A deposit-free online rental risk control system and method

Abstract

The present application discloses a risk control system and method for online rental without deposit, which includes: a data acquisition module for acquiring multi-dimensional credit data of the rental user's history within a preset time period; a social credit score calculation module for determining the social credit score data of the rental user based on the social relationship data of the rental user; a model construction module for building a random forest model; an optimization module for optimizing the parameters of the random forest model through the Bayesian algorithm to obtain a risk rating model for the rental user; a control management module for inputting the multi-dimensional credit data of the rental user to be evaluated within a preset time period into the rental user risk rating model to obtain the user's risk level, and determining the rental strategy based on the user's risk level. The present invention can combine multi-dimensional credit data to build a more comprehensive user portrait, and by introducing an intelligent risk assessment mechanism, a more accurate risk assessment of the rental user can be performed.

Description

Technical Field

[0001] The present invention relates to the technical field of online leasing, and in particular to a risk control system and method for online leasing without deposit. Background Art

[0002] With the popularity of online rental platforms and the rise of the sharing economy, user experience has become increasingly important, especially as the "deposit-free" rental model has become mainstream, but how to effectively control and manage risks in the rental process is still a technical challenge facing the industry. Platforms under the deposit-free model without effective risk control measures will face the following types of problems: User default or fraud: Users may disappear after renting items, or damage items, making it impossible for the platform to recover losses. Difficult credit assessment: Without traditional physical deposits, platforms need to rely on non-traditional methods (such as multidimensional data) to assess users' credit. Inaccurate risk control decisions: Without a scientific risk control model, the platform may make wrong risk control decisions, resulting in losses or user loss.

[0003] Currently, many online rental companies directly use credit data collected by third-party companies such as Sesame Credit and Tencent Credit to rate user credit. However, these ratings are often too general and do not fully consider the user's specific behavior and social relationships. For example, if a user has a record of returning items on time many times in the past, or has a strong connection with a social circle with good credit, then the user's credit can be considered good.

[0004] Therefore, the present invention proposes a deposit-free online rental risk control system and method, which can combine multi-dimensional credit data to build a more comprehensive and accurate user portrait, and conduct a more accurate risk assessment of rental users by introducing an intelligent risk assessment mechanism. Summary of the invention

[0005] The present invention provides a deposit-free online rental risk control system and method, which is used to solve the technical problem that the existing online rental risk management method does not fully consider the specific behavior and social relationship of users, resulting in the inability to accurately evaluate and manage the risk level of rental users.

[0006] In order to solve the above problems, the present invention provides a deposit-free online rental risk control system, comprising:

[0007] A data acquisition module, used to acquire multi-dimensional credit data of the rental user's history within a preset time period, wherein the multi-dimensional credit data includes the rental user's identity data, social relationship data, and platform credit score data;

[0008] A social credit score calculation module, used to determine the social credit score data of the rental user according to the social relationship data of the rental user;

[0009] A model building module, used to build a random forest model based on the identity data, social attribute data and platform credit score data of the rental user;

[0010] An optimization module, used to optimize the parameters of the random forest model by using a Bayesian algorithm to obtain a rental user risk rating model;

[0011] The control management module is used to input the multi-dimensional credit data of the rental user to be evaluated within a preset time period into the rental user risk rating model, obtain the risk level of the user, and determine the rental strategy based on the risk level of the user.

[0012] On the other hand, the present invention also provides a deposit-free online rental risk control method, comprising:

[0013] Acquire multi-dimensional credit data of the rental user's history within a preset time period, wherein the multi-dimensional credit data includes the rental user's identity data, social relationship data, and platform credit score data;

[0014] Determining the social credit score data of the leasing user according to the social relationship data of the leasing user;

[0015] Building a random forest model based on the rental user's identity data, social credit score data, and platform credit score data;

[0016] Optimizing the parameters of the random forest model by using a Bayesian algorithm to obtain a rental user risk rating model;

[0017] The multi-dimensional credit data of the rental user to be evaluated within a preset time period is input into the rental user risk rating model to obtain the risk level of the user, and the rental strategy is determined based on the risk level of the user.

[0018] Further, determining the social credit score data of the leasing user according to the social relationship data of the leasing user includes:

[0019] Establish a four-element relationship group based on the social relationship data of the rental user:

[0020] S = (P, R, W, T);

[0021] Determining a plurality of target contacts associated with the rental user and corresponding target relationship weights according to the four-element relationship group;

[0022] Obtaining the platform credit score of the target contact in a preset time period;

[0023] Calculating the social credit score of the rental user according to the platform credit score of the target contact and the target relationship weight;

[0024] Among them, P represents the set of rental users, R represents the set of contacts corresponding to the rental users, which represents the set of natural persons who have contact with the rental users within a preset time period; W represents the intimacy weight between the contacts and the rental users, which is represented by an adjacency matrix; T represents the clustering attribute weight, which represents the clustering relationship division between the contacts and the rental users in the social relationship.

[0025] Furthermore, the intimacy weight W matrix between the contact and the rental user is expressed as:

[0026]

[0027] Each element in the matrix

[0028] In the formula, m represents the sum of communication modes, α represents the intensity coefficient of the communication mode, and c represents the single communication volume.

[0029] Furthermore, the intensity coefficient α of the communication method is i Determined according to the clustering attribute weight T;

[0030] When the communication mode c is a telephone communication mode, if the rented user is the calling party, then If the leased user is the called party, then

[0031] Further, determining the social credit score data of the leasing user according to the social relationship data of the leasing user includes:

[0032] Obtaining platform credit score data of multiple target contacts of the rental user whose intimacy weight is greater than a preset relationship threshold in a preset time period;

[0033] The social credit score data of the rental user is obtained according to the platform credit score data and the intimacy weight of the target contact.

[0034] Furthermore, a random forest model is constructed based on the identity data, social credit score data and platform credit score data of the rental user, including:

[0035] Set the hyperparameters of the random forest model, including: number of decision tree models, maximum tree depth, minimum number of samples required for splitting, maximum number of leaf nodes, maximum number of features, and minimum number of samples for leaf nodes;

[0036] Extracting a plurality of samples from the multi-dimensional credit data of the rental user's history according to the hyperparameters using a bootsrap resampling method;

[0037] Decision tree modeling is performed for each bootsrap sample. Each decision tree is a classifier. The predictions of the multiple decision trees are combined for the input historical multi-dimensional credit data, and the prediction results are obtained by voting;

[0038] The historical multi-dimensional credit data set data is extracted with replacement for training, and the obtained random forest model is saved.

[0039] Furthermore, the random forest model is optimized by using a Bayesian algorithm to obtain a rental user risk rating model, including:

[0040] Establish the optimization objective function of Bayesian optimization;

[0041] Randomly iteratively selecting a hyperparameter combination in a hyperparameter search space of the random forest model;

[0042] Based on the hyperparameter combination, the random forest model is trained using classification parameters to obtain an optimized objective function value;

[0043] When the number of cycles reaches a preset number, the maximum value of the optimization objective function is used as the optimal random forest model hyperparameter to obtain a rental user risk rating model.

[0044] Furthermore, the method for intercepting the preset time period is:

[0045]

[0046] Among them, t is the preset time period, a and b are preset adjustment coefficients, and f is the frequency of use of the preset rental software by the rental user.

[0047] Further, determining a rental strategy based on the risk level of the user includes:

[0048] According to the preset rule engine and constraints, a no-deposit strategy is implemented for low-risk users; a non-deposit-free strategy is implemented for medium-risk users, and the first rental duration and price are adopted; for high-risk users, a rental strategy requiring supplementary review is implemented, and the second rental duration and price are adopted.

[0049] Compared with the prior art, the beneficial effects of the present invention include:

[0050] (1) By collecting multi-dimensional credit data, we can understand the credit status of rental users from multiple perspectives, which helps to build a more comprehensive and accurate user portrait. We have conducted detailed mining of social relationship data to obtain the trust and influence of users in their social circles, which has more comprehensive information value than traditional credit scoring.

[0051] (2) Bayesian optimization can automatically optimize the hyperparameters of the random forest through a probability model. The optimized random forest model can more accurately predict the user's credit risk, thereby providing the platform with more accurate risk control decisions and reducing the economic losses caused by user defaults.

[0052] (3) Assess each user's credit risk based on their recent multi-dimensional data and formulate personalized leasing strategies based on risk levels. This strategy does not rely on fixed rules or manual review and can quickly respond to changing user behaviors, thus achieving flexible risk control.

[0053] In summary, the present invention provides an efficient and accurate risk control system for the deposit-free rental platform by integrating multi-dimensional data, advanced machine learning models and optimization algorithms, which can effectively control risks without deposit guarantee. This risk control system can not only prevent potential credit defaults and losses, but also optimize rental strategies, improve user experience and the overall benefits of the rental platform. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] Figure 1 A schematic diagram of the structure of a deposit-free online rental risk control system provided by the present invention;

[0055] Figure 2 A schematic flow chart of a deposit-free online rental risk control method provided by the present invention. DETAILED DESCRIPTION

[0056] The preferred embodiments of the present invention are described in detail below in conjunction with the accompanying drawings, wherein the accompanying drawings constitute a part of this application and are used together with the embodiments of the present invention to illustrate the principles of the present invention, but are not used to limit the scope of the present invention.

[0057] Example 1

[0058] This embodiment provides a risk control system for online rental without deposit. Figure 1 As shown, Figure 1 : is a schematic diagram of the structure of the deposit-free online rental risk control system 100, including:

[0059] The data acquisition module 101 is used to acquire the multi-dimensional credit data of the rental user's history within a preset time period, wherein the multi-dimensional credit data includes the rental user's identity data, social relationship data, and platform credit score data;

[0060] A social credit score calculation module 102, configured to determine the social credit score data of the renting user according to the social relationship data of the renting user;

[0061] A model building module 103, used to build a random forest model based on the identity data, social attribute data and platform credit score data of the rental user;

[0062] An optimization module 104 is used to optimize the parameters of the random forest model by using a Bayesian algorithm to obtain a rental user risk rating model;

[0063] The control management module 105 is used to input the multi-dimensional credit data of the rental user to be evaluated within a preset time period into the rental user risk rating model to obtain the risk level of the user, and determine the rental strategy based on the risk level of the user.

[0064] The system provided in this embodiment can more comprehensively evaluate the credit status of users by combining multiple data sources such as multi-dimensional credit data, social relationship data, and platform credit scores. By using machine learning models and Bayesian optimization, risk assessment and model optimization can be performed efficiently, improving prediction accuracy. Different leasing strategies are automatically matched according to the user's risk level, improving user experience while protecting platform interests. In addition, it also has continuous optimization and self-adaptation capabilities: through Bayesian optimization and continuous learning, the system can respond to changes in the data environment and maintain long-term stable performance.

[0065] Example 2

[0066] This embodiment provides a risk control method for online rental without deposit. Figure 2 As shown, Figure 2 : is a flow chart of the risk control method for online rental without deposit, and the method comprises:

[0067] Step S101: Acquire multi-dimensional credit data of the rental user's history within a preset time period, wherein the multi-dimensional credit data includes the rental user's identity data, social relationship data, and platform credit score data;

[0068] Step S102: determining the social credit score data of the leasing user according to the social relationship data of the leasing user;

[0069] Step S103: constructing a random forest model based on the identity data, social credit score data and platform credit score data of the rental user;

[0070] Step S104: Optimizing the parameters of the random forest model using a Bayesian algorithm to obtain a rental user risk rating model;

[0071] Step S105: input the multi-dimensional credit data of the rental user to be evaluated within a preset time period into the rental user risk rating model to obtain the risk level of the user, and determine the rental strategy based on the risk level of the user.

[0072] The deposit-free online rental risk control method provided in this embodiment integrates multiple data dimensions (identity, social relationship, platform credit score, etc.) and uses random forest model and Bayesian optimization to provide an efficient and accurate risk control method, which can effectively solve the user credit assessment and risk management problems faced by the deposit-free online rental platform. On the basis of existing technologies, by introducing more data sources and optimization algorithms, more accurate risk control decisions are achieved, and the credit assessment and risk prevention capabilities of the platform are improved.

[0073] As a preferred embodiment, the method for intercepting the preset time period is:

[0074]

[0075] Among them, t is the preset time period, a and b are preset adjustment coefficients, and f is the frequency of use of the preset rental software by the rental user.

[0076] In this embodiment, the meaning of the above formula is: when the frequency f of the rental user using some monitored rental software in the past period of time is high, it means that the rental user is a regular user and has a certain user stickiness. From the perspective of the rental platform, high user stickiness can reflect that the customer is satisfied with the rental platform. At this time, the length of the interception time period can be appropriately shortened, the length of the rental data acquisition can be reduced, the hardware pressure can be reduced, and the processing speed can be increased, so that the user can get a better experience. On the contrary, when the frequency of the user using the rental platform in the past period of time is low, the risk is greater at this time, and the length of the interception time period to be calculated is longer to improve the security of the rental platform.

[0077] Since social relationship data can reflect the trust and influence of users in their social circles, it has more comprehensive information value than traditional credit scores. Whether the user's social circle is healthy, whether there are any breaches of contract or dishonesty, and their reputation in social networks will all affect the assessment of their social credit score. Therefore, in order to make full use of the social relationship data of rental users, when assessing the credit risk level of users, it is necessary to quantify social relationships into available social credit scores.

[0078] As a preferred embodiment, determining the social credit score data of the rental user according to the social relationship data of the rental user includes:

[0079] Establish a four-element relationship group based on the social relationship data of the rental user:

[0080] S = (P, R, W, T);

[0081] Determining a plurality of target contacts associated with the rental user and corresponding target relationship weights according to the four-element relationship group;

[0082] Obtaining the platform credit score of the target contact in a preset time period;

[0083] Calculating the social credit score of the rental user according to the platform credit score of the target contact and the target relationship weight;

[0084] Among them, P represents the set of rental users, R represents the set of contacts corresponding to the rental users, which represents the set of natural persons who have contact with the rental users within a preset time period; W represents the intimacy weight between the contacts and the rental users, which is represented by an adjacency matrix; T represents the clustering attribute weight, which represents the clustering relationship division between the contacts and the rental users in the social relationship.

[0085] Furthermore, the intimacy weight W matrix between the contact and the rental user is expressed as:

[0086]

[0087] Each element in the matrix

[0088] In the formula, m represents the sum of communication modes, α represents the intensity coefficient of the communication mode, and c represents the single communication volume.

[0089] The strength coefficient α of the communication method i Determined according to the clustering attribute weight;

[0090]

[0091] When the communication mode c is a telephone communication mode, if the rented user is the calling party, then If the leased user is the called party, then

[0092] For example, assuming that the preset time period is 1 month, first obtain the contacts corresponding to each rental user in the rental user set P within this month to form a contact set R. For the relationship between each contact and the rental user, the attributes of these relationships may be divided into "colleagues", "close relatives", "relatives", "life service personnel" (takeout, express delivery, sellers, etc.) relationships, etc. Different weights are assigned to the relationship between different contacts and rental users, such as a weight of 0.4 for close relatives, a weight of 0.2 for relatives, and 0.1 for other relationships. Secondly, statistics are taken for the communication method. If the telephone communication is the caller of the rental user, the possible scenario is that the intimate relationship is closer. On the contrary, it may be an inquiry from a life service staff (such as a caller of a courier or takeout). If these are assigned a higher weight, it may affect the final risk assessment result. For other instant communication methods, such as WeChat voice communication, the weight value can also be appropriately increased. Therefore, the quantification process of social relations fully considers the communication method, clustering attribute relationship, etc., and can quantify social relations more comprehensively, which is helpful for subsequent risk assessment.

[0093] The above method determines the "social intimacy circle" of the rental user, and through the analysis of social credit scores, potential high-risk users can be identified early. Those who frequently interact with low-credit users may have unstable behavior, and timely identification helps prevent risks.

[0094] As a preferred embodiment, determining the social credit score data of the rental user according to the social relationship data of the rental user includes:

[0095] Obtaining platform credit score data of multiple target contacts of the rental user whose intimacy weight is greater than a preset relationship threshold in a preset time period;

[0096] The social credit score data of the rental user is obtained according to the platform credit score data and the intimacy weight of the target contact.

[0097] In some embodiments, the processing of obtaining the social credit score data of the rental user based on the platform credit score data and intimacy weight of the target contact can be to obtain the average credit score of the target contact on multiple credit platforms, and then obtain the social credit score data of the rental user based on the target weight, and use the credit score of the rental user's close friend circle as its social credit score data.

[0098] Through the above processing of basic data, we obtain multi-dimensional evaluation data such as identity data, social credit score data and platform credit score data. As a preferred embodiment, a random forest model is constructed based on the identity data, social credit score data and platform credit score data of the rental user, including:

[0099] Set the hyperparameters of the random forest model, including: number of decision tree models, maximum tree depth, minimum number of samples required for splitting, maximum number of leaf nodes, maximum number of features, and minimum number of samples for leaf nodes;

[0100] Extracting a plurality of samples from the multi-dimensional credit data of the rental user's history according to the hyperparameters using a bootsrap resampling method;

[0101] Decision tree modeling is performed for each bootsrap sample. Each decision tree is a classifier. The predictions of the multiple decision trees are combined for the input historical multi-dimensional credit data, and the prediction results are obtained by voting;

[0102] The historical multi-dimensional credit data set data is extracted with replacement for training, and the obtained random forest model is saved.

[0103] As a specific embodiment, firstly, the hyperparameters of the random forest model are set, and the hyperparameters include: the number of decision tree models, the maximum tree depth, the minimum number of samples required for splitting, the maximum number of leaf nodes, the maximum number of features, and the minimum number of samples for leaf nodes;

[0104] For example: the search space for the maximum tree depth is [5, 30], the search space for the minimum number of samples required for splitting is [3, 50], the search space for the maximum number of leaf nodes is [5, 60], the search space for the maximum number of features is [0.1, 0.9], and the search space for the minimum number of samples for leaf nodes is [3, 8].

[0105] Assume that we already have the multi-dimensional rental data of historical users as shown in Table 1:

[0106] Table 1 Multidimensional rental data

[0107]

[0108] Use standardization (Z-score) or normalization (Min-Max Scaling) methods to standardize the values ​​for comparison between different features.

[0109] A bootstrap resampling method is used to extract multiple samples from the multidimensional credit data of the rental user's history according to the hyperparameters; bootstrap resampling is a statistical method that is usually used to generate multiple sample sets by randomly extracting data points from the original data (sampling with replacement) to create multiple different training sets. Each decision tree is trained on a different Bootstrap sample set, and each decision tree in the random forest will independently give a prediction label, and ultimately determine the category of the sample by majority voting. This diversity ultimately helps improve the stability and accuracy of the model.

[0110] The final predicted risk level y is obtained by voting among all trees:

[0111]

[0112] In the formula, h j (x) is the prediction result of the j-th tree for the x-th sample.

[0113] Bayesian optimization is an optimization method based on Bayesian theory, which is mainly used for high-dimensional and computationally expensive optimization problems. In hyperparameter tuning, it is usually necessary to find the optimal combination in a large number of hyperparameter spaces. The core advantage of Bayesian optimization is that it can gradually "learn" which hyperparameter combinations may bring better results during the search process, thereby reducing unnecessary computational overhead and avoiding blindly conducting extensive grid searches or random searches.

[0114] As a preferred embodiment, the random forest model is optimized by using a Bayesian algorithm to obtain a rental user risk rating model, including:

[0115] Establish the optimization objective function of Bayesian optimization;

[0116] Randomly iteratively selecting a hyperparameter combination in a hyperparameter search space of the random forest model;

[0117] Based on the hyperparameter combination, the random forest model is trained using classification parameters to obtain an optimized objective function value;

[0118] When the number of cycles reaches a preset number, the maximum value of the optimization objective function is used as the optimal random forest model hyperparameter to obtain a rental user risk rating model.

[0119] As a specific embodiment, the accuracy is selected as the optimization objective function in this embodiment to judge the model. The overall performance of the final rental user risk rating model can also be evaluated by drawing the ROC curve and calculating the AUC value.

[0120] In practice, we use Python and scikit-learn and the BayesianOptimization library to perform Bayesian optimization on the random forest model to show how to establish the optimization objective function, select the hyperparameter combination, and train the model to obtain the optimal hyperparameters.

[0121]

[0122]

[0123]

[0124] By using Bayesian optimization, we can obtain better model performance with less time and resource investment, and improve the model's predictive ability and stability. The optimized model can more accurately predict the user's credit risk, thereby providing the platform with more accurate risk control decisions and reducing the platform's economic losses caused by user defaults.

[0125] As a preferred embodiment, determining a rental strategy based on the risk level of the user includes:

[0126] According to the preset rule engine and constraints, a no-deposit strategy is implemented for low-risk users; a non-deposit-free strategy is implemented for medium-risk users, and the first rental duration and price are adopted; for high-risk users, a rental strategy requiring supplementary review is implemented, and the second rental duration and price are adopted.

[0127] In some embodiments, for users who rent without deposit, the rental fee can be further reduced and the available rental time can be increased based on their usage frequency to enhance user stickiness. For medium-risk users, the credit rating changes in the last N preset time periods can be further evaluated. If the change trend is from low risk to high risk, the second rental time and price can be used, which is the same as that for high-risk users.

[0128] It should be noted that for any credit leasing platform, user data can be obtained from the corresponding credit data platform through an interface, but a complete database will be established for the registered users of the leasing platform. Through the method of the present invention, the credit of registered users in the database can be regularly analyzed and clustered. In the process of external calls, it is usually only necessary to formulate three to four levels of risk levels and leasing policies to facilitate the convenience of the user end in using the data, so as to enhance the user's leasing experience.

[0129] For high-risk users, supplementary review may involve a manual review process or require uploading additional supporting materials. This process requires special attention to timeliness and accuracy, and further trade-offs between risk control and user experience. When conducting risk assessment according to preset rules, it is necessary to ensure the accuracy and timeliness of the risk assessment model.

[0130] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by any technician familiar with the technical field within the technical scope disclosed by the present invention should be covered within the protection scope of the present invention.

Claims

1. A deposit-free online rental risk control system, characterized in that: include; A data acquisition module, used to acquire multi-dimensional credit data of the rental user's history within a preset time period, wherein the multi-dimensional credit data includes the rental user's identity data, social relationship data, and platform credit score data; A social credit score calculation module, used to determine the social credit score data of the rental user according to the social relationship data of the rental user; A model building module, used to build a random forest model based on the identity data, social attribute data and platform credit score data of the rental user; An optimization module, used to optimize the parameters of the random forest model by using a Bayesian algorithm to obtain a rental user risk rating model; The control management module is used to input the multi-dimensional credit data of the rental user to be evaluated within a preset time period into the rental user risk rating model, obtain the risk level of the user, and determine the rental strategy based on the risk level of the user.

2. A deposit-free online rental risk control method, characterized in that: include: Acquire multi-dimensional credit data of the rental user's history within a preset time period, wherein the multi-dimensional credit data includes the rental user's identity data, social relationship data, and platform credit score data; Determining the social credit score data of the leasing user according to the social relationship data of the leasing user; Building a random forest model based on the rental user's identity data, social credit score data, and platform credit score data; Optimizing the parameters of the random forest model by using a Bayesian algorithm to obtain a risk rating model for rental users; The multi-dimensional credit data of the rental user to be evaluated within a preset time period is input into the rental user risk rating model to obtain the risk level of the user, and the rental strategy is determined based on the risk level of the user.

3. The risk control method for online rental without deposit according to claim 2 is characterized in that: Determining the social credit score data of the leasing user according to the social relationship data of the leasing user includes: Establish a four-element relationship group based on the social relationship data of the rental user: S = (P, R, W, T); Determining a plurality of target contacts associated with the rental user and corresponding target relationship weights according to the four-element relationship group; Obtaining the platform credit score of the target contact in a preset time period; Calculating the social credit score of the rental user according to the platform credit score of the target contact and the target relationship weight; Among them, P represents the set of rental users, R represents the set of contacts corresponding to the rental users, which represents the set of natural persons who have contact with the rental users within a preset time period; W represents the intimacy weight between the contacts and the rental users, which is represented by an adjacency matrix; T represents the clustering attribute weight, which represents the clustering relationship division between the contacts and the rental users in the social relationship.

4. The risk control method for online rental without deposit according to claim 3 is characterized in that: The intimacy weight W matrix between the contact and the rental user is expressed as: Each element in the matrix In the formula, m represents the total number of communication modes, α represents the intensity coefficient of the communication mode, and c represents the single communication volume.

5. The risk control method for online rental without deposit according to claim 4 is characterized in that: The strength coefficient α of the communication method i Determined according to the clustering attribute weight T; T={T1,T2,T3...T Z }, When the communication mode c is a telephone communication mode, if the rented user is the calling party, then If the leased user is the called party, then 6. The risk control method for online rental without deposit according to claim 3 is characterized in that: Determining the social credit score data of the leasing user according to the social relationship data of the leasing user includes: Obtaining platform credit score data of multiple target contacts of the rental user whose intimacy weight is greater than a preset relationship threshold in a preset time period; The social credit score data of the rental user is obtained according to the platform credit score data and the intimacy weight of the target contact.

7. The risk control method for online rental without deposit according to claim 2 is characterized in that: A random forest model is constructed based on the identity data, social credit score data and platform credit score data of the rental user, including: Set the hyperparameters of the random forest model, including: number of decision tree models, maximum tree depth, minimum number of samples required for splitting, maximum number of leaf nodes, maximum number of features, and minimum number of samples for leaf nodes; Extracting a plurality of samples from the multi-dimensional credit data of the rental user's history according to the hyperparameters using a bootsrap resampling method; Decision tree modeling is performed for each bootsrap sample. Each decision tree is a classifier. The predictions of the multiple decision trees are combined for the input historical multi-dimensional credit data, and the prediction results are obtained by voting. The historical multi-dimensional credit data set data is extracted with replacement for training, and the obtained random forest model is saved.

8. The risk control method for online rental without deposit according to claim 2 is characterized in that: The random forest model is optimized by using the Bayesian algorithm to obtain a rental user risk rating model, including: Establish the optimization objective function of Bayesian optimization; Randomly iteratively selecting a hyperparameter combination in a hyperparameter search space of the random forest model; Based on the hyperparameter combination, the random forest model is trained using classification parameters to obtain an optimized objective function value; When the number of cycles reaches a preset number, the maximum value of the optimization objective function is used as the optimal random forest model hyperparameter to obtain a rental user risk rating model.

9. The risk control method for online rental without deposit according to claim 2 is characterized in that: The method for intercepting the preset time period is: Among them, t is the preset time period, a and b are preset adjustment coefficients, and f is the frequency of use of the preset rental software by the rental user.

10. The risk control method for online rental without deposit according to claim 2, characterized in that: Determining a leasing strategy based on the risk level of the user includes: According to the preset rule engine and constraints, a no-deposit strategy is implemented for low-risk users; a non-deposit-free strategy is implemented for medium-risk users, and the first rental duration and price are adopted; for high-risk users, a rental strategy requiring supplementary review is implemented, and the second rental duration and price are adopted.

Images