A multi-dimensional dynamic scoring and real-time matching method for tourism service supply and demand.

Abstract

The application provides a multi-dimensional dynamic scoring and real-time matching method for tourism service supply and demand matching, which comprises standardizing request parameters and retrieving an initial supplier set; based on CPU core number and supplier scale, a slice is dynamically divided for calculation, qualification verification, multi-dimensional weighted basic scoring and dynamic strategy adjustment are performed to obtain a final dynamic score; a selected supplier is selected according to multi-level logic; an order is generated and a notice of elimination containing personalized improvement suggestions is pushed, a data feedback loop is formed; matching data is stored and strategy parameters are checked. The application can solve the defects of single scoring, low efficiency and poor adaptability of the prior art, realize multi-dimensional dynamic scoring and real-time matching, improve matching accuracy and efficiency, and continuously improve service quality through closed-loop optimization, which is suitable for different scene requirements.

Description

Technical Field

[0001] This application relates to the field of tourism service technology, and in particular to a multi-dimensional dynamic scoring and real-time matching method for tourism service supply and demand. Background Technology

[0002] With the digital transformation of the tourism industry, the matching of supply and demand for tourism services (such as car rental, tour guide services, and attraction-related services) is gradually moving towards online and intelligent models. Currently, mainstream tourism service supply and demand matching methods mostly adopt "single-dimensional sorting" or "simple weighted scoring" models, which have the following technical shortcomings: The scoring dimensions are singular and fixed, often relying solely on price or distance as the core sorting criteria, ignoring key dimensions such as service quality, historical performance, and vehicle condition. This leads to a disconnect between matching results and actual user needs, resulting in a poor user experience; the scoring logic is static, unable to dynamically adjust scoring weights and rules based on supplier status, market fluctuations, and user preferences, resulting in insufficient adaptability and easily leading to low matching efficiency in scenarios of supply-demand imbalance; The system suffers from low computational efficiency and lacks dynamic scheduling based on hardware resources. When the initial supplier pool is large, the matching process is too time-consuming, failing to meet the core requirement of "real-time" service in the tourism industry. Furthermore, it lacks effective feedback and attribution mechanisms, preventing unsuccessful candidates from understanding the specific reasons for their rejection and hindering targeted improvements. The system also cannot self-optimize based on historical matching data to form a closed loop. Updating strategy parameters is cumbersome; adjusting parameters such as scoring weights and benchmark prices requires system downtime and deployment, impacting system continuity. Moreover, it cannot achieve fine-grained configuration by city or time period, limiting its adaptability.

[0003] To address the aforementioned technical shortcomings, there is an urgent need for a technical solution that can achieve multi-dimensional dynamic scoring, real-time efficient matching, and closed-loop optimization, in order to solve the problems of inaccurate matching, low efficiency, and poor adaptability in the current supply and demand matching of tourism services. Summary of the Invention

[0004] In view of this, the embodiments of this application provide a multi-dimensional dynamic scoring and real-time matching method for tourism service supply and demand matching, which can overcome the defects of the existing technology in tourism service supply and demand matching, such as single scoring, low efficiency, poor adaptability, and lack of closed-loop optimization, to achieve accurate and real-time matching of supply and demand, improve user experience and supplier improvement efficiency, while ensuring the stability and adaptability of the system.

[0005] The technical solution of this application embodiment is implemented as follows: In a first aspect, embodiments of this application provide a multi-dimensional dynamic scoring and real-time matching method for tourism service supply and demand, comprising the following steps: The parameters of the original tourism service request are analyzed, the service time, location, and vehicle type requirements are standardized, and the initial set of suppliers is retrieved based on the city and service type index. Based on the number of CPU cores and the initial supplier set size, the calculation slices are dynamically divided. A progressive scoring is performed on the suppliers in each slice. Hard condition checks are performed through the qualification verification layer. Multi-dimensional weighted base scores are calculated for suppliers that pass the verification. The base scores are then dynamically fine-tuned through the strategy adjustment layer to obtain the final dynamic score. Suppliers that fail the qualification verification are marked as eliminated. The scoring results of all computational slices are aggregated to form a scoring list. The supplier with the highest comprehensive score is selected as the winning supplier according to a multi-level sorting logic, and the rest are candidate failures. The multi-level sorting logic includes at least one of whitelist priority, perfect price priority, and comprehensive score descending order. For the selected supplier, inventory locking and order generation are performed. For each candidate failure, a multi-dimensional feature comparison analysis is conducted to compare the score gap between it and the selected supplier on key evaluation dimensions, generate a personalized set of failure reasons, and send a selection notification to the selected supplier and a rejection notification containing improvement suggestions to the candidate failure through an asynchronous message channel. The input features, score details, and attribution results of this match are written to persistent storage, and a configuration check is triggered to ensure that the matching strategy parameters are up-to-date.

[0006] Secondly, embodiments of this application also provide a multi-dimensional dynamic scoring and real-time matching device for tourism service supply and demand matching, the device comprising: The initial screening module is used to parse the parameters of the original tourism service request, standardize the service time, location, and vehicle type requirements, and retrieve the initial set of suppliers based on the city and service type index. The scoring module is used to dynamically divide the calculation slices based on the number of CPU cores and the initial supplier set size. It performs progressive scoring on suppliers in each slice, conducts hard condition checks through the qualification verification layer, calculates multi-dimensional weighted base scores for suppliers that pass the verification, and then dynamically fine-tunes the base scores through the strategy adjustment layer to obtain the final dynamic score. Suppliers that fail the qualification verification are marked as eliminated. The decision module is used to aggregate the scoring results of all computation slices, form a scoring list, and select the supplier with the highest comprehensive score as the selected supplier according to a multi-level sorting logic, while the rest are candidate failures; wherein, the multi-level sorting logic includes at least one of whitelist priority, perfect price priority, and comprehensive score descending order; The feedback module is used to lock inventory and generate orders for the selected suppliers. For each candidate that fails, it compares the score difference between the candidate and the selected supplier on key evaluation dimensions through multi-dimensional feature comparison analysis, generates a personalized set of reasons for failure, and sends a selection notification to the selected supplier and a rejection notification containing improvement suggestions to the candidate that fails through an asynchronous message channel. The maintenance module is used to write the input features, scoring details, and attribution results of this match to persistent storage and trigger configuration checks to ensure that the matching strategy parameters are up-to-date.

[0007] Thirdly, embodiments of this application also provide an electronic device, including: a processor, a storage medium, and a bus, wherein the storage medium stores machine-readable instructions executable by the processor, and when the electronic device is running, the processor communicates with the storage medium via the bus, and the processor executes the machine-readable instructions to perform the multi-dimensional dynamic scoring and real-time matching method for tourism service supply and demand matching as described in any of the first aspects.

[0008] Fourthly, embodiments of this application also provide a computer-readable storage medium storing a computer program, wherein the computer program, when executed by a processor, performs the multi-dimensional dynamic scoring and real-time matching method for tourism service supply and demand matching as described in any of the first aspects.

[0009] The embodiments of this application have the following beneficial effects: By initializing and screening requests, the system standardizes the original request parameters and retrieves an initial supplier set based on a "city + service type" index. This effectively eliminates irrelevant suppliers, narrows the subsequent scoring range, and significantly improves the retrieval efficiency in the early stages of matching. Through hierarchical scoring combined with dynamic hardware load sharding for parallel processing, and progressive scoring, the system quickly eliminates unqualified suppliers and reduces invalid calculations through a qualification verification layer. Furthermore, by using non-linear segmented price scoring and dynamic strategy adjustment, it overcomes the shortcomings of existing technologies, such as single scoring dimensions and static logic, making the scoring more closely reflect actual supply and demand. Multi-level sorting logic prioritizes matching... The rationality and optimization of the matching ensure a high degree of alignment between the selected supplier and user needs, enhancing the user experience. Through multi-dimensional attribution and intelligent feedback, it not only enables rapid order generation and inventory locking but also provides personalized reasons for failure for unsuccessful candidates, building a data closed loop to support supplier service iteration and further optimize subsequent matching results. Through closed-loop data maintenance, it retains data from the entire matching process, ensuring that the matching strategy parameters are up-to-date, providing support for the stability and accuracy of subsequent matching. Overall, it achieves accurate and real-time matching of tourism service supply and demand, effectively solving the core problems of inaccurate matching, low efficiency, and poor adaptability in existing technologies. Attached Figure Description

[0010] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0011] Figure 1 This is a flowchart illustrating steps S101-S105 provided in the embodiments of this application; Figure 2 This is a schematic diagram of the structure of the multi-dimensional dynamic scoring and real-time matching device for tourism service supply and demand matching provided in the embodiments of this application; Figure 3 This is a schematic diagram of the composition structure of the electronic device provided in the embodiments of this application. Detailed Implementation

[0012] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. It should be understood that the accompanying drawings in this application are for illustrative and descriptive purposes only and are not intended to limit the scope of protection of this application. Furthermore, it should be understood that the schematic drawings are not drawn to scale. The flowcharts used in this application illustrate operations implemented according to some embodiments of this application. It should be understood that the operations in the flowcharts may not be implemented in sequence, and steps without logical contextual relationships may be reversed or implemented simultaneously. In addition, those skilled in the art, guided by the content of this application, may add one or more other operations to the flowcharts, or remove one or more operations from the flowcharts.

[0013] In the following description, references are made to “some embodiments,” which describe a subset of all possible embodiments. However, it is understood that “some embodiments” may be the same subset or different subsets of all possible embodiments and may be combined with each other without conflict.

[0014] Furthermore, the described embodiments are merely some, not all, of the embodiments of this application. The components of the embodiments of this application described and illustrated herein can typically be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely to illustrate selected embodiments of the application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0015] In the following description, the terms "first, second, third" are used merely to distinguish similar objects and do not represent a specific ordering of objects. It is understood that "first, second, third" may be interchanged in a specific order or sequence where permitted, so that the embodiments of this application described herein can be implemented in an order other than that illustrated or described herein.

[0016] It should be noted that the term "comprising" will be used in the embodiments of this application to indicate the presence of the features declared thereafter, but does not exclude the addition of other features.

[0017] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of this application and is not intended to limit this application.

[0018] See Figure 1 , Figure 1 This is a flowchart illustrating steps S101-S105 of the multi-dimensional dynamic scoring and real-time matching method for tourism service supply and demand matching provided in this application embodiment. Figure 1 Steps S101-S105 are explained below.

[0019] In step S101, the parameters of the original tourism service request are parsed, the service time, location, and vehicle type requirements are standardized, and the initial set of suppliers is retrieved based on the city and service type index. In step S102, based on the number of CPU cores and the initial supplier set size, the calculation slices are dynamically divided. Progressive scoring is performed on suppliers in each slice. Hard condition checks are performed through the qualification verification layer. Multi-dimensional weighted basic scores are calculated for suppliers that pass the verification. The basic scores are then dynamically fine-tuned through the strategy adjustment layer to obtain the final dynamic score. Suppliers that fail the qualification verification are marked as eliminated. In step S103, the scoring results of all calculated slices are aggregated to form a scoring list. The supplier with the highest comprehensive score is selected as the winning supplier according to the multi-level sorting logic, and the rest are candidate failures. The multi-level sorting logic includes at least one of whitelist priority, perfect price priority, and comprehensive score descending order. In step S104, inventory locking and order generation are performed for the selected supplier. For each candidate failure, a multi-dimensional feature comparison analysis is conducted to compare the score gap between it and the selected supplier on key evaluation dimensions, generate a personalized set of failure reasons, and send a selection notification to the selected supplier and a rejection notification containing improvement suggestions to the candidate failure through an asynchronous message channel. In step S105, the input features, scoring details, and attribution results of this match are written to persistent storage, and a configuration check is triggered to ensure that the matching strategy parameters are up-to-date.

[0020] First, this embodiment of the application filters out an initial set of suppliers that meet basic conditions, reducing subsequent invalid calculations. Specifically, the process involves parsing various parameters of the original tourism service request (including but not limited to service time, service location, vehicle type requirements, customer level, and service type), standardizing and cleaning these parameters (e.g., unifying time format, location coordinates, and vehicle type classification standards) to avoid matching errors caused by inconsistent parameters. Then, based on a joint index of "city + service type," it quickly retrieves suppliers (i.e., studios) within the platform that meet the requirements of that region and service category, forming the initial supplier set G_init. This solves the problem in existing technologies where the initial screening lacks clear standards and has an overly broad scope, leading to redundant subsequent calculations and providing efficient input for subsequent parallel scoring.

[0021] Next, based on the number of CPU cores and the size of the candidate set, independent computation slices are dynamically created. Then, a progressive scoring process is applied to suppliers within each slice, employing a logic of "veto power + weighted calculation + dynamic fine-tuning." (1) Qualification verification layer (Layer 1): This is a hard condition check. It uses short-circuit logic. If any condition is not met, it is immediately marked as eliminated and the subsequent scoring calculation is terminated, saving computing resources. The verification content includes blacklist, order status, inventory, and price reasonableness. (2) Matching score layer: This is only performed on suppliers who have passed the qualification verification. It calculates a multi-dimensional weighted base score S_base(g). The evaluation dimensions cover at least 10 dimensions, including price, service quality, vehicle condition, and geographical location. This solves the problem that the existing technology has insufficient scoring dimensions (only 3-5) and cannot comprehensively evaluate the supplier's capabilities. Among them, the price dimension adopts a non-linear segmented linear score and makes a special digital model for the consumer psychology of the tourism industry to distinguish the penalty for small premiums and large premiums. (3) Strategy adjustment layer: The basic score is dynamically fine-tuned to obtain the final dynamic score S_final(g). The adjustment factors include whitelist gain, perfect price incentive and continuous failure penalty. The supply ecosystem is balanced through dynamic compensation mechanism to avoid the Matthew effect and achieve deep integration of business strategy and scoring results to adapt to different scenario requirements.

[0022] Then, the scoring results of all calculated slices are aggregated to form a complete supplier scoring list. Subsequently, a multi-level sorting logic is executed, with the sorting priority being "whitelist priority > perfect price priority > comprehensive score descending order." This sorting logic ensures the service priority of high-quality suppliers (whitelist) while also considering price reasonableness and comprehensive service capabilities. Finally, the supplier with the highest comprehensive score is selected as the winning supplier (g_win), and the remaining suppliers that passed the qualification verification but were not selected are designated as candidate failures (G_fail). This solves the problem of existing technologies having a single sorting logic that cannot balance fairness and business needs. For example, it avoids the service quality degradation caused by simple price priority and the situation where high-quality suppliers on the whitelist are missed due to a single comprehensive score sorting.

[0023] Then, inventory locking (locking driver and vehicle resources for the corresponding service time) and order generation are implemented for the selected suppliers to ensure that they can fulfill their obligations. Subsequently, for each candidate that fails, a multi-dimensional feature comparison analysis is conducted. By comparing the score difference between the candidate and the selected supplier in each key evaluation dimension, a personalized set of reasons for failure is generated. Through an asynchronous message channel, a selection notification (including order details, fulfillment requirements, etc.) is pushed to the selected supplier, and a rejection notification is pushed to the candidate that fails. The rejection notification includes specific improvement suggestions (such as "insufficient price competitiveness, it is recommended to adjust the price to near the system benchmark price" and "service score is too low, it is recommended to improve historical fulfillment quality"), to achieve transparent feedback of matching results and build a data closed loop to iterate the quality of supplier services.

[0024] Finally, all the data from this matching (including input features, rating details of each supplier, attribution results, order generation information, etc.) is written to persistent storage for subsequent historical data statistics and model calibration. Then, a configuration hot update synchronization mechanism is triggered to check whether the current matching strategy parameters (such as rating weights, system benchmark prices, etc.) are the latest version, ensuring that subsequent matching tasks can use the latest strategy configuration and avoid strategy rigidity.

[0025] In some embodiments, the verification function of the qualification verification layer is: Layer1(g)=¬BlackList(g)∧ReceiveStatus(g)∧Stock(g)∧PriceRange(g); Wherein, BlackList(g) indicates whether supplier g is in the blacklist, ReceiveStatus(g) indicates whether supplier g has enabled order acceptance, Stock(g) indicates whether supplier g has available inventory during the service period, and PriceRange(g) indicates whether supplier g's price is within a reasonable range; The verification logic of PriceRange(g) is: P_quote(g)≤P_system×(1+δ), where P_quote(g) is the quote from supplier g, P_system is the system benchmark price, and δ is the floating tolerance, with a default value of 20%.

[0026] The qualification verification layer is the first layer (hard constraint layer) of the hierarchical scoring model. It uses Boolean logic to implement a "one-vote veto" mechanism, ensuring that only suppliers who meet all the hard conditions can enter the subsequent scoring stages, as detailed below: 1. Qualification verification function: Layer1(g) = ¬BlackList(g) ∧ ReceiveStatus(g) ∧ Stock(g) ∧ PriceRange(g), where “∧” represents logical AND, meaning that all four conditions must be met simultaneously for the qualification verification result to be true (pass the verification). If any one condition is not met, the result is false (marked as eliminated). Short-circuit logic is used, meaning that if the previous condition is not met, there is no need to check the subsequent conditions, which greatly improves the verification efficiency.

[0027] 2. Explanation of each verification condition: (1) ¬BlackList(g): indicates that supplier g is not in the blacklist set Ω_black. The blacklist set is maintained by the platform and includes suppliers that are restricted from accepting orders due to violations (such as false quotations, breach of contract, serious customer complaints, etc.). This condition ensures that violating suppliers cannot participate in matching and guarantees service quality.

[0028] (2) ReceiveStatus(g): This indicates that supplier g is currently in the order-accepting state (i.e., State(g)=ACTIVE). Suppliers can independently turn the order-accepting state on and off according to their own operational situation (such as insufficient inventory, staff rest, etc.). This condition ensures that orders are only assigned to suppliers that can provide services, avoiding invalid matching.

[0029] (3) Stock(g): indicates that supplier g has available inventory (including driver and vehicle resources) at time t corresponding to this service, i.e., Stock(g,t)>0. Tourism services are time-sensitive, and the inventory status will change dynamically over time. This condition ensures that the selected supplier can fulfill the contract on time and avoid order default due to insufficient inventory.

[0030] (4) PriceRange(g): indicates that the supplier g's price P_quote(g) is within the reasonable range preset by the system. The specific verification logic is P_quote(g)≤P_system×(1+δ), where the parameters are defined as follows: P_quote(g): The actual quote from supplier g for this tourism service order, submitted independently by the supplier based on its own costs, service quality, and other factors; P_system: The system benchmark price, which is determined by the platform based on factors such as market conditions, service categories, service difficulty, and regional differences, and serves as a benchmark for judging the reasonableness of the price. δ: Floating tolerance, with a default value of 20%, meaning the supplier's price cannot exceed 1.2 times the system's benchmark price. This can be dynamically adjusted based on business scenarios (such as holidays or urgent orders). This condition prevents suppliers from maliciously raising prices, protects customers' price rights, and provides suppliers with reasonable pricing flexibility.

[0031] The above method eliminates suppliers who do not meet the hard requirements in advance, saves subsequent complex computing resources, and ensures the feasibility of the matching results and the reasonableness of the price, thus solving the defects of existing technologies that do not distinguish between hard requirements and soft indicators and have too much invalid calculation.

[0032] In some embodiments, the calculation logic of the multi-dimensional weighted base score is: S_base(g)=Σ i w i ×f i (g,O), where w i Let Σw be the weight of the i-th evaluation dimension. i =1, f i (g,O) is the scoring function for the i-th evaluation dimension, which includes at least price, service quality, vehicle condition, geographical location, historical performance, and response speed. The scoring function for the price dimension is a piecewise linear function with non-linear penalties. Let the price ratio r = P_quote(g) / P_system, and the scoring logic be: When r ≤ 1.0, f_price(r) = 100; When 1.0 < r ≤ 1.05, f_price(r) = 100 λ1×(r 1.0); When 1.05 < r ≤ 1.10, f_price(r) = 100 λ1×0.05 λ2×(r 1.05); When r > 1.20, f_price(r) = 0; Where λ1 and λ2 are preset attenuation coefficients, with λ2 > λ1. The two-level attenuation coefficients achieve non-linear penalty for price premiums, in order to fit the consumer psychology model of the tour charter market users who are accepting of small premiums but extremely sensitive to large premiums.

[0033] The multi-dimensional weighted basic score is the second layer (matching score layer) of the hierarchical scoring model. Its core is to comprehensively and objectively evaluate the service capabilities of the supplier through multi-dimensional, weighted summation, thus addressing the shortcomings of existing technologies such as insufficient scoring dimensions and one-sided evaluation. The specific explanation is as follows: 1. Basic score calculation logic: S_base(g)=Σ i w i ×f i (g,O), where the meanings and requirements of each parameter are as follows: (1) w i Σw represents the weight of the i-th evaluation dimension and the sum of the weights of all evaluation dimensions. i =1, the weight value can be dynamically configured according to factors such as business scenario, supply and demand status, city differences, etc. For example, emergency orders can increase the weight of "response speed", and high-value orders can increase the weight of "service quality".

[0034] (2) f i (g,O): The scoring function for the i-th evaluation dimension. The input is the features of supplier g and the features of order O. The output is the standardized score for this dimension (usually 0-100 points). Different dimensions correspond to different scoring functions to ensure the relevance and objectivity of the scoring.

[0035] (3) Evaluation dimensions: There are at least 10 dimensions, covering the core aspects of the supplier's service capabilities, including price, service quality, vehicle condition, geographical location, historical performance, response speed, customer evaluation, service qualifications, language ability, customized service capabilities, etc. Compared with existing technologies (only 3-5 dimensions), the information utilization rate is increased from about 10% to 100%, which can comprehensively evaluate the supplier's overall capabilities.

[0036] 2. Price Dimension Scoring Function (Core Dimension Example): Price is one of the key factors in matching tourism services. This application's embodiment uses a piecewise linear function to balance price competitiveness with reasonable profits for the supplier. The specific logic is as follows: (1) Price ratio r: The core calculation parameter, defined as the ratio of the supplier’s quotation to the system’s benchmark price, i.e., r=P_quote(g) / P_system, is used to standardize the quotation differences between different orders and make the scores comparable.

[0037] (2) Segmented scoring logic: When r≤1.0, f_price(r)=100 points: This means that the supplier's quotation is not higher than the system benchmark price, and the price competitiveness is the strongest, so a full score incentive is given. When 1.0 < r ≤ 1.05, f_price(r) = 100 λ1×(r 1.0): This indicates that the quoted price is slightly higher than the system's benchmark price (within 5%), resulting in a slight decrease in price competitiveness. The score decreases linearly as the quoted price increases. λ1 is a preset attenuation coefficient (configured by the platform according to business needs to adjust the attenuation rate). When 1.05 < r ≤ 1.10, f_price(r) = 100 λ1×0.05 λ2×(r 1.05): This indicates that the quoted price is 5%-10% higher than the system benchmark price, further reducing price competitiveness. A two-level attenuation is adopted (λ2>λ1), that is, after the quoted price exceeds 1.05 times, the deceleration rate accelerates, increasing the penalty for high-priced quotes. When r > 1.20, f_price(r) = 0 points: This means the price is too high (exceeding the system benchmark price by 20%), which does not meet the requirements for price reasonableness. It is given 0 points directly. Even if other dimensions score highly, it is difficult to enter the top of the final ranking to avoid malicious high prices.

[0038] (3) Attenuation coefficients λ1 and λ2: These are preset parameters that are configured by the platform according to market conditions and business strategies. For example, λ1 can be configured to 1000 (i.e., 10 points are deducted for every 1% higher than the benchmark price) and λ2 can be configured to 2000 (i.e., 20 points are deducted for every 1.05 times higher than the benchmark price). These parameters can be adjusted in real time through a dynamic configuration mechanism.

[0039] In addition, service quality scores can also be calculated based on historical data with weighting, i.e., f_service(g)=α1×Score_history(g)+α2×Rate_complete(g)+α3×Score_response(g), where α1, α2, and α3 are weighting coefficients, Score_history(g) is the historical service score, Rate_complete(g) is the fulfillment completion rate, and Score_response(g) is the response speed score, further improving the comprehensiveness of multi-dimensional scoring.

[0040] The above approach enables a comprehensive and objective assessment of the supplier's overall capabilities, ensuring that the matching results take into account both price competitiveness and service quality, thereby improving the matching success rate.

[0041] In some embodiments, the final dynamic score calculation logic of the policy adjustment layer is as follows: S_final(g)=S_base(g)×β_whitelist(g)×β_price(g)×(1 γ_fail(g)); Wherein, β_whitelist(g) is the whitelist gain coefficient and β_whitelist(g) > 1.0, triggered when supplier g is a whitelist member; β_price(g) is the perfect price incentive coefficient, triggered when |P_quote(g) Triggered when P_system|<ε, where ε is a preset error threshold; γ_fail(g) is a penalty decay factor for recent consecutive PK failures, used to avoid the Matthew effect of the strong getting stronger in the supplier ecosystem, and to ensure system load balance and fair competition.

[0042] The strategy adjustment layer is the third layer of the hierarchical scoring model. Through adjustment factors, it deeply integrates business strategies with basic scores, achieving dynamic adaptation of scoring results. This addresses the shortcomings of existing technologies where scoring is disconnected from business strategies and cannot flexibly adapt to changes in scenarios. The specific explanation is as follows: 1. Final dynamic score calculation logic: S_final(g) = S_base(g) × β_whitelist(g) × β_price(g) × (1 γ_fail(g)), where S_base(g) is the multi-dimensional weighted base score, and β_whitelist(g), β_price(g), and γ_fail(g) are three adjustment factors that are adjusted by multiplication and directly affect the final score, thus affecting the ranking result.

[0043] 2. Explanation of each regulatory factor: (1) Whitelist gain coefficient β_whitelist(g): Its core function is to prioritize the order acquisition rights of high-quality suppliers. β_whitelist(g) > 1.0 (e.g., configured as 1.2). It is triggered only when supplier g is a member of the platform's whitelist (i.e., WhiteList(g) = True). For non-whitelist members, β_whitelist(g) = 1.0 (no gain is generated). Whitelist members are selected by the platform and are usually high-quality suppliers with high service quality, high fulfillment rate, and good customer reviews. This factor solves the problem that high-quality suppliers cannot obtain priority matching in the existing technology, and at the same time improves the customer experience (e.g., the comprehensive score of whitelist studio G is higher than other studios after gain, and it is successfully selected).

[0044] (2) Perfect Price Incentive Coefficient β_price(g): Its core function is to incentivize suppliers to quote prices close to the system benchmark price, thereby achieving a balance between price rationality and supplier profit. When the absolute value of the difference between the supplier's quote and the system benchmark price is less than the preset error threshold ε (i.e., |P_quote(g))... P_system|<ε, where ε is a preset parameter, for example, configured to 1 yuan), triggers this coefficient (β_price(g)>1.0, for example, configured to 1.1); otherwise β_price(g)=1.0 (no incentive is generated). This factor encourages suppliers to quote rationally, avoid maliciously low or high prices, and make the order transaction price closer to the reasonable market price. For example, Studio B, which quotes 280 yuan (consistent with the system benchmark price), can obtain this incentive coefficient, improving the final score.

[0045] (3) Consecutive failure penalty attenuation factor γ_fail(g): Its core function is to avoid the Matthew effect, ensure fair competition opportunities for novice suppliers and suppliers with recent consecutive failures, and promote the healthy development of the supplier ecosystem. γ_fail(g) is a value between 0 and 1, and is a function of the number of consecutive failures in recent competitive bidding by the supplier. The more consecutive failures, the larger γ_fail(g) is (the stronger the penalty), i.e., (1 The smaller γ_fail(g) is, the lower the final score; if there are no consecutive failures recently, then γ_fail(g) = 0, and no penalty is imposed (final score = base score × other gain coefficients). This factor solves the problem in existing technologies where suppliers with proximity and obvious advantages monopolize orders, making it difficult for novice suppliers to obtain opportunities, thus increasing the order share of novice suppliers.

[0046] 3. Configuration of adjustment factors: The specific values ​​of β_whitelist(g) and β_price(g), as well as the calculation logic of ε and γ_fail(g), can all be adjusted through a dynamic configuration mechanism without modifying the code or requiring downtime for deployment, thus adapting to the needs of different business scenarios.

[0047] The above approach enables flexible adaptation between business strategies and scoring results, ensuring the advantages of high-quality suppliers and suppliers with reasonable prices, while also ensuring fairness, promoting the healthy development of the supplier ecosystem, and improving the rationality of the matching results.

[0048] In some embodiments, the dynamic sharding logic for sensing hardware load is: K = min(C, N / θ_batch ), where K is the number of slices, C is the number of CPU cores, N is the initial supplier set size, and θ_batch is the preset maximum number of slices; this dynamic scheduling strategy is used to achieve millisecond-level matching calculation in scenarios with a large number of candidate suppliers, so as to solve the real-time problem of temporary vehicle use and emergency dispatch in tourism services; each slice is non-overlapping and the union of the slices is the initial supplier set.

[0049] The logic for calculating the number of slices is: K = min(C, N / θ_batch The meanings and requirements of each parameter are as follows: (1) K: The number of slices, i.e. the number of subtasks after splitting, and also the number of tasks executed in parallel; (2) C: CPU core count, i.e. the number of CPU cores of the server currently executing the scoring task. The number of slices cannot exceed the number of CPU cores to avoid increasing scheduling overhead due to too many tasks and to ensure the efficiency of parallel computing. (3) N: Initial supplier set size, i.e. the number of initial suppliers obtained by filtering in step S1 (the size of G_init). (4) θ_batch: Preset maximum number of slices, i.e., the upper limit of the number of suppliers in each computation slice. It is configured by the platform according to server performance and scoring task complexity (e.g., configured as 20) to avoid excessive computation time caused by excessively large slice size and to ensure that the computation time of each slice is balanced. (5) N / θ_batch : This means rounding up the result of N / θ_batch. That is, when N is not divisible by θ_batch, the number of suppliers in the last slice is less than θ_batch, but it ensures that all initial suppliers can be assigned to the slice.

[0050] The computation slices are mutually exclusive (i.e., the same supplier will not be assigned to multiple slices, avoiding duplicate calculations), and the union of all slices equals the initial supplier set (i.e., all initial suppliers can be assigned to a slice, avoiding omissions in calculations). This requirement ensures the accuracy of parallel computation, meaning that after aggregating the scoring results of all slices, a complete list of scores for all initial suppliers can be obtained.

[0051] This logic dynamically determines the number of slices by combining hardware resources (number of CPU cores) and task size (initial number of suppliers). This avoids both too few slices (which fails to fully utilize multi-core CPU resources, resulting in minimal efficiency gains) and too many slices (which increases scheduling overhead and actually reduces efficiency). For example, when the number of CPU cores is 8, the initial number of suppliers is 100, and the maximum number of slices per slice is 20... 100 / 20 =5, K=min(8,5)=5, that is, split into 5 slices, each slice has 20 suppliers, 5 of the 8 CPU cores are used to execute the scoring task in parallel, and 3 are used as backups to ensure computational efficiency.

[0052] By combining a parallel computing model, after the slice is divided, the scoring tasks of each slice are executed in parallel. The total execution time is T_total = T_split + max(T_compute(G_k)) + T_merge, where T_split is the time spent on task splitting (i.e., the time spent in this step), T_compute(G_k) is the time spent on scoring a single slice, and T_merge is the time spent on result aggregation. Through this parallel strategy, the scoring efficiency is significantly improved compared to serial computing.

[0053] The above method enables the reasonable division of scoring tasks, makes full use of multi-core CPU resources, greatly improves scoring efficiency, and meets the real-time matching needs of large-scale concurrent orders.

[0054] In some embodiments, the logic for generating the personalized failure reason set is as follows: R={d_j∈Dimensions|S_d(g_win) S_d(g_fail)>ε_diff∪F_hard(g_fail)}; Wherein, d_j is the evaluation dimension, S_d(g_win) is the score of the selected supplier in dimension d_j, S_d(g_fail) is the score of the candidate who failed in dimension d_j, ε_diff is the significant difference threshold, and F_hard(g_fail) is the set of failure items of the candidate who failed in the qualification verification layer. The system automatically identifies the difference in the core dimension scores between the selected and unselected suppliers through this logic, realizes multi-dimensional feature comparison analysis, and builds a data feedback closed loop to iterate the quality of supplier services.

[0055] The logic for generating the set of failure reasons is: R = {d_j ∈ Dimensions | S_d(g_win)} S_d(g_fail)>ε_diff∪F_hard(g_fail)}, this logic indicates that the set of failure reasons R contains two types of reasons—soft dimension difference reasons and hard condition failure reasons, which are explained in detail below: (1) Meaning of the symbols: d_j: Evaluation dimensions (i.e., 10+ evaluation dimensions, such as price, service quality, vehicle condition, etc.); Dimensions: The collection of all evaluation dimensions; S_d(g_win): The detailed score of the selected supplier g_win on dimension d_j; S_d(g_fail): The detailed score of the candidate failure g_fail on dimension d_j; ε_diff: Significant difference threshold, a preset parameter (e.g., configured as 10 points), used to determine whether the difference between the candidate failure and the selected supplier is significant in this dimension; F_hard(g_fail): The set of failure items of candidate failure g_fail in the qualification verification layer (Layer 1). That is, the supplier passes the qualification verification, but has flaws in some hard conditions (does not meet the elimination standard, but affects competitiveness), or fails to pass a certain hard condition but enters the candidate failure queue for other reasons (such as inventory threshold). ∪: Logical OR, indicating that the set of failure reasons R contains all reasons that satisfy either "significant difference in dimensional score" or "existence of a hard condition failure item".

[0056] The process for generating the failure reason is as follows: (1) First, compare the scores of the candidate failure g_fail and the selected supplier g_win on each evaluation dimension d_j, and calculate the score difference S_d(g_win). S_d(g_fail); (2) If the score difference of a certain dimension is greater than the significant difference threshold ε_diff, it indicates that the dimension is the main disadvantage of the candidate failure party. The dimension (and the corresponding gap description) is included in the failure reason set R (e.g., "the price dimension score is 15 points lower than the winning party, and the price competitiveness is insufficient"). (3) At the same time, check the performance of the candidate failure party in the qualification verification layer. If there is a hard condition failure item (such as the price is close to the upper limit of the reasonable range, the inventory is insufficient but temporarily replenished, etc.), the failure item is included in the failure reason set R (such as "the inventory is close to the critical value and the performance risk is high"). (4) The final set of personalized failure reasons R will be pushed to the candidate failure party along with the rejection notification, clearly informing them "where they are lacking" so as to facilitate targeted improvement.

[0057] For example, candidate D (ST10004) bids 300 yuan, and successful bidder G (ST10007) bids 270 yuan. Their price dimension scores are 85.7 and 96.4 respectively, with a score difference of 10.7 points. If ε_diff=10 points, then "price dimension score is 10.7 points lower than successful bidder, price competitiveness is insufficient" will be included in the failure reason set R.

[0058] The above methods enable transparency and personalization of the reasons for failure, helping suppliers to improve their services in a targeted manner, enhance their competitiveness, reduce studio complaint rates, and promote a positive cycle in the supplier ecosystem.

[0059] In some embodiments, when executing the multi-dimensional dynamic scoring and real-time matching method for tourism service supply and demand matching, a two-level caching architecture of L1 local memory and L2 distributed cache is adopted. L1 stores frequently accessed static basic information of suppliers, and L2 stores dynamic business status of suppliers. The cache access link is L1→L2→database. The Cache Aside mode and delayed double deletion strategy are adopted to ensure the consistency between the cache and the database.

[0060] This application's embodiment adopts an architecture combining L1 local memory caching and L2 distributed caching, storing different types of data in layers to achieve high-efficiency data access. The specific layering is as follows: L1 Local Memory Cache: Deployed in the local memory of each application server, it stores frequently accessed static basic information about suppliers. This information is updated infrequently but accessed frequently, including but not limited to supplier qualifications, service categories, whitelist / blacklist status, and fixed service scope. L1 cache has the fastest access speed (average 1ms), a hit rate of up to 85%, and can quickly respond to data access requests, reducing reliance on subsequent caches and databases.

[0061] L2 Distributed Cache: Deployed in a distributed cache cluster, it stores the dynamic business status of suppliers. This type of information is updated and accessed frequently, including but not limited to order status, real-time inventory, current price, recent response speed, and temporary violation records. The access speed of L2 cache is second best (average 5ms). It is triggered when L1 cache misses, and the hit rate can reach 95% (when L1 misses), further reducing the number of database accesses.

[0062] The cache access chain adopts a priority access chain of "L1→L2→database". When the system needs to access certain data, it first queries the L1 local memory cache. If the data is found, it returns the data directly. If L1 misses, it queries the L2 distributed cache. If L2 hits, the data is returned and synchronized to the L1 cache (improving the efficiency of subsequent accesses). If L2 also misses, it queries the underlying database, retrieves the data, returns it, and synchronizes it to the L1 and L2 caches (updating the cached data). This chain ensures high-efficiency data access and significantly reduces database I / O pressure.

[0063] The cache and database consistency guarantee adopts a combination of "Cache Aside mode" and "delayed double-delete strategy" to ensure that cached data can be synchronized in a timely manner after database data is updated, avoiding data inconsistency (such as matching errors caused by outdated cached data). The specific strategy is as follows: (1) Cache Aside mode: The core logic is "read cache, write database, delete cache", that is, when reading data, the cache is accessed first, and when writing data, the database is updated first, and then the corresponding cache data is deleted (to avoid inconsistency between cache data and database data). (2) Delayed double deletion strategy: After writing data and deleting cache, delay for a preset time (e.g., 100ms) and delete the corresponding cache data again to solve the problem of occasional data inconsistency caused by asynchronous cache update and database update (e.g., the first deletion operation has not taken effect, or other threads are reading cache data).

[0064] Compared to a single caching architecture, a two-level caching architecture balances access speed and data consistency. The L1 cache ensures extremely fast access to high-frequency static data, while the L2 cache ensures efficient access to dynamic data. At the same time, a consistency strategy is used to avoid data errors. In addition, the distributed L2 cache also has high availability. Even if a cache node fails, it will not affect the overall caching service, ensuring stable system operation.

[0065] The above methods significantly reduce database I / O pressure, improve data access speed, support the real-time matching needs of large-scale concurrent orders, and ensure data consistency, avoiding matching errors caused by outdated cache.

[0066] In some embodiments, the matching strategy parameters include a scoring dimension weight vector, a system benchmark price P_system, and a floating tolerance δ. Fine-grained configuration by city, time period, and vehicle type is supported, and hot updates of strategy parameters are achieved using version number control and a copy-on-write atomic update mechanism.

[0067] The range of matching strategy parameters includes three core categories, covering the entire process of scoring, verification, and adjustment, as detailed below: (1) Weight vector of scoring dimensions: i.e., the weight w of each evaluation dimension i The vector W = {w1, w2, ..., w} is formed. m} (m is the number of evaluation dimensions), used to adjust the proportion of different dimensions in the overall score; (2) System benchmark price P_system: that is, the price judgment benchmark, used to evaluate the reasonableness of the quotation and calculate the price dimension score; (3) Floating tolerance δ: that is, the upper limit adjustment parameter of the reasonable range of the quotation, which is used to control the maximum limit of the supplier's quotation; In addition, it also includes all parameters related to business strategy, such as attenuation coefficients λ1 and λ2, gain coefficients β_whitelist and β_price, error threshold ε, penalty factor γ_fail, and significant difference threshold ε_diff.

[0068] The strategy parameters support fine-grained configuration by city, time period, and vehicle type. This means that different parameter values ​​can be configured for different cities, different time periods (such as peak / off-peak, holidays / weekdays), and different vehicle types (such as 5-seater minivans and 7-seater SUVs) to adapt to the business needs of different scenarios. For example, the weight of "service quality" can be increased in first-tier cities, the weight of "inventory stability" can be increased during holidays, and the floating tolerance δ (controlling the upper limit of the price) can be reduced for high-value vehicles.

[0069] The strategy parameter hot update mechanism adopts a combination of "version number control" and "copy-on-write atomic update mechanism" to achieve hot updates of strategy parameters without modifying code, recompiling, or requiring downtime for deployment.

[0070] In summary, the embodiments of this application have the following beneficial effects: Its scoring dimensions are comprehensive and dynamically adjustable, employing multi-dimensional evaluation and a strategy adjustment layer to achieve dynamic fine-tuning of scores. It also supports fine-grained configuration and hot updates of strategy parameters by city, time period, and service type, effectively solving the shortcomings of existing technologies' single and static scoring. Matching efficiency is extremely high, achieved through dual optimization of "initial screening + parallel slice scoring," combined with a two-level caching architecture of L1 local memory and L2 distributed cache, significantly shortening matching time to meet the real-time requirements of tourism services. Simultaneously, a "one-vote veto" mechanism in the qualification verification layer reduces invalid calculations. Matching accuracy is outstanding, employing a multi-level sorting logic of "whitelist priority > perfect price priority > comprehensive score descending order," combined with multi-dimensional weighted scoring, ensuring a high degree of match between selected suppliers and user needs, significantly improving user experience. It can achieve closed-loop optimization, providing personalized improvement suggestions to failed candidates through an intelligent attribution mechanism, and adaptively optimizing scoring model parameters based on historical matching data and performance feedback to continuously improve matching quality. The system is extremely stable, employing parallel computing, two-level caching, and hot updates of strategy parameters to effectively reduce system bottlenecks and avoid the impact of downtime maintenance on services. Furthermore, through caching... The Aside mode and delayed double-delete strategy ensure data consistency between the cache and the database, significantly improving system reliability.

[0071] Based on the same inventive concept, this application also provides a multi-dimensional dynamic scoring and real-time matching device for tourism service supply and demand matching, which corresponds to the multi-dimensional dynamic scoring and real-time matching method for tourism service supply and demand matching in the first embodiment. Since the principle of the device in this application is similar to the multi-dimensional dynamic scoring and real-time matching method for tourism service supply and demand matching, the implementation of the device can refer to the implementation of the method, and the repeated parts will not be described again.

[0072] like Figure 2 As shown, Figure 2This is a schematic diagram of the structure of the multi-dimensional dynamic scoring and real-time matching device 200 for tourism service supply and demand matching provided in this application embodiment. The multi-dimensional dynamic scoring and real-time matching device 200 for tourism service supply and demand matching includes: The initial screening module 201 is used to parse the parameters of the original tourism service request, standardize the service time, location, and vehicle type requirements, and retrieve the initial set of suppliers based on the city and service type index. The scoring module 202 is used to divide the initial supplier set into multiple calculation slices, perform progressive scoring on suppliers in each slice, conduct hard condition checks through the qualification verification layer, calculate multi-dimensional weighted basic scores for suppliers that pass the verification, and then dynamically fine-tune the basic scores through the strategy adjustment layer to obtain the final dynamic score. Suppliers that fail the qualification verification are marked as eliminated. The decision module 203 is used to aggregate the scoring results of all computation slices, form a scoring list, and select the supplier with the highest comprehensive score as the selected supplier according to a multi-level sorting logic, while the rest are candidate failures; wherein, the multi-level sorting logic includes at least one of whitelist priority, perfect price priority, and comprehensive score descending order; Feedback module 204 is used to perform inventory locking and order generation for the selected supplier. For each candidate failure, it generates a personalized set of failure reasons by comparing its score difference with the selected supplier on key dimensions, and sends a selection notification to the selected supplier and a rejection notification containing improvement suggestions to the candidate failure through an asynchronous message channel. The maintenance module 205 is used to write the input features, scoring details and attribution results of this match to persistent storage and trigger configuration checks to ensure that the matching strategy parameters are up-to-date.

[0073] Those skilled in the art should understand that Figure 2 The functions of each unit in the multi-dimensional dynamic scoring and real-time matching device 200 for matching supply and demand of tourism services can be understood by referring to the relevant description of the aforementioned multi-dimensional dynamic scoring and real-time matching method for matching supply and demand of tourism services. Figure 2 The functions of each unit in the multi-dimensional dynamic scoring and real-time matching device 200 for matching supply and demand of tourism services shown can be realized by a program running on a processor or by specific logic circuits.

[0074] In one possible implementation, the verification function of the qualification verification layer is: Layer1(g)=¬BlackList(g)∧ReceiveStatus(g)∧Stock(g)∧PriceRange(g); Wherein, BlackList(g) indicates whether supplier g is in the blacklist, ReceiveStatus(g) indicates whether supplier g has enabled order acceptance, Stock(g) indicates whether supplier g has available inventory during the service period, and PriceRange(g) indicates whether supplier g's price is within a reasonable range; The verification logic of PriceRange(g) is: P_quote(g)≤P_system×(1+δ), where P_quote(g) is the quote from supplier g, P_system is the system benchmark price, and δ is the floating tolerance, with a default value of 20%.

[0075] In one possible implementation, the calculation logic of the multi-dimensional weighted base score is: S_base(g) = Σ i w i ×f i (g,O), where w i Let Σw be the weight of the i-th evaluation dimension. i =1, f i (g,O) is the scoring function for the i-th evaluation dimension, which includes at least price, service quality, vehicle condition, geographical location, historical performance, and response speed. The scoring function for the price dimension is a piecewise linear function. Let the price ratio r = P_quote(g) / P_system, and the scoring logic is as follows: When r ≤ 1.0, f_price(r) = 100; When 1.0 < r ≤ 1.05, f_price(r) = 100 λ1×(r 1.0); When 1.05 < r ≤ 1.10, f_price(r) = 100 λ1×0.05 λ2×(r 1.05); When r > 1.20, f_price(r) = 0; Where λ1 and λ2 are preset attenuation coefficients.

[0076] In one possible implementation, the final dynamic score calculation logic of the policy adjustment layer is as follows: S_final(g)=S_base(g)×β_whitelist(g)×β_price(g)×(1 γ_fail(g)); Wherein, β_whitelist(g) is the whitelist gain coefficient and β_whitelist(g) > 1.0, triggered when supplier g is a whitelist member; β_price(g) is the perfect price incentive coefficient, triggered when |P_quote(g) Triggered when P_system| < ε, where ε is a preset error threshold; γ_fail(g) is the penalty decay factor for recent consecutive PK failures.

[0077] In one possible implementation, the logic for calculating the slice segmentation is: K = min(C, N / θ_batch ), where K is the number of slices, C is the number of CPU cores, N is the initial supplier set size, and θ_batch is the preset maximum number of slices; the slices are mutually exclusive and their union is the initial supplier set.

[0078] In one possible implementation, the logic for generating the personalized failure reason set is as follows: R={d_j∈Dimensions|S_d(g_win) S_d(g_fail)>ε_diff∪F_hard(g_fail)}; Where d_j is the evaluation dimension, S_d(g_win) is the score of the selected supplier in dimension d_j, S_d(g_fail) is the score of the candidate failure in dimension d_j, ε_diff is the significant difference threshold, and F_hard(g_fail) is the set of failure items of the candidate failure in the qualification verification layer.

[0079] In one possible implementation, when executing the multi-dimensional dynamic scoring and real-time matching method for tourism service supply and demand matching, a two-level caching architecture of L1 local memory and L2 distributed cache is adopted. L1 stores frequently accessed static basic information of suppliers, and L2 stores dynamic business status of suppliers. The cache access link is L1→L2→database. The cache Aside mode and delayed double deletion strategy are adopted to ensure the consistency between the cache and the database.

[0080] In one possible implementation, the matching strategy parameters include a scoring dimension weight vector, a system benchmark price P_system, and a floating tolerance δ. It supports fine-grained configuration by city, time period, and vehicle type, and uses version number control and a copy-on-write atomic update mechanism to achieve hot updates of the strategy parameters.

[0081] The aforementioned multi-dimensional dynamic scoring and real-time matching device for tourism service supply and demand has the following beneficial effects: Its scoring dimensions are comprehensive and dynamically adjustable, employing multi-dimensional evaluation and a strategy adjustment layer to achieve dynamic fine-tuning of scores. It also supports fine-grained configuration and hot updates of strategy parameters by city, time period, and service type, effectively solving the shortcomings of existing technologies' single and static scoring. Matching efficiency is extremely high, achieved through dual optimization of "initial screening + parallel slice scoring," combined with a two-level caching architecture of L1 local memory and L2 distributed cache, significantly shortening matching time to meet the real-time requirements of tourism services. Simultaneously, a "one-vote veto" mechanism in the qualification verification layer reduces invalid calculations. Matching accuracy is outstanding, employing a multi-level sorting logic of "whitelist priority > perfect price priority > comprehensive score descending order," combined with multi-dimensional weighted scoring, ensuring a high degree of match between selected suppliers and user needs, significantly improving user experience. It can achieve closed-loop optimization, providing personalized improvement suggestions to failed candidates through an intelligent attribution mechanism, and adaptively optimizing scoring model parameters based on historical matching data and performance feedback to continuously improve matching quality. The system is extremely stable, employing parallel computing, two-level caching, and hot updates of strategy parameters to effectively reduce system bottlenecks and avoid the impact of downtime maintenance on services. Furthermore, through caching... The Aside mode and delayed double-delete strategy ensure data consistency between the cache and the database, significantly improving system reliability.

[0082] like Figure 3 As shown, Figure 3 This is a schematic diagram of the composition structure of the electronic device 300 provided in the embodiments of this application. The electronic device 300 includes: The device includes a processor 301, a storage medium 302, and a bus 303. The storage medium 302 stores machine-readable instructions executable by the processor 301. When the electronic device 300 is running, the processor 301 communicates with the storage medium 302 via the bus 303. The processor 301 executes the machine-readable instructions to perform the steps of the multi-dimensional dynamic scoring and real-time matching method for tourism service supply and demand matching described in this application embodiment.

[0083] In practical applications, the various components in the electronic device 300 are coupled together via bus 303. It is understood that bus 303 is used to achieve communication between these components. In addition to a data bus, bus 303 also includes a power bus, a control bus, and a status signal bus. However, for clarity, in... Figure 3 The general designated all buses as Bus 303.

[0084] The above-mentioned electronic devices have the following beneficial effects: Its scoring dimensions are comprehensive and dynamically adjustable, employing multi-dimensional evaluation and a strategy adjustment layer to achieve dynamic fine-tuning of scores. It also supports fine-grained configuration and hot updates of strategy parameters by city, time period, and service type, effectively solving the shortcomings of existing technologies' single and static scoring. Matching efficiency is extremely high, achieved through dual optimization of "initial screening + parallel slice scoring," combined with a two-level caching architecture of L1 local memory and L2 distributed cache, significantly shortening matching time to meet the real-time requirements of tourism services. Simultaneously, a "one-vote veto" mechanism in the qualification verification layer reduces invalid calculations. Matching accuracy is outstanding, employing a multi-level sorting logic of "whitelist priority > perfect price priority > comprehensive score descending order," combined with multi-dimensional weighted scoring, ensuring a high degree of match between selected suppliers and user needs, significantly improving user experience. It can achieve closed-loop optimization, providing personalized improvement suggestions to failed candidates through an intelligent attribution mechanism, and adaptively optimizing scoring model parameters based on historical matching data and performance feedback to continuously improve matching quality. The system is extremely stable, employing parallel computing, two-level caching, and hot updates of strategy parameters to effectively reduce system bottlenecks and avoid the impact of downtime maintenance on services. Furthermore, through caching... The Aside mode and delayed double-delete strategy ensure data consistency between the cache and the database, significantly improving system reliability.

[0085] This application also provides a computer-readable storage medium storing executable instructions. When the executable instructions are executed by at least one processor 301, the multi-dimensional dynamic scoring and real-time matching method for tourism service supply and demand matching described in this application is implemented.

[0086] In some embodiments, the storage medium may be a magnetic random access memory (FRAM), a read-only memory (ROM), or a programmable read-only memory (PROM). Erasable Programmable Read-Only Memory (EPROM) Electrically Erasable Programmable Read-Only Memory (EEPROM) Read-only memory, flash memory, magnetic surface storage, optical disc, or CD-ROM ROM, Compact Disc Read It can be a memory such as a memory only; or it can be a device that includes one or any combination of the above-mentioned memories.

[0087] In some embodiments, executable instructions may take the form of a program, software, software module, script, or code, written in any form of programming language (including compiled or interpreted languages, or declarative or procedural languages), and may be deployed in any form, including as a standalone program or as a module, component, subroutine, or other unit suitable for use in a computing environment.

[0088] As an example, executable instructions may, but do not necessarily, correspond to files in the file system. They may be stored as part of a file that holds other programs or data, for example, in one or more scripts in a HyperText Markup Language (HTML) document, in a single file dedicated to the program in question, or in multiple collaborating files (e.g., a file that stores one or more modules, subroutines, or code sections).

[0089] As an example, executable instructions can be deployed to execute on a single computing device, or on multiple computing devices located in one location, or on multiple computing devices distributed across multiple locations and interconnected via a communication network.

[0090] The aforementioned computer-readable storage media have the following beneficial effects: Its scoring dimensions are comprehensive and dynamically adjustable, employing multi-dimensional evaluation and a strategy adjustment layer to achieve dynamic fine-tuning of scores. It also supports fine-grained configuration and hot updates of strategy parameters by city, time period, and service type, effectively solving the shortcomings of existing technologies' single and static scoring. Matching efficiency is extremely high, achieved through dual optimization of "initial screening + parallel slice scoring," combined with a two-level caching architecture of L1 local memory and L2 distributed cache, significantly shortening matching time to meet the real-time requirements of tourism services. Simultaneously, a "one-vote veto" mechanism in the qualification verification layer reduces invalid calculations. Matching accuracy is outstanding, employing a multi-level sorting logic of "whitelist priority > perfect price priority > comprehensive score descending order," combined with multi-dimensional weighted scoring, ensuring a high degree of match between selected suppliers and user needs, significantly improving user experience. It can achieve closed-loop optimization, providing personalized improvement suggestions to failed candidates through an intelligent attribution mechanism, and adaptively optimizing scoring model parameters based on historical matching data and performance feedback to continuously improve matching quality. The system is extremely stable, employing parallel computing, two-level caching, and hot updates of strategy parameters to effectively reduce system bottlenecks and avoid the impact of downtime maintenance on services. Furthermore, through caching... The Aside mode and delayed double-delete strategy ensure data consistency between the cache and the database, significantly improving system reliability.

[0091] In the several embodiments provided in this application, it should be understood that the disclosed methods and electronic devices can be implemented in other ways. The device embodiments described above are merely illustrative. For example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods, such as: multiple units or components may be combined, or integrated into another system, or some features may be ignored or not executed. In addition, the coupling, direct coupling, or communication connection between the various components shown or discussed may be through some interfaces, and the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0092] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0093] In addition, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0094] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a processor-executable, non-volatile, computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, a platform server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, ROM, RAM, magnetic disks, or optical disks.

[0095] The above are merely specific embodiments of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A multi-dimensional dynamic scoring and real-time matching method for supply and demand matching of tourism services, characterized in that, Includes the following steps: The parameters of the original tourism service request are analyzed, the service time, location, and vehicle type requirements are standardized, and the initial set of suppliers is retrieved based on the city and service type index. Based on the number of CPU cores and the initial supplier set size, the calculation slices are dynamically divided. A progressive scoring is performed on the suppliers in each slice. Hard condition checks are performed through the qualification verification layer. Multi-dimensional weighted base scores are calculated for suppliers that pass the verification. The base scores are then dynamically fine-tuned through the strategy adjustment layer to obtain the final dynamic score. Suppliers that fail the qualification verification are marked as eliminated. The scoring results of all computational slices are aggregated to form a scoring list. The supplier with the highest comprehensive score is selected as the winning supplier according to a multi-level sorting logic, and the rest are candidate failures. The multi-level sorting logic includes at least one of whitelist priority, perfect price priority, and comprehensive score descending order. For the selected supplier, inventory locking and order generation are performed. For each candidate failure, a multi-dimensional feature comparison analysis is conducted to compare the score gap between it and the selected supplier on key evaluation dimensions, generate a personalized set of failure reasons, and send a selection notification to the selected supplier and a rejection notification containing improvement suggestions to the candidate failure through an asynchronous message channel. The input features, score details, and attribution results of this match are written to persistent storage, and a configuration check is triggered to ensure that the matching strategy parameters are up-to-date.

2. The method of claim 1, wherein, The verification function of the qualification verification layer is: Layer1(g)=¬BlackList(g)∧ReceiveStatus(g)∧Stock(g)∧PriceRange(g); Wherein, BlackList(g) indicates whether supplier g is in the blacklist, ReceiveStatus(g) indicates whether supplier g has enabled order acceptance, Stock(g) indicates whether supplier g has available inventory during the service period, and PriceRange(g) indicates whether supplier g's price is within a reasonable range; The verification logic of PriceRange(g) is: P_quote(g)≤P_system×(1+δ), where P_quote(g) is the quote from supplier g, P_system is the system benchmark price, and δ is the floating tolerance, with a default value of 20%.

3. The method of claim 1, wherein, The calculation logic for the multi-dimensional weighted base score is: S_base(g) = Σ i w i ×f i (g,O), where w i Let Σw be the weight of the i-th evaluation dimension. i =1, f i (g,O) is the scoring function for the i-th evaluation dimension, which includes at least price, service quality, vehicle condition, geographical location, historical performance, and response speed. The scoring function for the price dimension is a piecewise linear function with non-linear penalties. Let the price ratio r = P_quote(g) / P_system, and the scoring logic be: When r ≤ 1.0, f_price(r) = 100; f_price(r) = 100 when 1.0 < r < 1.05 λ1x(r 1.0); f_price(r) = 100 when 1.05 < r < 1.10 λ1 x 0.05 λ2 x (r 1.05); When r > 1.20, f_price(r) = 0; Where λ1 and λ2 are preset attenuation coefficients, with λ2 > λ1. The two-level attenuation coefficients achieve non-linear penalty for price premiums, in order to fit the consumer psychology model of the tour charter market users who are accepting of small premiums but extremely sensitive to large premiums.

4. The method of claim 1, wherein, The final dynamic score calculation logic of the strategy adjustment layer is as follows: S_final(g)=S_base(g)×β_whitelist(g)×β_price(g)×(1 γ_fail(g)); Wherein, β_whitelist(g) is the whitelist gain coefficient and β_whitelist(g) > 1.0, triggered when supplier g is a whitelist member; β_price(g) is the perfect price incentive coefficient, triggered when |P_quote(g) Triggered when P_system|<ε, where ε is a preset error threshold; γ_fail(g) is a penalty decay factor for recent consecutive PK failures, used to avoid the Matthew effect of the strong getting stronger in the supplier ecosystem, and to ensure system load balance and fair competition.

5. The method according to claim 1, characterized in that, The dynamic sharding logic for sensing hardware load is: K = min(C, N / θ_batch ), where K is the number of slices, C is the number of CPU cores, N is the initial supplier set size, and θ_batch is the preset maximum number of slices; this dynamic scheduling strategy is used to achieve millisecond-level matching calculation in scenarios with a large number of candidate suppliers, so as to solve the real-time problem of temporary vehicle use and emergency dispatch in tourism services; Each slice is non-overlapping and its union forms the initial supplier set.

6. The method according to claim 1, characterized in that, The logic for generating the personalized failure reason set is as follows: R={d_j∈Dimensions|S_d(g_win) S_d(g_fail)>ε_diff∪F_hard(g_fail)}; Wherein, d_j is the evaluation dimension, S_d(g_win) is the score of the selected supplier in dimension d_j, S_d(g_fail) is the score of the candidate who failed in dimension d_j, ε_diff is the significant difference threshold, and F_hard(g_fail) is the set of failure items of the candidate who failed in the qualification verification layer. The system automatically identifies the difference in the core dimension scores between the selected and unselected suppliers through this logic, realizes multi-dimensional feature comparison analysis, and builds a data feedback closed loop to iterate the quality of supplier services.

7. The method according to claim 1, characterized in that, The matching strategy parameters include a scoring dimension weight vector, a system benchmark price P_system, and a floating tolerance δ. It supports fine-grained configuration by city, time period, and vehicle type, and uses version number control and a copy-on-write atomic update mechanism to achieve hot updates of strategy parameters.

8. A multi-dimensional dynamic scoring and real-time matching device for tourism service supply and demand, characterized in that, The device includes: The initial screening module is used to parse the parameters of the original tourism service request, standardize the service time, location, and vehicle type requirements, and retrieve the initial set of suppliers based on the city and service type index. The scoring module is used to dynamically divide the calculation slices based on the number of CPU cores and the initial supplier set size. It performs progressive scoring on suppliers in each slice, conducts hard condition checks through the qualification verification layer, calculates multi-dimensional weighted base scores for suppliers that pass the verification, and then dynamically fine-tunes the base scores through the strategy adjustment layer to obtain the final dynamic score. Suppliers that fail the qualification verification are marked as eliminated. The decision module is used to aggregate the scoring results of all computation slices, form a scoring list, and select the supplier with the highest comprehensive score as the selected supplier according to a multi-level sorting logic, while the rest are candidate failures; wherein, the multi-level sorting logic includes at least one of whitelist priority, perfect price priority, and comprehensive score descending order; The feedback module is used to lock inventory and generate orders for the selected suppliers. For each candidate that fails, it compares the score difference between the candidate and the selected supplier on key evaluation dimensions through multi-dimensional feature comparison analysis, generates a personalized set of reasons for failure, and sends a selection notification to the selected supplier and a rejection notification containing improvement suggestions to the candidate that fails through an asynchronous message channel. The maintenance module is used to write the input features, scoring details, and attribution results of this match to persistent storage and trigger configuration checks to ensure that the matching strategy parameters are up-to-date.

9. An electronic device, characterized in that, include: The device includes a processor, a storage medium, and a bus. The storage medium stores machine-readable instructions executable by the processor. When the electronic device is running, the processor communicates with the storage medium via the bus, and the processor executes the machine-readable instructions to perform the multi-dimensional dynamic scoring and real-time matching method for tourism service supply and demand matching as described in 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, which, when executed by a processor, performs the multi-dimensional dynamic scoring and real-time matching method for tourism service supply and demand as described in any one of claims 1 to 7.

Images