Machine learning-based intelligent sorting method, system and equipment for out-calling of unhealthy assets

Abstract

The invention discloses a machine learning-based intelligent sorting method, system and equipment for out-calling of unhealthy assets, and the method comprises the steps: carrying out the data cleaning and data standardization of a data set containing out-calling records of the unhealthy assets, carrying out the feature selection of the extracted information, and creating a composite feature through the feature combination; training one or more types of models by using the composite features, and performing model fusion based on a Stacking integration method to obtain fusion probability prediction; constructing a comprehensive sorting index based on fusion probability prediction; sequencing the outbound clients in batches according to the comprehensive sequencing index, and updating a weight coefficient in the comprehensive sequencing index based on an outbound feedback effect; and dynamically adjusting the weight of the training sample based on the outbound feedback data so as to update the model in an incremental learning mode. Through the scheme of the invention, the accuracy and efficiency of sorting can be improved.

Description

Technical Field

[0001] The present application relates to the field of electrical data processing, and in particular to a method, system and device for intelligent sorting of outbound calls for non-performing assets based on machine learning. Background Art

[0002] NPL outbound call sorting has become an important part of financial institutions' management and disposal of NPLs, which is closely related to improving asset recovery rates, optimizing resource allocation, and reducing operating costs. However, not all NPLs have the same recovery potential and urgency, which leads to the issue of distinguishing high-value targets from low-value targets. High-value targets are usually borrowers with higher repayment ability or willingness, while low-value targets may face more serious financial difficulties or lack repayment willingness. Therefore, accurately identifying and prioritizing high-value targets is crucial to improving the efficiency and effectiveness of NPL disposal. NPL outbound call sorting involves analyzing and evaluating a large amount of borrower data to determine the optimal outbound call sequence. This process can not only improve the collection success rate, but also significantly reduce unnecessary manpower and time costs, thereby optimizing overall resource allocation. At present, the outbound call sorting methods of NPL are mainly based on traditional credit scoring models, historical repayment record analysis, and simple rule-based sorting. These methods determine outbound call priority by evaluating factors such as the borrower's credit status, overdue time, and amount of debt. Although these methods are effective to a certain extent, they have obvious limitations, such as the inability to fully consider changes in the borrower's real-time financial situation, ignoring the impact of the external economic environment on repayment ability, and difficulty in capturing the borrower's behavioral patterns and psychological characteristics.

[0003] In recent years, with the rapid development of big data and artificial intelligence technologies, more advanced methods for outbound call sorting of non-performing assets have gradually attracted attention. Machine learning algorithms, especially deep learning models, have become a research focus because of their ability to process and analyze massive multidimensional data. These methods build a more comprehensive borrower profile by integrating multi-source data, including borrowers' social media activities, consumption behaviors, changes in employment status, etc. By learning patterns in historical data through complex algorithmic models, these methods can more accurately predict borrowers' repayment probability and optimal contact time. In actual application scenarios, due to the limited high-quality labeled data, model training often requires techniques such as data enhancement and transfer learning to improve the generalization ability of the model. However, existing methods still face challenges in dealing with data imbalance, feature engineering, and model interpretability. For example, traditional random oversampling or undersampling techniques may lead to model overfitting or information loss.

[0004] Therefore, there is an urgent need for a technical solution to improve the accuracy and efficiency of sorting. Summary of the invention

[0005] In order to solve the shortcomings of the prior art, the embodiments of the present application provide a method, system and device for intelligent sorting of non-performing asset outbound calls based on machine learning. The present application solves the technical problems that the accuracy and efficiency of the prior art cannot be taken into account at the same time.

[0006] The embodiment of the present application provides a method for intelligent sorting of outbound calls for non-performing assets based on machine learning, comprising: performing data cleaning and data standardization on a data set containing outbound call records of non-performing assets to extract information including time features, amount features, behavior features, outbound call features and / or credit features; performing feature selection on the extracted information, and creating composite features through feature combination; using the composite features to train one or more types of models, and performing model fusion based on the Stacking integration method to obtain a fusion probability prediction; wherein the model includes logistic regression, random forest, gradient boosting decision tree and / or support vector machine; constructing a comprehensive sorting index based on the fusion probability prediction; sorting outbound customers in batches according to the comprehensive sorting index, and updating the weight coefficient in the comprehensive sorting index based on the outbound call feedback effect; dynamically adjusting the weight of the training sample based on the outbound call feedback data to update the model through incremental learning.

[0007] In one possible implementation, the data set containing the outbound call records of non-performing assets is cleaned and standardized, including: removing duplicate data and unifying the data format; using the mean, median or mode filling for numerical features; using the mode filling or creating unknown categories for categorical features; using Outliers are identified using the principle or interquartile range method to be deleted or adjusted based on anomaly thresholds; numerical features are standardized.

[0008] In a possible implementation, feature selection is performed on the extracted information, and composite features are created by feature combination, including: calculating the correlation coefficient matrix between features to remove highly correlated features; calculating the variance between features to remove features with variances below a variance threshold; using a random forest model to sort the feature importance to select a target number of features; performing polynomial combination of numerical features and / or cross-combination of categorical features to construct composite features.

[0009] In a possible implementation, the method for obtaining the feature importance includes: ,in, Representation characteristics The importance of represents a node in a decision tree, Represents the child nodes on the node, Indicates that for all features The node on which the split occurs To sum, Representation Node The number of samples in represents the total number of samples, Representation Node The entropy of express Child nodes on a node The entropy of Represents the left and right child nodes, express Child nodes on a node The number of samples in .

[0010] In a possible implementation, a comprehensive ranking index is constructed based on fusion probability prediction, including: ,in, represents the comprehensive ranking index, represents the weight coefficient, represents the fused probability prediction of repayment, Indicates the overdue amount. Represents the customer value score.

[0011] In a possible implementation, outbound customers are sorted in batches according to the comprehensive sorting index, and the weight coefficient in the comprehensive sorting index is updated based on the outbound feedback effect, including: dividing the outbound customers into multiple batches based on the comprehensive sorting index; for each batch, the outbound customer with the highest comprehensive sorting index is selected using a first probability, and the outbound customer is randomly selected using a second probability; wherein the method for determining the first probability includes: ,in, represents the first probability, represents the initial probability, represents the attenuation coefficient, Indicates the current batch number; sets the corresponding prior distribution for each weight coefficient in the comprehensive ranking index, and updates the distribution parameters according to the feedback effect of each outbound call to adjust the weight coefficient.

[0012] In a possible implementation, the distribution parameter is updated according to the feedback effect of each outbound call, including: ,in, and Indicates The prior Beta distribution parameters of the weight coefficients, Indicates The success rate corresponding to the weight coefficient.

[0013] In a possible implementation, dynamically adjusting the weight of the training sample based on the outbound call feedback data includes: ,in, represents the adjusted sample weight, represents the sample weight before adjustment, represents the actual label, represents the predicted label, Represents the learning rate.

[0014] The embodiment of the present application also provides an intelligent sorting system for outbound calls of non-performing assets based on machine learning, including: a data pre-processing unit, a prediction unit and a sorting unit; wherein the data pre-processing unit is used to perform data cleaning and data standardization on a data set containing outbound call records of non-performing assets to extract information including time features, amount features, behavior features, outbound call features and / or credit features; feature selection is performed on the extracted information, and composite features are created by feature combination; the prediction unit is used to train one or more types of models using composite features, and perform model fusion based on the Stacking integration method to obtain a fusion probability prediction; the sorting unit is used to construct a comprehensive sorting index based on the fusion probability prediction; outbound call customers are sorted in batches according to the comprehensive sorting index, and the weight coefficient in the comprehensive sorting index is updated based on the outbound call feedback effect; the prediction unit is also used to dynamically adjust the weight of the training sample based on the outbound call feedback data to update the model by incremental learning.

[0015] An embodiment of the present application also provides an intelligent sorting device for outbound calls of non-performing assets based on machine learning, comprising: a processor, a memory, and a system bus; wherein the processor and the memory are connected via the system bus; the memory is used to store one or more programs, and the one or more programs include instructions, which, when executed by the processor, enable the processor to execute the method described in the above embodiment.

[0016] In the intelligent sorting method, system and device for outbound calls for non-performing assets provided above, the embodiments of the present application can achieve intelligent sorting of non-performing assets through multi-dimensional data analysis, dynamic feature engineering, multi-model fusion and real-time feedback optimization, which not only improves the outbound call efficiency and recovery rate, but also can continuously optimize according to real-time feedback, and has strong adaptability and scalability. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0018] Figure 1 A flowchart of a method for intelligent sorting of outbound calls for non-performing assets based on machine learning provided in an embodiment of the present application; Figure 2 A schematic block diagram of a non-performing asset outbound intelligent sorting system based on machine learning provided in an embodiment of the present application. DETAILED DESCRIPTION

[0019] Various exemplary embodiments of the present application will now be described in detail with reference to the accompanying drawings. It should be noted that unless otherwise specifically stated, the relative arrangement of components and steps, numerical expressions and numerical values ​​set forth in these embodiments do not limit the scope of the present application.

[0020] Those skilled in the art will appreciate that the terms "first", "second" and the like in the embodiments of the present application are only used to distinguish different steps, devices or modules, etc., and do not represent any specific technical meaning, nor do they represent the necessary logical order between them. It should also be understood that in the embodiments of the present application, "multiple" may refer to two or more, and "at least one" may refer to one, two or more. It should also be understood that for any component, data or structure mentioned in the embodiments of the present application, in the absence of explicit limitation or contrary revelation given in the context, it can generally be understood as one or more. In addition, the term "and / or" in the present application is only a kind of association relationship describing the associated objects, indicating that there can be three kinds of relationships, for example, A and / or B, which can represent: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " in the present application generally indicates that the associated objects before and after are an "or" relationship. It should also be understood that the description of each embodiment in the present application emphasizes the differences between the embodiments, and the same or similar parts can refer to each other. For the sake of brevity, they will not be repeated one by one.

[0021] At the same time, it should be understood that, for ease of description, the sizes of the various parts shown in the drawings are not drawn according to the actual proportional relationship. The following description of at least one exemplary embodiment is actually only illustrative and is by no means intended to limit the present application and its application or use. The techniques, methods and devices known to ordinary technicians in the relevant fields may not be discussed in detail, but where appropriate, the techniques, methods and devices should be considered as part of the specification. It should be noted that similar numbers 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 discussed in subsequent drawings.

[0022] In order to make the purpose, technical solution and advantages of the embodiments of the present application clearer, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0023] Figure 1 A flow chart of an intelligent sorting method for outbound calls of non-performing assets based on machine learning provided for an embodiment of the present application. It should be noted that the disposal of non-performing assets is one of the important challenges facing financial institutions. Traditional outbound call sorting methods for non-performing assets often rely on manual experience, which is inefficient and prone to deviations. The present application proposes an intelligent sorting method for outbound calls of non-performing assets based on machine learning, which realizes intelligent sorting of non-performing assets through multi-dimensional data analysis and model training, and improves outbound call efficiency and recovery rate.

[0024] like Figure 1 As shown, at step S01, data cleaning and data standardization are performed on a data set containing outbound call records of non-performing assets to extract information including time features, amount features, behavior features, outbound call features and / or credit features.

[0025] The datasets containing outbound call records of non-performing assets may include but are not limited to: Basic customer information: age, gender, occupation, income level, etc.; Loan information: loan amount, loan term, overdue period, overdue amount, etc.; Historical transaction data: repayment records, consumption habits, etc.; Credit score: internal score, external credit score, etc.; Contact information: phone number, address, etc.; Outbound call records: historical outbound call times, connection rate, call duration, etc.

[0026] After obtaining the data authorized by the customer, the data set containing the outbound call records of non-performing assets is cleaned and standardized. This includes: removing duplicate data and unifying the data format; using the mean, median or mode to fill in numerical features; using the mode to fill in categorical features or creating unknown categories; using The outliers are identified by the principle or interquartile range method, and deleted or adjusted according to the abnormal threshold. The abnormal threshold can be set by the relevant business personnel according to their business experience; the numerical features are standardized. In order to eliminate the dimensional differences between different features, the Z-score standardization method can be used to standardize the numerical features: , where x is the original value, μ is the mean, and σ is the standard deviation.

[0027] It should be understood that, for extracting information including time features, amount features, behavior features, outbound call features and / or credit features, the following content may be extracted: Time-related features: number of days overdue, time since last payment, remaining term of loan; Amount-related characteristics: ratio of overdue amount to total loan amount, ratio of average monthly repayment amount to income; Behavior-related characteristics: historical repayment times, historical overdue times, and longest continuous normal repayment periods; Outbound call related features: historical outbound call times, historical connection rate, average call duration; Credit-related characteristics: internal credit score, external credit score change trend.

[0028] At step S102, feature selection is performed on the extracted information, and composite features are created by feature combination. Specifically, the following steps are included: calculating the correlation coefficient matrix between features to remove highly correlated features; calculating the variance between features to remove features with variances below a variance threshold; using a random forest model to sort the importance of features to select a target number of features; performing polynomial combination on numerical features and / or cross-combination on categorical features to construct composite features.

[0029] The determination of the high correlation and variance thresholds can be flexibly adjusted according to business needs. The method for obtaining the importance of the features includes: ,in, Representation characteristics The importance of represents a node in a decision tree, Represents the child nodes on the node, Indicates that for all features The node on which the split occurs To sum, Representation Node The number of samples in represents the total number of samples, Representation Node The entropy of express Child nodes on a node The entropy of Represents the left and right child nodes, express Child nodes on a node The number of samples in .

[0030] The importance of a feature is measured by calculating the average reduction in impurity (such as Gini impurity or entropy) when the feature is used as a split node in all trees. For example, the calculation steps may include the following: For each tree in the random forest, traverse each internal node in the tree, if the node uses the feature Perform the split, calculate the change in impurity before and after the split, and multiply the change by the node weight ( ), and added to the importance score of the feature. The results of all trees are averaged to get the final feature importance score.

[0031] The specific implementation can be achieved by training a random forest model, traversing all internal nodes for each tree, and calculating the change in impurity before and after the split for each node. According to the split features, the importance scores are accumulated. The importance scores of all features are normalized. The features are sorted according to the importance scores. Finally, the top-k features are selected.

[0032] In step S103, one or more models are trained using composite features, and the models are fused based on the Stacking integration method to obtain a fusion probability prediction.

[0033] In one embodiment, the following four different types of models can be selected as basic models: logistic regression (LR), random forest (RF), gradient boosted decision tree (GBDT), support vector machine (SVM). Each basic model is trained and the k-fold cross validation method is used to evaluate the model performance. Taking logistic regression as an example, its objective function is: , where m is the number of samples and n is the number of features. is the prediction function, and λ is the regularization parameter.

[0034] Next, the Stacking integration method is used for model fusion: the prediction results of the basic model are used as new features, and a meta-model (such as LightGBM) is trained as the final classifier. The prediction probability calculation formula after fusion is: ,in, is the prediction function of the meta-model, are the predicted probabilities of each basic model.

[0035] In step S104, a comprehensive ranking index is constructed based on the fusion probability prediction. In one implementation scenario, it can be constructed as: ,in, represents the comprehensive ranking index, represents the weight coefficient, represents the fused probability prediction of repayment, Indicates the overdue amount. Represents the customer value score.

[0036] In an implementation scenario, R can be subdivided into several objective and quantifiable indicators. For example, it can include but is not limited to the following factors: historical repayment records, current economic conditions, asset conditions, income stability, and communication and cooperation with the company. The quantification scheme of the indicators can be set as: Historical repayment record (maximum 100 points), number of on-time repayments in the past 12 months: 5 points each time, maximum 60 points, time since the most recent overdue payment: within 3 months (0 points), 3-6 months (10 points), 6-12 months (20 points), 12 months and above (40 points).

[0037] Current economic situation (out of 100 points), employment status: full-time (50 points), part-time (30 points), unemployed (0 points). Monthly income level: <3,000 yuan (10 points), 3,000-5,000 yuan (20 points), 5,000-10,000 yuan (30 points), 10,000 yuan and above (50 points).

[0038] Asset status (out of 100 points), real estate: own real estate (50 points), no own real estate (0 points). Vehicles: own car (30 points), no car (0 points). Other convertible assets value: <10,000 yuan (5 points), 10,000-50,000 yuan (10 points), 50,000 yuan and above (20 points).

[0039] Income stability (full score 100 points), current working years: <1 year (10), 1-3 years (20), 3-5 years (30), 5 years and above (40). Occupation type: civil servant / public institution (30), state-owned enterprise employee (25), private enterprise formal employee (20), freelance (10), other (5). Whether there is a side job: yes (30), no (0).

[0040] Communication and cooperation (full score 100 points), answering rate: 80% (40 points), 60-80% (30 points), 40-60% (20 points), <40% (10 points). Attitude: active cooperation (40), general (20), passive resistance (0). Whether to actively provide a repayment plan: yes (20), no (0).

[0041] Calculation of customer value score R: R = (b1 * historical repayment record + b2 * current economic situation + b3 * asset situation + b4 * income stability + b5 * communication and cooperation) / (b1 + b2 + b3 + b4 + b5). Among them, the weights of b1 to b5 can be adjusted according to actual conditions, for example: b1 = 0.3, b2 = 0.2, b3 = 0.15, b4 = 0.2, b5 = 0.15. In the end, the score range of R is 0-100 points.

[0042] This quantitative scheme can be adjusted according to actual conditions. For example, some sub-items can be added or deleted, or the score distribution of each item can be adjusted. At the same time, the score can be evaluated and adjusted regularly (such as quarterly) according to the actual collection effect to ensure its continued effectiveness.

[0043] At step S105, outbound customers are sorted in batches according to the comprehensive sorting index, and the weight coefficient in the comprehensive sorting index is updated based on the outbound feedback effect. This includes: dividing the outbound customers into multiple batches based on the comprehensive sorting index; for each batch, the outbound customer with the highest comprehensive sorting index is selected using a first probability, and the outbound customer is randomly selected using a second probability; wherein the method for determining the first probability includes: ,in, represents the first probability, represents the initial probability, represents the attenuation coefficient, Indicates the current batch number; sets the corresponding prior distribution for each weight coefficient in the comprehensive ranking index, and updates the distribution parameters according to the feedback effect of each outbound call to adjust the weight coefficient.

[0044] Specifically, after making outbound calls in batches, a prior distribution (such as Beta distribution) is set for each weight coefficient in the comprehensive ranking index. After each outbound call, the distribution parameters are updated according to the actual effect, and new weight coefficients are sampled from the updated distribution. Among them, the distribution parameters are updated according to the feedback effect of each outbound call, including: ,in, and Indicates The prior Beta distribution parameters of the weight coefficients, Indicates The success rate corresponding to the weight coefficients. For example, in the above-mentioned embodiment, The three weight coefficients are updated.

[0045] In step S106, the weight of the training sample is dynamically adjusted based on the outbound call feedback data to update the model through incremental learning. The weight of the training sample is dynamically adjusted based on the outbound call feedback data, including: ,in, represents the adjusted sample weight, Indicates the sample weight before adjustment, which can be initialized in the range of 0.75 to 1.25. represents the actual label, represents the predicted label, Indicates the learning rate. At the same time, online learning algorithms (such as Online Gradient Descent) can be used to incrementally update the model: ,in, For the model parameters, η is the learning rate, L is the loss function, and are the features and labels of new samples.

[0046] In addition, in order to be able to regularly re-evaluate the importance of features, the feature set can be dynamically adjusted: the SHAP (SHapley Additive exPlanations) value of each feature is calculated, the features are ranked according to the SHAP value, the top k features are retained, and low-importance features are removed or merged. The SHAP value calculation formula is: , where N is the set of all features, S is the subset that does not contain feature i, is the predicted value of the model on the feature subset S.

[0047] In summary, the embodiments of the present application comprehensively characterize customer characteristics through multi-dimensional data fusion, combined with multi-dimensional data such as basic customer information, loan information, transaction data, credit score, etc. Through feature extraction, selection and combination, a rich feature set is constructed, and the feature importance is dynamically adjusted according to feedback. The Stacking integration method is adopted to integrate different types of machine learning models to improve prediction accuracy. Combined with the dynamic sorting strategy, a balance between exploration and utilization is achieved, and the sorting weight is dynamically adjusted. Through sample weight update, incremental learning and dynamic evaluation of feature importance, continuous optimization of the model is achieved. In short, the intelligent sorting method for outbound calls of non-performing assets based on machine learning proposed in the present application scheme realizes intelligent sorting of non-performing assets through multi-dimensional data analysis, dynamic feature engineering, multi-model fusion and real-time feedback optimization. This method not only improves the efficiency and recovery rate of outbound calls, but also can be continuously optimized according to real-time feedback, and has strong adaptability and scalability.

[0048] Figure 2 A schematic block diagram of a non-performing asset outbound intelligent sorting system based on machine learning provided in an embodiment of the present application. It should be understood that the system shown in the figure is exemplary and not restrictive. This means that the system architecture involved is not limited to a specific form or design, but is presented as an example. In other words, the architecture shown in the figure can be regarded as a way of expression to clearly describe related concepts and relationships, and does not exclude other forms of architecture. Therefore, when interpreting the architecture in the picture, it should be understood that the model is flexible and diverse, and its purpose is to provide an exemplary description rather than a restrictive provision on a specific form.

[0049] The embodiment of the present application includes a data pre-processing unit 201, a prediction unit 202 and a sorting unit 203; wherein the data pre-processing unit 201 is used to clean and standardize a data set containing outbound call records of non-performing assets to extract information including time features, amount features, behavior features, outbound call features and / or credit features; feature selection is performed on the extracted information, and composite features are created by feature combination; the prediction unit 202 is used to train one or more models using composite features, and to fuse the models based on the Stacking integration method to obtain a fusion probability prediction; the sorting unit 203 is used to construct a comprehensive sorting index based on the fusion probability prediction; outbound call customers are sorted in batches according to the comprehensive sorting index, and the weight coefficient in the comprehensive sorting index is updated based on the outbound call feedback effect; the prediction unit 202 is also used to dynamically adjust the weight of the training sample based on the outbound call feedback data to update the model by incremental learning.

[0050] Furthermore, an embodiment of the present application also provides an intelligent sorting device for outbound calls of non-performing assets based on machine learning, including: a processor, a memory, and a system bus; the processor and the memory are connected via the system bus; the memory is used to store one or more programs, and the one or more programs include instructions, which, when executed by the processor, enable the processor to execute any of the above methods.

[0051] Furthermore, an embodiment of the present application also provides a computer program product, which, when executed on a terminal device, enables the terminal device to execute any of the above methods.

[0052] It can be known from the description of the above implementation mode that those skilled in the art can clearly understand that all or part of the steps in the above-mentioned embodiment method can be implemented by means of software plus a necessary general hardware platform. Based on such an understanding, the technical solution of the present application can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product can be stored in a storage medium such as ROM / RAM, a disk, an optical disk, etc., including several instructions for enabling a computer device (which can be a personal computer, a server, or a network communication device such as a media gateway, etc.) to execute the methods described in the various embodiments of the present application or certain parts of the embodiments.

[0053] It should be noted that the various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments, and the same or similar parts between the various embodiments can be referred to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part description.

[0054] It should also be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the presence of other identical elements in the process, method, article or device including the elements.

[0055] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. The intelligent sorting method for outbound calls of non-performing assets based on machine learning is characterized by: include: Perform data cleaning and data standardization on the dataset containing outbound call records of non-performing assets to extract information including time characteristics, amount characteristics, behavior characteristics, outbound call characteristics and / or credit characteristics; Perform feature selection on the extracted information and create composite features through feature combination; Use composite features to train one or more models, and perform model fusion based on the Stacking ensemble method to obtain fusion probability prediction; wherein the model comprises logistic regression, random forest, gradient boosted decision tree and / or support vector machine; Constructing a comprehensive ranking index based on fusion probability prediction; Sort outbound customers in batches according to the comprehensive ranking index, and update the weight coefficient in the comprehensive ranking index based on the outbound call feedback effect; The weights of training samples are dynamically adjusted based on outbound call feedback data to update the model through incremental learning.

2. The intelligent sorting method according to claim 1, characterized in that: in, Data cleaning and data standardization of the dataset containing outbound call records of non-performing assets, including: Remove duplicate data and unify data formats; For numerical features, use mean, median or mode filling; For categorical features, use the majority to fill or create unknown categories; use Principle or interquartile range method to identify outliers for deletion or adjustment based on outlier thresholds; Normalize numerical features.

3. The intelligent sorting method according to claim 1, characterized in that: in, Perform feature selection on the extracted information and create composite features through feature combination, including: Calculate the correlation coefficient matrix between features to remove highly correlated features; Calculate the variance between features to remove features whose variance is lower than the variance threshold; Ranking feature importance using a random forest model to select the target number of features; Composite features are constructed by performing polynomial combinations on numerical features and / or cross combinations on categorical features.

4. The intelligent sorting method according to claim 3, characterized in that: The method for obtaining the feature importance includes: ， in, Representation characteristics The importance of represents a node in a decision tree, Represents the child nodes on the node, Indicates that for all features The node on which the split occurs To sum, Representation Node The number of samples in represents the total number of samples, Representation Node The entropy of express Child nodes on a node The entropy of Represents the left and right child nodes, express Child nodes on a node The number of samples in .

5. The intelligent sorting method according to claim 1, characterized in that: in, Based on the fusion probability prediction, a comprehensive ranking index is constructed, including: ， in, represents the comprehensive ranking index, represents the weight coefficient, represents the fused probability prediction of repayment, Indicates the overdue amount. Represents the customer value score.

6. The intelligent sorting method according to claim 1, characterized in that: in, Outbound customers are sorted in batches according to the comprehensive sorting index, and the weight coefficients in the comprehensive sorting index are updated based on the outbound call feedback effect, including: Divide outbound customers into multiple batches based on comprehensive ranking indicators; For each batch, the outbound customer with the highest comprehensive ranking index is selected using the first probability, and the outbound customer is randomly selected using the second probability; wherein the method for determining the first probability includes: ， in, represents the first probability, represents the initial probability, represents the attenuation coefficient, Indicates the current batch number; A corresponding prior distribution is set for each weight coefficient in the comprehensive ranking index, and the distribution parameters are updated according to the feedback effect of each outbound call to adjust the weight coefficient.

7. The intelligent sorting method according to claim 6, characterized in that: in, Update distribution parameters based on the feedback effect of each outbound call, including: ， in, and Indicates The prior Beta distribution parameters of the weight coefficients, Indicates The success rate corresponding to the weight coefficient.

8. The intelligent sorting method according to claim 1, characterized in that: in, Dynamically adjust the weights of training samples based on outbound call feedback data, including: ， in, represents the adjusted sample weight, represents the sample weight before adjustment, represents the actual label, represents the predicted label, Represents the learning rate.

9. The intelligent outbound sorting system for non-performing assets based on machine learning is characterized by: include: Data pre-processing unit, prediction unit and sorting unit; among them, The data pre-processing unit is used to perform data cleaning and data standardization on the data set containing the outbound call records of non-performing assets to extract information including time features, amount features, behavior features, outbound call features and / or credit features; perform feature selection on the extracted information, and create composite features through feature combination; The prediction unit is used to train one or more models using composite features, and perform model fusion based on the Stacking integration method to obtain a fusion probability prediction; The ranking unit is used to construct a comprehensive ranking index based on the fusion probability prediction; sort the outbound customers in batches according to the comprehensive ranking index, and update the weight coefficient in the comprehensive ranking index based on the outbound call feedback effect; The prediction unit is also used to dynamically adjust the weight of the training samples based on the outbound call feedback data, so as to update the model through incremental learning.

10. Intelligent outbound sorting equipment for non-performing assets based on machine learning, characterized in that: include: A processor, a memory, and a system bus; wherein the processor and the memory are connected via the system bus; The memory is used to store one or more programs, wherein the one or more programs include instructions, and when the instructions are executed by the processor, the processor executes the method according to any one of claims 1 to 8.

Images