Data risk identification method and system for home decoration consumption installment business

Abstract

The invention discloses a data risk identification method and system for a home decoration consumption installment service, and relates to the technical field of consumption risks. The method comprises the following steps: constructing a scene cross index space based on user and home decoration information structured analysis; performing multi-dimensional cross risk identification according to the space to obtain a first risk value; acquiring a merchant performance record and a progress image, and introducing time parameter dynamic evaluation to obtain a second risk value; and fusing the first risk value and the second risk value to carry out time sequence link analysis, and outputting risk results of each staging node. The technical problems that an existing home decoration consumption staging risk identification method is single in data dimension and lacks scene cross validation and dynamic time sequence analysis, so that risk identification is not accurate, and scene specific risks and staging node risk evolution cannot be effectively captured are solved, and the purpose of improving the risk identification efficiency through multi-source data cross validation and dynamic time sequence risk analysis is achieved. Risk identification accuracy and timeliness are improved, and specific risks of home decoration scenes and staging node risks are effectively prevented and controlled.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of consumer risk, in particular to a data risk identification method and system for home decoration consumer installment business. BACKGROUND

[0002] With the upgrading of residents' consumption and the popularity of whole house customization and other home decoration modes, the amount of home decoration consumption continues to rise, the single decoration expenditure is large, and more and more consumers choose to complete the payment of home decoration through installment payment. Home decoration consumer installment business has become an important business form of financial institutions and home decoration platforms. However, compared with traditional consumer installment, home decoration installment business has the characteristics of high amount, long period, multiple participants, and complex performance process. Its business chain usually involves consumers, home decoration merchants, material suppliers, and financial service institutions, and other parties, and the risk sources show the characteristics of diversification, cross and dynamic. For example, the user may have inconsistent identity information, and the credit information quality does not match the actual consumption capacity; there may be abnormal deviation between the home decoration contract amount, decoration area, material configuration and regional market average; the home decoration merchant may have the risk of delay, high complaint rate or default rate in the historical project; and there may be problems such as progress lag, engineering quality dispute, and fund diversion in the construction process.

[0003] The existing home decoration consumer installment risk control method mainly uses user static credit data or single contract information as the core for approval, focuses on pre-loan risk judgment, lacks in-depth analysis of the logical consistency between decoration scene data, and lacks dynamic risk continuous tracking and installment node risk evolution evaluation mechanism in the construction process. Especially in the phased lending mode, if the precise risk identification and evolution prediction of each installment node cannot be performed, the risk may be concentrated and exploded in the later period, increasing the probability of capital loss. SUMMARY

[0004] The present application provides a data risk identification method and system for home decoration consumer installment business, which solves the technical problems of single data dimension, lack of scene cross verification and dynamic time sequence analysis in the existing home decoration consumer installment risk identification method, leading to inaccurate risk identification and inability to effectively capture scene-specific risk and installment node risk evolution.

[0005] In a first aspect, the present application provides a data risk identification method for home decoration consumer installment business, the method comprising: The user information, home decoration consumption related information are subjected to data structural analysis to construct a scene cross-consistent index space; multi-dimensional cross-index risk identification is performed according to the scene cross-consistent index space to obtain a first risk value; historical performance records and decoration progress image data of the home decoration business are collected, and a time parameter is introduced to perform dynamic risk evaluation to obtain a second risk value; time sequence risk link analysis is performed according to the first risk value and the second risk value, and risk evolution is performed on each installment node to obtain a risk identification result of the installment node.

[0006] In a second aspect of the present application, a data risk identification system for home decoration consumption installment business is provided, and the system comprises: A data analysis module: based on user information and home decoration consumption related information, data structural analysis is performed to construct a scene cross-consistent index space; a first risk identification module: according to the scene cross-consistent index space, multi-dimensional cross-index risk identification is performed to obtain a first risk value; a second risk identification module: historical performance records and decoration progress image data of the home decoration business are collected, and a time parameter is introduced to perform dynamic risk evaluation to obtain a second risk value; a risk evolution module: according to the first risk value and the second risk value, time sequence risk link analysis is performed, and risk evolution is performed on each installment node to obtain a risk identification result of the installment node.

[0007] The one or more technical solutions provided in the present application have at least the following technical effects or advantages: First, the user identity, credit situation, and home decoration related information such as decoration contract, budget, and real estate are subjected to structural processing, the scattered data is converted into standardized parameters, and a cross-consistency index system between different elements is constructed to depict the rationality and matching degree of the current home decoration scene. Then, on the basis of the index system, a plurality of cross-indices are subjected to comprehensive analysis and weighted calculation to form a first risk value reflecting the overall static risk level of the user and the home decoration project. Then, the historical performance of the home decoration business and the construction progress image data of the current project are introduced, and dynamic evaluation is performed in combination with the time factor to obtain a second risk value reflecting the risk change trend of the construction stage. Finally, time sequence correlation analysis is performed on the static risk result and the dynamic risk result, a risk evolution link is constructed, each node of the installment lending is gradually deduced and judged, and a risk identification result of each installment node is output, so that phased and traceable risk management of the whole process of home decoration installment is realized. BRIEF DESCRIPTION OF DRAWINGS

[0008] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed to be used in the embodiments will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without any creative effort based on these drawings.

[0009] Figure 1 A flowchart of a data risk identification method for home decoration consumption installment business provided by the embodiments of the present application is shown.

[0010] Figure 2 A system structure diagram of a data risk identification system for home decoration consumption installment business provided by the embodiments of the present application is shown.

[0011] Legend of the drawings: data analysis module 11, first risk identification module 12, second risk identification module 13, risk evolution module 14. DETAILED DESCRIPTION

[0012] In order to further illustrate the technical means and effects adopted by the present application to achieve the predetermined application purpose, the specific embodiments, structures, features and effects according to the present application will be described in detail below with reference to the drawings and preferred embodiments.

[0013] Embodiment one, as shown in the present application, a data risk identification method for home decoration consumption installment business is provided, wherein the method comprises: Figure 1 Based on user information and home decoration consumption related information, data structure analysis is performed to construct a scene cross-consistency index space.

[0014] In the embodiments of the present application, first, user side information and home decoration consumption side information are obtained, wherein the user side information at least includes user identification information, user identity information, credit information; the home decoration consumption side information at least includes house property information related to residence, decoration contract text information, home decoration budget list information, quotation details, material category and brand information, engineering cycle agreement, etc. Then, the above data is subjected to structured analysis processing, that is, user identity information, credit factor parameters and other fields are extracted from the user information, home decoration parameter fields are extracted from the house property information, decoration contract information and home decoration budget information, and then all the extracted fields are subjected to standardization processing to construct a multi-dimensional structured data vector. Subsequently, based on the structured data vector, basic scene elements are extracted, and a scene cross-consistency index space is constructed through processes of two-by-two cross combination and consistency deviation value mapping. Each dimension in the scene cross-consistency index space corresponds to one cross-consistency index, and the current home decoration project corresponds to a risk coordinate point in the space, which provides an input basis for subsequent multi-dimensional cross-index risk identification.

[0015] ​Further, based on user information, home decoration consumption related information, data structure analysis is carried out, including: The user identification information is collected by face scanning; the user identity information, credit information, property information, decoration contract information and home decoration budget information are collected; the user identity information and credit information are verified based on the user identification information to obtain user information; the user identity information and credit factor parameter field of the user information are extracted; the property information, decoration contract information and home decoration budget information are taken as home decoration consumption related information, and text semantic analysis is carried out to extract home decoration parameter fields; the extracted user information and parameter fields of the home decoration consumption related information are standardized to construct a multi-dimensional structured data vector.

[0016] Preferably, first, the terminal device such as a mobile terminal or a counter device calls the camera to perform face scanning, collects user living body image data, and generates user face feature vector by face feature extraction algorithm as user identification information, while collecting user identity information, credit information, property information, decoration contract information and home decoration budget information, wherein the user identity information includes name, ID number, contact information, etc.; the credit information includes credit limit, debt ratio, historical overdue record, etc.; the property information includes house address, building area, property attribute, evaluation value, etc. Then, the user identity information and credit information are verified based on the user identification information, that is, the face feature vector in the user identification information is compared with the ID photo for similarity, to determine whether it is the same person, and then the identity information and credit subject are verified. When the verification is passed, the verification result is marked in the user information record to form a verified user information data set. Then, the user information data set is extracted according to the preset field name to obtain the user identity information field and the credit factor parameter field (such as credit score, debt-income ratio, historical overdue times, credit utilization rate, etc.), wherein the user identity information field includes age, marital status, household attribute, etc.; the credit factor parameter field includes credit score, debt-income ratio, historical overdue times, credit utilization rate, etc. By summarizing these fields, a structured user feature field set is formed.

[0017] After that, in the aspect of home decoration consumption related information processing, the real estate information, decoration contract information and home decoration budget information are input into the semantic analysis unit as contract text data. The semantic analysis of the contract text data is performed by combining text segmentation, entity recognition and rule matching. Specifically, the obtained decoration contract text is preprocessed, including removing special symbols, unifying coding format, deleting redundant spaces and invalid characters, and then using Chinese segmentation algorithm, such as dictionary matching or statistical model based segmentation method, to segment the contract text and split the continuous text into a sequence of words with semantic meaning. Then, the segmentation result is processed by entity recognition based on a named entity recognition model, which can be constructed based on conditional random field (CRF), bidirectional long short-term memory network (BiLSTM) or pre-trained language model, for identifying key entity types related to home decoration scenarios, including amount entity, area entity, time period entity, address entity, material category entity, etc., such as identifying “120 square meters” in “the decoration area is 120 square meters” as an area entity, and identifying “15 million yuan” in “the total price is 15 million yuan” as an amount entity. After completing the entity recognition, a rule matching mechanism is introduced to correct and supplement the recognition result, which can be based on pre-set regular expression rules, keyword trigger rules or context window rules, such as determining the decoration total amount field by including “total price” “contract amount” “engineering cost” and other keywords; matching the area field by “area” “building area” “construction area” and other keywords; matching the engineering period field by “period” “construction period” “start to completion” and other keywords. For the identified amount and area values, context semantics can also be combined to classify and judge the fields to avoid mis-extraction. Then, the entity recognition result and the rule matching result are fused and verified, and when they are consistent, the field value is directly confirmed; when there is a conflict, the final structured field data is output according to the pre-set priority rule or confidence threshold. Through the above process, the decoration area parameter, decoration total amount parameter, material category parameter, unit cost parameter, engineering period parameter and construction address parameter in the contract text data are accurately extracted, providing technical support for subsequent standardization processing and risk modeling.

[0018] After obtaining these fields, the system standardizes these fields, converts the decoration area into a standard area unit such as square meters, converts the total decoration amount into a standard currency unit such as RMB, maps the material category to a preset material classification label system such as economy, standard, high-end, or according to the main material / auxiliary material category code, converts the engineering period into a standard time unit such as days, and encodes the construction address according to the administrative division, for subsequent regional statistical analysis. For missing values or outliers, pre-set rules can be used for filling or truncation. Finally, the above standardized fields are integrated with the previously extracted user identity fields and credit factor fields, arranged in a pre-set field order, and a multi-dimensional structured data vector is constructed, which at least includes user credit dimension, property attribute dimension, decoration amount dimension, area dimension, unit cost dimension, engineering period dimension, material category dimension, and regional dimension, etc., for subsequent scene cross-consistency calculation and risk identification model input, to realize the structured data foundation construction of home decoration consumption installment business.

[0019] Further, the construction of the scene cross-consistency index space includes: Based on the multi-dimensional structured data vector, extract the basic scene elements; construct a multi-dimensional scene feature vector according to the basic scene elements, cross-combine the multi-dimensional scene feature vector, and establish a cross-consistency feature matrix; calculate the consistency deviation value for each cross relationship in the cross-consistency feature matrix, and map the consistency deviation value to a standard risk interval to construct the scene cross-consistency index space.

[0020] Preferably, after completing the construction of the multi-dimensional structured data vector, first, the structured data vector is grouped and the elements are extracted, which are divided into three types of element sets: user credit sub-vector, asset support sub-vector, and home decoration project sub-vector. Specifically, the user credit score, debt-income ratio, historical overdue times, credit utilization rate, etc. are extracted from the multi-dimensional structured data vector as user identity and credit factors and added to the user credit sub-vector; the housing evaluation value, housing area, housing property attribute, housing use type, etc. are extracted as housing valuation factors and housing attribute factors and added to the asset support sub-vector; the decoration area value, total decoration amount value, unit area cost value, engineering period days, material category code, and construction address administrative division code are added to the home decoration project sub-vector, thereby forming an original feature set containing M basic scene elements. Then, the above basic scene elements are arranged in a pre-set order to construct a multi-dimensional scene feature vector F=(f1, f2, …, fM), where each f m i ​are standardized numerical or coded basic scene elements. Then, the multidimensional scene feature vectors are pairwise cross combined to establish a cross-consistency feature matrix G = (g ij ), where g ij represents the cross-consistency relationship value between the ith basic scene element and the jth basic scene element, i≠j. For the cross relationship between different types of elements, different calculation models are adopted: when two elements are continuous numerical variables, a proportional function or a difference function is used for calculation, for example, the total decoration amount and the property assessment value are calculated to obtain the proportion R1 = total decoration amount / property assessment value; the relative deviation rate R2 = (current unit cost - regional average) / regional average is calculated between the unit area cost and the regional historical average; the unit area construction period value R3 = construction period / decoration area is calculated between the construction period and the decoration area; when the cross elements contain category variables, a matching score function or a distribution deviation function is used for calculation, for example, the material category code is matched and scored with the budget amount structure, or the KL divergence value is calculated by comparing the probability of the material category in the historical normal sample with the current sample. Through the above calculation, a matrix G containing several cross relationship indicators is obtained, and each item g ij represents a scene consistency feature.

[0021] Then, the consistency deviation value D ij of each cross relationship indicator in the cross-consistency feature matrix is calculated. ij Specifically, the current cross indicator value is compared with the reference value μ ij established based on the historical normal performance samples, and the standardized deviation calculation formula D ij = |g ij - μ ij | / σ ij is adopted, where σ ij is the standard deviation of the corresponding cross relationship in the historical samples. The deviation value D ij calculated is converted by a preset risk mapping function, for example, a Sigmoid function or a linear segmented function is used to map the deviation value to the standard risk interval of 0-1 to obtain the standardized cross-consistency indicator S ij . When S ij is close to 0, it indicates high consistency, and when S ij is close to 1, it indicates high abnormality. Finally, all standardized cross-consistency indicators S ij are used as different dimensional coordinates to form an N-dimensional scene cross-consistency indicator space, where N is the total number of cross relationship indicators, and N is greater than 2. The current home decoration installment project corresponds to a risk coordinate vector P = (S1, S2, …, SN) in the N-dimensional indicator space. n), and the overall scene consistency deviation degree can be obtained by calculating the weighted distance between the risk coordinate vector and the preset normal scene center vector, thereby providing a quantitative input for subsequent comprehensive calculation of the first risk value, and realizing fine description of the rationality and abnormality degree of the home decoration consumption installment scene.

[0022] Further, the basic scene elements include: user identity and credit factors, house value factors and house attribute factors, decoration area parameters, decoration total amount parameters, unit area cost parameters, engineering cycle parameters, material category parameters.

[0023] Optionally, the basic scene elements refer to a set of core variables reflecting the user qualification status, asset support ability, and decoration project rationality characteristics in the home decoration consumption installment business, including user identity and credit factors, house value factors and house attribute factors, decoration area parameters, decoration total amount parameters, unit area cost parameters, engineering cycle parameters, material category parameters, wherein the user identity and credit factors are used to represent the subject authenticity and repayment ability of the applicant, and specifically include identity attribute information and credit risk indicators after verification, which are used to describe the user's performance ability and default tendency; the house value factors and house attribute factors are used to reflect the asset guarantee degree and the basic conditions of the decoration scene, the house value factors include house evaluation value, market valuation interval, etc. numerical variables, and the house attribute factors include house area, house type, property nature, house age, etc. attribute information, which are used to measure the matching degree between decoration investment and asset value; the decoration area parameter is used to represent the actual construction or the decoration area size agreed in the contract, which is an important indicator for measuring the engineering scale; the decoration total amount parameter is used to represent the total decoration fee agreed in the contract, which is a core variable for measuring the fund demand scale; the unit area cost parameter is usually calculated by the decoration total amount and the decoration area, which is used to reflect the decoration unit price level and can be compared with the regional market average to judge the reasonableness of the quotation; the engineering cycle parameter is used to represent the total construction period required for completing the decoration agreed in the contract or planned, which is usually measured in days, and is used to analyze the matching degree between the engineering scale and the construction time arrangement; the material category parameter is used to represent the material grade or type used in decoration, such as economy type, standard type or high-end type, or is coded according to main materials and auxiliary materials, which is used to measure the rationality between the budget structure and the material configuration. The above various elements jointly constitute the basic scene element system of the home decoration installment business, and through unified standardization and modeling processing of these elements, the fund scale, asset support ability, construction scale and material configuration level of the home decoration project can be comprehensively described, thereby providing basic data support for subsequent scene cross-consistency analysis and risk identification.

[0024] Further, the cross-consistency feature matrix at least includes: a proportion relationship between the property valuation and the total decoration amount; a matching relationship between the decoration area and the unit area cost; a matching relationship between the engineering period and the decoration area size; a matching relationship between the material category and the budget amount structure; and a deviation relationship between the total decoration amount and the regional historical decoration mean value distribution.

[0025] Optionally, the cross-consistency feature matrix is used to depict the logical matching degree and the rationality relationship between different core elements in the home decoration project, and at least includes the proportion relationship between the property valuation and the total decoration amount, the matching relationship between the decoration area and the unit area cost, the matching relationship between the engineering period and the decoration area size, the matching relationship between the material category and the budget amount structure, and the deviation relationship between the total decoration amount and the regional historical decoration mean value distribution. For the proportion relationship between the property valuation and the total decoration amount, the property valuation and the total decoration contract amount are obtained, the asset ratio R1=total decoration contract amount / property valuation is calculated, and a reference interval is established based on historical normal performance samples. When R1 is in the reference interval, it is determined to be a reasonable match, and if it is out of the interval, the relative deviation rate is calculated, so as to quantify the matching degree between the decoration investment and the asset value. For the matching relationship between the decoration area and the unit area cost, the decoration area A and the unit area cost P are obtained, and the matching degree between the decoration area and the unit area cost is quantified by calculating the relative deviation rate of the decoration area A and the unit area cost P. unit (P unit= Total decoration contract amount / decoration area), combined with the historical unit cost average and standard deviation of the construction address belonging to the region, calculate the unit cost deviation Z2 = (unit area cost - historical unit cost average) / standard deviation, and at the same time, construct an area segmented function to establish a corresponding mapping between the area interval and the reasonable unit price interval. If the current sample falls into a mismatched interval, record the matching deviation value to measure the consistency between the area size and the unit price level. For the matching relationship between the project period and the decoration area size, obtain the project period and the decoration area, calculate the unit area period index R3 = project period / decoration area, and compare it with the average unit period of the historical similar area interval to obtain the deviation rate D3 = unit area period index - average unit period | / average unit period, which is used to determine whether the construction time arrangement is suitable for the project size and identify the abnormal risks caused by too short or too long period. For the matching relationship between the material category and the budget amount structure, map the material category parameter to a preset classification label C, such as economy, standard, and high-end, and calculate the amount proportion structure of the main material, auxiliary material, and labor cost in the budget list. Then, based on the typical budget structure distribution vector corresponding to different material categories in the historical samples, calculate the distance value D4 between the current structure vector and the reference vector, such as using Euclidean distance or cosine similarity, to determine whether the material grade and the fund allocation structure are consistent. For the deviation relationship between the total decoration amount and the regional historical decoration average value distribution, determine the administrative region to which the construction address belongs, extract the total decoration amount average value and distribution interval under the same area interval or similar house type in the regional historical database, and calculate the deviation value D5 = | total decoration contract amount - total decoration amount average value | / total decoration amount average value between the current total decoration amount and the regional average value, or calculate its quantile position in the historical distribution, which is used to measure whether the current quotation deviates significantly from the regional normal level. Through the above calculations, D1 to D5 cross-consistency deviation indexes are obtained, which are arranged in a preset order to form elements in the cross-consistency feature matrix. Through normalization processing of each deviation value and mapping to a unified risk interval, the cross-consistency feature matrix is used as the basic data for constructing the scene cross-consistency index space, which realizes the quantitative description of the rationality and abnormality of the home decoration installment scene, and improves the identification accuracy and early warning ability of potential risks such as abnormal quotation, unreasonable budget structure, and insufficient asset support.

[0026] Further, the scene cross-consistency index space is an N-dimensional index space, where each dimension corresponds to a cross-consistency index, and N is a positive integer greater than 2. Based on the consistency index of each dimension, a risk coordinate point is obtained in the scene cross-consistency index space. By calculating the distance between the risk coordinate point and the preset normal scene center point, the scene consistency deviation value is determined.

[0027] Optionally, the scene cross-consistency index space is constructed as an N-dimensional index space, where N is a positive integer greater than 2, each dimension corresponds to a cross-consistency index, and is used to uniformly carry the risk quantification results of different cross relationships. Based on each cross deviation index obtained in the cross-consistency feature matrix, the system generates a standardized consistency index value after normalizing and risk mapping each deviation value, wherein each standardized consistency index value falls within a preset risk interval and corresponds to cross-consistency index dimensions such as the consistency of the total amount of decoration and the proportion of the property valuation, the matching consistency of the decoration area and the unit area cost, the matching consistency of the engineering period and the decoration area scale, the matching consistency of the material category and the budget amount structure, and the deviation consistency of the total amount of decoration and the regional historical mean value distribution. Then, the consistency index value of the current home decoration installment project in each dimension is arranged in a preset dimension order to form a risk coordinate point P, which is used to represent the position of the project in the scene cross-consistency index space. At the same time, a preset normal scene center point C is established based on historical normal performance samples, wherein the coordinate values of the center point can be the mean or median of the corresponding dimension in the normal samples, which represents the typical risk coordinate of the normal home decoration installment scene. To quantify the deviation degree of the current project relative to the normal scene, the distance between the risk coordinate point P and the normal scene center point C is calculated, which can be calculated using the Euclidean distance or the weighted Euclidean distance. The weight of the weighted Euclidean distance can be set according to the influence intensity of the dimension on the default result in the historical samples. The calculated distance D is used as the scene consistency deviation value, which represents the overall cross-consistency deviation degree of the current project; the greater the distance D, the more significant the deviation from the normal scene, and the higher the abnormal risk; when the distance D is within a preset threshold range, it indicates that the overall consistency of the scene cross relationship is good, which provides a quantitative basis for subsequent comprehensive risk scoring and first risk value generation.

[0028] According to the scene cross-consistency index space, multi-dimensional cross-index risk identification is performed to obtain a first risk value.

[0029] In one embodiment, after the construction of the scene cross-consistency index space, multi-dimensional cross-index risk identification is performed on the current home decoration installment project according to the index space, a comprehensive consistency risk score is calculated by weighting, and then the comprehensive risk score value is mapped to a preset risk level interval to form a first risk value. This first risk value represents the overall consistency risk level of the current home decoration installment project in the static scene level, mainly reflects the comprehensive risk degree formed by the multi-dimensional cross relationship of user credit status, asset support ability, and decoration parameter matching rationality, and provides a basic risk state input for subsequent dynamic risk assessment and installment node risk evolution analysis.

[0030] Further, according to the scene cross-consistency index space, multi-dimensional cross-index risk identification is performed to obtain a first risk value, which includes: Obtain the consistency index values ​​of each dimension in the cross-consistency index space of the scenario, and establish the risk feature vector of the current home decoration project; assign weights to the risk feature vector, wherein the weights are determined based on the difference in the impact of each consistency index on the default result in the historical home decoration phase sample data; calculate the comprehensive consistency risk score based on the weighted risk feature vector, and generate the first risk value.

[0031] Optionally, first extract the consistency index values ​​of the current home renovation installment project across various dimensions from the scenario cross-consistency index space. The index space is usually an N-dimensional space, then read the corresponding standardized consistency indices S1, S2, ..., S... n And construct a risk feature vector F=(S1,S2,…,S) according to a preset dimension order. n ), where each dimension S i All values ​​are mapped to a unified risk range and are used to characterize the degree of anomaly in the corresponding cross-relationships, thus forming a vector representation that can depict the overall static scenario risk characteristics of the current project. Subsequently, weights are assigned to the risk feature vectors. Specifically, a training dataset is constructed based on historical home renovation phased sample data. This dataset includes the values ​​of each consistency indicator and the corresponding actual default outcome labels. Statistical analysis is used to calculate the influence strength of each consistency indicator on the default outcome. For example, a logistic regression model is used to obtain the absolute value of the regression coefficients of each indicator, or information gain, IV value, and feature importance score are used to measure its ability to distinguish between defaulted and non-defaulted samples. Weights w are then assigned to each dimension based on the influence strength. i The weights are then normalized to a sum of 1, resulting in a weight vector W = (w1, w2, ..., w...) that corresponds one-to-one with each consistency index. n The weighting is determined by the index, where the more significant the impact on the default outcome, the higher the weight. After weighting, a weighted linear fusion method is used to calculate the comprehensive consistency risk score. The comprehensive consistency risk score is then mapped to a preset risk score range to generate a first risk value. This first risk value reflects the overall static risk level of the current home decoration project in the cross-consistency dimension of the scenario, providing an initial risk status basis for subsequent dynamic risk assessment and phased node risk evolution analysis.

[0032] Collect historical performance records and renovation progress image data of home decoration merchants, introduce time parameters to conduct dynamic risk assessment, and obtain a second risk value.

[0033] In one embodiment, after completing a static risk assessment based on scenario cross-consistency, the performance data of the target home improvement merchant in historical home improvement installment projects is retrieved from the home improvement platform database or installment business system. This performance data includes at least indicators such as historical project completion rate, average delay days or delay rate, complaint rate, phased default rate, and historical installment fund recovery status. Statistical analysis of the above data is performed to calculate the merchant's performance stability index. Simultaneously, image data of the current home improvement project's construction progress uploaded by merchants or users at each construction stage is collected. A preset construction stage recognition model is used to identify the image data, determining the current construction stage category, such as plumbing and electrical stage, masonry and carpentry stage, painting stage, etc., and extracting the upload timestamp information corresponding to the image data. Subsequently, based on the construction schedule and planned time nodes of each stage as stipulated in the contract, the current actual construction stage is compared with the planned construction stage to calculate the construction progress deviation rate. Then, the introduced time parameters are combined with the construction progress deviation rate to perform time-series coupling analysis, calculate the dynamic risk probability value, and map this dynamic risk probability value to a preset risk scoring range to generate a second risk value. This second risk value is used to characterize the dynamic performance risk status of the home decoration project during the construction process. It can reflect the real-time risk level formed by the merchant's historical performance stability and the current construction progress change trend, providing dynamic risk input for subsequent phase node risk evolution analysis.

[0034] Furthermore, by collecting historical performance records of home improvement merchants and video data of renovation progress, and introducing time parameters for dynamic risk assessment, a second risk value is obtained, including: The system collects performance data from home improvement merchants in historical installment projects, including completion rate, delay rate, complaint rate, and default rate, and calculates a merchant performance stability index. It also collects construction progress image data uploaded at each construction stage of the current home improvement project, identifies the current construction stage category using a construction stage identification model, and extracts construction timestamp information. Based on a preset construction plan cycle, it compares the identified current construction stage with the planned stage to calculate the construction progress deviation rate. A time parameter is introduced to couple the performance data with the construction progress deviation rate in a time-series correlation, and a dynamic risk probability assessment is performed based on the time-series coupled performance deviation to obtain a dynamic risk probability. Finally, a second risk value is generated based on the dynamic risk probability.

[0035] Preferably, the process begins by retrieving the target home improvement merchant's performance data from historical home improvement installment projects from the home improvement platform or installment business system. This data includes completion rate, delay rate, complaint rate, and default rate. The above indicators are then statistically aggregated and normalized to construct a merchant performance stability index U, which characterizes the overall reliability of the merchant's performance in historical projects. The completion rate is used as a positive factor, while the delay rate, complaint rate, and default rate are used as negative factors, weighted and integrated to ensure that a higher U value indicates more stable merchant performance and lower basic risk. Subsequently, image data of the current home improvement project's construction progress, uploaded by the merchant or user at each construction stage, is collected. After quality verification of the image data, it is input into a construction stage identification model to identify the current construction stage category K, such as plumbing and electrical, masonry and carpentry, painting, and installation. Simultaneously, the upload timestamp t corresponding to the image data is extracted, forming a construction progress observation sequence of stage category - timestamp.

[0036] For the construction phase identification model, a multi-stage identification structure using convolutional neural networks and temporal modeling modules can be adopted to achieve accurate classification of decoration progress image data. Specifically, firstly, an image input layer is constructed to perform unified preprocessing on the uploaded decoration site images, including scaling the image resolution to 224×224 pixels, normalizing pixel values ​​to the [0,1] range, and performing data augmentation to enhance the model's adaptability to different shooting angles and lighting conditions. After the input layer, a convolutional feature extraction backbone network is constructed, which can use a ResNet-50 structure as the base network. This network contains a 7×7 convolutional kernel (stride 2, output channels 64) and a max pooling layer, followed by four residual module groups stacked sequentially. Each group contains 3, 4, 6, and 3 residual blocks, respectively. Each residual block contains 1×1, 3×3, and 1×1 convolutional structures, and is combined with a BatchNormalization layer and a ReLU activation function to extract multi-scale spatial features. After the ResNet backbone outputs a global feature map, a global average pooling layer compresses the feature map into a one-dimensional feature vector, which is then fed into a fully connected layer for dimensionality compression. For example, the first fully connected layer outputs 512-dimensional features, using ReLU as the activation function and adding a Dropout layer (dropout rate of 0.5) to prevent overfitting. Next, an output classification layer is set, with the number of fully connected output nodes equal to the number of construction stage categories K, and a Softmax function is used to output the probability distribution of each stage. To improve the temporal consistency of stage identification, a lightweight temporal modeling module can be further introduced. When multiple images of the same project are uploaded at multiple time points, the image feature vectors within a continuous time window are input into an LSTM module for sequence modeling. This LSTM can be set to a single-layer or two-layer structure, with 256 hidden layer units. The input is a sequence of 512-dimensional feature vectors from consecutive frames, and the output is a stage prediction sequence. This temporal module avoids misjudgment of a single image that could lead to abnormal stage jumps, thereby improving the stability of construction stage identification.

[0037] During model training, labeled historical construction image data is used as the training set, where each image or image sequence corresponds to a specific construction stage label. The cross-entropy loss function is used, and the Adam optimization algorithm is employed as the optimizer. The initial learning rate can be set to 0.001, and a learning rate decay strategy is used, for example, decreasing to 0.1 times the original rate every 10 epochs. During training, the batch size can be set to 32 or 64, and the number of training epochs can be 50-100. Early stopping control is implemented based on the validation set accuracy. After model training is complete, it is deployed to the system. When the system receives new decoration progress image data, image preprocessing is performed first, then features are extracted through a convolutional backbone network, and the current construction stage category and corresponding probability value are output through a classification layer. If the time-series module is enabled, the stage prediction results within the historical time window are combined for smoothing, and the final construction stage category is output. Simultaneously, timestamp information is extracted from the image metadata to form a stage category-time correspondence, providing basic data for subsequent construction progress deviation rate calculation.

[0038] Then, the planned phase sequence K is obtained according to the preset construction plan cycle. plan(t) The current construction stage K and the planned stage K will be identified based on the planned completion time nodes for each stage. plan(t)The construction progress deviation rate E(t) is calculated by comparison. The construction progress deviation rate can be quantified by the relative difference between the actual completion time of the stage and the planned completion time of the stage, or by the proportion of stages that are behind or ahead of the current stage. The larger E(t) is, the more obvious the deviation of the construction progress from the plan. After obtaining the merchant performance stability index U and the construction progress deviation rate E(t), a time parameter is introduced for time-series correlation coupling processing to construct a dynamic risk probability model. This dynamic risk probability model includes at least the influence function f(U) of performance stability on the risk benchmark value, the influence function g(E(t)) of the construction progress deviation rate on the risk increment value, and the risk accumulation factor h(t) that increases over time. The risk benchmark value influence function f(U) is used to map the merchant performance stability to the basic risk level. For example, the lower U is, the higher the output of f(U) to increase the benchmark risk. The risk increment value influence function g(E(t)) is used to map the construction deviation to the stage risk increment. For example, the larger E(t) is, the higher the output of g(E(t)) to amplify the risk increment. The risk accumulation factor h(t) is used to characterize the cumulative effect of risk over time. When the construction deviation occurs at multiple consecutive time points or the deviation change rate continues to rise, the value of h(t) is increased to reflect the risk superposition. Based on the above model, a dynamic risk probability value P(t) is calculated at the current time t. For example, it is fused and output according to P(t) = Φ(f(U) + g(E(t)) × h(t)), where Φ is a probability mapping function used to constrain the fusion result to the interval between 0 and 1, thus obtaining the dynamic risk probability of the current project at time t. Finally, a second risk value is generated based on the dynamic risk probability value P(t). Specifically, P(t) is converted into a unified risk scoring interval according to a preset scoring mapping rule to obtain the second risk value, which is used to characterize the dynamic performance risk level under the combined effects of the merchant's historical performance stability, current construction progress deviation, and time accumulation effect, thereby realizing real-time assessment and early warning of risks in the phased construction process of home decoration.

[0039] Based on the first risk value and the second risk value, a time-series risk chain analysis is performed to analyze the risk evolution of each phase node and obtain the risk identification results of the phase node.

[0040] In one embodiment, after obtaining the first and second risk values, the first risk value is first used as the initial risk state value of the project, representing the static basic risk level of home decoration installment payments during the project initiation or approval stage. The second risk value is used as a dynamically updated risk state value that changes with the construction progress, representing the performance risk state that changes over time during construction. Subsequently, the entire home decoration installment cycle is divided into multiple installment nodes according to the installment loan plan, with each installment node corresponding to a construction stage, and a risk state sequence consistent with the time sequence is established. Then, the risk evolution value between adjacent installment nodes is calculated based on a pre-built risk state transition model and compared with a preset risk threshold. When the risk evolution value of a node exceeds the preset threshold, it is determined that there is an abnormal risk in that installment node, which may trigger control measures such as manual review, suspension of loan disbursement, or enhanced supervision; when the risk is within an acceptable range, the node risk is determined to be controllable. Finally, the risk identification results of each installment node are output, realizing full-process, node-based, and traceable risk management of home decoration consumer installment business from approval to each stage of loan disbursement, improving the foresight and refinement of risk control.

[0041] Furthermore, based on the first risk value and the second risk value, a time-series risk chain analysis is performed to analyze the risk evolution of each phase node, obtaining the risk identification results of the phase node, including: A risk state sequence is constructed by using a first risk value as the initial risk state value and a second risk value as the dynamic risk state value. Based on the installment loan plan, the entire home renovation installment cycle is divided into multiple installment nodes, each corresponding to a construction stage. A risk state transition model is constructed to calculate the risk evolution value between adjacent installment nodes. Based on the risk state sequence, the risk evolution value is calculated using the risk state transition model. The risk identification result for each installment node is output based on its risk evolution value. Specifically, when the risk evolution value of a certain installment node exceeds a preset threshold, the corresponding installment node is determined to have abnormal risk. When the overall default rate of merchants increases or the deviation change rate continuously exceeds a preset threshold, a risk accumulation factor is added, and this risk accumulation factor is superimposed with the merchant performance stability index to generate a phased risk increment, thereby increasing the risk benchmark value for subsequent installment nodes.

[0042] Preferably, the first risk value is set as the initial risk status value R0 at the time of phased approval or project initiation, which is used to characterize the basic risk level of the home decoration project at the static scene consistency level. The second risk value is used as the dynamic risk status value R that is updated in real time with the construction progress. d(t) Where t represents the corresponding construction stage or time node. Based on the above two risk inputs, a risk state sequence R(t) = {R0, R...} is constructed in chronological order. d(t1) R d(t2) …R d(tk)This is used to describe the trajectory of risk changes throughout the entire phased project lifecycle. Subsequently, based on the installment payment plan, the entire home renovation phased project lifecycle is divided into multiple phase nodes N1, N2, ..., N... k Each phase corresponds to a specific construction stage or phased acceptance milestone, and a corresponding time marker t is assigned to each milestone. i Next, a risk state transition model is constructed based on the risk state sequence to characterize the transmission relationship of risk between adjacent stage nodes. Specifically, the risk state value R of the i-th stage node can be set. i Based on the risk state R of the previous node {i-1} Current dynamic risk value R d(ti) and risk accumulation factor A i Jointly determined, for example, through a weighted update formula R i =α·R {i-1} +β·R d(ti) +γ·A i The calculation is performed, where α, β, and γ are preset coefficients used to control the continuity of historical risks, the degree of impact of current dynamic risks, and the degree of amplification of cumulative risks, thereby calculating the risk evolution value ΔR between adjacent stage nodes. i =R i -R {i-1} This quantifies the magnitude of risk changes. During the risk evolution process, the system continuously monitors the overall merchant default rate trend and the construction progress deviation rate. When an increase in the overall merchant default rate or a construction progress deviation rate exceeding a preset threshold at multiple consecutive time points is detected, a risk enhancement mechanism is triggered, increasing the risk accumulation factor A. i This risk accumulation factor can be calculated using a time-increasing function or a function of consecutive anomalies, and then superimposed on the merchant's performance stability index U. For example, A i =λ·(number of anomalies) + η·(1-U), thus generating a phased risk increment and increasing the risk benchmark value of subsequent phase nodes, causing the risk to exhibit an amplified transmission effect in the chain. Finally, based on the risk evolution value R calculated at each phase node... i By comparing the risk evolution value or cumulative risk status value at a certain installment stage with a preset risk threshold, the system determines that there is an abnormal risk at that stage, which may trigger control measures such as risk warnings, suspension of loan disbursement, or enhanced review. If the risk does not exceed the threshold, the risk at that stage is considered controllable. Through the above-mentioned risk status transfer and accumulation mechanism, the system achieves temporal evolution analysis and refined identification of risks at each stage throughout the entire home renovation installment process, improving the continuity and foresight of risk management.

[0043] In summary, the embodiments of this application have at least the following technical effects: First, based on user information and home decoration consumption-related information, data is structured and analyzed to construct a scenario-based cross-consistency indicator space. Next, multi-dimensional cross-indicator risk identification is performed based on this space to obtain a first risk value. Then, historical performance records of home decoration merchants and decoration progress image data are collected, and time parameters are introduced for dynamic risk assessment to obtain a second risk value. Finally, based on the first and second risk values, a time-series risk chain analysis is performed to analyze the risk evolution of each phase node, obtaining the risk identification results for each phase node. This solves the technical problems of existing home decoration consumption phase risk identification methods, such as single data dimensions, lack of scenario cross-validation and dynamic time-series analysis, leading to inaccurate risk identification and an inability to effectively capture scenario-specific risks and phase node risk evolution. It achieves the technical effect of improving the accuracy and timeliness of risk identification through multi-source data cross-validation and dynamic time-series risk analysis, effectively preventing scenario-specific risks and phase node risks in home decoration.

[0044] Example 2, based on the same inventive concept as the data risk identification method for home decoration installment payment business in the foregoing examples, such as... Figure 2 As shown, this application provides a data risk identification system for home decoration installment payment business, wherein the system includes: Data parsing module 11: Performs structured data parsing based on user information and home decoration consumption-related information to construct a scenario cross-consistency indicator space; First risk identification module 12: Performs multi-dimensional cross-indicator risk identification based on the scenario cross-consistency indicator space to obtain a first risk value; Second risk identification module 13: Collects historical performance records of home decoration merchants and decoration progress image data, introduces time parameters for dynamic risk assessment, and obtains a second risk value; Risk evolution module 14: Performs time-series risk link analysis based on the first risk value and the second risk value, performs risk evolution for each phase node, and obtains the risk identification results for each phase node.

[0045] Furthermore, the data parsing module 11 is used to perform the following methods: User identification information is collected through facial scanning; user identity information, credit information, property information, renovation contract information, and home improvement budget information are collected; user information verification is performed on the user identity information and credit information based on the user identification information to obtain user information; user identity information and credit factor parameter fields are extracted from the user information; the property information, renovation contract information, and home improvement budget information are used as home improvement consumption-related information, and text semantic parsing is performed to extract home improvement parameter fields; the extracted user information and home improvement consumption-related parameter fields are standardized to construct a multi-dimensional structured data vector.

[0046] Furthermore, the data parsing module 11 is used to perform the following methods: Based on the multi-dimensional structured data vector, basic scene elements are extracted; multi-dimensional scene feature vectors are constructed according to the basic scene elements; the multi-dimensional scene feature vectors are combined in pairs to establish a cross-consistency feature matrix; consistency deviation value is calculated for each cross relationship in the cross-consistency feature matrix, and the consistency deviation value is mapped to a standard risk range to construct the scene cross-consistency index space.

[0047] Furthermore, the data parsing module 11 is used to perform the following methods: The basic scenario elements include: user identity and credit factors, property valuation factors and housing attribute factors, decoration area parameters, total decoration amount parameters, unit area cost parameters, project cycle parameters, and material category parameters.

[0048] Furthermore, the data parsing module 11 is used to perform the following methods: The cross-consistency feature matrix includes at least the following: the ratio of property valuation to total renovation cost; the matching relationship between renovation area and cost per unit area; the matching relationship between project cycle and renovation area scale; the matching relationship between material categories and budget amount structure; and the deviation relationship between total renovation cost and the regional historical average renovation distribution.

[0049] Furthermore, the data parsing module 11 is used to perform the following methods: The scenario cross-consistency index space is an N-dimensional index space, where each dimension corresponds to a cross-consistency index, and N is a positive integer greater than 2. Based on the consistency index of each dimension, a risk coordinate point corresponds to the scenario cross-consistency index space. The scenario consistency deviation value is determined by calculating the distance between the risk coordinate point and the preset normal scenario center point.

[0050] Furthermore, the first risk identification module 12 is used to perform the following method: Obtain the consistency index values ​​of each dimension in the cross-consistency index space of the scenario, and establish the risk feature vector of the current home decoration project; assign weights to the risk feature vector, wherein the weights are determined based on the difference in the impact of each consistency index on the default result in the historical home decoration phase sample data; calculate the comprehensive consistency risk score based on the weighted risk feature vector, and generate the first risk value.

[0051] Furthermore, the second risk identification module 13 is used to perform the following method: The system collects performance data from home improvement merchants in historical installment projects, including completion rate, delay rate, complaint rate, and default rate, and calculates a merchant performance stability index. It also collects construction progress image data uploaded at each construction stage of the current home improvement project, identifies the current construction stage category using a construction stage identification model, and extracts construction timestamp information. Based on a preset construction plan cycle, it compares the identified current construction stage with the planned stage to calculate the construction progress deviation rate. A time parameter is introduced to couple the performance data with the construction progress deviation rate in a time-series correlation, and a dynamic risk probability assessment is performed based on the time-series coupled performance deviation to obtain a dynamic risk probability. Finally, a second risk value is generated based on the dynamic risk probability.

[0052] Furthermore, the risk evolution module 14 is used to perform the following methods: A risk state sequence is constructed by using a first risk value as the initial risk state value and a second risk value as the dynamic risk state value. Based on the installment loan plan, the entire home renovation installment cycle is divided into multiple installment nodes, each corresponding to a construction stage. A risk state transition model is constructed to calculate the risk evolution value between adjacent installment nodes. Based on the risk state sequence, the risk evolution value is calculated using the risk state transition model. The risk identification result for each installment node is output based on its risk evolution value. Specifically, when the risk evolution value of a certain installment node exceeds a preset threshold, the corresponding installment node is determined to have abnormal risk. When the overall default rate of merchants increases or the deviation change rate continuously exceeds a preset threshold, a risk accumulation factor is added, and this risk accumulation factor is superimposed with the merchant performance stability index to generate a phased risk increment, thereby increasing the risk benchmark value for subsequent installment nodes.

[0053] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any modifications, equivalent changes, and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A data risk identification method for home decoration installment payment business, characterized in that, include: Based on user information and home decoration consumption-related information, data structure analysis is performed to construct a consistent indicator space for cross-scenario interactions; Based on the scenario-based cross-consistency indicator space, multi-dimensional cross-indicator risk identification is performed to obtain a first risk value; Collect historical performance records and renovation progress image data of home decoration merchants, introduce time parameters to conduct dynamic risk assessment, and obtain a second risk value; Based on the first risk value and the second risk value, a time-series risk chain analysis is performed to analyze the risk evolution of each phase node and obtain the risk identification results of the phase node.

2. The data risk identification method for home decoration installment payment business according to claim 1, characterized in that, Data is structured and analyzed based on user information and home decoration consumption information, including: User identification information is collected through facial scanning; Collect user identity information, credit information, property information, renovation contract information, and home improvement budget information; Based on the user identification information, the user identity information and credit information are verified to obtain user information; Extract user identity information and credit factor parameter fields from the user information; The property information, renovation contract information, and home decoration budget information are treated as home decoration consumption-related information. Text semantic parsing is performed to extract home decoration parameter fields. The extracted user information and home decoration consumption-related parameter fields are standardized to construct a multi-dimensional structured data vector.

3. The data risk identification method for home decoration installment payment business according to claim 2, characterized in that, The constructed scenario cross-consistency index space includes: Based on the multi-dimensional structured data vector, basic scene elements are extracted; Based on the basic scene elements, construct a multi-dimensional scene feature vector, and perform pairwise cross-combination of the multi-dimensional scene feature vectors to establish a cross-consistency feature matrix. For each cross relationship in the cross-consistency feature matrix, calculate the consistency deviation value and map the consistency deviation value to the standard risk interval to construct the scenario cross-consistency index space.

4. The data risk identification method for home decoration installment payment business according to claim 3, characterized in that, The basic scenario elements include: user identity and credit factors, property valuation factors and housing attribute factors, decoration area parameters, total decoration amount parameters, unit area cost parameters, project cycle parameters, and material category parameters.

5. The data risk identification method for home decoration installment payment business according to claim 3, characterized in that, The cross-consistency feature matrix includes at least: The relationship between property valuation and total renovation cost; The relationship between the decoration area and the cost per unit area; The relationship between project duration and renovation area size; Matching relationship between material categories and budget amount structure; Relationship between total renovation cost and the historical average renovation cost in the region.

6. The data risk identification method for home decoration installment payment business according to claim 3, characterized in that, The scenario cross-consistency index space is an N-dimensional index space, where each dimension corresponds to a cross-consistency index, and N is a positive integer greater than 2. Each dimension of the consistency index corresponds to a risk coordinate point in the scenario cross-consistency index space. The scenario consistency deviation value is determined by calculating the distance between the risk coordinate point and the preset normal scenario center point.

7. The data risk identification method for home decoration installment payment business according to claim 3, characterized in that, Based on the aforementioned scenario cross-consistency indicator space, multi-dimensional cross-indicator risk identification is performed to obtain a first risk value, including: Obtain the consistency index values ​​of each dimension in the cross-consistency index space of the scenario, and establish the risk feature vector of the current home decoration project; The risk feature vector is weighted, where the weights are determined based on the difference in the impact of each consistency indicator on the default outcome in historical home renovation phase sample data. A comprehensive consistency risk score is calculated based on the weighted risk feature vector, generating a first risk value.

8. The data risk identification method for home decoration installment payment business according to claim 1, characterized in that, Collect historical performance records and renovation progress video data from home improvement merchants, introduce time parameters for dynamic risk assessment, and obtain a second risk value, including: Collect performance data of home improvement merchants in historical home improvement installment projects. The performance data includes completion rate, delay rate, complaint rate and default rate, and calculate the merchant performance stability index. Collect renovation progress image data uploaded at each construction stage of the current home renovation project, identify the current construction stage category through the construction stage recognition model, and extract construction timestamp information; Based on the preset construction plan cycle, the current construction stage is compared with the planned stage to calculate the construction progress deviation rate. By introducing a time parameter, the performance data and the construction progress deviation rate are coupled in a time series. Based on the performance deviation of the time series coupling, a dynamic risk probability assessment is performed to obtain the dynamic risk probability. The second risk value is generated based on the dynamic risk probability.

9. The data risk identification method for home decoration installment payment business according to claim 1, characterized in that, Based on the first risk value and the second risk value, a time-series risk chain analysis is performed to analyze the risk evolution of each phase node, obtaining the risk identification results of the phase node, including: A risk state sequence is constructed by using the first risk value as the initial risk state value and the second risk value as the dynamic risk state value. Based on the installment loan plan, the entire home renovation installment cycle is divided into multiple installment nodes, each corresponding to a construction stage. A risk state transition model is constructed to calculate the risk evolution value between adjacent installment nodes. Based on the risk state sequence, a risk evolution value is calculated using a risk state transition model. The risk identification result of each phase node is output according to the risk evolution value of each phase node. When the risk evolution value of a certain phase node exceeds a preset threshold, it is determined that the corresponding phase node has abnormal risks. When the overall default rate of merchants increases or the deviation change rate continuously exceeds a preset threshold, a risk accumulation factor is added, and the risk accumulation factor is superimposed with the merchant performance stability index to generate a phased risk increment, thereby increasing the risk benchmark value of subsequent phase nodes.

10. A data risk identification system for home decoration installment payment business, characterized in that, The data risk identification method for home decoration consumer installment business as described in any one of claims 1-9 includes: Data parsing module: Based on user information and home decoration consumption-related information, it performs structured data parsing and constructs a consistent indicator space for cross-scenarios; First risk identification module: Based on the scenario cross-consistency indicator space, perform multi-dimensional cross-indicator risk identification to obtain the first risk value; The second risk identification module collects historical performance records and renovation progress image data of home decoration merchants, introduces time parameters to conduct dynamic risk assessment, and obtains the second risk value. Risk Evolution Module: Based on the first risk value and the second risk value, perform time-series risk link analysis, perform risk evolution on each phase node, and obtain the risk identification results of the phase node.

Images