Intelligent pricing method and system for shared vehicle rental platform, and related equipment

Abstract

The invention provides an intelligent pricing method and system for a shared vehicle rental platform and related equipment, and the method comprises the steps: predicting a rental demand of a target geographic region in a target time period based on historical operation data through a preset demand prediction model; inputting the lease demand as a decision context into a preset reinforcement learning pricing model; selecting a target price strategy from a plurality of preset price strategies through the reinforcement learning pricing model, and executing the target price strategy on the shared vehicle in the target geographic area; according to a leasing result generated after execution of the target price strategy, calculating a reward value of the pricing decision; and updating the reinforcement learning pricing model and the demand prediction model by using the decision context, the target price strategy and the reward value. The invention provides an intelligent pricing method which is based on historical data driving, considers geographic area consideration and forms closed-loop feedback, pricing is flexible, and adaptability is higher.

Description

Technical Field

[0001] This invention relates to the field of shared vehicle technology, and in particular to an intelligent pricing method, system and related equipment for a shared vehicle rental platform. Background Technology

[0002] Currently, most two-wheeled electric vehicle rental platforms adopt fixed pricing or simple time-sharing pricing strategies, such as billing by the hour or by the day, lacking comprehensive consideration of factors such as real-time market supply and demand and geographical areas. Existing technology results in rigid pricing, poor user experience, and users facing the same price at different times and locations, lacking personalized incentives, and unable to predict future demand and adjust pricing strategies in advance, making them weak in responding to emergencies.

[0003] Therefore, existing technologies still need to be improved and developed. Summary of the Invention

[0004] This invention provides an intelligent pricing method, system, and related equipment for a shared vehicle rental platform. The main objective of this invention is to solve the technical problems mentioned in the background section of the prior art.

[0005] The first aspect of this invention provides an intelligent pricing method for a shared vehicle rental platform, comprising: Based on historical operational data and through a pre-set demand forecasting model, predict the rental demand of the target geographical area within the target time period; The rental demand is used as the decision context input into a preset reinforcement learning pricing model; The reinforcement learning pricing model selects a target pricing strategy from a set of preset pricing strategies and applies it to shared vehicles within the target geographic area. Calculate the reward value for this pricing decision based on the leasing results generated after the execution of the target price strategy; The reinforcement learning pricing model and the demand forecasting model are updated using the decision context, the target price strategy, and the reward value.

[0006] In an optional embodiment of the first aspect of the present invention, predicting the rental demand of a target geographical area within a target time period based on historical operational data and a preset demand prediction model includes: Acquire historical operational data, including vehicle dynamic data, environmental parameter data, user demand data, and platform operational data; The target geographic area is divided into multiple standard geographic grids using the Geohash algorithm, and the historical operational data is preprocessed into grids. Data cleaning and feature engineering are performed on the regional historical operational data for each of the standard geographic grids to generate multiple standardized feature vectors; Multiple standardized feature vectors are input into a hybrid prediction model, which outputs the predicted order demand and vehicle idle probability values ​​for each geographic grid within the target time period. The hybrid prediction model combines a gradient boosting decision tree model for extracting feature correlations and a long short-term memory network model for capturing feature temporal dependencies. In an optional embodiment of the first aspect of the present invention, the step of inputting the rental demand as a decision context into a preset reinforcement learning pricing model includes: Based on the rental demand, a state vector for the decision context is constructed. The state vector includes the predicted order demand of the target geographic grid, the real-time vehicle idle rate of the target geographic grid, the time period code of the current time period, the feature code of the weather conditions, and the average battery charge of vehicles in the target geographic grid. The state vector is input into a preset reinforcement learning pricing model.

[0007] In an optional embodiment of the first aspect of the present invention, the step of selecting a target pricing strategy from a plurality of preset pricing strategies through the reinforcement learning pricing model and executing it on shared vehicles within the target geographical area includes: In the reinforcement learning pricing model, a target price adjustment coefficient is selected from a discrete action space using a probability sampling algorithm. The discrete action space is defined as a preset price coefficient set, and each coefficient in the preset price coefficient set corresponds to a discount or premium strategy for the base price. The probability sampling algorithm uses the Thompson sampling algorithm or the upper bound confidence algorithm. The execution price is generated based on the selected target price adjustment factor and the base price; The execution price is used as a target price strategy and applied in real time to the shared vehicle rental service within the target geographic grid.

[0008] In an optional embodiment of the first aspect of the present invention, calculating the reward value for this pricing decision based on the leasing results generated after the execution of the target price strategy includes: Based on the rental results, the immediate revenue of the reward function, the vehicle distribution balance, and user satisfaction are obtained: The comprehensive reward value is obtained by weighting and summing the immediate revenue, the vehicle distribution balance, and the user satisfaction.

[0009] In an optional embodiment of the first aspect of the present invention, the vehicle distribution balance is obtained as follows: if the target price strategy successfully guides the user to ride the vehicle from the high-density area to the low-density area, a positive vehicle distribution balance value is assigned to the current pricing decision; otherwise, a negative vehicle distribution balance value is assigned to the current pricing decision.

[0010] In an optional embodiment of the first aspect of the present invention, updating the reinforcement learning pricing model and the demand forecasting model using the decision context, the target price strategy, and the reward value includes: Based on the decision context, the target price strategy, and the reward value, a reinforcement learning training sample triplet is constructed. The training sample triplet includes a decision context state vector, a target price adjustment coefficient, and a comprehensive reward value. The reinforcement learning model is updated using incremental learning, and the model parameters are adjusted using the gradient descent algorithm so that the reward estimate approximates the actual observed comprehensive reward value. The demand forecasting model is updated in conjunction with the data. When the average error between the predicted demand and the actual order volume exceeds a threshold over N consecutive periods, the demand forecasting model is fine-tuned using newly added operational data. Update the weight hyperparameters and dynamically adjust the weight coefficients in the reward function based on the changing trends of user satisfaction feedback indicators.

[0011] A second aspect of the present invention provides an intelligent pricing system for a shared vehicle rental platform, the intelligent pricing system for the shared vehicle rental platform comprising: The demand forecasting module is used to predict the rental demand of a target geographical area within a target time period based on historical operational data and a preset demand forecasting model. The demand input module is used to input the rental demand as a decision context into a preset reinforcement learning pricing model; The strategy generation and execution module is used to select a target price strategy from a plurality of preset price strategies through the reinforcement learning pricing model, and execute it on shared vehicles within the target geographical area. The reward calculation module is used to calculate the reward value of this pricing decision based on the leasing results generated after the execution of the target price strategy. The model update module is used to update the reinforcement learning pricing model and the demand forecasting model using the decision context, the target price strategy, and the reward value.

[0012] A third aspect of the present invention provides an intelligent pricing device for a shared vehicle rental platform, the intelligent pricing device for the shared vehicle rental platform comprising: a memory and at least one processor, wherein the memory stores instructions, and the memory and the at least one processor are interconnected via a line; The at least one processor invokes the instructions in the memory to cause the smart pricing device of the shared vehicle rental platform to execute the smart pricing method of the shared vehicle rental platform as described in any one of the first aspects of the present invention.

[0013] A fourth aspect of the present invention provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the intelligent pricing method of the shared vehicle rental platform as described in any one of the first aspects of the present invention.

[0014] Beneficial Effects: This invention provides an intelligent pricing method, system, and related equipment for a shared vehicle rental platform. The method includes: predicting rental demand in a target geographical area within a target time period based on historical operational data and a preset demand prediction model; inputting the rental demand as a decision context into a preset reinforcement learning pricing model; selecting a target pricing strategy from multiple preset pricing strategies through the reinforcement learning pricing model and executing it on shared vehicles within the target geographical area; calculating the reward value for this pricing decision based on the rental results generated after the execution of the target pricing strategy; and updating the reinforcement learning pricing model and the demand prediction model using the decision context, the target pricing strategy, and the reward value. This invention provides an intelligent pricing method based on historical data, taking into account geographical considerations, and forming a closed-loop feedback loop, resulting in flexible and more adaptable pricing. Attached Figure Description

[0015] Figure 1 This is a schematic diagram illustrating one embodiment of the main steps of an intelligent pricing method for a shared vehicle rental platform according to the present invention; Figure 2 This is a schematic diagram of an embodiment of the core interactive framework of the intelligent pricing method for a shared vehicle rental platform according to the present invention; Figure 3 This is a schematic diagram of an embodiment of the intelligent pricing system for a shared vehicle rental platform according to the present invention; Figure 4 This is a schematic diagram of an embodiment of an intelligent pricing device for a shared vehicle rental platform according to the present invention. Detailed Implementation

[0016] The terms "first," "second," "third," "fourth," etc. (if present) in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" or "having" and any variations thereof are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0017] For ease of understanding, the specific process of the embodiments of the present invention is described below. Please refer to [link / reference]. Figure 1 The first aspect of this invention provides an intelligent pricing method for a shared vehicle rental platform, comprising: S100. Based on historical operational data and using a preset demand forecasting model, predict the rental demand of the target geographical area within a target time period; In an optional embodiment of step S100 of the present invention, predicting the rental demand of the target geographical area within a target time period based on historical operational data and using a preset demand forecasting model includes: Historical operational data is acquired, including vehicle dynamic data, environmental parameter data, user demand data, and platform operation data. Specifically, this invention systematically collects at least four major categories of data: vehicle-specific data (e.g., GPS location, battery level, and vehicle status (including idle / in use / faulty)), external environmental data (e.g., weather conditions, date type, and work period division), market demand data (e.g., historical order spatiotemporal distribution, real-time request heatmaps, and search behavior logs), and platform operation data (e.g., regional vehicle idle rate statistics, dispatch cost records, and promotional activity information). This historical operational data provides comprehensive and real-time intelligence for subsequent predictions and decision-making.

[0018] The Geohash algorithm is used to divide the target geographic area into multiple standard geographic grids, and the historical operational data is preprocessed into grids. This invention uses the Geohash algorithm to implement regional division, which can better realize regional differentiated pricing.

[0019] For each of the standard geographic grids, the regional historical operation data is cleaned and feature engineering is performed to generate multiple standardized feature vectors. In this invention, before the data is input into the model, the regional historical operation data obtained after the regional division also needs to be cleaned (processing missing values ​​and outliers) and feature engineering is performed to generate standardized feature vectors for model training and prediction.

[0020] Multiple standardized feature vectors are input into a hybrid prediction model, which outputs predicted order demand and vehicle idle probability for each geographic grid within a target time period. The hybrid prediction model combines a gradient boosting decision tree model for extracting feature correlations with a long short-term memory network model for capturing feature temporal dependencies. In this invention, demand forecasting employs a hybrid prediction model combining an ensemble learning model (such as XGBoost / LightGBM) and a time series model (such as LSTM). Model input includes preprocessed historical feature data (e.g., order volume in the same time period, region, and similar weather conditions over the past 30 days). Model output includes estimated order demand and vehicle idle probability for different geographic grids (Geohash) within a specific future time period (e.g., the next hour). Training is conducted using supervised training with historical data, employing mean squared error (MSE) as a loss function, and optimized through gradient descent. This model can capture non-linear relationships. This invention's hybrid prediction model solves the problem of traditional methods being unable to predict the future, making pricing more than just guesswork.

[0021] S200. Input the rental demand as a decision context into a preset reinforcement learning pricing model; In an optional embodiment of step S200 of the present invention, inputting the rental demand as a decision context into the preset reinforcement learning pricing model includes: constructing a state vector of the decision context based on the rental demand, the state vector including the predicted order demand of the target geographic grid, the real-time vehicle idle rate of the target geographic grid, the time period code of the current time period, the feature code of the weather conditions, and the average battery charge of vehicles in the target geographic grid; inputting the state vector into the preset reinforcement learning pricing model. The pricing model is the key to the present invention, and its essence is a reinforcement learning framework based on Contextual Multi-Armed Bandit (Contextual MAB), which aims to find the optimal balance between exploration and utilization. The state (S_t) is defined as a feature vector, including the predicted demand D_pred for the current grid, the real-time vehicle idle rate S_ratio for the current grid, the current time period T, the weather condition W, and the average battery charge B_avg. This invention uses the prediction results of XGBoost+LSTM as the state input of Contextual MAB, realizing the leap from "pricing based on historical data" to "pricing based on future predictions", making the decision more forward-looking and accurate.

[0022] S300. Using the reinforcement learning pricing model, select a target pricing strategy from a set of preset pricing strategies and execute it on shared vehicles within the target geographical area; In an optional embodiment of step S300 of the present invention, selecting a target pricing strategy from a set of preset pricing strategies and executing it on shared vehicles within the target geographical area using the reinforcement learning pricing model includes: In the reinforcement learning pricing model, a target price adjustment coefficient is selected from a discrete action space using a probabilistic sampling algorithm. The discrete action space is defined as a preset set of price coefficients, where each coefficient corresponds to a discount or premium strategy on the base price. The probabilistic sampling algorithm employs either the Thompson sampling algorithm or an upper confidence bound algorithm. An action (A_t) is defined as a price adjustment coefficient α. The system presets a base price P_base, and the final price P_final = P_base × α, where α is a discrete set of actions, such as [0.7, 0.9, 1.0, 1.2, 1.5], representing 70% off, 90% off, original price, 1.2x premium, and 1.5x premium, respectively. Based on the current state S_t, the system selects a price adjustment coefficient α from the action space using either the Thompson sampling algorithm or an upper confidence bound algorithm.

[0023] The execution price is generated based on the selected target price adjustment factor and the base price; the execution price P_final = P_base × α is generated based on the selected factor.

[0024] The execution price is used as the target pricing strategy and applied in real time to the shared vehicle rental service within the target geographic grid. This step involves using P_final as the target pricing strategy and applying it in real time to the shared two-wheeled electric vehicle rental service within the target geographic grid.

[0025] S400. Based on the leasing results generated after the execution of the target price strategy, calculate the reward value for this pricing decision. In this invention, the reward (Reward, R_t) is the revenue feedback obtained by the system after executing action A_t (i.e., pricing α). The design of the reward function R_t is crucial, as it considers not only short-term revenue but also long-term operational goals. R_t= (α×P_base×I_ordered)+β×I_balanced+γ×I_user_satisfaction. I_ordered: A binary indicator function, equal to 1 if the vehicle is rented within Δt after pricing, and 0 otherwise. This part represents the immediate revenue.

[0026] I_balanced: Balancing factor. A positive reward is given if the pricing strategy successfully guides users to move vehicles from high-density areas to low-density areas. This part is used to optimize vehicle distribution and reduce scheduling costs.

[0027] I_user_satisfaction: User satisfaction factor. Modeled using indirect indicators such as user ratings, complaint rates, and long-term retention rates. It avoids deteriorating user experience due to excessively high prices. β and γ are weighted hyperparameters determined by operational strategies.

[0028] In an optional embodiment of step S400 of the present invention, calculating the reward value of this pricing decision based on the leasing results generated after the execution of the target price strategy includes: Based on the rental results, the immediate revenue, vehicle distribution balance, and user satisfaction of the reward function are obtained. The reward function used in this invention is designed with a triple reward mechanism: immediate revenue, vehicle distribution balance, and user satisfaction. The reward function R_t not only includes short-term revenue, but also creatively incorporates two long-term operational indicators: vehicle distribution balance and user satisfaction. This achieves a strategic upgrade from "single pursuit of revenue maximization" to "pursuing the comprehensive optimization of platform health (revenue, operating costs, and user retention)".

[0029] The comprehensive reward value is obtained by weighting and summing the immediate revenue, the vehicle distribution balance, and the user satisfaction. Comprehensive reward value R_t = (Immediate order revenue) + β × (Vehicle distribution balance) + γ × (User satisfaction).

[0030] In this invention, the immediate revenue is directly related to whether an order is generated after pricing. The vehicle distribution balance is obtained by assigning a positive vehicle distribution balance value to the pricing decision if the target price strategy successfully guides users to ride vehicles from high-density areas to low-density areas; otherwise, a negative vehicle distribution balance value is assigned. The satisfaction level incorporates long-term indicators such as user ratings and retention rates to avoid harming the user experience due to excessive pursuit of high prices.

[0031] S500: Using the decision context, the target price strategy, and the reward value, update the reinforcement learning pricing model and the demand forecasting model. This invention forms a complete, automated loop of data collection, prediction, decision-making, execution, and result feedback. The system possesses self-learning and evolution capabilities, automatically adapting to market changes, achieving intelligent and automated operations, and reducing the cost of manual intervention.

[0032] In an optional embodiment of step S500 of the present invention, updating the reinforcement learning pricing model and the demand forecasting model using the decision context, the target price strategy, and the reward value includes: Based on the decision context, the target price strategy, and the reward value, a reinforcement learning training sample triplet is constructed. The training sample triplet includes a decision context state vector, a target price adjustment coefficient, and a comprehensive reward value. Each pricing decision (i.e., the state-action-reward triplet) and its actual results (whether a transaction is completed, user rating, vehicle final location, etc.) are fed back.

[0033] The reinforcement learning model is updated using incremental learning, and the model parameters are adjusted using the gradient descent algorithm to make the reward estimate approximate the actual observed comprehensive reward value. The reinforcement learning model is then updated with (S_t, A_t, R_t) data to optimize the next decision. The algorithm gradually learns that in situations like "few vehicles at the subway entrance during evening rush hour on a rainy day," choosing α=1.5 yields a higher reward; while in situations like "many vehicles in the park on a weekday afternoon," choosing α=0.9 promotes faster traffic flow, thus obtaining a higher comprehensive reward.

[0034] The demand forecasting model is updated in a coordinated manner. When the average error between the predicted demand and the actual order volume exceeds a threshold over N consecutive periods, the demand forecasting model is fine-tuned using new operational data. Regularly fine-tuning the demand forecasting model with new data can adapt it to market changes.

[0035] The weight hyperparameters are updated, and the weight coefficients (i.e., coefficients β and γ) in the reward function are dynamically adjusted based on the changing trends of user satisfaction feedback indicators. In this invention, the system establishes a closed-loop feedback mechanism where the results of each pricing decision (whether an order is generated, user feedback, vehicle movement trajectory) are recorded and used to periodically fine-tune the demand forecasting model and update the reward estimation of the reinforcement learning model, enabling the system to adapt to market changes and shifts in user behavior patterns.

[0036] In summary, the core interactive framework of the intelligent pricing method for the shared vehicle rental platform of this invention can be as follows: Figure 2As shown, the core methods include: a real-time demand forecasting mechanism based on multi-source data fusion, a dynamic pricing algorithm based on machine learning and reinforcement learning, a user behavior feedback and model self-optimization system, and differentiated pricing strategies based on regional and time periods. The intelligent pricing method of the vehicle rental platform based on this invention can achieve the following technical effects: maximized revenue (through reinforcement learning dynamic pricing, the pricing strategy is directly optimized to maximize long-term cumulative revenue, which can increase platform revenue by 15%-30% compared to a fixed-price strategy), precise supply and demand matching (based on high-precision demand forecasting, demand is effectively regulated through price levers to smooth peak pressure, fill off-peak demand, and significantly improve vehicle turnover rate), automation and intelligence (the entire process requires no manual intervention; the system automatically makes optimal decisions based on data, significantly reducing operational labor costs and improving decision-making efficiency), multi-objective optimization (the reward function comprehensively considers revenue, vehicle distribution, and user experience, avoiding user churn caused by a "one-size-fits-all" high-price strategy, and achieving healthy and sustainable operation), and strong adaptability (the algorithm can automatically learn and adapt to external factors such as urban expansion, seasonal changes, and changes in the competitive landscape, maintaining the effectiveness of the strategy).

[0037] See Figure 3 The second aspect of the present invention provides an intelligent pricing system for a shared vehicle rental platform, the intelligent pricing system for the shared vehicle rental platform comprising: Demand forecasting module 10 is used to predict the rental demand of a target geographical area within a target time period based on historical operational data and a preset demand forecasting model. The demand input module 20 is used to input the rental demand as a decision context into a preset reinforcement learning pricing model; The strategy generation and execution module 30 is used to select a target price strategy from a plurality of preset price strategies through the reinforcement learning pricing model, and execute it on shared vehicles within the target geographical area. The reward calculation module 40 is used to calculate the reward value of this pricing decision based on the leasing results generated after the execution of the target price strategy; The model update module 50 is used to update the reinforcement learning pricing model and the demand forecasting model using the decision context, the target price strategy, and the reward value.

[0038] In an optional embodiment of the second aspect of the present invention, the demand forecasting module includes: The data acquisition unit is used to acquire historical operational data, which includes vehicle dynamic data, environmental parameter data, user demand data, and platform operational data. The grid processing unit is used to divide the target geographic area into multiple standard geographic grids using the Geohash algorithm, and to perform gridded preprocessing on the historical operational data. The data and processing unit is used to perform data cleaning and feature engineering on the regional historical operational data of each of the standard geographic grids to generate multiple standardized feature vectors. The demand forecasting unit is used to input multiple standardized feature vectors into a hybrid forecasting model and output the predicted order demand and vehicle idle probability values ​​for each geographic grid within the target time period. The hybrid forecasting model combines a gradient boosting decision tree model for extracting feature correlations and a long short-term memory network model for capturing feature temporal dependencies. In an optional embodiment of the second aspect of the present invention, the demand input module includes: The state vector construction unit is used to construct a state vector of the decision context based on the rental demand. The state vector includes the predicted order demand of the target geographic grid, the real-time vehicle idle rate of the target geographic grid, the time period code of the current time period, the feature code of the weather conditions, and the average battery charge of vehicles in the target geographic grid. A state vector input unit is used to input the state vector into a preset reinforcement learning pricing model.

[0039] In an optional embodiment of the second aspect of the present invention, the strategy generation and execution module includes: The price adjustment coefficient acquisition unit is used to select a target price adjustment coefficient from a discrete action space through a probability sampling algorithm in the reinforcement learning pricing model. The discrete action space is defined as a preset price coefficient set, and each coefficient in the preset price coefficient set corresponds to a discount or premium strategy of the base price. The probability sampling algorithm adopts the Thompson sampling algorithm or the upper bound confidence algorithm. An execution price calculation unit is used to generate an execution price based on the selected target price adjustment factor and the base price; The pricing strategy application unit is used to apply the execution price as a target pricing strategy to the shared vehicle rental service within the target geographic grid in real time.

[0040] In an optional embodiment of the second aspect of the present invention, the reward calculation module includes: The reward component calculation unit is used to obtain the immediate revenue of the reward function, the vehicle distribution balance, and user satisfaction based on the rental results. The weighted summation calculation unit is used to obtain a comprehensive reward value by weighted summation of the instantaneous revenue, the vehicle distribution balance, and the user satisfaction.

[0041] In an optional embodiment of the second aspect of the present invention, the vehicle distribution balance is obtained as follows: if the target price strategy successfully guides the user to ride the vehicle from the high-density area to the low-density area, a positive vehicle distribution balance value is assigned to this pricing decision; otherwise, a negative vehicle distribution balance value is assigned to this pricing decision.

[0042] In an optional embodiment of the second aspect of the present invention, the model update module includes: The triplet construction unit is used to construct reinforcement learning training sample triplets based on the decision context, the target price strategy, and the reward value. The training sample triplet includes a decision context state vector, a target price adjustment coefficient, and a comprehensive reward value. The reinforcement learning model adjustment unit is used to update the reinforcement learning model using an incremental learning approach and adjust the model parameters through a gradient descent algorithm so that the reward estimate approximates the actual observed comprehensive reward value. The demand forecasting model adjustment unit is used to update the demand forecasting model in a coordinated manner. When the average error between the predicted demand and the actual order volume exceeds the threshold within N consecutive periods, the demand forecasting model is fine-tuned using newly added operational data. The reward function adjustment unit is used to update the weight hyperparameters and dynamically adjust the weight coefficients in the reward function based on the changing trend of user satisfaction feedback indicators.

[0043] Figure 4 This is a schematic diagram of the structure of an intelligent pricing device for a shared vehicle rental platform according to an embodiment of the present invention. The intelligent pricing device for this shared vehicle rental platform can vary significantly due to differences in configuration or performance. It may include one or more processors 50 (central processing units, CPUs) (e.g., one or more processors) and memory 60, and one or more storage media 70 (e.g., one or more mass storage devices) for storing applications or data. The memory and storage media can be temporary or persistent storage. The program stored in the storage media may include one or more modules (not shown in the diagram), each module including a series of instruction operations on the intelligent pricing device of the shared vehicle rental platform. Furthermore, the processor may be configured to communicate with the storage media and execute the series of instruction operations stored in the storage media on the intelligent pricing device of the shared vehicle rental platform.

[0044] The intelligent pricing device of the shared vehicle rental platform of this invention may also include one or more power supplies 80, one or more wired or wireless network interfaces 90, one or more input / output interfaces 100, and / or one or more operating systems, such as Windows Server, Mac OS X, Unix, Linux, FreeBSD, etc. Those skilled in the art will understand that... Figure 4 The smart pricing device structure of the shared vehicle rental platform shown does not constitute a limitation on the smart pricing device of the shared vehicle rental platform. It may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0045] The present invention also provides a computer-readable storage medium, which can be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium, wherein the computer-readable storage medium stores instructions that, when the instructions are executed on a computer, cause the computer to perform the steps of the intelligent pricing method of the shared vehicle rental platform.

[0046] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working process of the system or system / unit described above can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0047] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part 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, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0048] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A smart pricing method for a shared vehicle rental platform, characterized in that, include: Based on historical operational data and through a pre-set demand forecasting model, predict the rental demand of the target geographical area within the target time period; The rental demand is used as the decision context input into a preset reinforcement learning pricing model; The reinforcement learning pricing model selects a target pricing strategy from a set of preset pricing strategies and applies it to shared vehicles within the target geographic area. Calculate the reward value for this pricing decision based on the leasing results generated after the execution of the target price strategy; The reinforcement learning pricing model and the demand forecasting model are updated using the decision context, the target price strategy, and the reward value.

2. The intelligent pricing method for a shared vehicle rental platform according to claim 1, characterized in that, The prediction of rental demand in a target geographical area within a target time period based on historical operational data and a pre-set demand forecasting model includes: Acquire historical operational data, including vehicle dynamic data, environmental parameter data, user demand data, and platform operational data; The target geographic area is divided into multiple standard geographic grids using the Geohash algorithm, and the historical operational data is preprocessed into grids. Data cleaning and feature engineering are performed on the regional historical operational data for each of the standard geographic grids to generate multiple standardized feature vectors; Multiple standardized feature vectors are input into a hybrid prediction model, which outputs the predicted order demand and vehicle idle probability values ​​for each geographic grid within the target time period. The hybrid prediction model combines a gradient boosting decision tree model for extracting feature correlations and a long short-term memory network model for capturing feature temporal dependencies.

3. The intelligent pricing method for a shared vehicle rental platform according to claim 1, characterized in that, The step of inputting the rental demand as a decision context into a preset reinforcement learning pricing model includes: Based on the rental demand, a state vector for the decision context is constructed. The state vector includes the predicted order demand of the target geographic grid, the real-time vehicle idle rate of the target geographic grid, the time period code of the current time period, the feature code of the weather conditions, and the average battery charge of vehicles in the target geographic grid. The state vector is input into a preset reinforcement learning pricing model.

4. The intelligent pricing method for the shared vehicle rental platform according to claim 3, characterized in that, The step of selecting a target pricing strategy from a set of preset pricing strategies using the reinforcement learning pricing model and applying it to shared vehicles within the target geographical area includes: In the reinforcement learning pricing model, a target price adjustment coefficient is selected from a discrete action space using a probability sampling algorithm. The discrete action space is defined as a preset price coefficient set, and each coefficient in the preset price coefficient set corresponds to a discount or premium strategy for the base price. The probability sampling algorithm uses the Thompson sampling algorithm or the upper bound confidence algorithm. The execution price is generated based on the selected target price adjustment factor and the base price; The execution price is used as a target price strategy and applied in real time to the shared vehicle rental service within the target geographic grid.

5. The intelligent pricing method for a shared vehicle rental platform according to claim 1, characterized in that, The calculation of the reward value for this pricing decision based on the leasing results generated after the execution of the target price strategy includes: Based on the rental results, the immediate revenue of the reward function, the vehicle distribution balance, and user satisfaction are obtained: The comprehensive reward value is obtained by weighting and summing the immediate revenue, the vehicle distribution balance, and the user satisfaction.

6. The intelligent pricing method for a shared vehicle rental platform according to claim 5, characterized in that, The vehicle distribution balance is obtained as follows: if the target price strategy successfully guides the user to ride the vehicle from the high-density area to the low-density area, a positive vehicle distribution balance value is assigned to this pricing decision; otherwise, a negative vehicle distribution balance value is assigned to this pricing decision.

7. The intelligent pricing method for a shared vehicle rental platform according to claim 1, characterized in that, The step of updating the reinforcement learning pricing model and the demand forecasting model using the decision context, the target price strategy, and the reward value includes: Based on the decision context, the target price strategy, and the reward value, a reinforcement learning training sample triplet is constructed. The training sample triplet includes a decision context state vector, a target price adjustment coefficient, and a comprehensive reward value. The reinforcement learning model is updated using incremental learning, and the model parameters are adjusted using the gradient descent algorithm so that the reward estimate approximates the actual observed comprehensive reward value. The demand forecasting model is updated in conjunction with the data. When the average error between the predicted demand and the actual order volume exceeds a threshold over N consecutive periods, the demand forecasting model is fine-tuned using newly added operational data. Update the weight hyperparameters and dynamically adjust the weight coefficients in the reward function based on the changing trends of user satisfaction feedback indicators.

8. An intelligent pricing system for a shared vehicle rental platform, characterized in that, The intelligent pricing system of the shared vehicle rental platform includes: The demand forecasting module is used to predict the rental demand of a target geographical area within a target time period based on historical operational data and a preset demand forecasting model. The demand input module is used to input the rental demand as a decision context into a preset reinforcement learning pricing model; The strategy generation and execution module is used to select a target price strategy from a plurality of preset price strategies through the reinforcement learning pricing model, and execute it on shared vehicles within the target geographical area. The reward calculation module is used to calculate the reward value of this pricing decision based on the leasing results generated after the execution of the target price strategy. The model update module is used to update the reinforcement learning pricing model and the demand forecasting model using the decision context, the target price strategy, and the reward value.

9. A smart pricing device for a shared vehicle rental platform, characterized in that, The smart pricing device of the shared vehicle rental platform includes: a memory and at least one processor, wherein the memory stores instructions, and the memory and the at least one processor are interconnected via a line; The at least one processor invokes the instructions in the memory to cause the smart pricing device of the shared vehicle rental platform to execute the smart pricing method of the shared vehicle rental platform as described in any one of claims 1-7.

10. A computer-readable storage medium storing a computer program thereon, characterized in that, When the computer program is executed by the processor, it implements the intelligent pricing method of the shared vehicle rental platform as described in any one of claims 1-7.

Images