trained using the multi-dimensional feature data to obtain a price prediction model. Thus, the deep learning-based price prediction model can accurately predict the fuel price of candidate gas stations, helping users plan the best refueling time in advance. At the same time, it can also improve the accuracy of refueling recommendations, so that the recommendation results can better meet economic needs.

[0027] In another embodiment of this application, a refueling recommendation scheme can predict a user's refueling behavior within a preset time interval based on the user's historical behavior. After calculating coupon recommendation information, coupons are issued to the target user based on the coupon recommendation information. Through a matching mechanism based on user behavior prediction, more suitable coupons can be issued to users with needs, which significantly improves the usage rate of coupons after issuance and enables users to obtain refueling services at a more favorable price.

[0028] In another embodiment of this application, a refueling recommendation scheme can obtain service efficiency indicators, service attitude indicators, facility quality indicators, and price accuracy indicators of candidate gas stations. Then, standardized scores corresponding to the service efficiency indicators, service attitude indicators, facility quality indicators, and price accuracy indicators are calculated respectively. Based on the standardized scores, the service quality information of the candidate gas stations is calculated, thereby establishing a real-time service quality monitoring and evaluation mechanism. This enables timely and accurate evaluation of the service quality of gas stations, faster detection of service anomalies, and more accurate data support for refueling recommendations, resulting in a more accurate overall recommendation effect.

[0029] In another embodiment of this application, a refueling recommendation scheme provides that, for data to be processed about the same data object obtained from multiple data sources, the data to be processed from different data sources can be converted into the same data format according to mapping rules. Then, the credibility coefficient of each data source is calculated, and the data to be processed for each data object is fused according to the credibility coefficient and data source weight of each data source. Thus, heterogeneous data from different refueling service providers can be effectively fused through a unified data mapping and conversion mechanism, resulting in a significant improvement in data quality and better accuracy of recommendation results. Attached Figure Description

[0030] Other features, objects, and advantages of this application will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings: Figure 1 A flowchart illustrating a refueling recommendation method provided in an embodiment of this application; Figure 2 A flowchart illustrating the process of training and applying a price prediction model using the scheme of this application embodiment; Figure 3 A flowchart illustrating the coupon matching process using the solution described in this application embodiment; Figure 4 A flowchart illustrating the process of heterogeneous data fusion using the scheme described in this application embodiment; Figure 5 A flowchart illustrating the interaction between various parties during the refueling process, based on the solution provided in the embodiments of this application; The same or similar reference numerals in the accompanying drawings represent the same or similar parts. Detailed Implementation

[0031] The present application will now be described in further detail with reference to the accompanying drawings.

[0032] In a typical configuration of this application, the terminal and the service network devices each include one or more processors (CPUs), input / output interfaces, network interfaces, and memory.

[0033] Memory may include non-persistent storage in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0034] Computer-readable media include permanent and non-permanent, removable and non-removable media, which can store information by any method or technology. Information can be computer program instructions, data structures, program devices, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, read-only optical disc (CD-ROM), digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transfer medium that can be used to store information accessible by a computing device.

[0035] This application provides a refueling recommendation method that no longer relies on simple user preferences. Instead, it calculates and integrates multiple influencing factors, such as fuel price predictions, coupon recommendations, and service quality information of candidate gas stations, to achieve more personalized refueling service recommendations. This results in recommendations with better convenience, economy, and service quality, effectively improving recommendation accuracy and better meeting users' actual refueling needs.

[0036] In practical scenarios, the execution subject of this method can be a user device, a network device, or a device composed of user devices and network devices integrated through a network, or it can be an application running on the aforementioned devices. The user device includes, but is not limited to, various terminal devices such as computers, mobile phones, and tablets; the network device includes, but is not limited to, network hosts, single network servers, multiple network server sets, or cloud computing-based computer sets. Here, the cloud consists of a large number of hosts or network servers based on cloud computing, where cloud computing is a type of distributed computing, consisting of a virtual computer composed of a group of loosely coupled computer sets.

[0037] Figure 1 This application illustrates a refueling recommendation method provided by an embodiment of the present application. The method includes at least the following processing steps: Step S101: Obtain the user's refueling request information.

[0038] The refueling demand information is generated based on the user's need to refuel their vehicle. It can be generated based on the refueling-related operations actually entered by the user in the refueling service platform. For example, if the user performs a search for gas stations through the refueling service entry point in the refueling service platform, the refueling service platform can obtain the user's refueling demand information based on the refueling-related operations entered by the user.

[0039] Furthermore, refueling demand information can also be generated based on partially relevant information predictions, including predictions based on fuel prices or user behavior information. Therefore, obtaining user refueling demand information can also include: predicting the optimal refueling time based on fuel price forecasts and generating user refueling demand information based on the optimal refueling time; or predicting the user's expected refueling time based on user behavior information and generating user refueling demand information based on the expected refueling time. For example, if fuel price forecasts show that fuel prices will be lowest in 3 days, then 3 days later can be considered the optimal refueling time, thus generating refueling demand information for 3 days later. Similarly, if user behavior information predicts that a user will refuel every Thursday evening, then corresponding refueling demand information can be generated before Thursday.

[0040] Step S102: Based on the refueling demand information, filter candidate gas stations.

[0041] In real-world scenarios, refueling demand information can include any information related to refueling services, such as fuel type: 92#, 95#, 98# gasoline or diesel; refueling quantity requirement: expected refueling amount or quantity (e.g., fill up, add 200 yuan); time requirement: whether refueling is urgent, acceptable waiting time; price preference: price sensitivity, whether accepting a longer distance for a lower price; and service preference: fuel brand preference, service quality requirements. This refueling demand information allows refueling service platforms to more accurately filter candidate gas stations.

[0042] When filtering candidate gas stations, the process can be based on the content of the refueling demand information, the user's location, and the current status of the gas stations. For example, if user Zhang San needs to refuel with 92# gasoline in BB District of City AA, the first step is to filter by location, calculating the distance between all gas stations in City AA and the user, and filtering for gas stations within a 10km radius. This results in 50 candidate gas stations. Next, the operating status filter removes gas stations that are closed or temporarily closed at the time of refueling, resulting in 45 candidate gas stations. Then, the fuel type filter removes gas stations that do not offer 92# gasoline, resulting in 40 candidate gas stations. Finally, the service availability filter removes gas stations with equipment malfunctions, resulting in 38 candidate gas stations. After completing all the filtering criteria, the candidate gas stations corresponding to the current refueling demand information are obtained.

[0043] Step S103: Calculate the fuel price prediction information, coupon recommendation information, service quality information, and distance information of the candidate gas stations.

[0044] In some embodiments of this application, when calculating the fuel price prediction information of candidate gas stations, we can first obtain time characteristic information, regional characteristic information, supply and demand characteristic information, market characteristic information, and historical oil price characteristic information of the candidate gas stations to construct multi-dimensional characteristic data.

[0045] The time-related features refer to time information related to fuel price changes, which in real-world scenarios may include hourly, weekly, monthly, holiday, and peak refueling periods. Geographical features refer to geographic information related to fuel price changes, which in real-world scenarios may include city, region, regional competition intensity, and transportation convenience. Supply-demand features refer to fuel supply and demand information related to fuel price changes, which in real-world scenarios may include fuel inventory levels, surrounding demand density, and historical sales volume. Market features refer to market information related to fuel price changes, which in real-world scenarios may include benchmark oil prices, international oil price trends, and policy factors. Historical oil price features refer to historical data on fuel price changes, which in real-world scenarios may include historical price sequences, price volatility, and seasonality factors. Constructing multi-dimensional feature data for model training using these related features allows the dataset to more comprehensively reflect various potential influencing factors of oil price changes, thus improving the effectiveness of model training.

[0046] After obtaining multi-dimensional feature data, a deep learning model based on LSTM (Long Short-Term Memory) or Transformer architecture is constructed. This model is then trained using the multi-dimensional feature data to obtain a price prediction model. This model can then be used to obtain fuel price predictions for candidate gas stations. This solution, by utilizing a deep learning-based price prediction model, can accurately obtain fuel price predictions for candidate gas stations, helping users plan the best refueling time in advance. It also improves the accuracy of refueling recommendations, ensuring that the recommendations better meet economic needs.

[0047] In real-world scenarios, the fuel price forecast information can include fuel prices over a future period, such as a predicted fuel price for the next 7 days. By using these fuel price forecasts, the optimal time to refuel can be identified in advance when fuel prices are lowest, thus helping refueling service platforms to better recommend fuel options.

[0048] Figure 2 The process flow for training and applying a price prediction model using the scheme of this application is shown, including the following processing steps: Step S201: Obtain historical data related to fuel prices, such as time characteristic information, regional characteristic information, supply and demand characteristic information, market characteristic information, and historical oil price characteristic information.

[0049] Step S202: Perform feature engineering processing to extract the feature vectors of the corresponding information.

[0050] Step S203: Construct multi-dimensional data based on the extracted feature vectors to serve as the training dataset for the model. In this dataset, 70% of the data is used as the training set, 15% as the validation set, and 15% as the test set.

[0051] Step S204: Construct a deep learning model using an LSTM or Transformer architecture. In this embodiment, if an LSTM model structure is used, its specific model structure can be as follows: Input layer: Used to input multi-dimensional feature vectors, including feature vectors generated from time feature information, regional feature information, supply and demand feature information, market feature information, and historical oil price feature information.

[0052] LSTM layer 1: contains 64 hidden units and is used to process time series features.

[0053] LSTM layer 2: Includes 32 hidden units, used to extract long-term dependencies.

[0054] Dropout layer: Used to prevent model overfitting.

[0055] Fully connected layer: Used to output fuel price prediction information.

[0056] If the price prediction model adopts the Transformer architecture, its specific model structure can be as follows: Input layer: used for feature embedding.

[0057] Location coding: Used for time location coding.

[0058] Multi-head attention layer: used to capture the relationships between features.

[0059] Feedforward neural network layer: used for nonlinear transformations.

[0060] Output layer: Used to output fuel price forecast information.

[0061] Step S205: Train the deep learning model using the training set. During training, the Adam optimizer is used, with a learning rate set to 0.001 and a batch size set to 32.

[0062] Step S206: Perform model validation on the trained deep learning model to determine whether the model performance meets the standard. If the result is negative, proceed to step S207; if the result is positive, proceed to step S208.

[0063] Step S207: Optimize the hyperparameters of the model. For example, in real-world scenarios, grid search algorithms and Bayesian optimization algorithms can be used.

[0064] Step S208: Save the current deep learning model, complete model training, and obtain the price prediction model.

[0065] Step S209: Obtain real-time data including time characteristic information, regional characteristic information, supply and demand characteristic information, market characteristic information, and historical oil price characteristic information.

[0066] Step S210 involves extracting features from the real-time data and inputting it into the price prediction model to obtain fuel price prediction information, i.e., the fuel price trend over a future period. This fuel price prediction information can be used to identify the best time to refuel, or for subsequent calculations of comprehensive recommendation scores.

[0067] This embodiment can also construct a real-time fuel price prediction and update mechanism based on the above-mentioned price prediction scheme, making the price prediction results more accurate. Specifically, this mechanism may include: updating the price prediction results at preset time intervals (e.g., hourly); triggering retraining of the price prediction model when the deviation between the actual price and the predicted price exceeds a certain threshold (e.g., 5%); and dynamically adjusting prediction parameters in conjunction with real-time market information. For example, real-time market information may include news related to international oil price changes, refined oil price adjustment notices, supply and demand changes caused by holidays and special events, changes in pricing strategies of refueling service providers, changes in gas station inventory, and other factors affecting fuel prices. By adjusting these parameters in real time, this scheme can obtain more accurate predicted fuel prices.

[0068] In some embodiments of this application, a user's potential refueling behavior within a future period can be predicted using user behavior prediction. Simultaneously, coupon recommendation information can be used to determine and distribute coupons suitable for that user, thereby increasing coupon usage. Therefore, this embodiment can also predict a user's expected refueling time based on user behavior information, then calculate coupon recommendation information, and distribute coupons to the target user based on the coupon recommendation information. The target user is a user whose expected refueling time falls within a preset time interval, meaning that the target user is predicted to refuel their vehicle within a future period (e.g., 3 days, 7 days, etc.).

[0069] For example, in this embodiment, a machine learning model can be used to predict a user's refueling behavior over the next 7 days. By acquiring historical data on user behavior, feature extraction is performed to obtain features such as historical refueling frequency, refueling time patterns, refueling volume patterns, and geographical location patterns. These features are then input into models such as XGBoost (eXtreme Gradient Boosting) and LightGBM (Light Gradient Boosting Machine) for behavior prediction. This yields prediction information such as the user's refueling probability, expected refueling volume, preferred fuel type, and expected refueling time. This information can determine the user's potential refueling needs in the near future, thereby deciding whether to issue coupons to the user. For example, in a real-world scenario, coupons for refueling can be issued to the user 1-3 days in advance if a refueling need is predicted.

[0070] In real-world scenarios, the timing of issuing coupons to users can be determined based on user activity levels. For example, coupons can be issued one day in advance for highly active users, two days in advance for average users, and three days in advance for inactive users. Activity levels can be correlated with the frequency with which users use the gas station service platform's client application, allowing for more personalized coupon distribution and ensuring that users receive and use the coupons as often as possible, thus avoiding ineffective coupon distribution.

[0071] When issuing coupons to the target user, coupon recommendation information can be calculated. This recommendation information indicates whether a coupon is suitable for the user. Therefore, coupons can be issued to the target user based on this recommendation information, ensuring that suitable coupons are given to users with actual refueling needs, thereby increasing coupon usage and achieving higher commercial value. For example, if the coupon recommendation information in this embodiment is represented by a score, and the recommendation scores for the three coupons are 90, 85, and 40 points respectively, then the two coupons with the higher scores can be selected to issue to the target user, ensuring coupon usage while also achieving higher commercial value.

[0072] When calculating coupon recommendation information, multiple influencing factors can be considered for a comprehensive calculation. This ensures that the final coupon recommendation information better reflects the coupon's value and increases its usage rate. Specifically, the calculation process includes determining the match between the coupon and the user, the coupon's commercial value, and its usage rate.

[0073] The matching degree between coupons and users can be calculated based on user preference information and coupon condition information. User preference information can include data-driven information such as user profiles and user behavior, including price sensitivity, brand preference, and geographic location preference. It can also be user-defined needs, such as the type of fuel needed and the brand of the chosen refueling service provider. Coupon condition information can include applicable fuel types, usage thresholds, validity period, discount amount, and other information related to the conditions of use.

[0074] When calculating the matching degree, user preference information and coupon condition information can be extracted from multiple dimensions, and the matching degree of each dimension can be calculated. Then, the overall matching degree between the coupon and the user is obtained by weighting the results. The specific calculation method can be expressed as follows: Matching degree = Σ(Dimension matching degree i × Dimension weight i) For example, to determine the match between user Zhang San and a specific coupon, the overall matching degree between the coupon and the user can be calculated using four dimensions: fuel type, amount, discount rate, and geographical location, as detailed below: Dimension 1, Fuel Type Matching. User preference: 92# gasoline, coupon applicable: 92# gasoline, -Fuel Type Matching = 1.0 (Perfect Match), Dimension Weight = 0.3.

[0075] Dimension 2, Amount Matching. Average historical fuel purchases by users: 300 yuan; coupon threshold: 200 yuan; Amount Matching = min(300 / 200, 1.5) = 1.0 (exceeds the threshold, easy to use); Weight = 0.25.

[0076] Dimension 3, Discount Matching Degree. Discount amount: 15%, User price sensitivity: 0.9, Discount matching degree = 0.15 × 10 × 0.9 = 1.35, after normalization to 1.0, weight = 0.25.

[0077] Dimension 4, Geographic Location Matching. The user's geographic location preference is City AA, and the coupon is valid throughout City AA. Matching score = 0.9 (easy to use), weight = 0.2.

[0078] Therefore, the matching degree between the coupon and the user can be calculated as 1.0×0.3 + 1.0×0.25 + 1.0×0.25 + 0.9×0.2 = 0.98.

[0079] The commercial value of a coupon can be calculated based on the expected transaction amount after using the coupon and the platform's revenue rate. Specifically, commercial value = expected transaction amount × platform revenue rate, where expected transaction amount = expected fuel volume × expected price, and platform revenue rate represents the percentage of revenue the refueling service platform receives when a user uses the coupon. The actual revenue rate can be determined based on the coupon type and the supplier's revenue sharing ratio.

[0080] The probability of coupon usage can be calculated based on its relevant usage features. These relevant usage features refer to feature vectors extracted from any information related to coupon usage, such as coupon value, usage threshold, remaining validity period, historical usage rate, and user price sensitivity. In practical scenarios, to improve the accuracy of usage probability prediction, logistic regression or neural network models can be used. The model can use the sigmoid function as the activation function to output the coupon usage probability value, specifically expressed as: Usage Probability = sigmoid(feature vector × weight vector), where the feature vector can be relevant usage features extracted based on information such as coupon value, usage threshold, remaining validity period, historical usage rate, and user price sensitivity.

[0081] After obtaining the coupon's matching degree, commercial value, and probability of use, coupon recommendation information can be calculated based on these factors and a timeliness factor. The timeliness factor is positively correlated with the coupon's validity period; that is, the shorter the coupon's validity period, the higher the timeliness factor value. For example, in this embodiment, the following formula can be used for calculation: Coupon Recommendation Information = Matching Degree × Commercial Value × Probability of Use × Timeliness Factor, where the timeliness factor = 1 - (Current Time - Coupon Issuance Time) / Coupon Validity Period.

[0082] Furthermore, some embodiments of this application can also set coupon push strategies based on the matching degree between coupons and users, thereby better prompting users to use suitable coupons and improving coupon usage rates. Specifically, in the solution of this embodiment, when obtaining users' refueling needs information, a prompt message about target coupons can be pushed to users based on the matching degree between coupons and users. The target coupon is a coupon whose matching degree meets the push conditions. The specific push conditions can be set according to the needs of the actual business scenario, such as the highest matching degree, higher than a preset value, or ranked higher, etc., without specific limitations here.

[0083] For example, if user Zhang San currently has three coupons and it's predicted that he will need to refuel in two days, the refueling service platform can send a system notification message to Zhang San upon receiving this information: "You have one highly compatible coupon that is about to expire." This notification will then focus on coupon A, the one with the highest compatibility, prompting the user to use it for refueling. This approach can further increase coupon usage and enhance business value.

[0084] Figure 3 The process flow for coupon matching using the scheme in the embodiments of this application is shown, including the following processing steps: Step S301: Obtain historical behavioral data related to refueling.

[0085] Step S302: Construct a user behavior prediction model and predict refueling behavior. This can predict the user's refueling probability, expected refueling amount, preferred fuel type, and expected refueling time in the next N days.

[0086] Step S303: Obtain a list of available coupons.

[0087] Step S304: Calculate the matching degree between the user and the coupon. Based on the matching of user preference information and coupon condition information, calculate the matching degree value.

[0088] Step S305: Calculate the commercial value of the coupon based on the expected transaction amount multiplied by the platform's profit rate.

[0089] Step S306: Use a logistic regression / neural network model to predict the probability of coupon usage.

[0090] Step S307: Calculate a recommendation score representing the coupon recommendation information by considering factors such as matching degree, commercial value, probability of use, and timeliness. This coupon recommendation information can be used to distribute suitable coupons to users or to calculate the overall recommendation score in subsequent steps.

[0091] Step S308: Sort according to the recommended scores.

[0092] Step S309: Select the top N coupons in the sorted list.

[0093] Step S310: Dynamically distribute coupons to target users at appropriate times.

[0094] When calculating the service quality information of candidate gas stations, the solution in this embodiment first obtains the candidate gas station's service efficiency index, service attitude index, facility quality index, and price accuracy index. The service efficiency index can be determined based on information related to the gas station's service efficiency, such as refueling waiting time, service completion time, and equipment response time. The service attitude index can be determined based on information related to the gas station's service attitude, such as staff professionalism, service enthusiasm, and problem-solving ability. The facility quality index can be determined based on information related to the gas station's facility quality, such as fuel quality, equipment availability, and environmental hygiene. The price accuracy index represents the consistency between the gas station's actual price and the price displayed on the platform, and can be calculated using relevant price information.

[0095] The information used to calculate the aforementioned indicators can be obtained through real-time data collection. For example, in real-world scenarios, service data such as refueling wait times and equipment status can be collected in real time using IoT (Internet of Things) devices. Some data can also be obtained by collecting user feedback and evaluations in the platform's client application. Additionally, gas station evaluation data can be obtained through third-party data sources.

[0096] After obtaining these indicators, standardized scores can be calculated for the service efficiency, service attitude, facility quality, and price accuracy indicators, respectively. Then, based on the standardized scores, the service quality information of the candidate gas stations can be calculated. This constructs a multi-dimensional real-time service quality monitoring mechanism to promptly and accurately assess the service quality of gas stations, more quickly detect service anomalies, and provide more accurate data support for gas station recommendations, resulting in a more accurate overall recommendation effect.

[0097] To improve the accuracy of service quality information assessment, a weighted calculation can be performed by combining dimensional weights and credibility weights during the calculation process. Based on the standardized scores, dimensional weights, and credibility weights corresponding to service efficiency indicators, service attitude indicators, facility quality indicators, and price accuracy indicators, a weighted calculation is performed to obtain the service quality information of the candidate gas stations. Here, dimensional weights refer to the weights of the dimensions corresponding to each type of indicator, which can be set and adjusted according to the needs of the actual scenario. The credibility weights are positively correlated with the reliability of the data source for the corresponding indicator; the higher the reliability of the data source, the higher the credibility weight. For example, in this embodiment, the following credibility weight settings can be used: The system data of the gas station service platform comes from data automatically recorded by the system, including order time, payment amount, and price comparison results, with a credibility weight set at 1.0.

[0098] IoT device data, which comes from data automatically collected by IoT devices, includes refueling waiting time, device status, refueling volume, etc., with a credibility weight set at 0.9-0.95.

[0099] User-generated feedback data comes from user ratings in the platform's client application, such as service attitude, and the credibility weight is set at 0.7-0.8.

[0100] Third-party data, which is evaluation data from third-party data sources, such as the environmental hygiene of gas stations, has a credibility weight set at 0.6-0.7.

[0101] User-reported data comes from specific data filled in and reported by users, such as price information reported by users, and the credibility weight is set to 0.5-0.6.

[0102] Taking information about a gas station obtained in a real-world scenario as an example, the calculation process for service quality information is explained: First, regarding the relevant information for each indicator dimension, the collected raw data is as follows: Dimension 1, Service Efficiency. Waiting time: 5 minutes; Data source: Automatically recorded by IoT devices. Dimension 2, Service Attitude. User rating: 4.5 out of 5, data source: user App (application) ratings; Dimension 3, Facility Quality. Equipment Status: Normal; Data Source: IoT Device Monitoring. Dimension 4, Price Accuracy. Price accuracy: 100% consistent; data source: platform system comparison.

[0103] First, calculate the standardized score for each dimension corresponding to each indicator. The calculation formula is as follows: Standardized score = (actual value - minimum value) / (maximum value - minimum value) × 100.

[0104] Therefore, the standardized score for the service efficiency index is calculated as follows: Assuming the best waiting time is 0 minutes and the worst waiting time is 30 minutes, and the current waiting time is 5 minutes, the standardized score for the service efficiency index is (30-5) / (30-0)×100=83.3 points. The standardized score for the service attitude index is calculated as follows: with a maximum score of 5 points and a minimum score of 0 points, the standardized score for the service attitude index is 4.5 / 5×100=90 points. The standardized score for the facility quality index is calculated as follows: 100 points for normal equipment and 0 points for abnormal equipment. Since the current equipment status is normal, the standardized score for the facility quality index is 100 points. Similarly, the standardized score for price accuracy is 100 points.

[0105] In this embodiment, the dimensional weights and credibility weights for each dimension are set as follows: Service efficiency dimension, dimension weight: 0.3, credibility weight: 0.9; Service attitude dimension, dimension weight: 0.3, credibility weight: 0.8; Facility quality dimension, dimension weight: 0.25, credibility weight: 0.95; Price accuracy dimension, dimension weight: 0.15, credibility weight: 1.0.

[0106] Based on the above weighting information, the following formula is used for calculation: Service quality information = Σ(dimensional weight i × standardized score i × credibility weight i) / Σ(dimensional weight i × credibility weight i) = 82.84 / 0.8975 = 92.3.

[0107] In calculating service quality information, an anomaly detection mechanism can be established to provide early warnings and timely handling of service anomalies. In this embodiment, statistical or machine learning methods can be used to detect service quality anomalies. For example, an anomaly warning can be triggered when the service score is below two standard deviations of the historical average; the isolated forest algorithm can be used to detect service anomaly patterns; and anomaly warnings can be triggered when the waiting time exceeds 30 minutes or the price deviation exceeds 5%. Using these methods, anomalies in gas station service quality can be detected quickly and effectively.

[0108] When an abnormality in service quality is detected, an early warning notification can be sent to relevant personnel at the gas station immediately. At the same time, a problem handling process can be established to track the problem resolution status and update the service quality score in real time based on the problem handling results, thereby achieving rapid early warning and handling of service quality issues.

[0109] When calculating the distance information of candidate gas stations, it can be based on the distance between the candidate gas station's location and the user's location. In real-world scenarios, the driving path distance between the two can be calculated, rather than the straight-line distance, thus more accurately reflecting the distance-related factors affecting the user's refueling experience.

[0110] Step S104: Calculate the comprehensive recommendation score of the candidate gas stations based on the fuel price prediction information, coupon recommendation information, service quality information, and distance information.

[0111] When calculating the overall recommendation score, the weights of various factors can be considered, specifically including price weight, discount weight, service weight, and distance weight. In practical scenarios, to calculate an overall recommendation score that better reflects the user's actual needs, the price weight, discount weight, service weight, and distance weight can be determined based on user preference information.

[0112] For example, for price-sensitive users, gas stations with lower predicted prices are prioritized, with other factors weighted appropriately. The weighting could be set as follows: price 60%, discounts 20%, service 10%, and distance 10%. Similarly, for convenience-oriented users, the nearest gas station is prioritized, with price and distance considered to some extent. Therefore, the weighting could be set as follows: price 20%, discounts 10%, service 20%, and distance 50%. In real-world scenarios, different weighting ratios can be set according to the user's actual preferences to achieve more personalized refueling service recommendations and improve the relevance of the recommendations to the user.

[0113] After determining the price weight, discount weight, service weight, and distance weight based on user preference information, a weighted calculation can be performed based on the fuel price prediction information, coupon recommendation information, service quality information, and distance information, combined with the corresponding price weight, discount weight, service weight, and distance weight, to obtain the comprehensive recommendation score for the candidate gas stations. For example, the formula for calculating the comprehensive recommendation score in this embodiment is as follows: The overall recommendation score is calculated as follows: Price Prediction Weight × Price Score + Coupon Weight × Discount Score + Service Quality Weight × Quality Score + Distance Weight × Distance Score. The price score, discount score, quality score, and distance score are derived from fuel price prediction information, coupon recommendation information, service quality information, and distance information, respectively. These factors are converted and calculated according to certain rules. For example, the price score can be calculated using the following formula: Price score = (Highest price - Current gas station price) / (Highest price - Lowest price) × 100 The highest price is the highest price among all candidate gas stations, the lowest price is the lowest price among all candidate gas stations, and the current gas station price is the discounted price of the candidate gas station currently being calculated. For example, if user Zhang San needs to fill up with 200 yuan worth of 92# gasoline, candidate gas station A can use a coupon to reduce the price by 20 yuan for purchases over 200 yuan, resulting in a discounted price of 7.5 yuan / liter. The lowest price among the candidate gas stations is 7.0 yuan / liter, and the highest price is 7.8 yuan / liter. Therefore, the price score for gas station A can be calculated as (7.8-7.5) / (7.8-7.0)×100=37.5 points.

[0114] For each candidate gas station, a similar method can be used to calculate the price score, discount score, quality score, and distance score. Then, by combining their respective weights, a weighted calculation can be performed to obtain the comprehensive recommendation score for each candidate gas station.

[0115] Step S105: Based on the comprehensive recommendation score, determine recommended gas stations from the candidate gas stations and push the recommended gas stations to the user.

[0116] In a real-world scenario, a personalized recommendation list can be generated for each user based on their overall recommendation score. This personalized recommendation list can include at least one recommended gas station. The gas station service platform can then send this personalized recommendation list to the user, thus achieving accurate gas station recommendations.

[0117] Because this solution does not rely on simple user preferences, but instead calculates and integrates various influencing factors such as fuel price predictions, coupon recommendations, and service quality information of candidate gas stations, it achieves more personalized refueling service recommendations by combining these multi-dimensional factors. This results in recommendations that are more convenient, economical, and offer better service quality, effectively improving recommendation accuracy and better meeting users' actual refueling needs.

[0118] In some embodiments of this application, a recommendation weight adjustment mechanism is also provided, which allows the recommendation weight of each candidate gas station to be dynamically adjusted before the final recommendation of gas stations is pushed to the user, so that service quality information can further affect the final recommendation result and ensure that the user can obtain a better service experience.

[0119] The specific principle of this recommendation weight adjustment mechanism is as follows: when determining recommended gas stations from the candidate gas stations based on the comprehensive recommendation score and pushing the recommended gas stations to the user, if the service quality information of the candidate gas stations meets the recommendation weight adjustment conditions, the comprehensive recommendation score of the candidate gas stations is adjusted. Thus, based on the adjusted comprehensive recommendation score, the recommended gas stations are determined from the candidate gas stations and pushed to the user.

[0120] The weight adjustment condition can be set such that the value corresponding to the service quality information is within a preset value range. Thus, the recommendation weight coefficient can be determined based on the value range of the service quality information of the candidate gas station. Then, the comprehensive recommendation score is multiplied by the recommendation weight coefficient to obtain the adjusted comprehensive recommendation score.

[0121] For example, in this embodiment, service quality information can be converted into a corresponding service quality score (0-5 points). When the service quality score is in the range of 4.5-5.0, it indicates excellent service quality, and the recommendation weight coefficient is determined to be 1.2; when the service quality score is in the range of 3.5-4.4, it indicates good service quality, and the recommendation weight coefficient is determined to be 1.0; when the service quality score is in the range of 2.5-3.4, it indicates average service quality, and the recommendation weight coefficient is determined to be 0.85; when the service quality score is in the range of 2.0-2.4, it indicates poor service quality, and the recommendation weight coefficient is determined to be 0.7; when the service quality score is less than 2.0, it indicates very poor service quality, and the recommendation weight coefficient is determined to be 0. If gas station A has a service quality score of 4.7 and an unadjusted overall recommendation score of 80, while gas station B has a service quality score of 3.1 and an unadjusted overall recommendation score of 90. After adjusting the recommendation weight coefficient, gas station A's final comprehensive recommendation score is 80 × 1.2 = 96, while gas station B's final comprehensive recommendation score is 90 × 0.85 = 76.5. At this point, gas station A's comprehensive recommendation score will exceed that of gas station B, making gas station A's recommendation priority higher.

[0122] In the actual application scenario of this application, some of the data that needs to be processed will come from different refueling service providers. These providers are independent data sources, so there will be data heterogeneity issues. For example, different providers have different data formats for gas station location information and oil price information, making it difficult to use this heterogeneous data directly.

[0123] To address this issue, some embodiments of this application also provide a data fusion mechanism. After obtaining data on the same data object from multiple data sources, the data from different data sources can be converted into the same data format according to mapping rules. Then, the credibility coefficient of each data source is calculated. Based on the credibility coefficient and data source weight of each data source, the data on the same data object is fused to obtain fused data. Thus, a unified data mapping and transformation mechanism can effectively fuse heterogeneous data from different refueling service providers, significantly improving data quality and increasing the accuracy of recommendation results.

[0124] The mapping rules can be set according to the actual scenario. For example, location information can be uniformly converted to coordinate values ​​in the WGS84 global positioning system. Price information can be concatenated into a unified data format by combining JSON parsing with field reorganization. This allows for the acquisition of data in the same format for processing. Furthermore, since the data content of the same data object from different data sources may differ, it is necessary to fuse the data for the same data object to obtain more accurate fused data, thereby improving the accuracy of subsequent processing results.

[0125] To improve the accuracy of data fusion, the solution in this application introduces a data source credibility coefficient. This coefficient can be obtained by acquiring the data quality score, supplier historical performance, data update frequency, and reporting normality rate of each data source. Then, based on these factors, the credibility coefficient of each data source is calculated. Furthermore, the data source weight can be set based on different needs in the actual scenario, considering factors such as data source size, historical data quality, and user preferences.

[0126] The data quality score of the data source is used to represent the quality of the data provided by the data source this time. It can be obtained by weighting the data integrity, timeliness, accuracy and consistency of the data to be processed.

[0127] Data integrity can be determined by the ratio of the actual number of fields in the acquired data to the standard number of fields. For example, if the actual number of fields in the acquired data is 80, while the standard number of fields that should be acquired is 100, then the data integrity = 80 / 100 = 0.8.

[0128] Data timeliness can be calculated using the current time, the last data update time, and the preset maximum allowable delay. For example, if the current time is 14:00 on 2025-01-01, the data update time is 13:50 on 2025-01-01, and the maximum allowable delay is 60 minutes, then according to the timeliness calculation formula: Data Timeliness = 1 - (Current Time - Update Time) / Maximum Allowable Delay, we can calculate that Data Timeliness = 1 - (14:00 - 13:50) / 60 minutes = 1 - 0.167 = 0.833.

[0129] Data accuracy can be determined based on the proportion of abnormal data in the acquired data to be processed. For example, if the proportion of abnormal data in the acquired data to be processed is 5%, then the data accuracy can be calculated as 1 - 0.05 = 0.95. In this embodiment, abnormal data detection can be implemented using statistical methods or machine learning methods. For example, it can identify data exceeding the normal range based on the 3σ principle, use the isolated forest algorithm to detect abnormal data, or identify abnormal data based on business rules (such as price fluctuations exceeding 20% ​​being considered abnormal).

[0130] Data consistency can be determined based on the difference between the current data to be processed and the historical average of these data. For example, if a data source reports a price of 7.5 yuan / liter this time, and the historical average price is 7.3 yuan / liter, then according to the data consistency calculation formula: Data Consistency = 1 - (Current Data - Historical Average) / Historical Average, the data consistency is calculated as 1 - (7.5 - 7.3) / 7.3 = 1 - 0.027 = 0.973.

[0131] The data quality score of the data source is obtained by weighting the data integrity, timeliness, accuracy, and consistency of the data to be processed. Taking the aforementioned scenario as an example, if the weights for data integrity, timeliness, accuracy, and consistency are set to 0.3, 0.3, 0.2, and 0.2 respectively, the data quality score of the data source can be calculated as: 0.8 × 0.3 + 0.833 × 0.3 + 0.95 × 0.2 + 0.973 × 0.2 = 0.8745.

[0132] The historical performance of the data source's suppliers can be determined based on historical data accuracy and a time decay factor. The time decay factor indicates that the accuracy of data from more recent times has a higher weight in relation to the current supplier's historical performance. Therefore, the historical performance of the data source's suppliers can accurately reflect the data accuracy of the data source when it provided data as a supplier recently.

[0133] The data update frequency can be determined based on the ratio of the actual number of updates to the expected number of updates from the data source. For example, if the agreement between a gas station service platform and the data source stipulates that price information needs to be updated every 15 minutes, then the expected number of updates in one hour is 4. If the data source actually updates twice in a certain hour, then the data update frequency can be calculated as 2 / 4 = 0.5.

[0134] The reporting success rate is determined based on the historical anomaly reporting ratio of the data source. The lower the historical anomaly reporting ratio, the higher the reporting success rate. For example, if the anomaly reporting ratio of a data source in the last 30 days is 3%, then its reporting success rate in those 30 days = 1 - 0.03 = 0.97.

[0135] After calculating the data quality score, supplier historical performance, data update frequency, and reporting normality rate, the credibility coefficient of each data source can be calculated based on these factors.

[0136] In this embodiment, when fusing the data of each identical data object based on the reliability coefficient and weight of each data source, the following calculation formula can be used: fused data value = Σ(value of data source i × reliability coefficient of data source i × weight of data source i) / Σ(reliability coefficient of data source i × weight of data source i). Therefore, by introducing a reliability coefficient, this scheme can more accurately quantify the reliability of each data source, providing a better calculation basis for data fusion and improving the accuracy of data fusion.

[0137] During data fusion, when the differences between the data to be processed from multiple data sources are too large and exceed the threshold, the weighted median method can be used to identify abnormal data. Depending on the actual degree of difference, manual review and confirmation or direct removal can be selected to ensure the normality of the fused data and thus improve the accuracy of the data fusion results.

[0138] Furthermore, to obtain higher-quality data to be processed, an incremental update strategy can be adopted during data updates, synchronizing only the changed data to improve update efficiency. A data version management mechanism can also be established to support data rollback, preventing service disruptions caused by abnormal data updates. Additionally, after data fusion, further verification can be performed to ensure the accuracy of the merged data. In practical scenarios, data verification in this step can include the following aspects: business rule verification, such as verifying whether prices are within a reasonable range (e.g., 5-10 yuan / liter); logical consistency verification, such as whether the price relationship between different oil products at the same gas station is normal (95# ≥ 92#); cross-data verification, such as whether prices are consistent with inventory and operating status; and time-series consistency verification, whether price changes conform to historical trends. These verification methods ensure the accuracy of the merged data and avoid data that does not conform to common sense.

[0139] Figure 4 The process flow for heterogeneous data fusion using the scheme in the embodiments of this application is shown, including the following processing steps: Step S401: Connect to the data interfaces of multiple data sources and obtain the data to be processed from the data interfaces. In this embodiment, the data sources can be different suppliers providing refueling services.

[0140] Step S402: Identify data format differences.

[0141] Step S403: Apply data mapping rules.

[0142] Step S404: Perform data format conversion using data mapping rules.

[0143] Step S405: Check if there is any abnormal data. If abnormal data is found, proceed to step S406; otherwise, proceed to step S407.

[0144] Step S406: Process the abnormal data, including marking abnormal data and filtering abnormal data. After processing is completed, proceed to step S407.

[0145] Step S407: Calculate the credibility coefficient of the data source.

[0146] Step S408: Perform data fusion processing.

[0147] Step S409: Perform data verification.

[0148] Step S410: Determine whether the verification passes. If it fails, proceed to step S411; if it passes, proceed to step S412.

[0149] Step S411, data rollback.

[0150] Step S412: Store the fused data and update it incrementally.

[0151] Furthermore, after determining a recommended gas station from the candidate gas stations based on the comprehensive recommendation score and pushing the recommended gas station to the user, the solution of this application embodiment can also obtain the user's refueling order information based on the recommended gas station, then generate a reservation order based on the refueling order information, and assign the recommended gas station to the fuel pump position in the reservation time period corresponding to the reservation order, while sending order confirmation information and navigation information to the user, thereby guiding the user to the recommended gas station for refueling.

[0152] Figure 5 This paper illustrates the interaction flow between various parties involved in refueling recommendations and subsequent refueling processes, based on the solution provided in this application embodiment. The process involves users, a refueling service platform, and a refueling service provider. The user is the refueling service user, which in a real-world scenario could be a car owner needing refueling. The refueling service platform is the operator of the refueling service, which in a real-world scenario could be the platform's operating company. The refueling service provider is the actual provider of the refueling service, which in a real-world scenario could be a gas station providing the refueling service. The entire interaction includes the following steps: Step S1: The user searches for nearby gas stations and generates refueling request information.

[0153] Step S2: The platform requests data from the supplier.

[0154] Step S3: The supplier returns the relevant data required for the refueling recommendation.

[0155] In step S4, the platform merges data from different suppliers and combines it with other relevant data it has acquired to perform recommendation calculations and determine recommended gas stations.

[0156] Step S5: Push a list of recommended gas stations to the user.

[0157] In step S6, the user places an order and pays based on the recommendation. During this process, the user can select a gas station and confirm the quantity or amount of fuel. The platform locks in the current discounted price and available coupons, generates a reservation order, assigns a time slot and fuel pump location, and simultaneously sends an order confirmation and navigation information to the driver. Upon arrival at the gas station, the platform automatically recognizes the arrival and alerts the gas station staff.

[0158] Step S7: Gas station staff provide refueling services to users.

[0159] Step S8: After refueling, the platform automatically calculates the cost based on the actual amount of fuel dispensed. Users can then pay the cost and obtain an invoice.

[0160] In step S9, the platform transfers the remaining funds to the supplier after deducting its own revenue, according to the agreement with the supplier.

[0161] In step S10, car owners evaluate the gas station service. The platform collects the evaluation data to update service quality information and recommendation weight coefficients. At the same time, it can optimize user profiles and personalized recommendation algorithms based on the data during the service process.

[0162] Based on another aspect of this application, embodiments of this application also provide a refueling recommendation device, which includes a memory for storing computer program instructions and a processor for executing the computer program instructions, wherein when the computer program instructions are executed by the processor, the device is triggered to execute the aforementioned refueling recommendation method.

[0163] In particular, the methods and / or embodiments in this application can be implemented as computer software programs. For example, embodiments of this disclosure include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowchart. When the computer program is executed by a processing unit, it performs the functions defined in the methods of this application.

[0164] It should be noted that the computer-readable medium described in this application can be a computer-readable signal medium or a computer-readable storage medium, or any combination thereof. A computer-readable medium can be, for example,—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this application, a computer-readable medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0165] In this application, a computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. The computer-readable signal medium may also be any computer-readable medium other than a computer-readable storage medium, capable of transmitting, propagating, or transmitting a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium may be transmitted using any suitable medium, including but not limited to: wireless, wireline, optical fiber, RF, etc., or any suitable combination thereof.

[0166] Computer program code for performing the operations of this application can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, and conventional procedural programming languages ​​such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0167] The flowcharts or block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of devices, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-specific system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0168] In another aspect, this application also provides a computer-readable medium, which may be included in the device described in the above embodiments; or it may exist independently and not assembled into the device. The aforementioned computer-readable medium carries one or more computer program instructions, which may be executed by a processor to implement the methods and / or technical solutions of the various embodiments of this application.

[0169] It should be noted that this application can be implemented in software and / or a combination of software and hardware, for example, using an application-specific integrated circuit (ASIC), a general-purpose computer, or any other similar hardware device. In some embodiments, the software program of this application can be executed by a processor to implement the steps or functions described above. Similarly, the software program of this application (including related data structures) can be stored in a computer-readable recording medium, such as RAM memory, a magnetic or optical drive, a floppy disk, or similar devices. Furthermore, some steps or functions of this application can be implemented in hardware, for example, as circuitry that cooperates with a processor to perform the various steps or functions.

[0170] It will be apparent to those skilled in the art that this application is not limited to the details of the exemplary embodiments described above, and that this application can be implemented in other specific forms without departing from the spirit or essential characteristics of this application. Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of this application is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be embraced within this application. No reference numerals in the claims should be construed as limiting the scope of the claims. Furthermore, it is clear that the word "comprising" does not exclude other units or steps, and the singular does not exclude the plural. Multiple units or devices recited in the apparatus claims may also be implemented by a single unit or device in software or hardware. The terms "first," "second," etc., are used to indicate names and do not indicate any particular order.

Claims

1. A method for recommending refueling services, characterized in that, The method includes: Obtain users' refueling needs information; Based on the refueling demand information, filter candidate gas stations; Calculate fuel price forecasts, coupon recommendations, service quality, and distance information for candidate gas stations; Based on the fuel price forecast information, coupon recommendation information, service quality information, and distance information, calculate the comprehensive recommendation score for the candidate gas stations; Based on the comprehensive recommendation score, recommended gas stations are determined from the candidate gas stations, and the recommended gas stations are pushed to the user.

2. The method according to claim 1, characterized in that, Calculate fuel price forecast information for candidate gas stations, including: Obtain temporal, regional, supply and demand, market, and historical fuel price characteristics of candidate gas stations to construct multidimensional feature data; Construct a deep learning model based on LSTM or Transformer architecture, and train the deep learning model using the multidimensional feature data to obtain a price prediction model; The price prediction model is used to obtain fuel price prediction information for candidate gas stations.

3. The method according to claim 1, characterized in that, The method further includes: Predict users' expected refueling time based on user behavior information; Calculate coupon recommendation information and issue coupons to the target users based on the coupon recommendation information, wherein the target users are users whose expected refueling time is within a preset time range.

4. The method according to claim 1 or 3, characterized in that, Calculate coupon recommendation information, including: Calculate the matching degree between coupons and users based on user preference information and coupon conditions; Calculate the commercial value of the coupon based on the expected transaction amount after using the coupon and the platform's revenue rate; Calculate the probability of using the coupon based on its relevant usage characteristics; Based on the coupon's matching degree, commercial value, probability of use, and timeliness factor, coupon recommendation information is calculated, wherein the timeliness factor is positively correlated with the coupon's validity period.

5. The method according to claim 4, characterized in that, The method further includes: When obtaining users' refueling needs, based on the matching degree between coupons and users, a prompt message about the target coupon is pushed to the user, wherein the target coupon is a coupon whose matching degree meets the push conditions.

6. The method according to claim 1, characterized in that, Calculate the service quality information of candidate gas stations, including: Obtain service efficiency indicators, service attitude indicators, facility quality indicators, and price accuracy indicators for candidate gas stations; Calculate the standardized scores corresponding to the service efficiency index, service attitude index, facility quality index, and price accuracy index respectively; Based on the standardized score, the service quality information of the candidate gas stations is calculated.

7. The method according to claim 6, characterized in that, Based on the standardized score, the service quality information of the candidate gas stations is calculated, including: Based on the standardized scores, dimensional weights, and credibility weights corresponding to service efficiency indicators, service attitude indicators, facility quality indicators, and price accuracy indicators, a weighted calculation is performed to obtain the service quality information of the candidate gas stations. The credibility weight is positively correlated with the reliability of the data source for the corresponding indicator.

8. The method according to claim 1, characterized in that, Based on the comprehensive recommendation score, recommended gas stations are determined from the candidate gas stations, and the recommended gas stations are pushed to the user, including: If the service quality information of a candidate gas station meets the recommendation weight adjustment conditions, the overall recommendation score of the candidate gas station will be adjusted. Based on the adjusted comprehensive recommendation score, recommended gas stations are determined from the candidate gas stations and pushed to the user.

9. The method according to claim 8, characterized in that, If the service quality information of a candidate gas station meets the recommendation weight adjustment conditions, the overall recommendation score of the candidate gas station is adjusted, including: The recommendation weight coefficient is determined based on the numerical range of the service quality information of the candidate gas stations; The adjusted comprehensive recommendation score is obtained by multiplying the comprehensive recommendation score by the recommendation weight coefficient.

10. The method according to claim 1, characterized in that, Based on the fuel price forecast information, coupon recommendation information, service quality information, and distance information, a comprehensive recommendation score for the candidate gas stations is calculated, including: Price weight, discount weight, service weight, and distance weight are determined based on user preference information; Based on the fuel price forecast information, coupon recommendation information, service quality information, and distance information, and combined with the corresponding price weight, discount weight, service weight, and distance weight, a weighted calculation is performed to obtain the comprehensive recommendation score of the candidate gas stations.

11. The method according to claim 1, characterized in that, The method further includes: Retrieve data to be processed from multiple data sources for the same data object; Based on the mapping rules, the data to be processed from different data sources is converted into the same data format; Calculate the credibility coefficient of each data source; Based on the credibility coefficient and weight of each data source, the data to be processed for each identical data object is fused to obtain fused data.

12. The method according to claim 11, characterized in that, Calculate the credibility coefficient for each data source, including: Obtain data quality scores, supplier historical performance, data update frequency, and reporting success rate for each data source; The credibility coefficient of each data source is calculated based on the data quality score, the supplier's historical performance, the data update frequency, and the reporting normality rate.

13. The method according to claim 12, characterized in that, Obtain the data quality scores for each data source: Based on the data integrity, timeliness, accuracy, and consistency of the data to be processed, a weighted calculation is performed to obtain the data quality score for each data source.

14. The method according to claim 1, characterized in that, Obtaining users' refueling needs information includes at least one of the following: Based on the user's input of refueling-related actions, obtain the user's refueling demand information; The best time to refuel is predicted based on fuel price forecasts, and the user's refueling demand information is generated based on the best time to refuel. Predict the user's expected refueling time based on user behavior information, and generate the user's refueling demand information based on the expected refueling time.

15. The method according to claim 1, characterized in that, After determining recommended gas stations from the candidate gas stations based on the comprehensive recommendation score and pushing the recommended gas stations to the user, the process further includes: Obtain the user's refueling order information based on the recommended gas stations; Based on the refueling order information, a reservation order is generated, and the recommended gas station is assigned to the fuel nozzle position for the reservation time period corresponding to the reservation order. Send order confirmation and navigation information to users.

16. A refueling recommendation device, wherein, The device includes a memory for storing computer program instructions and a processor for executing the computer program instructions, wherein when the computer program instructions are executed by the processor, the device is triggered to perform the method of any one of claims 1 to 15.

17. A computer-readable medium having stored thereon computer program instructions that can be executed by a processor to implement the method as claimed in any one of claims 1 to 15.

Images