Intelligent lease management method and system

Abstract

The invention provides an intelligent lease management method and system, and the method comprises the steps: building a non-parametric model based on the historical lease behavior data of a user and the state data of equipment, so as to capture the nonlinear behavior characteristics of the user and the equipment; lease management is modeled as a dynamic game between a lease platform and a user, and a risk-controllable rent pricing and tenancy term strategy is generated through a game theory framework; and constructing a risk aversion utility function of the rental platform to dynamically balance platform income and risk. The non-parametric model is constructed to accurately capture the non-linear behavior characteristics between the user and the equipment, the resource allocation of the lease market is optimized through the dynamic game between the lease platform and the user to generate the lease strategy, and meanwhile, the risk aversion utility function is constructed, so that the lease platform is ensured to maximize the revenue, and the lease efficiency is improved. And potential risks can be effectively controlled.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of lease management, and particularly relates to an intelligent lease management method and system. BACKGROUND

[0002] Lease management is widely used in many fields. For example, in the real estate field, it covers the leasing of houses, apartments, office buildings, etc.; in the transportation field, it includes the leasing of cars and ships; and in the equipment leasing field, it involves the leasing of various equipment such as machinery, construction, medical, and office equipment. With the continuous development of leasing business in various fields and the intensification of market competition, traditional lease management methods have gradually exposed many problems and are difficult to meet the needs of modern leasing business.

[0003] Traditional lease management often relies on experience in pricing and lease term strategy formulation, lacking precise analysis of market dynamics and user needs, which leads to a series of problems, especially in the equipment leasing field. For example, in equipment leasing, it is difficult to dynamically adjust prices according to the actual use of equipment and market demand. This results in the inability to raise prices to achieve higher profits when the market demand for certain equipment is strong, and the inability to lower prices to attract customers and improve equipment utilization when demand is low. At the same time, the lease term is fixed and cannot be adjusted flexibly according to the needs and use characteristics of different users, resulting in blind resource allocation and unstable profits.

[0004] Moreover, traditional lease management has great difficulty in balancing revenue and risk. When formulating leasing strategies, risk factors are not fully considered, resulting in a higher risk of default and business risk while pursuing high revenue. For example, to attract more customers, the rent price may be reduced or the lease term limit may be relaxed, but this may increase the probability of user default and cause losses to the enterprise. At the same time, the traditional method lacks a mechanism for dynamically adjusting risk strategies, and cannot adjust risk preferences in a timely manner according to market fluctuations, leaving the enterprise in a passive position when facing market changes. SUMMARY

[0005] Therefore, the purpose of the present application is to provide an intelligent lease management method and system to solve the above-mentioned problems.

[0006] According to the intelligent lease management method proposed by the present application, the method comprises:

[0007] Based on user historical leasing behavior data and equipment state data, a non-parametric model is constructed to capture the nonlinear behavior characteristics of users and equipment;

[0008] The lease management is modeled as a dynamic game between the leasing platform and the user, and the risk-controllable rent pricing and lease term strategy is generated through the game theory framework;

[0009] Construct a risk-averse utility function for the rental platform to dynamically balance platform revenue and risk.

[0010] Furthermore, the construction of a non-parametric model based on user historical rental behavior data and device status data to capture the non-linear behavioral characteristics of users and devices includes:

[0011] A kernel function is used to map historical user rental behavior data and device status data to a high-dimensional feature space using a Gaussian kernel. The user rental behavior data includes rental device frequency, payment delay, and device usage time, while the device status data includes failure rate and wear rate.

[0012] Based on the feature vector mapped by the kernel function, kernel ridge regression is used to predict users' future rental behavior, including the probability of renewal and the probability of default.

[0013] By modeling the dynamic changes of equipment utilization and market supply and demand time series data through Gaussian processes, probabilistic predictions are generated, namely, the predicted value of equipment leasing demand in the future.

[0014] Furthermore, the prediction of future user rental behavior using kernel ridge regression based on the feature vector mapped by the kernel function includes:

[0015] The objective function of kernel ridge regression is defined by combining user behavior feature vectors and rental behavior labels. The objective function is:

[0016]

[0017] Where, x u，i Let y be the behavioral feature vector of the i-th user, including rental frequency and payment delay. u，i Let w be the rental behavior label for the i-th user, including the renewal probability label and the default probability label, and w be the weight vector used for linear combination of features. Let λ be the eigenvector after kernel function mapping. reg For regularization parameters;

[0018] Solving the dual problem using the kernel trick yields the optimal dual weight α. * And predict users' future rental behavior tags:

[0019]

[0020] in, To predict the future rental behavior of the i-th user, K(x) is used as a label, including a renewal probability label and a default probability label. u，i x u，j ) is a kernel function used to calculate the similarity between the i-th user and the j-th user.

[0021] Further, the lease management is modeled as a dynamic game between the leasing platform and the users, and risk-controllable rent pricing and lease term strategies are generated through a game theory framework, including:

[0022] The lease management is modeled as a Stackelberg game, with the leasing platform as the leader and the users as the followers;

[0023] The platform strategy space is defined as a combination of rent pricing and lease term, and the user strategy space is defined as the amount of leasing;

[0024] The revenue functions are defined, including the platform revenue function and the user revenue function;

[0025] The Stackelberg equilibrium is solved by backward induction to generate the optimal rent pricing and lease term strategies of the leasing platform.

[0026] Further, the revenue functions are defined, including:

[0027] The platform revenue function is defined in combination with the actual amount of leasing q i and the predicted default probability P of the users, as follows:

[0028]

[0029] where U1 is the total revenue of the leasing platform, p i is the rent of the i-th user, q i is the actual amount of leasing of the i-th user, which is obtained by solving the game equilibrium by backward induction, c v is the unit default cost, P(d i >T) is the predicted default probability of the i-th user, which is obtained by kernel ridge regression prediction, i.e., the probability of not returning the device within the lease term T, d i is the actual lease duration of the i-th user, c m is the unit device maintenance cost, and n is the total number of users;

[0030] The user revenue function is defined in combination with the user leasing utility and the deposit occupancy cost, as follows:

[0031]

[0032] where U2 is the total revenue of the users, u i is the leasing utility of the i-th user, δ is the unit time deposit occupancy cost rate, and δ·d i is the deposit occupancy cost of the i-th user.

[0033] Further, the Stackelberg equilibrium is solved by reverse induction to generate the optimal rent pricing and lease term strategy of the leasing platform, comprising:

[0034] For the user, the optimal lease quantity q is selected according to the fixed platform strategy G * : Wherein, q * is the optimal lease quantity vector of the user, that is, the lease quantity selected by the user according to the utility maximization principle, U2(G, q) is the user's income function under the given platform strategy G, and the platform strategy G includes the rent p and the lease term T;

[0035] For the leasing platform, the optimal platform leasing strategy G * is selected according to the optimal lease quantity response q * : Wherein, G * is the optimal rent pricing and lease term strategy vector of the leasing platform, and U1(G, q * (G)) is the platform income function under the given optimal lease quantity response q * (G) of the user;

[0036] The optimal lease quantity of the user and the optimal platform leasing strategy are iteratively solved until the Stackelberg equilibrium (G * , q * ) is converged, which is used for dynamic rent p pricing and lease term T adjustment of the leasing platform.

[0037] Further, the risk-averse utility function of the leasing platform is constructed to dynamically balance the platform income and risk, comprising:

[0038] A risk aversion coefficient is introduced to construct the risk-averse utility function of the leasing platform:

[0039] Q=E[R]-λ risk ·Var(R),

[0040] Wherein, Q is the risk-averse utility of the leasing platform, E[R] is the expected value of the platform income, Var(R) is the variance of the platform income, λ risk is the risk aversion coefficient, and λ risk ≥0;

[0041] The risk aversion coefficient is dynamically adjusted according to market fluctuations.

[0042] Further, the construction of the risk-averse utility function of the platform further comprises:

[0043] The expected value E[R] and the variance Var(R) of the platform income are calculated, and the formula is:

[0044]

[0045] wherein E[q i ] is the expected value of the i-th user's rental quantity, and Var(q i ) is the variance of the i-th user's rental quantity.

[0046] Further, the risk-averse coefficient is dynamically adjusted according to market fluctuations, including:

[0047] If the default rate rises or the predicted equipment rental demand in the future period is lower than the current rental demand, and the difference exceeds the preset demand threshold, the risk-averse coefficient is increased to prefer low-risk strategies, including shortening the rental period and increasing the deposit, wherein the default rate is the average of the default probabilities of all users, and the current rental demand is the actual observed rental demand in the current market;

[0048] If the current rental demand is close to or exceeds the predicted equipment rental demand in the future period, the risk-averse coefficient is reduced to pursue higher returns.

[0049] The present application also provides an intelligent rental management system for implementing the above intelligent rental management method, the system comprising:

[0050] A user rental behavior module for constructing a non-parametric model based on user historical rental behavior data and equipment state data to capture the nonlinear behavior characteristics of users and equipment;

[0051] A rental strategy generation module for modeling the rental management as a dynamic game between the rental platform and the users, and generating risk-controllable rental pricing and rental period strategies through a game theory framework;

[0052] A revenue and risk balance module for constructing a risk-averse utility function of the rental platform to dynamically balance platform revenue and risk.

[0053] In summary, the intelligent leasing management method of the present application firstly constructs a non-parametric model based on user historical leasing behavior data and equipment state data. Since it does not depend on specific function form assumption, it can flexibly fit complex data patterns and accurately capture the nonlinear behavior characteristics between users and equipment, thereby improving the prediction accuracy of user future leasing behavior (such as renewal, default probability), enabling the leasing platform to allocate resources to generate reasonable leasing strategies and prevent risks, and also enabling personalized leasing services for different users to improve user satisfaction and loyalty. Then, the leasing management is modeled as a dynamic game between the leasing platform and the user, considering the interaction of both parties' strategies, and a reasonable leasing strategy is developed that takes into account the platform's revenue and the user's interests. In the game process, various uncertain factors and risks can be evaluated and controlled, and the leasing platform can dynamically adjust the leasing strategy according to the market environment and user leasing behavior to reduce user default and market risk. At the same time, a risk-averse utility function is introduced to comprehensively consider the platform's expected revenue and variance to quantify the platform's revenue and risk, enabling the leasing platform to more intuitively evaluate the revenue and risk characteristics of different leasing strategies and adjust the leasing strategy in real time according to the utility function changes, thereby improving the timeliness and accuracy of leasing decision-making and bringing significant economic benefits and operational stability to the leasing platform.

[0054] Additional aspects and advantages of the application will be described in the following description, will become apparent from the following description, or will be learned by practicing the application. BRIEF DESCRIPTION OF DRAWINGS

[0055] The above and / or additional aspects and advantages of the present application will become apparent and be readily understood from the following description, taken in conjunction with the accompanying drawings, in which:

[0056] Figure 1 A flowchart of the intelligent leasing management method of the present application embodiment one;

[0057] Figure 2 A system block diagram of the intelligent leasing management system of the present application embodiment two. DETAILED DESCRIPTION

[0058] In order to facilitate the understanding of the present application, the present application will be described in more detail below with reference to the relevant drawings. The drawings show several embodiments of the present application. However, the present application can be realized in many different forms and is not limited to the embodiments described herein. On the contrary, the purpose of providing these embodiments is to make the disclosure of the present application more thorough and comprehensive.

[0059] It is to be understood that where an element such as a layer, region or substrate is described as being "on" another element, it can be directly on the other element or intervening elements can also be present. Where an element such as a layer, region or substrate is described as being "connected" to another element, it can be directly connected to the other element or intervening elements can also be present. As used herein, the term "and / or" includes any and all combinations of one or more of the associated listed items.

[0060] 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 in the description herein is for describing particular embodiments only and is not intended to be limiting of the application. As used herein, the term "and / or" includes any and all combinations of one or more of the associated listed items.

[0061] Embodiment One

[0062] Referring to Figure 1 The present application provides an intelligent leasing management method, which comprises steps S101-S103:

[0063] S101, based on user historical leasing behavior data and equipment state data, a non-parametric model is constructed to capture the nonlinear behavior characteristics of users and equipment.

[0064] Further optionally, the non-parametric model is constructed based on the user historical leasing behavior data and equipment state data to capture the nonlinear behavior characteristics of users and equipment, comprising:

[0065] The historical user leasing behavior data and equipment state data are mapped by kernel function to map the data to high-dimensional feature space by Gaussian kernel, wherein the user leasing behavior data includes leasing equipment frequency, payment delay and equipment use time, and the equipment state data includes failure rate and loss rate;

[0066] Based on the feature vector mapped by the kernel function, the future leasing behavior of the user is predicted by kernel ridge regression, including the probability of renewal and the probability of default;

[0067] The time series data of equipment utilization and market supply and demand are modeled by Gaussian process to generate probabilistic prediction, i.e. the predicted value of the leasing demand of the equipment in the future period of time.

[0068] It can be understood that the kernel function mapping of the historical user rental behavior data (such as rental equipment frequency, payment delay, and equipment use time length, etc.) and the equipment state data (such as failure rate and loss rate, etc.) can convert the nonlinear relationship in the original data into a linear relationship in a high-dimensional space using a Gaussian kernel, which can mine the hidden nonlinear relationship in the user rental behavior and equipment state data, for example, there may be a nonlinear correlation between the rental equipment frequency and the default probability.

[0069] Based on the kernel mapped feature vector, the future rental behavior of the user is predicted, such as the renewal probability, the default probability, etc. by kernel ridge regression. Kernel ridge regression is a method that combines kernel trick and ridge regression. It maps data to a high-dimensional space through a kernel function, and then performs linear regression in the high-dimensional space. Kernel ridge regression can handle data in a high-dimensional feature space, effectively capturing nonlinear features in the data, thereby improving the prediction accuracy of the user's future behavior. For example, by accurately predicting the user's renewal probability, the rental platform can help prepare equipment allocation and marketing strategies in advance, and by accurately predicting the default probability, the rental platform can take appropriate risk prevention measures.

[0070] For time series data such as equipment utilization and market supply and demand, Gaussian process is used to model its dynamic changes to generate probabilistic predictions, i.e. the predicted value of the rental demand of the equipment in the future period (such as the 95% confidence interval of the rental demand of the equipment in the next 30 days). Gaussian process is a non-parametric model based on Bayesian theory, which can model time series data and give uncertainty estimates of prediction results. Unlike traditional point prediction, Gaussian process can give the probability distribution of future equipment rental demand, such as 95% confidence interval, which enables the rental platform to better understand the uncertainty of demand and thus develop more flexible and robust rental strategies.

[0071] And because equipment utilization and market supply and demand conditions change over time, Gaussian process can dynamically capture these changes and adjust the prediction results in a timely manner to provide more accurate rental decision-making basis for the rental platform.

[0072] Further optionally, the future rental behavior of the user is predicted based on the kernel mapped feature vector by kernel ridge regression, including:

[0073] The target function of the kernel ridge regression is defined in combination with the user behavior feature vector and the rental behavior label, and the target function is:

[0074]

[0075] Where x u，i is the behavior feature vector of the i-th user, y u，iis the rental behavior label of the i-th user, including the renewal probability label and the default probability label, w is the weight vector used to linearly combine the features, is the feature vector mapped by the kernel function, and λ reg is the regularization parameter.

[0076] The dual problem is solved by the kernel trick to obtain the optimal dual weight α * , and the future rental behavior label of the i-th user is predicted as:

[0077]

[0078] wherein, is the predicted future rental behavior label of the i-th user, including the renewal probability label and the default probability label, K(x u，i , x u，j ) is the kernel function used to calculate the similarity between the i-th user and the j-th user.

[0079] It is understood that the objective function of the kernel ridge regression is defined in combination with the user behavior feature vector and the rental behavior label (such as the renewal probability and the default probability), and the objective function is wherein, x u，i is the behavior feature vector of the i-th user, including the rental history record (such as the number of rentals, the distribution of rental periods, the renewal record, etc.), the device usage time (such as the single rental duration, the cumulative usage time, etc.), the payment behavior (such as the payment delay frequency, the payment method, the historical overdue situation, etc.), the rental device type (such as the device type, the degree of newness, etc.), the device maintenance and damage record (such as the number of damage times of the device, the maintenance request frequency, etc.), etc., y u，i is the rental behavior label of the i-th user, such as the renewal probability label and the default probability label, y u，i ∈ [0, 1], n is the number of users, w is the weight vector used to linearly combine the features, is the feature vector mapped by the kernel function, and λ reg is the regularization parameter.

[0080] The in the objective function is the mean square error loss function, which functions to minimize the square error between the predicted value and the true rental behavior label y u，i , so that the model can fit the training data as accurately as possible and improve the prediction accuracy. λ reg ||w|| 2 The introduction of ||w|| can prevent the model from overfitting during the training process, so that it can have a good prediction effect when facing new user data.

[0081] To solve the weight vector w, the original problem can be transformed into a dual problem. First, rewrite the objective function as: where Y is the label vector, Y = [y u，1 , y u，2 , …, y u，n ] T , is the design matrix,

[0082] By solving the optimal solution of w, we can get: w = (Φ T Φ+λ reg I) -1 Φ T Y.

[0083] Using the kernel trick, replace Φ T Φwith the kernel matrix Therefore, the weight vector w can be represented as: w = Φ T (ΦΦ T +λ reg I) -1 Y.

[0084] Further, introduce the dual variable α, such that: α = (ΦΦ T +λ reg I) -1 Y, so that the weight vector w can be represented as: w = Φ T α.

[0085] Substitute w = Φ T α into the objective function to get the objective function of the dual problem: Simplify as: where K = ΦΦ T .

[0086] By taking the derivative of α and setting the derivative to zero, we can get the optimal solution, i.e. the optimal dual weight: α * = (K+λ reg I) -1 Y.

[0087] Once the optimal dual weight α * is obtained, it can be used to predict the rental behavior label of each user. The formula is: where is the predicted future rental behavior label of the i-th user, such as the probability of renewal label, the probability of default label, etc., and K(x u，i , x u，j ) is the kernel function used to measure the test point (i.e. the behavior feature vector x u，i of the i-th user).similarity between the training points (i.e., the behavior feature vector x u，j ) of the jth user, the higher the similarity, the larger the value of the kernel function, indicating that the distance between the two points in the feature space is closer. Using kernel function calculation can convert the nonlinear problem in the original feature space into a linear problem in the high-dimensional feature space. The inner product can be calculated directly in the original feature space without explicitly calculating the high-dimensional feature vector Avoiding complex calculations in high-dimensional space.

[0088] The optimal dual weight α * obtained by solving the dual problem is used to predict the future rental behavior label of each user This prediction method can consider the information of all training samples, and make prediction according to the similarity between the user behavior feature vector and the training sample, so that the prediction result is more accurate and reliable.

[0089] Further optionally, the time series data of equipment utilization and market supply and demand are used to model the dynamic changes by Gaussian process to generate probabilistic prediction, i.e., the predicted value of equipment demand in the future period of time, including:

[0090] The time series data of equipment utilization and market supply and demand are denoted as observation data Where x d，i is the feature vector of the ith observation point, and y d，i is the equipment demand of the ith observation point.

[0091] A subset is selected from the observation data D as training data D train , which is used to train the Gaussian process model, denoted as

[0092] The prior distribution of Gaussian process is defined, and the covariance function is Gaussian kernel:

[0093]

[0094] Where K(x d , x' d ) is the covariance function, which is used to measure the similarity between the feature vectors x d and x' d of the two observation points in the time series data, and σ gp is the kernel bandwidth parameter of Gaussian process, which is used to control the decay rate of similarity;

[0095] Based on the training data D train , the mean and variance of the posterior distribution are calculated to obtain the predicted value and uncertainty of the equipment rental demand, and the formula is:

[0096]

[0097] where μ(x d，* ) is the posterior mean of the prediction point x d，* , i.e., the predicted value of the equipment rental demand, σ 2 (x d，* ) is the posterior variance of the prediction point x d，* , i.e., the uncertainty of the equipment rental demand, k * is the covariance vector of the test point x d，* and the training points x d，i ∈ D train , K is the covariance matrix between the training points, is the noise variance, and I is the identity matrix.

[0098] It is understood that the time series data of equipment utilization and market supply and demand are recorded as observation data where x d，i is the feature vector of the i-th observation point (such as time stamp, equipment type, market area, etc.), y d，i is the equipment demand of the i-th observation point, and a subset is selected from the observation data D as training data, denoted as D train .

[0099] The prior distribution of the Gaussian process is defined, where the covariance function is the Gaussian kernel: where K(x d , x’ d ) is the covariance function, which is used to measure the similarity of the feature vectors x d and x’ d of two observation points in the time series data. The higher the similarity, the larger the covariance value, indicating that the two observation points are closer in the feature space. This similarity measurement helps the model capture local features and trends in the data. σ gp is the kernel bandwidth parameter of the Gaussian process, which can control the speed of similarity decay with distance. By adjusting the value of σ gp , the model can be flexibly adapted to the characteristics of different data sets, making it better fit the data. For example, for a data set with relatively flat changes, the value of σ gp can be appropriately increased; for a data set with relatively sharp changes, the value of σ gp can be appropriately reduced.

[0100] Based on the training data D train , the mean and variance of the posterior distribution are calculated to obtain the predicted value and uncertainty of the equipment rental demand, and the formula is: where μ(x d，* ) is the posterior mean of the prediction point x d，*the posterior mean, i.e., the predicted value of the equipment rental demand, σ 2 (x d，* ) is the posterior variance of the prediction point x d，* , i.e., the uncertainty of the equipment rental demand, k * is the covariance vector of the test point x d，* and the training points x d，i ∈ D train , K is the covariance matrix between the training points, is the noise variance, representing the noise level of the observed data, and I is the identity matrix.

[0101] By calculating the mean and variance of the posterior distribution, not only the predicted value of the equipment rental demand μ(x d，* ) can be obtained, but also the uncertainty of the predicted equipment rental demand σ 2 (x d，* ) can be obtained. This probabilistic prediction method provides more comprehensive information for the rental platform, enabling the rental platform to better cope with the uncertainty of demand.

[0102] The introduction of the noise variance takes into account the noise in the observed data, enabling the model to more accurately estimate the predicted value and uncertainty. In practical applications, there is often some noise in the observed data. By reasonably setting the value of , the robustness and prediction accuracy of the model can be improved.

[0103] S102, model the rental management as a dynamic game between the rental platform and the users, and generate risk-controllable rent pricing and rental period strategies through the game theory framework.

[0104] Further optionally, the rental management is modeled as a game problem between the rental platform and the rental users through the game theory framework to dynamically generate the rental strategy, including:

[0105] modeling the rental management as a Stackelberg game, with the rental platform as the leader and the users as the followers;

[0106] defining the platform strategy space as a combination of rent pricing and rental period, and defining the user strategy space as rental quantity;

[0107] defining the revenue function, including the platform revenue function and the user revenue function;

[0108] solving the Stackelberg equilibrium through backward induction to generate the optimal rent pricing and rental period strategy of the rental platform.

[0109] It can be understood that in the leasing market, the leasing platform usually has the initiative to make rules and strategies, and the user makes a response according to the conditions provided by the platform. The embodiment models the leasing management as a Stackelberg game, with the leasing platform as the leader and the user as the follower. Thus, the Stackelberg game model provides a clear framework for analyzing the strategic interaction between the leasing platform and the user. The leasing platform acts first, and the user makes the optimal response according to the strategy of the leasing platform. This sequential decision-making process facilitates the step-by-step analysis and optimization of leasing strategies.

[0110] At the same time, the platform strategy space is defined as the combination of rent pricing and lease period, and the user strategy space is defined as the leasing quantity. Rent pricing and lease period are two key factors in leasing management, directly affecting the platform's revenue and the user's leasing decision. Defining the platform strategy space as the combination of rent pricing and lease period can comprehensively consider the core decision variables of the platform in the leasing business. The user's leasing decision is usually based on factors such as rent pricing and lease period. Defining the user strategy space as the leasing quantity can clearly reflect the user's behavior choice under different platform strategies, providing a basis for subsequent revenue function definition and equilibrium solving.

[0111] The Stackelberg equilibrium is solved by reverse induction to generate the optimal rent pricing and lease period strategy of the leasing platform. That is, starting from the optimal response of the follower user, the optimal leasing strategy of the leader leasing platform is gradually derived. Through this method, the optimal rent pricing and lease period strategy of the leasing platform under given market conditions can be obtained.

[0112] And because the leasing market is dynamic, solving the Stackelberg equilibrium by reverse induction can adjust the platform's leasing strategy in real time according to market conditions. When market demand, competition, and other factors change, the platform can re-analyze the game and solve the equilibrium to adapt to market changes and maintain a competitive advantage.

[0113] Further optionally, the definition of the revenue function includes:

[0114] Combining the actual leasing quantity q of the user i and the predicted default probability P of the user, the platform revenue function is defined as:

[0115]

[0116] where U1 is the total revenue of the leasing platform, p i is the rent of the i-th user, q i is the actual leasing quantity of the i-th user, obtained by solving the game equilibrium by reverse induction, c v is the unit default cost, P(d iT) is the predicted default probability of the i-th user, which is predicted by kernel ridge regression, i.e., the probability of not returning the device within the rental period T, d i is the actual rental duration of the i-th user, c m is the unit device maintenance cost, and n is the total number of users;

[0117] The user benefit function is defined by combining the user rental utility and the deposit occupation cost:

[0118]

[0119] where U2 is the total benefit of the user, u i is the rental utility of the i-th user, δ is the unit time deposit occupation cost rate, δ·d i is the deposit occupation cost of the i-th user.

[0120] It is understandable that the benefit function is a core concept in game theory, and by defining the platform benefit and the user benefit, the decision-making goal in rental management can be quantified. By quantifying the benefit function, it is helpful to evaluate and compare different rental strategies. Considering the platform benefit and the user benefit at the same time reflects the fairness and sustainability of intelligent rental management. In the game process, both the rental platform and the user will pursue the maximization of their own benefits, and through reasonable benefit function design, both parties can reach a relatively balanced state in the game, achieving a win-win or multi-win situation.

[0121] Specifically, the platform benefit function is where p i ·q i represents the rental income obtained by the platform from the i-th user, which is one of the main sources of platform benefit, and the rental income p i is the product of the actual rental quantity q i of the user, which directly reflects the direct economic benefit of the platform in the rental transaction.c v ·P(d i >T) represents the cost of the platform due to user default, c v is the unit default cost, and P(d i >T) is the predicted probability of the i-th user not returning the device within the rental period T (i.e., the default probability), which is predicted by kernel ridge regression. This part of the cost reflects the risk faced by the platform, and the higher the default probability, the greater the loss the platform may suffer.c m ·q i represents the maintenance cost of the platform due to user rental of the device, c m is the unit device maintenance cost, and q iThe larger the rental amount, the higher the maintenance cost that the platform needs to invest. By quantifying the rental income, default cost and equipment maintenance cost, the leasing platform can more accurately evaluate the pros and cons of different leasing strategies, so as to make more reasonable leasing decisions.

[0122] In the Stackelberg game model, the leasing platform acts as the leader, and the definition of its revenue function directly affects the platform's leasing strategy selection. Through the platform revenue function, the platform can maximize its own revenue while considering user response, and to some extent, balance user interests, improve the stability and sustainability of the leasing market.

[0123] The user revenue function is Wherein, u i ·q i represents the utility that the user obtains from the rental equipment, such as the value of equipment use, rental utility u i ·q i The product of the actual rental amount q i ·q i represents the rental cost that the user needs to pay, which is the deduction of user revenue, the higher the rental, the lower the user's revenue. δ·d i ·q i represents the user's deposit occupation cost, δ is the unit time deposit occupation cost rate, d i is the actual rental duration of the i-th user, the deposit occupation cost is proportional to the rental duration and the rental amount, the longer the rental duration and the larger the rental amount, the higher the user's deposit occupation cost.

[0124] The user revenue function considers the rental utility, rental cost and deposit occupation cost from the user's perspective, and can accurately reflect the actual interests of the user in the leasing process. The user can evaluate the impact of different leasing strategies on himself according to the revenue function, so as to make a leasing decision that is more in line with his own interests. This helps to improve the efficiency of the leasing market and reduce unnecessary disputes and losses.

[0125] Further optionally, the solving the Stackelberg equilibrium by reverse induction, generating the optimal rental pricing and rental period strategy of the leasing platform, comprises:

[0126] For users, given the platform strategy G, the optimal rental amount q * is selected: Wherein, q * is the optimal rental amount vector of the user, that is, the number of rentals selected by the user according to the principle of maximizing his own utility, U2(G, q) is the user revenue function under the given platform strategy G, and the platform strategy G includes the rental price p and the rental period T;

[0127] For the leasing platform, according to the optimal leasing quantity response q * (G) of the user, the optimal platform leasing strategy G * is selected Where G * is the optimal rent pricing and lease period strategy vector of the leasing platform, and U1(G, q * (G)) is the platform revenue function under the given optimal leasing quantity response q * (G) of the user.

[0128] Iterative solution of the optimal leasing quantity of the user and the optimal platform leasing strategy until the Stackelberg equilibrium (G * , q * ) is converged, and the dynamic rent p pricing and lease period T adjustment for the leasing platform is used.

[0129] Understandably, for the user, the optimal leasing quantity q * is selected under the fixed platform strategy G (including rent p and lease period T), and the formula is: The user as a follower will select the optimal leasing quantity according to the principle of maximizing his own utility under the known platform strategy, which can accurately reflect the behavior characteristics of the user in the leasing market.

[0130] And the optimal leasing quantity q * selected by the user is a function of the platform strategy G, that is, q * = q * (G). The leasing platform can adjust its own strategy according to the response of the user to maximize its own revenue.

[0131] Therefore, for the leasing platform, the optimal platform leasing strategy G * is selected according to the optimal leasing quantity response q * (G) of the user Where G * is the optimal rent pricing and lease period strategy vector of the leasing platform, and U1(G, q * (G)) is the platform revenue function under the given optimal leasing quantity response q * (G) of the user. The platform as a leader selects the rent pricing and lease period strategy that maximizes its own revenue based on considering the user's response. The platform dominates in the game, and through reasonable leasing strategy adjustment, the platform can maximize its own interests while meeting the needs of the user.

[0132] Through iterative solution of the above two steps (i.e., the user selects the optimal leasing quantity and the platform selects the optimal platform leasing strategy), the equilibrium solution (G * , q *). In the equilibrium state, neither the leasing platform nor the user can improve their own benefits by unilaterally changing the strategy.

[0133] Since the leasing market is dynamic, and the optimal leasing strategy of the platform is based on the optimal leasing quantity response of the user, it means that the platform fully considers the feedback information of the leasing market when formulating the leasing strategy. The platform will continuously adjust the rent pricing and lease period strategy according to the market situation. Through iterative solution of Stackelberg equilibrium, the platform can obtain the optimal leasing strategy in real time and dynamically adjust according to the market changes, so as to improve the operation efficiency and competitiveness of the platform.

[0134] S103, constructing a risk-averse utility function of the leasing platform to dynamically balance the platform benefits and risks.

[0135] Further optionally, the constructing a risk-averse utility function of the leasing platform to dynamically balance the platform benefits and risks comprises:

[0136] Introducing a risk aversion coefficient to construct a risk-averse utility function of the leasing platform:

[0137] Q = E[R] - λ risk ·Var(R),

[0138] Wherein, Q is the risk-averse utility of the leasing platform, E[R] is the expected value of the platform benefits, Var(R) is the variance of the platform benefits, λ risk is the risk aversion coefficient, λ risk ≥ 0;

[0139] Adjusting the risk aversion coefficient according to market fluctuations.

[0140] Understandably, through the risk-averse utility function Q = E[R] - λ risk ·Var(R), the expected value and variance of the platform benefits are combined to quantify the benefits and risks of the platform. The expected value E[R] of the benefits reflects the average benefits that the platform may obtain in the leasing business, and the variance Var(R) of the benefits measures the degree of fluctuation of the benefits, that is, the size of the risk. In this way, the platform can more intuitively evaluate the benefits and risk characteristics of different leasing strategies.

[0141] The introduction of the risk aversion coefficient λ risk allows the platform to adjust the importance of risk according to its risk preference. When λ risk is large, the platform is more sensitive to risk and tends to choose leasing strategies with lower risk but relatively stable benefits; when λ risk is small, the platform has a higher tolerance for risk and may focus more on pursuing high benefits even if it faces certain risks.

[0142] Since the leasing market is dynamic, by dynamically adjusting the risk aversion coefficient, the platform can flexibly adjust its risk preference according to the actual situation of the market. When the market fluctuates greatly and the risk is high, the platform can increase λ risk , reduce the tolerance to risk; when the market is relatively stable and the risk is low, the platform can reduce λ risk , appropriately increase the ability to withstand risk, in order to pursue higher returns.

[0143] Further optionally, the risk-averse utility function of the construction platform further comprises:

[0144] The expected value E[R] and the variance Var(R) of the platform revenue are calculated, and the formula is:

[0145]

[0146] Wherein, E[q i ] is the expected value of the i-th user's leasing amount, and Var(q i ) is the variance of the i-th user's leasing amount.

[0147] It can be understood that the expected value formula of the platform revenue is Wherein, p i ·E[q i ] represents the expected rental income obtained by the platform from the i-th user. The product of the rental p i and the expected value E[q i ] of the user's leasing amount reflects the expected income of the platform from the user under the consideration of the uncertainty of the user's leasing amount.c v ·P(d i >T) represents the expected default cost of the platform due to the i-th user's default, c v is the unit default cost, and P(d i >T) is the probability of the i-th user not returning the device within the lease period T, which reflects the cost caused by the default risk faced by the platform.c m ·E[q i ] represents the expected maintenance cost of the platform due to the i-th user's leasing of the device, c m is the unit device maintenance cost, which is proportional to the expected value E[q i ] of the user's leasing amount, and reflects the expected maintenance cost of the platform in the leasing business due to the use of the device.

[0148] By calculating the expected value of the platform's revenue, the expected revenue of the platform in the leasing business can be accurately quantified, so that the platform can more clearly understand its expected economic situation under different leasing conditions. The expected revenue E[R] is an important part of the risk-averse utility function, which reflects the platform's target level when pursuing revenue. In the utility function, the platform will consider risk factors based on the expected revenue to develop the optimal leasing strategy.

[0149] The variance formula of the platform's revenue is Wherein, represents the risk impact of the fluctuation of the user's leasing quantity q i on the platform's rental income, and the variance Var(q i ) of the leasing quantity reflects the uncertainty of the user's leasing quantity. The square of the rental price p i multiplied by the variance amplifies the impact of the fluctuation of the leasing quantity on the risk of rental income. represents the risk impact of the uncertainty of the user's default probability on the platform's default cost, and P(d i >T)·(1-P(d i >T)) is the variance of the default probability, multiplied by the square of the unit default cost reflects the impact of the fluctuation of the default probability on the risk of the platform's cost.

[0150] By calculating the variance of the platform's revenue, the degree of fluctuation of the platform's revenue, i.e. the risk, is quantified. The larger the variance, the higher the uncertainty of the platform's revenue and the greater the risk, which helps the platform more intuitively understand the risk situation it faces. The revenue variance Var(R) is a key indicator for measuring risk in the risk-averse utility function, and the platform can adjust the risk in the utility function according to the size of the variance and its risk aversion degree, so as to develop a leasing strategy that is more in line with its risk preference.

[0151] Further optionally, the dynamically adjusting the risk aversion coefficient according to market fluctuations comprises:

[0152] If the default rate rises or the predicted device leasing demand in a future period of time is lower than the current leasing demand, and the difference exceeds a preset demand threshold, the risk aversion coefficient is increased to prefer low-risk strategies, including shortening the lease period and increasing the deposit, wherein the default rate is the average of the default probabilities of all users, and the current leasing demand is the actual observed leasing demand in the current market;

[0153] If the current leasing demand is close to or exceeds the predicted device leasing demand in a future period of time, the risk aversion coefficient is reduced to pursue higher revenue.

[0154] It can be understood that for the case of insufficient demand: if the default rate increases (i.e. the current default rate is higher than the historical average default rate) or the lower limit of the 95% confidence interval of the predicted future equipment rental demand for a period of time is lower than the current rental demand, and the difference exceeds the preset demand threshold, the uncertainty of the predicted demand σ can also be considered at the same time 2 ( posterior variance), when the uncertainty is large, appropriately relax the difference threshold requirement or make a decision after further analyzing the market trend; if the uncertainty is small, increase the risk aversion coefficient λ risk , to preferentially select a low-risk strategy, including shortening the rental period and increasing the deposit, wherein the default rate is the average of the default probabilities of all users, and the current rental demand is the actual observed rental demand in the current market (such as order quantity, consultation quantity, etc.);

[0155] For the case of high demand: if the current rental demand is close to or exceeds the upper limit of the 95% confidence interval of the predicted future equipment rental demand for a period of time, the uncertainty of the predicted demand σ can also be considered at the same time 2 , when the uncertainty is large, the risk aversion coefficient λ is reduced with caution risk , and the market changes are closely monitored; if the uncertainty is small, the risk aversion coefficient λ is reduced risk , to pursue higher returns.

[0156] In summary, the intelligent leasing management method of the present application first constructs a non-parametric model based on historical leasing behavior data of users and equipment state data. Because it does not rely on specific function form assumptions, it can flexibly fit complex data patterns, accurately capture the nonlinear behavior characteristics between users and equipment, improve the prediction accuracy of future leasing behavior (such as renewal, default probability) of users, enable the leasing platform to allocate resources to generate reasonable leasing strategies and prevent risks, and also provide personalized leasing services for different users to improve user satisfaction and loyalty. Then, the leasing management is modeled as a dynamic game between the leasing platform and the user, considering the interaction of both parties' strategies, formulating a reasonable leasing strategy that takes into account the platform's revenue and the user's interests, and evaluating and controlling various uncertain factors and risks during the game process. The leasing platform can dynamically adjust the leasing strategy according to the market environment and user leasing behavior to reduce user defaults and market risks. At the same time, a risk-averse utility function is introduced to comprehensively consider the expected value and variance of platform revenue to quantify platform revenue and risk, so that the leasing platform can more intuitively evaluate the revenue and risk characteristics of different leasing strategies, and real-time monitor and adjust the leasing strategy according to the utility function change, thereby improving the timeliness and accuracy of leasing decision-making, bringing significant economic benefits and operational stability to the leasing platform.

[0157] Embodiment two

[0158] Please refer to Figure 2The intelligent leasing management system comprises:

[0159] The user leasing behavior module is configured to construct a non-parametric model based on user historical leasing behavior data and equipment state data to capture nonlinear behavior characteristics of the user and the equipment.

[0160] The leasing strategy generation module is configured to model the leasing management as a dynamic game between the leasing platform and the user, and generate a risk-controllable rent pricing and lease period strategy through a game theory framework.

[0161] The revenue and risk balance module is configured to construct a risk-averse utility function of the leasing platform to dynamically balance platform revenue and risk.

[0162] Further, the user leasing behavior module is further configured to:

[0163] The historical user leasing behavior data and equipment state data are mapped by a kernel function to map the data to a high-dimensional feature space through a Gaussian kernel, and the user leasing behavior data includes leasing equipment frequency, payment delay, and equipment use time length, and the equipment state data includes failure rate and loss rate.

[0164] Based on the feature vector mapped by the kernel function, the future leasing behavior of the user is predicted through kernel ridge regression, including the probability of renewal and the probability of default.

[0165] The time series data of equipment utilization and market supply and demand are modeled by a Gaussian process to generate a probabilistic prediction, i.e., a predicted value of the leasing demand of the equipment in the future period of time.

[0166] Further, the user leasing behavior module is further configured to:

[0167] The target function of the kernel ridge regression is defined in combination with the user behavior feature vector and the leasing behavior label, and the target function is:

[0168]

[0169] where x u，i is the behavior feature vector of the i u，i th user, including the leasing frequency, the payment delay, y is the feature vector mapped by the kernel function, and λ reg is a regularization parameter.

[0170] The dual problem is solved by the kernel trick to obtain the optimal dual weight α * , and the future leasing behavior label of the user is predicted as:

[0171]

[0172] wherein, is the predicted future rental behavior label of the i-th user, including the renewal probability label, the default probability label, K(x u，i , x u，j ) is a kernel function used to calculate the similarity between the i-th user and the j-th user.

[0173] Further optionally, the rental strategy generation module is further configured to:

[0174] model the rental management as a Stackelberg game, with the rental platform as the leader and the user as the follower;

[0175] define the platform strategy space as the combination of the rent pricing and the rental period, and define the user strategy space as the rental quantity;

[0176] define the revenue function, including the platform revenue function and the user revenue function;

[0177] solve the Stackelberg equilibrium by reverse induction to generate the optimal rent pricing and rental period strategy of the rental platform.

[0178] Further optionally, the rental strategy generation module is further configured to:

[0179] combine the actual rental quantity q i of the user and the predicted default probability P of the user to define the platform revenue function:

[0180]

[0181] wherein, U1 is the total revenue of the rental platform, p i is the rent of the i-th user, q i is the actual rental quantity of the i-th user, which is obtained by solving the game equilibrium by reverse induction, c v is the unit default cost, P(d i >T) is the predicted default probability of the i-th user, which is obtained by kernel ridge regression prediction, i.e., the probability of not returning the device within the rental period T, d i is the actual rental duration of the i-th user, c m is the unit device maintenance cost, and n is the total number of users;

[0182] combine the user rental utility and the deposit occupancy cost to define the user revenue function:

[0183]

[0184] wherein, U2 is the total revenue of the user, u iThe rental utility of the ith user is δ, and the deposit occupation cost rate per unit time is δ·d i The deposit occupation cost of the ith user is δ·d

[0185] Further, the rental strategy generation module is further configured to:

[0186] For a user, the fixed platform strategy G, the optimal rental quantity q * is selected. Wherein, q * is the optimal rental quantity vector of the user, that is, the rental quantity selected by the user according to the utility maximization principle, and U2(G, q) is the user revenue function under the given platform strategy G. The platform strategy G includes the rent p and the rental period T.

[0187] For the rental platform, the optimal platform rental strategy G * is selected according to the optimal rental quantity response q * (G) of the user. Wherein, G * is the optimal rent pricing and rental period strategy vector of the rental platform, and U1(G, q * (G)) is the platform revenue function under the given optimal rental quantity response q * (G) of the user.

[0188] The optimal rental quantity of the user and the optimal platform rental strategy are iteratively solved until the Stackelberg equilibrium (G * , q * ) is converged, and the dynamic rent p pricing and rental period T adjustment of the rental platform are used.

[0189] Further, the revenue risk balance module is further configured to:

[0190] A risk aversion coefficient is introduced to construct a risk-averse utility function of the rental platform:

[0191] Q = E[R] - λ risk ·Var(R),

[0192] Wherein, Q is the risk-averse utility of the rental platform, E[R] is the expected value of the platform revenue, Var(R) is the variance of the platform revenue, λ risk is the risk aversion coefficient, and λ risk ≥ 0.

[0193] The risk aversion coefficient is dynamically adjusted according to market fluctuations.

[0194] Further, the revenue risk balance module is further configured to:

[0195] The expected value E[R] and the variance Var(R) of the platform revenue are calculated, and the formula is:

[0196]

[0197] wherein E[q i ] is the expected value of the i-th user's rental quantity, and Var(q i ) is the variance of the i-th user's rental quantity.

[0198] Further optionally, the return-risk balance module is further configured to:

[0199] if the default rate rises or the predicted equipment rental demand in a future period of time is lower than the current rental demand, and the difference exceeds a preset demand threshold, then increase the risk-aversion coefficient to prefer a low-risk strategy, including shortening the rental period, increasing the deposit, wherein the default rate is the average of the default probabilities of all users, and the current rental demand is the actual observed rental demand in the current market;

[0200] if the current rental demand is close to or exceeds the predicted equipment rental demand in a future period of time, then decrease the risk-aversion coefficient to pursue higher returns.

[0201] The above-described embodiments only express several embodiments of the present application, and the description is more specific and detailed, but it should not be understood as a limitation on the scope of the patent of the present application. It should be noted that for ordinary skilled persons in the art, without departing from the concept of the present application, a number of modifications and improvements can be made, which are all within the scope of protection of the present application. Therefore, the protection scope of the patent of the present application should be subject to the appended claims.

Claims

1. An intelligent lease management method, characterized by, The method comprises: Based on the user historical leasing behavior data and equipment state data, a non-parametric model is constructed to capture the nonlinear behavior characteristics of users and equipment; Modeling the leasing management as a dynamic game between the leasing platform and the user, generating a risk-controllable rent pricing and lease period strategy through the game theory framework; Constructing a risk-averse utility function of the leasing platform to dynamically balance the platform revenue and risk. 2.The intelligent lease management method of claim 1, wherein, The method comprises: Kernel function mapping is performed on historical user leasing behavior data and equipment state data to map the data to a high-dimensional feature space through a Gaussian kernel, wherein the user leasing behavior data includes leasing equipment frequency, payment delay, and equipment use time length, and the equipment state data includes failure rate and wear rate; Based on the feature vectors after kernel function mapping, the future leasing behavior of the user is predicted through kernel ridge regression, including the probability of renewal and the probability of default; The time series data of equipment utilization and market supply and demand are modeled through a Gaussian process to generate a probabilistic prediction, i.e., a predicted value of the leasing demand of the equipment in the future period of time. 3.The intelligent lease management method of claim 2, wherein, The method comprises: A target function of the kernel ridge regression is defined in combination with the user behavior feature vector and the leasing behavior label, and the target function is: wherein x u，i is the behavior feature vector of the i-th user, y u，i is the rental behavior label of the i-th user, including the renewal probability label and the default probability label, w is a weight vector used for linear combination of features, is the feature vector after kernel function mapping, λ reg is a regularization parameter; By solving the dual problem through the kernel trick, the optimal dual weights α * and predict the future rental behavior label of the user: wherein, is the predicted future rental behavior label of the i-th user, including the renewal probability label, the default probability label, K(x u，i , x u，j ) is a kernel function used to calculate the similarity between the i-th user and the j-th user. 4.The intelligent lease management method of claim 1, wherein, The method comprises: The leasing management is modeled as a Stackelberg game, with the leasing platform as the leader and the user as the follower; The platform strategy space is defined as a combination of rent pricing and lease period, and the user strategy space is defined as the leasing amount; A revenue function is defined, including a platform revenue function and a user revenue function; A Stackelberg equilibrium is solved through a backward induction method to generate an optimal rent pricing and lease period strategy of the leasing platform. 5.The intelligent lease management method according to claim 2 or 4, characterized in that, The method comprises: Combining the actual rental quantity q of the user i and the predicted probability of default P of the user, define the platform revenue function: wherein U1 is the total revenue of the leasing platform, p i is the rental of the i-th user, q i is the actual rental quantity of the i-th user, obtained by solving the game equilibrium through reverse induction, c v is the unit default cost, P(d i ) is the predicted default probability of the i-th user, obtained by kernel ridge regression, i.e., the probability of not returning the device within the rental period T, d i is the actual rental duration of the i-th user, c m is the unit device maintenance cost, and n is the total number of users. A user revenue function is defined in combination with the user leasing utility and the deposit occupancy cost: wherein U2 is the total revenue of the user, u i is the rental utility of the i-th user, δ is the rate of deposit occupation cost per unit time, and δ·d i is the deposit occupation cost of the i-th user. 6.The intelligent lease management method of claim 5, wherein, The method comprises: For the user, the fixed platform strategy G, select the optimal lease amount q * : Where q * is the optimal lease amount vector of the user, that is, the lease amount selected by the user according to the utility maximization principle, U2(G, q) is the user's income function under the given platform strategy G, and the platform strategy G includes the rent p and the lease period T; For the leasing platform, according to the optimal leasing quantity response q * (G) of the user, the optimal platform leasing strategy G * is selected Wherein, G * is the optimal rental pricing and lease term strategy vector of the leasing platform, U1(G, q * (G)) is the platform revenue function under the given user optimal leasing quantity response q * (G) Iteratively solve the user's optimal leasing quantity and the optimal platform leasing strategy until converging to Stackelberg equilibrium (G * , q * ), for dynamic rent p pricing and lease term T adjustment of the leasing platform. 7.The intelligent lease management method of claim 5, wherein, A risk-averse utility function of the leasing platform is constructed by introducing a risk-averse coefficient: The risk-averse coefficient is dynamically adjusted according to market fluctuations. Q = E[R] - λ risk • Var(R), wherein Q is the risk-averse utility of the leasing platform, E[R] is the expected value of the platform revenue, Var(R) is the variance of the platform revenue, λ risk is the risk-averse coefficient, λ risk ≥ 0; The method comprises: 8.The intelligent lease management method of claim 7, wherein, An expected value E[R] and a variance Var(R) of the platform revenue are calculated, and the formula is: The method comprises: where E[q i ] is the expected value of the i-th user's rental quantity, and Var(q i ) is the variance of the i-th user's rental quantity. 9.The intelligent lease management method of claim 7, wherein, If the default rate rises or the predicted leasing demand of the equipment in the future period of time is lower than the current leasing demand, and the difference exceeds a preset demand threshold, the risk-averse coefficient is increased to prefer a low-risk strategy, including shortening the lease period and increasing the deposit, wherein the default rate is the average of the default probabilities of all users, and the current leasing demand is the actual observed leasing demand in the current market. ​ If the current rental demand is close to or exceeds the predicted equipment rental demand in the future period of time, the risk aversion coefficient is reduced to pursue higher income.

10. An intelligent lease management system for implementing the intelligent lease management method according to any one of claims 1 to 9, characterized by, The system comprises: A user rental behavior module: for constructing a non-parametric model based on user historical rental behavior data and equipment state data to capture the nonlinear behavior characteristics of users and equipment; A rental strategy generation module: for modeling the rental management as a dynamic game between the rental platform and the user, and generating a risk-controllable rental pricing and rental period strategy through a game theory framework; A revenue and risk balance module: for constructing a risk-averse utility function of the rental platform to dynamically balance platform revenue and risk.

Images