An online mobile phone rental business risk control system and method based on a big data model

Abstract

The application provides an online mobile phone rental business risk control system and method based on a big data model, applied to the technical field of computers, which immediately collects social network data and system running state data of a new user when the new user initiates a rental request, and constructs an initial social relationship model; based on the system running state data, the running performance of the current system is calculated in real time, and the average risk evaluation delay obtained by the calculation module and the current system load are used by the updating module to dynamically adjust and update the previously constructed social relationship model, so as to ensure the stability and response speed of the system; the updated social relationship model is used to evaluate the risk of the new user, and a risk evaluation result is obtained, and whether the rental operation is executed is determined according to the result. Therefore, the scheme has the beneficial effects of reflecting the dynamic social relationship of the user in real time and ensuring the fast response of risk evaluation.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of computer, in particular to an online mobile phone rental business risk control system and method based on a big data model. BACKGROUND

[0002] With the rapid development of mobile Internet and e-commerce, online mobile phone rental services are becoming increasingly popular. In order to attract a wider user base, platforms are constantly simplifying the registration process and lowering the rental threshold. However, this also presents new challenges for risk control, especially when dealing with new users who are using the rental service for the first time. Due to the lack of historical rental behavior data, traditional risk assessment methods are difficult to apply effectively. To address the fraud risk of new users, online mobile phone rental platforms often use big data models to build real-time risk control systems. These systems rely on analyzing new users' social network data, building user relationship graphs, and assessing the risk characteristics of users and their social relationship networks in real time when new users initiate rental requests.

[0003] However, the social relationship network of new users is not static, but evolves dynamically with their social interactions and behaviors. For example, new users may temporarily join some new social groups due to rental needs, or interact more frequently with certain specific users. These dynamic changes in social relationships may contain important risk signals. In order to more accurately assess risks, the system needs to be able to capture and analyze these dynamic user social relationships in real time. However, frequently updating and analyzing large new user social relationship networks can result in significant computational overhead, potentially leading to increased risk assessment delays and failing to meet the needs of high concurrency and low latency application scenarios for online rental businesses. In particular, during rental peak periods, a large number of users simultaneously initiate rental requests, and the system needs to ensure low latency while considering real-time updates of new user social relationship networks, which becomes a key bottleneck restricting the performance of risk control systems. How to reflect user dynamic social relationships in real time while ensuring fast response of risk assessment under limited computing resources has become a technical problem that needs to be solved for online mobile phone rental platform risk control systems.

[0004] Therefore, when online mobile phone rental platforms perform real-time risk control for new users, in scenarios where user social relationships are dynamically changing and the platform is facing high-concurrency rental requests, how to design a technical solution that can adaptively adjust the user social relationship network update strategy according to the system running state, so as to ensure low latency and high throughput of risk assessment while taking into account the effective use of user dynamic social relationship information, thereby improving the accuracy and efficiency of new user fraud risk identification? SUMMARY

[0005] In view of the above deficiencies of the prior art, the present application provides an online mobile phone rental business risk control system and method based on a big data model, which is applied to the field of computer technology and has the beneficial effects of being able to reflect user dynamic social relationships in real time and guarantee the rapid response of risk assessment.

[0006] In a first aspect, an online mobile phone rental business risk control system based on a big data model comprises:

[0007] An acquisition module is configured to extract social network data and system running state data of a new user when the new user makes a rental request;

[0008] A construction module is configured to construct an initial social relationship model of the new user according to the social network data;

[0009] A calculation module is configured to calculate a current system running performance based on the system running state data, wherein the current system running performance at least includes an average risk assessment delay and a current system load;

[0010] An update module is configured to update the social relationship model according to the average risk assessment delay and the current system load to obtain an updated social relationship model;

[0011] A risk assessment module is configured to perform risk assessment on the new user according to the updated social relationship model to obtain a risk assessment result, and perform a rental operation according to the risk assessment result.

[0012] The application provides an online mobile phone rental business risk control system based on a big data model. When a new user initiates a rental request, the system immediately collects the social network data and system running state data of the new user through an acquisition module. The social network data is the basis for building a user social relationship model, and the system running state data, such as the average risk assessment delay and the current system load, provides a basis for subsequent dynamic adjustment of the social relationship model. A building module uses the acquired social network data to build an initial social relationship model for the new user. The initial social relationship model is a preliminary depiction of the new user's social relationship network and provides basic user portrait information for subsequent risk assessment. A calculation module calculates the running performance of the current system in real time based on the system running state data, and the specific indicators include the average risk assessment delay and the current system load. These two indicators directly reflect the efficiency of the system in processing risk assessment requests and the occupation of system resources. An updating module dynamically adjusts and updates the previously built social relationship model according to the average risk assessment delay and the current system load obtained by the calculation module. This means that the system can adaptively adjust the risk assessment strategy according to its own running status. When the system resources are tight or the risk assessment delay is high, the updating strategy of the social relationship model can be adjusted to ensure the stability and response speed of the system. A risk assessment module uses the updated social relationship model to assess the risk of the new user and obtains a risk assessment result, and decides whether to perform a rental operation according to the result. Since the risk assessment is based on a dynamically updated social relationship model, it can more accurately reflect the real-time risk status of the user. At the same time, the dynamic updating mechanism of the system running state also ensures the efficiency and stability of the risk control system in a high-concurrency scenario. Therefore, the present scheme dynamically updates the social relationship model of the user based on the system running state data, and realizes the adaptive adjustment of the risk control system. This mechanism can ensure the accuracy of risk assessment while considering the running efficiency of the system, effectively solving the technical problem of new user risk control in online mobile phone rental business. The modules work together to realize the complete process from data acquisition, model building, performance monitoring, model updating to risk assessment, forming a closed-loop adaptive risk control system with the beneficial effects of reflecting the dynamic social relationship of the user in real time and ensuring the fast response of risk assessment.

[0013] Further, the initial social relationship model includes a basic social relationship graph and a dynamic social relationship graph of the new user; and the updating module includes:

[0014] A first updating unit: when the average risk delay exceeds a preset delay threshold, the updating frequency of the dynamic social relationship graph is increased to obtain the updated social relationship model.

[0015] The second updating unit: when the current system load exceeds a preset load threshold, the updating frequency of the basic social relationship graph is reduced to obtain the updated social relationship model.

[0016] The application provides an online mobile phone rental business risk control system based on a big data model, which realizes differential updating control of a basic social relationship graph and a dynamic social relationship graph, can adaptively adjust an updating strategy according to a system running state, guarantees timeliness of risk assessment, takes into account system performance, and improves efficiency and resource utilization of the risk control system.

[0017] Further, the updating module further comprises:

[0018] The third updating unit:

[0019] for calculating a first ratio of the average risk delay to the preset delay threshold and a second ratio of the current system load to the preset load threshold when the average risk delay exceeds the preset delay threshold and the current system load exceeds the preset load threshold;

[0020] calculating a first updating frequency adjustment amount of the dynamic social relationship graph according to the first ratio and a second updating frequency adjustment amount of the basic social relationship graph according to the second ratio;

[0021] adjusting the updating frequency of the dynamic social relationship graph according to the first updating frequency adjustment amount and adjusting the updating frequency of the basic social relationship graph according to the second updating frequency adjustment amount, so as to obtain the updated social relationship model.

[0022] The application provides an online mobile phone rental business risk control system based on a big data model. A third updating unit is activated when the average risk delay and the current system load both exceed the threshold value. It first calculates the ratio of the average risk delay to the delay threshold value (first ratio) and the ratio of the current system load to the load threshold value (second ratio). These two ratios reflect the degree of system delay and load exceeding the threshold. Then, the first ratio is used to determine the updating frequency adjustment amount of the dynamic social relationship graph, and the second ratio is used to determine the updating frequency adjustment amount of the basic social relationship graph. This means that the more serious the average risk delay exceeds the threshold, the greater the updating frequency adjustment amount of the dynamic social relationship graph; the more serious the current system load exceeds the threshold, the greater the updating frequency adjustment amount of the basic social relationship graph. Finally, the system adjusts the updating frequency of the dynamic social relationship graph and the basic social relationship graph according to the calculated adjustment amounts, respectively, to obtain the updated social relationship model. In this way, the system can finely adjust the updating frequency of the two social relationship graphs according to the specific exceeding degree of the average risk delay and the current system load, so as to more effectively balance the timeliness of risk assessment and the consumption of system resources when the average risk delay and the current system load are both high, and improve the overall performance of the system.

[0023] Further, the construction module comprises:

[0024] The first obtaining unit is configured to obtain the attention relationship data and the interaction behavior data of the new user on multiple social platforms according to the social network data; determine the social group to which the new user belongs according to the attention relationship data and the interaction behavior data, and extract the size, activity level and member credit level of the social group;

[0025] The first construction unit is configured to establish a basic social relationship graph according to the size, activity level and member credit level;

[0026] The second construction unit is configured to obtain the browsing behavior and search keywords of the new user on the rental platform, and establish a dynamic social relationship graph according to the browsing behavior and search keywords;

[0027] The third construction unit is configured to fuse the basic social relationship graph and the dynamic social relationship graph to obtain the initial social relationship model.

[0028] The application provides an online mobile phone rental business risk control system based on a big data model. A first acquisition unit mines the characteristics of a community to which a new user belongs, and uses the overall portrait of the community to assist in risk assessment of the individual user. The size, activity level and member credit level of the community can reflect the risk level of the social environment in which the user is located, providing an important reference basis for subsequent risk assessment; a first construction unit converts abstract community characteristics into a structured basic social relationship graph, providing quantifiable data support for subsequent risk assessment modules; a second construction unit captures the dynamic behavior of the new user in the rental scenario, and integrates the rental intention and preference of the new user into the social relationship model; and a third construction unit fuses the basic social relationship graph and the dynamic social relationship graph to obtain an initial social relationship model. By fusing two different types of social relationship graphs, the initial social relationship model can more accurately depict the social risk characteristics of the user, providing a more reliable data basis for subsequent risk assessment, thereby improving the accuracy and effectiveness of risk assessment.

[0029] Further, the second construction unit comprises:

[0030] A first recording unit is configured to record the browsing behavior of the new user on the rental platform, extract the browsed mobile phone model, rental duration, payment method, and coupon usage, to form a browsing behavior sequence;

[0031] A second recording unit is configured to record the search keywords of the new user on the rental platform, and perform word segmentation processing on the search keywords to obtain a keyword set;

[0032] A fourth construction unit is configured to construct a two-part graph of "user-mobile phone model" according to the browsing behavior sequence and the keyword set;

[0033] A fifth construction unit is configured to calculate the similarity between the new user and other users according to the two-part graph, select the N users with the highest similarity as the social neighbors of the new user, and construct a dynamic social relationship graph.

[0034] Further, the fourth construction unit comprises:

[0035] A prediction unit is configured to predict the potential interest of the new user in other mobile phone models based on the browsing behavior sequence and the keyword set using a collaborative filtering algorithm, and add the mobile phone models with potential interest to the browsing behavior sequence to update the browsing behavior sequence;

[0036] A first construction subunit is configured to construct a two-part graph of "user-mobile phone model" according to the updated browsing behavior sequence and the keyword set.

[0037] Further, the calculation module comprises:

[0038] The first calculation unit obtains an average risk assessment delay according to the average response time of the risk assessment request in a preset time period based on the system running state data.

[0039] The second calculation unit obtains the current system load by performing weighted summation on the system CPU utilization, the memory occupancy rate, and the disk I / O based on the system running state data system CPU utilization, memory occupancy rate, and disk I / O.

[0040] Further, the risk assessment module comprises:

[0041] The first extraction unit extracts the social group credit score, the social group activity level, and the social group size of the social group to which the new user belongs from the updated social relationship model, and the social neighbor similarity, the potential interest matching degree, and the social neighbor credit risk of the new user in the leasing platform.

[0042] The third calculation unit calculates the risk score of the new user according to the social group credit score, the social group activity level, the social group size, the social neighbor similarity, the potential interest matching degree, and the social neighbor credit risk.

[0043] The risk assessment unit divides the risk level of the new user according to the risk score, and performs a leasing operation according to the risk level.

[0044] Further, the risk assessment unit comprises:

[0045] The first collection unit collects historical leasing data of all users in a preset time period, and the historical leasing data comprises user risk score, user actual default situation, and market environment data.

[0046] The grade division unit divides the risk level of the user by using a clustering algorithm based on the historical leasing data, obtains a plurality of risk levels and corresponding risk score ranges, and calculates the default probability of each risk level according to the actual default situation of the user corresponding to the risk level.

[0047] The leasing unit determines the risk level to which the new user belongs according to the risk score of the new user, sets the leasing amount, the rent, and the insurance fee according to the default probability corresponding to the risk level, and performs a leasing operation.

[0048] In a second aspect, a risk control method for an online mobile phone leasing business based on a big data model is applied to the risk control system for the online mobile phone leasing business based on the big data model, and the method comprises the following steps:

[0049] S1: when a new user initiates a rental request, extract the social network data of the new user and system running state data;

[0050] S2: construct an initial social relationship model of the new user according to the social network data;

[0051] S3: calculate the current system running performance based on the system running state data, the current system running performance at least including average risk assessment delay and current system load;

[0052] S4: update the social relationship model according to the average risk assessment delay and the current system load, to obtain an updated social relationship model;

[0053] S5: perform risk assessment on the new user according to the updated social relationship model, to obtain a risk assessment result, and perform a rental operation according to the risk assessment result.

[0054] Beneficial effects: the online mobile phone rental business risk control system and method based on big data model proposed in the present application, by collecting the social network data of the new user and the system running state data immediately when the new user initiates a rental request, and constructing an initial social relationship model; the initial social relationship model is a preliminary description of the social relationship network of the new user, providing basic user portrait information for subsequent risk assessment; based on the system running state data, the running performance of the current system is calculated in real time, and the specific indicators include average risk assessment delay and current system load; the update module dynamically adjusts and updates the previously constructed social relationship model according to the average risk assessment delay and the current system load obtained by the calculation module, and when the system resource is tight or the risk assessment delay is high, the update strategy of the social relationship model can be adjusted to ensure the stability and response speed of the system; the updated social relationship model is used to perform risk assessment on the new user, to obtain a risk assessment result, and whether to perform a rental operation is determined according to the result. Therefore, the present scheme realizes the adaptive adjustment of the risk control system by using the system running state data and dynamically updating the social relationship model of the user based on these data. The various modules work cooperatively to realize the complete process from data acquisition, model construction, performance monitoring, model updating to risk assessment, forming a closed-loop adaptive risk control system, which has the beneficial effects of reflecting the dynamic social relationship of the user in real time and ensuring the fast response of risk assessment. BRIEF DESCRIPTION OF DRAWINGS

[0055] Figure 1 The structure diagram of the online mobile phone rental business risk control system based on big data model proposed in the present application.

[0056] Figure 2A flowchart of an online mobile phone rental business risk control method based on a big data model is provided for the present application.

[0057] Label explanation: 101, acquisition module; 102, construction module; 103, calculation module; 104, update module; 105, risk assessment module. DETAILED DESCRIPTION

[0058] The technical solutions in the embodiments of the present application will be described clearly and completely below in combination with the drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, not all the embodiments. The components of the embodiments of the present application described and indicated in the drawings herein can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the claimed present application, but only represents selected embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present application.

[0059] It should be noted that: similar labels and letters represent similar items in the following drawings, so once an item is defined in one drawing, it does not need to be further defined and explained in subsequent drawings. At the same time, in the description of the present application, the terms "first, second" and the like are only used to distinguish the description, and cannot be understood as indicating or implying relative importance.

[0060] When the online mobile phone rental platform performs real-time risk control for new users, in the scenario where the user social relationship is dynamically changing and the platform is facing high concurrent rental requests, how to design a technical solution that can adaptively adjust the user social relationship network update strategy according to the system running state, so as to ensure low latency and high throughput of risk assessment, while taking into account the effective use of user dynamic social relationship information, thereby improving the accuracy and efficiency of new user fraud risk identification? In order to solve this problem, the present application provides an online mobile phone rental business risk control system and method based on a big data model, as follows:

[0061] In a first aspect, an online mobile phone rental business risk control system based on a big data model comprises:

[0062] Acquisition module 101: for extracting social network data of a new user and system running state data when the new user's rental request;

[0063] Construction module 102: for constructing an initial social relationship model of the new user according to the social network data;

[0064] The computing module 103 is configured to calculate a current system performance based on the system running state data, the current system performance including at least an average risk assessment delay and a current system load.

[0065] The updating module 104 is configured to update the social relationship model according to the average risk assessment delay and the current system load, to obtain an updated social relationship model.

[0066] The risk assessment module 105 is configured to perform risk assessment on a new user according to the updated social relationship model, to obtain a risk assessment result, and perform a leasing operation according to the risk assessment result.

[0067] The obtaining module 101 can obtain public social data of the new user from multiple social platform interfaces, such as the friend relationship, group information and interaction record of the user on the social media. The system running state data can be collected from a system monitoring component, including the average response time of the risk assessment request and the CPU utilization of the server.

[0068] The constructing module 102 can use a graph database technology, such as Neo4j, to store the social relationship data of the user as a graph structure, in which the nodes represent the user individuals or social groups including the new user, and the edges represent the relationships between the user individuals including the new user. The initial social relationship model can include a basic social relationship graph and a dynamic social relationship graph of the new user, the basic social relationship graph can be constructed based on the long-term social relationship of the user, and the dynamic social relationship graph can be constructed based on the recent social behavior of the user.

[0069] The computing module 103 can periodically poll the system performance indicators, the average risk assessment delay can be calculated as the average response time of all risk assessment requests in a preset time period, and the current system load can be calculated as the weighted sum of the CPU utilization, the memory occupancy and the disk I / O.

[0070] The updating module 104 can include multiple updating units, the first updating unit can increase the updating frequency of the dynamic social relationship graph when the average risk delay exceeds a delay threshold; the second updating unit can reduce the updating frequency of the basic social relationship graph when the current system load exceeds a load threshold; and the third updating unit can finely adjust the updating frequency of the dynamic social relationship graph and the basic social relationship graph according to the degree of delay and load exceeding the threshold when both the average risk delay and the current system load exceed the respective thresholds.

[0071] The risk assessment module 105 can extract the features required for risk assessment from the updated social relationship model, such as community credit score, community activity, and social neighbor similarity. The risk assessment module can use a machine learning model, such as a classification model or a regression model, to calculate the risk score of the user based on the extracted features, and divide the user into different risk levels according to the risk score. The lease operation can be differentiated according to the risk level of the user, for example, for high-risk users, the lease request can be rejected or the rent and insurance fees can be increased, and for low-risk users, more favorable lease conditions can be provided. In this way, the system can quickly obtain relevant data when a new user initiates a lease request, build and update the social relationship model, and perform real-time risk assessment, ensuring the efficiency and safety of the lease business.

[0072] In some embodiments, when a new user initiates a mobile phone lease request, the acquisition module 101 immediately acquires the user's social network data from social platform APIs such as WeChat and Weibo, and acquires the average risk assessment delay within the last 5 minutes from the system monitoring center, which is 0.8 seconds, and the current system load is 70%.

[0073] The construction module 102 uses the JGraphT graph theory library to construct an initial social relationship model containing 1000 nodes based on the social attention relationship and interaction behavior of the new user.

[0074] The calculation module 103 calculates the average risk assessment delay and the current system load.

[0075] The update module 104 detects that the average risk assessment delay exceeds the preset threshold of 0.5 seconds, and the current system load exceeds the preset threshold of 60%, triggering the third update unit. The third update unit calculates the ratio of the average risk delay to the threshold as 1.6, and the ratio of the current system load to the threshold as about 1.17. According to these two ratios, the update frequency of the dynamic social relationship graph is increased by 20%, and the update frequency of the basic social relationship graph is reduced by 10%.

[0076] The risk assessment module 105 extracts the average credit score of the community where the user is located, the risk level of the user's social neighbors, and other features based on the updated social relationship model, and uses a pre-trained gradient boosting decision tree model to calculate the user's risk score. The risk assessment unit divides the user into a medium risk level according to the risk score, and sets the corresponding lease amount, rent and insurance fees, and finally executes the lease operation, allowing the user to lease a mobile phone but requiring the payment of a certain proportion of insurance fees. Through the above steps, the system realizes adaptive adjustment of the social relationship model update strategy according to the system running state, and realizes risk control while ensuring low delay of risk assessment and taking into account the effective use of dynamic social relationship information of new users.

[0077] Further, the initial social relationship model includes a basic social relationship graph and a dynamic social relationship graph of the new user; the updating module includes:

[0078] The first updating unit: when the average risk delay exceeds the preset delay threshold, the updating frequency of the dynamic social relationship graph is increased to obtain an updated social relationship model;

[0079] The second updating unit: when the current system load exceeds the preset load threshold, the updating frequency of the basic social relationship graph is reduced to obtain an updated social relationship model.

[0080] Further, the initial social relationship model is divided into a basic social relationship graph and a dynamic social relationship graph. The basic social relationship graph depicts relatively stable social associations of users, and the dynamic social relationship graph reflects real-time behavior and interest changes of users.

[0081] The updating module 104 includes a first updating unit and a second updating unit. The first updating unit is responsible for monitoring the average risk delay. When the average risk delay exceeds the preset delay threshold, the updating frequency of the dynamic social relationship graph is increased. The dynamic social relationship better reflects the recent behavior changes of users, and has a greater real-time influence on risk assessment. Therefore, when the system delay is high, the updating of the dynamic social relationship graph is prioritized to improve the timeliness of risk assessment. The second updating unit is responsible for monitoring the current system load. When the system load exceeds the preset load threshold, the updating frequency of the basic social relationship graph is reduced. The basic social relationship is relatively stable, and reducing the updating frequency of the basic social relationship graph can reduce system resource consumption and relieve system pressure. Through the differentiated updating strategy, the system can adaptively adjust the updating behavior according to the running state, while ensuring the timeliness of risk assessment, and taking into account the system performance.

[0082] In some specific embodiments, the preset delay threshold is set to 200 milliseconds, and the preset load threshold is set to 80% of CPU utilization. In the initial state, the updating frequency of the dynamic social relationship graph is once per minute, and the updating frequency of the basic social relationship graph is once per hour. When the system monitors that the average risk assessment delay exceeds 200 milliseconds, the first updating unit increases the updating frequency of the dynamic social relationship graph to once every 30 seconds. When the system monitors that the CPU utilization exceeds 80%, the second updating unit reduces the updating frequency of the basic social relationship graph to once every two hours. In this way, the system can dynamically adjust the updating strategy of different social relationship graphs according to the real-time running state, to ensure the efficiency and accuracy of risk assessment.

[0083] Further, the updating module further includes:

[0084] The third updating unit:

[0085] calculating a first ratio of the average risk latency to the preset latency threshold and a second ratio of the current system load to the preset load threshold when the average risk latency exceeds the preset latency threshold and the current system load exceeds the preset load threshold;

[0086] calculating a first update frequency adjustment amount of the dynamic social relationship graph according to the first ratio and a second update frequency adjustment amount of the base social relationship graph according to the second ratio;

[0087] adjusting the update frequency of the dynamic social relationship graph according to the first update frequency adjustment amount and adjusting the update frequency of the base social relationship graph according to the second update frequency adjustment amount, thereby obtaining an updated social relationship model.

[0088] The third updating unit operates when both the average risk latency and the current system load exceed their respective preset thresholds. The first ratio is the average risk latency divided by the preset latency threshold. The second ratio is the current system load divided by the preset load threshold. Thus, the first ratio and the second ratio respectively quantify the extent to which the average risk latency and the current system load exceed the thresholds.

[0089] As a preferred implementation, the first update frequency adjustment amount can be directly proportional to the first ratio, for example, the first update frequency adjustment amount can be set as the product of the first ratio and a preset coefficient. Similarly, the second update frequency adjustment amount of the base social relationship graph is determined according to the second ratio, and the second update frequency adjustment amount can be directly proportional to the second ratio. Subsequently, the update frequency of the dynamic social relationship graph is adjusted according to the first update frequency adjustment amount, for example, if the first update frequency adjustment amount is positive, the update frequency of the dynamic social relationship graph is increased; if the first update frequency adjustment amount is negative, the update frequency of the dynamic social relationship graph is decreased. The update frequency of the base social relationship graph is adjusted according to the second update frequency adjustment amount. Finally, by adjusting the update frequencies of the dynamic social relationship graph and the base social relationship graph, an updated social relationship model is obtained.

[0090] Specifically, when the average risk delay exceeds the delay threshold, it indicates that the system risk assessment response speed is slow, and the updating frequency of the dynamic social relationship graph needs to be accelerated to capture the user dynamic social relationship changes more timely, thereby improving the accuracy of risk assessment. When the current system load exceeds the load threshold, it indicates that the system resources are tight, and the updating frequency of the basic social relationship graph needs to be reduced to reduce the system resource consumption and ensure the stable operation of the system. When both exceed the threshold, the third updating unit can finely measure the degree of delay and load exceeding the threshold by calculating the first ratio and the second ratio, and calculate the updating frequency adjustment amount of the dynamic social relationship graph and the basic social relationship graph according to the two ratios, respectively. Therefore, the system can more reasonably allocate computing resources, while ensuring low delay of risk assessment, and taking into account reducing system load. Through this adaptive updating frequency adjustment mechanism, the risk control system can maintain high performance and efficiency in the complex scenario of high concurrency and dynamic changes of user social relationships.

[0091] Further, the construction module comprises:

[0092] The first obtaining unit is configured to obtain the attention relationship data and the interaction behavior data of the new user on the plurality of social platforms according to the social network data, and determine the belonging community of the new user according to the attention relationship data and the interaction behavior data, and extract the scale, the activity and the member credit level of the belonging community;

[0093] The first construction unit is configured to establish the basic social relationship graph according to the scale, the activity and the member credit level;

[0094] The second construction unit is configured to obtain the browsing behavior and the search keyword of the new user on the rental platform, and establish the dynamic social relationship graph according to the browsing behavior and the search keyword;

[0095] The third construction unit is configured to fuse the basic social relationship graph and the dynamic social relationship graph to obtain the initial social relationship model.

[0096] The attention relationship data can include the fan list and the attention list of the user on the social platform, and the interaction behavior data can include the behavior records such as the like, the comment and the forward of the user. Based on these data, the first obtaining unit can determine the community to which the new user belongs, for example, through a community discovery algorithm, analyze the attention relationship and the interaction frequency between users, and divide the users with close relationship into the same community. After determining the community, the first obtaining unit further extracts the scale of the community, for example, the number of users in the community; the activity of the community, for example, the frequency of posting and interacting of the users in the community; and the credit level of the members of the community, for example, which can be calculated based on the historical behavior data of the users in the community through a credit scoring model.

[0097] The first construction unit is configured to construct a basic social relationship graph based on the community features extracted by the first acquisition unit, such as the community size, activity level, and member credit level. The basic social relationship graph can be understood as a static representation of user social relationships, which reflects the stable social connections and community affiliations of users in the social network. For example, the basic social relationship graph can be represented in a graph structure, where nodes represent users and edges represent social relationships between users. The weights of the edges can be assigned based on community features.

[0098] The browsing behavior can include information such as the mobile phone models browsed by the user on the rental platform, the browsing duration, the clicked links, and the like. The search keywords are the keywords input by the user in the search box of the rental platform. The second construction unit establishes a dynamic social relationship graph based on these browsing behaviors and search keywords. The dynamic social relationship graph aims to capture the real-time interests and intentions of users in the rental scenario and reflect the dynamic changes in social relationships of users. For example, the dynamic social relationship graph can be constructed based on the mobile phone models browsed by the user and the keywords searched, a bipartite graph between users and mobile phone models is constructed, and the similarity between users is calculated based on the bipartite graph, so as to discover the social neighbors of users and construct the dynamic social relationship graph. The third construction unit is responsible for fusing the basic social relationship graph and the dynamic social relationship graph to obtain the final initial social relationship model. The fusion method can be various, such as weighted merging of the two social relationship graphs or using a more complex graph neural network model for fusion. Through fusion, the initial social relationship model can comprehensively consider the static social relationships and dynamic rental behaviors of users, thereby more comprehensively and accurately depicting the social risk features of users.

[0099] Specifically, when a new user initiates a rental request, the system first collects the social network data of the user from multiple social platforms through the first acquisition unit, such as public data or authorized data of each network platform. The collected data includes the user's follow list, fan list, circle dynamic, interaction record, etc. The first acquisition unit cleanses and analyzes these data, and extracts the user's follow relationship data and interaction behavior data.

[0100] Then, the first acquisition unit uses a community discovery algorithm, such as the Louvain algorithm or the K-means algorithm, to divide the user into different communities based on the user's follow relationship and interaction behavior data. For each community, the first acquisition unit counts the number of users in the community as the community size, calculates the average posting frequency and interaction frequency of users in the community as the community activity level, and calculates the average credit level or credit score of community members based on the historical credit data of users in the community.

[0101] The first construction unit receives the community characteristics such as the community size, the activity level and the member credit level output by the first acquisition unit, and constructs a basic social relationship graph. For example, a graph with users as nodes can be constructed, and if two users belong to the same community, an edge is established between the two user nodes, and the weight of the edge can be assigned according to the community size, the activity level and the member credit level. The larger the community size, the higher the activity level and the higher the member credit level, the higher the edge weight.

[0102] The second construction unit monitors the browsing behavior and search keywords of the user on the rental platform in real time. For example, the user browses the rental page of a mobile phone with model X for 5 minutes and searches for the keyword "student discount mobile phone rental". The second construction unit records the browsing behavior sequence and search keyword set of the user. Then, the second construction unit constructs a dynamic social relationship graph based on the browsing behavior sequence and search keyword set of the user. For example, a bipartite graph of "user-mobile phone model" can be constructed, and the mobile phone models browsed or searched by the user are connected with the user nodes, and the weight of the edge can be determined according to the browsing time, search frequency and other factors. Further, based on the bipartite graph, the similarity between users can be calculated using a collaborative filtering algorithm or a graph embedding algorithm, the top N users with the highest similarity are selected as the social neighbors of the new user, and a dynamic social relationship graph is constructed.

[0103] The third construction unit fuses the basic social relationship graph and the dynamic social relationship graph to obtain an initial social relationship model. For example, the basic social relationship graph and the dynamic social relationship graph can be weighted and averaged, and the weight can be adjusted according to the actual application scenario and data. The fused initial social relationship model is used in the subsequent risk assessment module to provide more comprehensive and accurate social relationship features for risk assessment.

[0104] Further, the second construction unit includes:

[0105] The first recording unit is used to record the browsing behavior of the new user on the rental platform, extract the mobile phone model, rental duration, payment method and coupon usage to form a browsing behavior sequence;

[0106] The second recording unit is used to record the search keywords of the new user on the rental platform and perform word segmentation processing on the search keywords to obtain a keyword set;

[0107] The fourth construction unit is used to construct a bipartite graph of "user-mobile phone model" according to the browsing behavior sequence and the keyword set;

[0108] The fifth construction unit is used to calculate the similarity between the new user and other users according to the bipartite graph, select the top N users with the highest similarity as the social neighbors of the new user, and construct a dynamic social relationship graph.

[0109] Among them, the collected data includes the mobile phone models browsed by the user, the user's choice of rental duration, the user's choice of payment method, and the user's coupon information. After integration, these information is arranged in chronological order to form the user's browsing behavior sequence.

[0110] When the user inputs keywords for search, the system records these keywords and performs word segmentation processing on the keywords. The purpose of word segmentation processing is to decompose the user's input search sentence into independent and meaningful words, thereby obtaining a keyword set for subsequent analysis of the user's focus on specific mobile phone models.

[0111] The bipartite graph is a special graph structure that contains two types of nodes, namely user nodes and mobile phone model nodes in this scheme. According to the browsing behavior sequence generated by the first recording unit and the keyword set generated by the second recording unit, a connection relationship is established between the user nodes and the mobile phone model nodes. For example, if a user browses a certain mobile phone or searches for keywords related to a certain mobile phone, an edge is established between the user node and the mobile phone model node. The weight of the edge can be determined according to factors such as the number of views, the viewing duration, the frequency of keyword searches, etc.

[0112] Similarity calculation can use various algorithms, such as collaborative filtering algorithm, cosine similarity algorithm, etc. By comparing the behavior patterns of users in the bipartite graph, i.e., the similarity of the mobile phone models they browse and search, the similarity between users is quantified. After calculating the similarity between the new user and other users, the top N users with the highest similarity are selected as the new user's social neighbors. These social neighbors are considered to have similar interest preferences and risk characteristics as the new user. Finally, based on the information of these social neighbors, a dynamic social relationship graph of the new user is constructed. The dynamic social relationship graph can reflect the new user's interest preferences and potential social connections on the rental platform.

[0113] Specifically, the first recording unit monitors the user's browsing behavior on the rental platform in real time. For example, the user browses a mobile phone with model X, selects a 12-month rental duration, uses an electronic platform payment, and receives a discount coupon. These information will be recorded to form a browsing behavior record. When the user browses different mobile phone models multiple times and performs other related operations, the system will arrange these records in chronological order to generate the user's browsing behavior sequence.

[0114] For the second recording unit, when the user inputs "recommended mobile phone with good camera" in the search box, the system will record this search sentence and use word segmentation technology, such as jieba word segmentation, to decompose the sentence into keywords such as "camera", "good", "mobile phone", and "recommendation". After removing stop words, the keyword set {camera, mobile phone, recommendation} is obtained.

[0115] The fourth construction unit receives the browsing behavior sequence and the keyword set as input and starts constructing the bipartite graph. Assuming that user A has browsed mobile phone models M1 and M2, and the keyword set is {K1, K2}, where K1 is related to M1 and K2 is related to M3. In the bipartite graph, the user A node will establish a connection with the mobile phone model M1 node, the mobile phone model M2 node, and the mobile phone model M3 node. The weight of the edge can be set according to the number of browsing times and the keyword relevance. For example, the more the number of browsing times and the higher the keyword relevance, the greater the weight of the edge. The fifth construction unit uses the constructed bipartite graph to calculate the user similarity using the collaborative filtering algorithm. The collaborative filtering algorithm analyzes the behavior patterns of users in the bipartite graph, such as commonly browsed mobile phone models and commonly searched keywords. By comparing the common behaviors between users, the similarity scores between users are calculated. The higher the score, the more similar the users are. A threshold is set or the N users with the highest similarity are selected as the social neighbors of the new user. For example, the top 5 users with the highest similarity are selected as the social neighbors of the new user, and these neighbors are included in the dynamic social relationship graph of the new user. In this way, the dynamic social relationship graph is constructed for subsequent risk assessment.

[0116] Further, the fourth construction unit includes:

[0117] The prediction unit is configured to predict the potential interest of the new user in other mobile phone models based on the browsing behavior sequence and the keyword set using the collaborative filtering algorithm, and add the mobile phone models with potential interest to the browsing behavior sequence for updating the browsing behavior sequence.

[0118] The first construction sub-unit is configured to construct a "user-mobile phone model" bipartite graph based on the updated browsing behavior sequence and the keyword set.

[0119] The prediction unit can use the collaborative filtering algorithm to analyze the browsing behavior sequence and the keyword set of the new user. The collaborative filtering algorithm identifies other users with similar behavior patterns to the new user to predict mobile phone models that the new user may be interested in but has not yet explicitly expressed. The prediction unit then adds these predicted mobile phone models to the browsing behavior sequence of the new user to expand the browsing behavior sequence and thus more comprehensively reflect the potential interest of the new user.

[0120] The first construction sub-unit is configured to receive the updated browsing behavior sequence and the keyword set from the prediction unit and construct a "user-mobile phone model" bipartite graph based on these data. In this bipartite graph, users and mobile phone models are two types of nodes, and the browsing behavior or search keywords of a user for a specific mobile phone model constitute the edges connecting the user nodes and the mobile phone model nodes.

[0121] Through the mining and supplement of the potential interests of the user by the prediction unit, a more complete and accurate "user-mobile phone model" bipartite graph can be constructed, which provides a more effective data basis for subsequent risk assessment, and thus improves the performance of the risk control system.

[0122] Further, the calculation module comprises:

[0123] The first calculation unit: obtaining the average response time of the risk assessment request in the preset time period according to the system running state data, and obtaining the average risk assessment delay;

[0124] The second calculation unit: according to the system CPU utilization rate, the memory occupancy rate and the disk I / O of the system running state data, and performing weighted summation on the system CPU utilization rate, the memory occupancy rate and the disk I / O, to obtain the current system load.

[0125] The first calculation unit is configured to extract the average time consumed for completing the risk assessment request in the preset time period from the system running state data, and the average time is the average risk assessment delay. The average risk assessment delay can directly reflect the time consumption of the risk assessment process.

[0126] The second calculation unit is configured to monitor the system CPU utilization rate, the memory occupancy rate and the disk I / O in the system running state data. These indicators are key parameters for measuring the system resource usage. By performing weighted summation operation on these three indicators, a comprehensive value representing the current system load level can be obtained.

[0127] The weighted summation method allows adjusting the weights according to the different influences of different resources on the system performance, so that the evaluation result of the system load can more accurately reflect the actual running pressure of the system. As a preferred embodiment, the CPU utilization rate can be set to a weight of 0.5, the memory occupancy rate can be set to a weight of 0.3, and the disk I / O can be set to a weight of 0.2. Then, the three indicators are multiplied by their corresponding weights and summed up to obtain the current system load value.

[0128] Further, the risk assessment module comprises:

[0129] The first extraction unit: extracting the social group credit score, the social group activity level and the social group size of the social group to which the new user belongs, and the social neighbor similarity, the potential interest matching degree and the social neighbor credit risk of the new user in the leasing platform from the updated social relationship model;

[0130] The third calculation unit: calculating the risk score of the new user according to the social group credit score, the social group activity level, the social group size, the social neighbor similarity, the potential interest matching degree and the social neighbor credit risk;

[0131] The risk assessment unit divides the risk level of the new user according to the risk score, and performs the leasing operation according to the risk level.

[0132] Among them, the risk features cover two aspects of community and social neighbors. In the community aspect, community credit score, community activity and community size are extracted to reflect the overall risk situation of the community where the new user is located. In the aspect of social neighbors, the similarity of social neighbors, the potential interest matching degree and the credit risk of social neighbors of the new user on the leasing platform are extracted to focus on the risk information of the social neighbors directly associated with the new user.

[0133] The risk score is a quantitative representation of the risk level of the new user. The risk assessment unit divides the risk level of the new user according to the risk score. The risk level division enables the system to take differentiated leasing operations based on the risk degree.

[0134] In some embodiments, the first extraction unit can be further configured to automatically learn deeper risk features from the social relationship model using a graph neural network. The third calculation unit can use a machine learning model, such as a gradient boosting decision tree model, to calculate the risk score to improve the accuracy of the risk score. The risk assessment unit can establish a dynamic risk level adjustment mechanism to adjust the risk level division threshold and the leasing strategy in real time according to the user's leasing behavior and market environment changes to adapt to the changing risk control needs.

[0135] Further, the risk assessment unit includes:

[0136] The first collection unit collects historical leasing data of all users within a preset time period, and the historical leasing data includes user risk score, user actual default situation and market environment data;

[0137] The grade division unit divides the risk level of the user based on the historical leasing data using a clustering algorithm to obtain multiple risk levels and corresponding risk score ranges, and calculates the default probability of each risk level according to the actual default situation of the user corresponding to the risk level;

[0138] The leasing unit determines the risk level to which the new user belongs according to the risk score of the new user, and sets the leasing amount, rent and insurance fee according to the default probability corresponding to the risk level, and performs the leasing operation.

[0139] Among them, the first mobile unit can retrieve and integrate the historical leasing data of all users within a preset time period from a database or data warehouse. The historical leasing data components can include the risk score obtained by the user in the risk assessment process, the actual default situation record of the user after the end of the leasing, and the market environment data during the leasing, such as macroeconomic indicators, industry policy changes, etc.

[0140] The function implementation of the grade division unit can adopt K-Means clustering algorithm, DBSCAN clustering algorithm, etc. to perform clustering analysis on the historical leasing data collected by the first mobile phone unit. The clustering analysis result is to obtain multiple user risk grades, and each risk grade corresponds to a risk score range. The grade division unit will also statistically calculate the default probability of each risk grade according to the actual default situation of the users in each risk grade. The leasing unit can find the mapping relationship between the risk grade and the risk score range preset by the grade division unit according to the risk score of the new user, so as to determine the risk grade to which the new user belongs.

[0141] Further, the leasing unit can obtain the default probability corresponding to the risk grade, and based on the default probability, combine the preset risk pricing strategy to calculate and set the leasing limit, rent and insurance fee of the new user.

[0142] In some specific embodiments, the preset time period can be set as the last year to ensure the timeliness of the historical leasing data. The clustering algorithm selects K-Means algorithm, and the number of risk grades is divided into five grades, which are low risk, lower risk, medium risk, higher risk and high risk. After completing the clustering, the grade division unit statistically calculates the ratio of the number of users in default in each risk grade to the total number of users in the grade, which is used as the default probability of the risk grade. For example, for the high-risk grade, if it is statistically obtained that the total number of users in the grade is 1000, and the number of users in default is 200, then the default probability of the high-risk grade is calculated as 20%. When setting the leasing limit, the leasing unit can set the leasing limit to a lower level for the users in the high-risk grade, and the rent and insurance fee can be correspondingly increased to cover the higher default risk. For the users in the low-risk grade, the leasing limit can be appropriately relaxed, and the rent and insurance fee can be appropriately reduced to improve the user's leasing willingness.

[0143] The second aspect is a risk control method for online mobile phone leasing business based on a big data model, which is applied to any one of the above-mentioned risk control systems for online mobile phone leasing business based on a big data model. The method comprises the following steps:

[0144] S1: When a new user's leasing request is received, extract the new user's social network data and system running state data;

[0145] S2: Construct an initial social relationship model of the new user according to the social network data;

[0146] S3: Based on the system running state data, calculate the current system running performance, which at least includes the average risk assessment delay and the current system load;

[0147] S4: updating the social relationship model according to the average risk assessment delay and the current system load to obtain an updated social relationship model;

[0148] S5: performing risk assessment on the new user according to the updated social relationship model to obtain a risk assessment result, and performing a leasing operation according to the risk assessment result.

[0149] In step S1, the system responds to a new user leasing request, and the data extraction module is started to obtain the social network activity information of the new user from the database or API interface, such as the friend relationship, group information, interaction record, etc. of the social platform, at the same time, the system monitoring module collects system running state data, such as CPU utilization, memory usage, average response time of risk assessment request, etc.

[0150] In step S2, the model construction module uses the social network data extracted in step S1 to construct an initial social relationship model of the new user. The initial social relationship model can be stored in a graph database, the nodes represent user individuals including the new user, and the edges represent the social relationships between the user individuals including the new user, and the weight of the edge can represent the closeness of the relationship.

[0151] In step S3, the performance calculation module analyzes the system running state data to calculate the running performance indicators of the current system, including at least the average risk assessment delay and the current system load. The average risk assessment delay can be obtained by counting the processing time of risk assessment requests in a predetermined period, and the current system load can be obtained by weighted sum of CPU utilization, memory occupancy and disk I / O, etc. system resource utilization.

[0152] In step S4, the model updating module adjusts the updating strategy of the social relationship model according to the average risk assessment delay and the current system load calculated in step S3. For example, when the average risk assessment delay exceeds the threshold or the system load is too high, the model updating frequency can be reduced or the updating algorithm can be simplified to reduce the system overhead and ensure the response speed of risk assessment.

[0153] In step S5, the risk assessment module uses the updated social relationship model in step S4 to perform risk assessment on the new user to obtain a risk score or risk level, and automatically performs a leasing operation according to the risk assessment result, such as approving or rejecting the lease, or adjusting the lease amount, rent or insurance fee.

[0154] Specifically, the model updating process of step S4 is an adaptive adjustment strategy. When the system is in good condition, for example, the average risk assessment delay is low and the system load is not high, a higher frequency or more complex algorithm can be used to update the social relationship model to more accurately reflect the changes in user social relationships and improve the accuracy of risk assessment. Conversely, when the system pressure increases, for example, the average risk assessment delay exceeds the preset delay threshold or the current system load exceeds the preset load threshold, the model update frequency can be reduced or the update algorithm can be simplified, for example, the number of social relationship dimensions considered or the number of iterations of the update can be reduced, to prioritize low delay and high throughput of risk assessment.

[0155] As a preferred embodiment, the model updating module can include multiple updating units, for example, a first updating unit increases the update frequency of the dynamic social relationship graph when the average risk delay exceeds the delay threshold; a second updating unit reduces the update frequency of the basic social relationship graph when the current system load exceeds the load threshold; and a third updating unit considers the ratio of the average risk delay and the current system load when both exceed the corresponding threshold, and fine-tunes the update frequency adjustment amount of the dynamic social relationship graph and the basic social relationship graph, to achieve more flexible model updating control. In this way, the system can dynamically balance between system performance and risk assessment accuracy, ensuring risk control effectiveness while optimizing system resource utilization.

[0156] In this document, relational terms such as first and second and the like can be used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any such actual relationship or order between such entities or actions.

[0157] The above only describes the embodiments of the present application and does not limit the protection scope of the present application. For those skilled in the art, the present application can have various modifications and changes. Any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. An online mobile phone rental business risk control system based on a big data model, characterized in that, The method comprises the following steps: an acquisition module is configured to extract social network data of a new user and system running state data when a lease request of the new user is received; a construction module is configured to construct an initial social relationship model of the new user according to the social network data; a calculation module is configured to calculate a current system running performance based on the system running state data, wherein the current system running performance comprises at least an average risk assessment delay and a current system load; an updating module is configured to update the social relationship model according to the average risk assessment delay and the current system load to obtain an updated social relationship model; the initial social relationship model comprises a basic social relationship graph and a dynamic social relationship graph of the new user; the updating module comprises: a first updating unit is configured to increase an updating frequency of the dynamic social relationship graph when the average risk assessment delay exceeds a preset delay threshold to obtain the updated social relationship model; a second updating unit is configured to decrease an updating frequency of the basic social relationship graph when the current system load exceeds a preset load threshold to obtain the updated social relationship model; the updating module further comprises: a third updating unit is configured to: calculate a first ratio of the average risk assessment delay to the preset delay threshold and a second ratio of the current system load to the preset load threshold when the average risk assessment delay exceeds the preset delay threshold and the current system load exceeds the preset load threshold; calculate a first updating frequency adjustment amount of the dynamic social relationship graph according to the first ratio and a second updating frequency adjustment amount of the basic social relationship graph according to the second ratio; adjust the updating frequency of the dynamic social relationship graph according to the first updating frequency adjustment amount and adjust the updating frequency of the basic social relationship graph according to the second updating frequency adjustment amount to obtain the updated social relationship model; a risk assessment module is configured to perform risk assessment on the new user according to the updated social relationship model to obtain a risk assessment result and perform a lease operation according to the risk assessment result.

2. The online mobile phone rental business risk control system based on a big data model according to claim 1, characterized in that, The construction module comprises: a first acquisition unit is configured to acquire attention relationship data and interactive behavior data of the new user on multiple social platforms according to the social network data, determine a social group to which the new user belongs according to the attention relationship data and the interactive behavior data, and extract a size, an activity level and a member credit level of the social group; a first construction unit is configured to establish a basic social relationship graph according to the size, the activity level and the member credit level; a second construction unit is configured to acquire browsing behavior and search keywords of the new user on a lease platform and establish a dynamic social relationship graph according to the browsing behavior and the search keywords; a third construction unit is configured to fuse the basic social relationship graph and the dynamic social relationship graph to obtain the initial social relationship model.

3. The online mobile phone rental business risk control system based on a big data model according to claim 2, characterized in that, The second construction unit comprises: The first recording unit is configured to record the browsing behavior of the new user on the rental platform, extract the browsed mobile phone model, rental duration, payment method, and coupon usage, and form a browsing behavior sequence. The second recording unit is configured to record the search keywords of the new user on the rental platform, and perform word segmentation processing on the search keywords to obtain a keyword set. The fourth construction unit is configured to construct a bipartite graph of "user-mobile phone model" based on the browsing behavior sequence and the keyword set. The fifth construction unit is configured to calculate the similarity of the new user and other users based on the bipartite graph, select the N users with the highest similarity as the social neighbors of the new user, and construct a dynamic social relationship graph.

4. The online mobile phone rental business risk control system based on a big data model according to claim 3, characterized in that, The fourth construction unit includes: The prediction unit is configured to predict the potential interest of the new user in other mobile phone models based on the browsing behavior sequence and the keyword set using a collaborative filtering algorithm, and add the mobile phone models with potential interest to the browsing behavior sequence for updating the browsing behavior sequence. The first construction sub-unit is configured to construct a bipartite graph of "user-mobile phone model" based on the updated browsing behavior sequence and the keyword set.

5. The online mobile phone rental business risk control system based on a big data model according to claim 1, characterized in that, The calculation module includes: The first calculation unit is configured to obtain the average response time of risk assessment requests in a preset time period based on the system running state data, and obtain the average risk assessment delay. The second calculation unit is configured to obtain the current system load based on the system running state data, system CPU utilization, memory occupancy, and disk I / O, and perform weighted summation on the system CPU utilization, memory occupancy, and disk I / O.

6. The online mobile phone rental business risk control system based on a big data model according to claim 1, characterized in that, The risk assessment module includes: The first extraction unit is configured to extract the social group credit score, social group activity, and social group size of the new user's social group from the updated social relationship model, as well as the social neighbor similarity, potential interest matching degree, and social neighbor credit risk of the new user on the rental platform. The third calculation unit is configured to calculate the risk score of the new user based on the social group credit score, social group activity, social group size, social neighbor similarity, potential interest matching degree, and social neighbor credit risk. The risk assessment unit is configured to divide the risk level of the new user based on the risk score, and perform a rental operation based on the risk level.

7. The online mobile phone rental business risk control system based on a big data model according to claim 1, characterized in that, The risk assessment unit includes: The first collection unit is configured to collect historical rental data of all users in a preset time period, including user risk score, actual default situation, and market environment data. The grade division unit is configured to divide users into risk grades based on the historical rental data using a clustering algorithm, obtain multiple risk grades and corresponding risk score ranges, and calculate the default probability of each risk grade based on the actual default situation of users corresponding to the risk grade. The rental unit is configured to determine the risk grade of the new user based on the risk score of the new user, set the rental amount, rent, and insurance fee based on the default probability corresponding to the risk grade, and perform a rental operation.

8. An online mobile phone rental service risk control method based on a big data model, applied to the online mobile phone rental service risk control system based on a big data model in any one of claims 1-7, characterized in that, The method comprises the steps of: S1: extracting social network data of a new user and system running state data when a lease request of the new user is received; S2: constructing an initial social relationship model of the new user according to the social network data; S3: calculating a current system running performance based on the system running state data, wherein the current system running performance at least comprises an average risk assessment delay and a current system load; S4: updating the social relationship model according to the average risk assessment delay and the current system load to obtain an updated social relationship model; S5: performing risk assessment on the new user according to the updated social relationship model to obtain a risk assessment result, and performing a lease operation according to the risk assessment result.

Images