Risk user identification method and system for railway ticketing system

By adopting a risk user identification method based on differential privacy in the railway ticketing system, combining homomorphic encryption and differential privacy algorithms, a risk user identification model of the railway ticketing system was constructed, which solved the problem of user data leakage risk, and realized the data and Improved model security and recognition effect.

CN119089316APending Publication Date: 2024-12-06TIELV TECH CO LTD +1
View PDF 7 Cites 0 Cited by

Patent Information

Application Number
CN202411113102.4
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-14
Publication Date
2024-12-06

AI Technical Summary

Technical Problem

The prior art lacks a risk user identification method based on a privacy protection model for railway ticketing systems, resulting in the risk of user data leakage, making it difficult to effectively identify risk users while ensuring data security.

Method used

A risk user identification method based on differential privacy is adopted, and through feature construction, encryption and model construction steps, combined with homomorphic encryption and differential privacy algorithms, a risk user identification model for the railway ticketing system is built to ensure the security of data and models.

Benefits of technology

It effectively reduces the risk of user data leakage, improves the security of multi-source data building risk user identification models, ensures the protection of user personal information, and improves the accuracy of risk user identification.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119089316A_ABST
    Figure CN119089316A_ABST
Patent Text Reader

Abstract

The invention discloses a risk user identification method and system for a railway ticketing system, and the method comprises the steps: constructing ticket buying user features and generating ticket buying user risk tags for the multi-source sample data of the railway ticketing system; after privacy information of the risk tag is encrypted by adopting an encryption method, ticket buying user features are screened by adopting a feature screening algorithm, and the screened ticket buying user features are fused to generate a user training data set; using the user training data set to construct a risk user identification model based on differential privacy; and inputting a to-be-identified ticket buying user into the risk user identification model, and outputting a prediction identification result of the risk user. According to the risk user identification method oriented to the railway ticketing system, the personal information of the ticket buying user can be protected in the whole process from the data set to the model, and the safety of the data of the ticket buying user is also protected while the risk user identification service requirement is met.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of user behavior characteristics, and in particular to a risk user identification method and system for a railway ticketing system. Background Art

[0002] At present, the railway Internet ticketing system is the main ticketing channel for passenger transportation. With the increase in passengers' willingness to travel, some lines are undercapacitated. Third-party platforms use the contradiction between supply and demand to launch various ticket grabbing services, which has brought great pressure to the Internet ticketing system. In order to ensure the stability of the system and the fairness of passengers' ticket purchases, it is urgent to identify risky users and protect the ticket purchase rights of normal users. In order to better identify risky users, it is necessary to collect user behavior trajectory information, but in the process of data use, sharing, and analysis, there is a risk of leakage of user personal privacy information. In order to meet the requirements of data security, it is necessary to strengthen data security protection from two aspects: one is the protection of data sets, and the original data information cannot be leaked; the other is the protection of statistical information. In order to analyze the characteristics of risky users, it is necessary to collect a large amount of user information, which inevitably leads to two problems.

[0003] With the advent of the data intelligence era, data-based risk prevention and control has become more popular. The use of multi-source data is conducive to improving the effectiveness of risk user identification models, but the use of multi-party data assets poses the risk of privacy information leakage. Currently, user data sets can be processed in an encrypted and anonymized manner, while the corresponding statistical information is mostly protected by noise perturbation.

[0004] The risk user identification model can use the widely used random forest algorithm. As an integrated algorithm, random forest has a good classification effect, but its nodes may leak the user's statistical information. A new privacy protection model, the differential privacy model, is proposed in the prior art. This model can achieve a high level of privacy protection with minimal noise.

[0005] However, there is still a lack of research on risky user identification methods based on privacy protection models for railway ticketing systems. Therefore, it is urgent to propose a risky user identification method based on a privacy protection model, and use differential privacy to improve the security of the random forest algorithm to reduce the risk of data information leakage, so as to protect user personal information from the entire process from data set to model, and protect the security of user data while meeting the business needs of identifying risky users. Summary of the invention

[0006] In order to solve the defects of the above-mentioned prior art, the embodiment of the present application provides a new method for identifying risky users in a railway ticketing system to reduce the risk of user data information leakage.

[0007] In a first aspect, an embodiment of the present application provides a method for identifying risky users in a railway ticketing system, the method comprising:

[0008] Feature construction steps: construct the characteristics of ticket purchasing users and generate risk labels for ticket purchasing users based on the multi-source sample data of the railway ticketing system. The multi-source sample data includes: attribute characteristics, access characteristics and travel characteristics of ticket purchasing users;

[0009] Feature screening and fusion steps: Based on the risk label of the ticket purchasing user, the privacy information of the user label is encrypted by an encryption method, and then the features of the ticket purchasing user are screened by a feature screening algorithm. The screened features of the ticket purchasing user are fused to generate a user training data set;

[0010] Model building steps: Use user training datasets to build a risk user identification model based on differential privacy;

[0011] Risky user identification step: input the ticket purchasing user to be identified into the risky user identification model and output the user's risk prediction result.

[0012] As an optional implementation manner, in the first aspect of the present invention, the feature construction step further includes:

[0013] Steps for constructing third-party data features: Use third-party data features to supplement the characteristics of ticket purchasing users. User characteristics include: basic characteristics of ticket purchasing users and third-party data characteristics.

[0014] As an optional implementation, in the first aspect of the present invention, the feature construction step further includes:

[0015] Attribute feature construction step: extract risk user identification features based on the identity information registered by the ticket purchasing user;

[0016] Access feature construction step: extract risk user identification features based on at least one of the device environment information used by the ticket purchasing user to access, the content features accessed by the user, and the time features of the user's access;

[0017] Travel feature construction steps: Extract risk user identification features based on the travel-related behavioral characteristics of ticket purchasing users.

[0018] As an optional implementation, in the first aspect of the present invention, the feature screening and fusion steps include:

[0019] Feature screening step: After encrypting the risk label of the ticket purchasing user using the homomorphic encryption algorithm, the information value IV of the attribute feature, access feature and travel feature is calculated respectively to complete the screening of user features;

[0020] Feature fusion step: after feature screening, the user features are fused to form a user data set, which includes user features and user risk labels.

[0021] As an optional implementation, in the first aspect of the present invention, the feature screening step comprises:

[0022] Label encryption step: Encrypt the risk label of each ticket purchasing user using a homomorphic encryption algorithm;

[0023] IV value calculation steps: calculate the IV value for each user group of ticket purchasing users, sort and filter the IV values ​​respectively, and output the minimum data set of valid feature data.

[0024] As an optional implementation, in the first aspect of the present invention, the model building step further includes:

[0025] The random forest algorithm is used to build the model, and the differential privacy algorithm is combined to improve the security of the model. The risk user identification model uses the CART decision tree as a single decision tree of the random forest, selects the split value of the continuity attribute according to the quantile, uses the Laplace mechanism to add noise to the sample set and leaf nodes, and selects the best split attribute through the exponential mechanism. The evaluation function of the exponential mechanism selects the Gini index.

[0026] In a second aspect, an embodiment of the present application provides a risk user identification system for a railway ticketing system, using the risk user identification method for a railway ticketing system as described above, the system includes:

[0027] Data providing module: used to provide multi-source sample data of the railway ticketing system, build the characteristics of ticket purchasing users based on the multi-source sample data and generate risk labels for ticket purchasing users, wherein the multi-source sample data includes: data on attribute characteristics, access characteristics and travel characteristics of ticket purchasing users;

[0028] Label ownership module: It is used to encrypt the privacy information of user labels based on the risk labels of ticket purchasing users by using encryption methods, and then send it to the data providing module. It uses feature screening algorithms to screen the features of ticket purchasing users, and fuses the screened features of ticket purchasing users to generate a user training data set.

[0029] Model building module: used to build a risk user identification model based on differential privacy using user training datasets;

[0030] Risky user identification module: used to input the ticket purchasing users to be identified into the risky user identification model and output the risk prediction results of the users.

[0031] As an optional implementation, in the second aspect of the present invention, the data providing module further includes:

[0032] Third-party data providing module: used to supplement the characteristics of ticket purchasing users with third-party data characteristics. User characteristics include: basic characteristics of ticket purchasing users and third-party data characteristics.

[0033] In a third aspect, an embodiment of the present application provides an electronic device, comprising: a memory storing executable program code; a processor coupled to the memory; the processor calls the executable program code stored in the memory to execute the risk user identification method for the railway ticketing system as described above.

[0034] In a fourth aspect, an embodiment of the present application provides a computer storage medium, wherein the computer storage medium stores computer instructions, and when the computer instructions are called, are used to execute the risk user identification method for the railway ticketing system as described above.

[0035] Compared with the related prior art, it has the following outstanding beneficial effects:

[0036] 1. The present invention supports the risk user identification model training of multi-source data, and uses the IV algorithm to pre-screen the data features of multiple parties without integrating all the features of multiple parties;

[0037] 2. The present invention combines third-party data to construct basic features of users, enriching the training features of ticket purchasing users;

[0038] 3. The present invention de-identifies important data and uses homomorphic encryption technology to calculate IV values, thereby protecting the security of ticket purchase user data;

[0039] 4. The present invention uses a random forest classification model combined with differential privacy protection to protect the security of the model;

[0040] 5. Security protection is added to the data set and model of the present invention, which improves the overall security of building risk user identification models using multi-source data. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0042] Figure 1 This is a schematic diagram of a risk user identification method for a railway ticketing system according to an embodiment of the present invention;

[0043] Figure 2 This is a schematic diagram of a risk user identification method for a railway ticketing system according to another embodiment of the present invention;

[0044] Figure 3 A schematic diagram of a process flow diagram for constructing features of an embodiment of the present invention;

[0045] Figure 4 This is a schematic diagram of the homomorphic encryption method according to an embodiment of the present invention;

[0046] Figure 5 This is a schematic diagram of a risk user identification system for a railway ticketing system according to an embodiment of the present invention;

[0047] Figure 6 Schematic diagram of computer hardware of the present invention.

[0048] In the above picture:

[0049] 101 data providing module 102 label owning module

[0050] 103 Model building module 104 Risk user identification module. DETAILED DESCRIPTION

[0051] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0052] The terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish different objects rather than to describe a specific order. In addition, the term "and / or" in this article is only a description of the association relationship between related objects, indicating that three relationships may exist. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone, where A and B can be singular or plural. In addition, the character " / " in this article generally indicates that the previous and next related objects are in an "or" relationship, but it may also indicate an "and / or" relationship. Please refer to the previous and next context for specific understanding.

[0053] In the present invention, "at least one" means one or more, and "more than one" means two or more. "At least one of the following" or similar expressions refers to any combination of these items, including any combination of single or plural items. For example, at least one of a, b, or c can be represented by: a, b, c, ab, ac, bc, or abc, where a, b, c can be single or multiple.

[0054] It should also be understood that in various embodiments of the present invention, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0055] In the several embodiments provided by the present invention, it should be understood that the disclosed devices, apparatuses and methods can be implemented in other ways. For example, the device embodiments described above are only schematic, for example, the division of the units is only a logical function division, and there may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another device, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0056] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0057] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0058] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0059] In order to make the above features and effects of the present invention more clearly and understandably described, embodiments are given below and described in detail with reference to the accompanying drawings. This specification discloses one or more embodiments that include the features of the present invention. The disclosed embodiments are only for illustration. The scope of protection of the present invention is not limited to the disclosed embodiments, and the present invention is defined by the attached claims.

[0060] The following is a system embodiment corresponding to the above method embodiment. This embodiment can be implemented in conjunction with the above embodiment. The relevant technical details mentioned in the above embodiment are still valid in this embodiment. In order to reduce repetition, they are not repeated here. Accordingly, the relevant technical details mentioned in this embodiment can also be applied in the above embodiment.

[0061] The present invention aims to provide a risk user identification method for a railway ticketing system using railway multi-source data. First, the basic features of users are constructed by combining data from multiple systems with third-party data. Then, the model builder filters the features of different data providers based on the labels of the user group. In this way, the model builder only needs to obtain effective features without seeing all the feature data of the data provider, which can minimize the shared data. In the process of feature screening, homomorphic encryption is used to protect the user's label from being known by the user provider to prevent the leakage of user label information. Finally, the user data set after multi-party data fusion is used to train a random forest model based on differential privacy. Noise is added to the model through the Laplace mechanism and the exponential mechanism, which further improves the security of the training model. The following is a detailed description in conjunction with specific embodiments.

[0062] Embodiment 1

[0063] See also Figure 1 , Figure 1 This is a flow chart of a method for identifying risky users in a railway ticketing system according to an embodiment of the present invention. Figure 1 The risk user identification method for the railway ticketing system can be applied to the embodiment of the present application to provide a risk user identification system for the railway ticketing system, and can also be applied to electronic devices, which is not limited by the embodiment of the present invention. The risk user identification method for the railway ticketing system includes:

[0064] S101, feature construction step: constructing features of ticket purchasing users and generating risk labels for ticket purchasing users based on multi-source sample data of the railway ticketing system, wherein the multi-source sample data includes one or more of attribute features, access features, and travel features of ticket purchasing users, which are not specifically limited in the embodiment of the present invention;

[0065] In the embodiment of the present invention, optionally, the multi-source sample data of the railway ticketing system can be acquired in real time or at a fixed time, which is not specifically limited in the embodiment of the present invention;

[0066] The feature construction step S101 further includes:

[0067] Third-party data feature construction step: third-party data features are used to supplement the ticket purchasing user features. The user features include: one or more of the basic features of the ticket purchasing user and the third-party data features, which are not specifically limited in the embodiment of the present invention.

[0068] The feature construction step S101 further includes:

[0069] Attribute feature construction step: extract risk user identification features based on the identity information registered by the ticket purchasing user;

[0070] Access feature construction step: extract risk user identification features based on at least one of the device environment information used by the ticket purchasing user to access, the content features accessed by the user, and the time features of the user's access;

[0071] Travel feature construction steps: Extract risk user identification features based on the travel-related behavioral characteristics of ticket purchasing users.

[0072] S102, feature screening and fusion step: based on the risk label of the ticket purchasing user, after encrypting the privacy information of the user label by using an encryption method, the features of the ticket purchasing user are screened by using a feature screening algorithm, and the screened features of the ticket purchasing user are fused to generate a user training data set;

[0073] The feature screening and fusion step S102 includes:

[0074] Feature screening step: After encrypting the risk label of the ticket purchasing user using the homomorphic encryption algorithm, calculate the information value IV of the attribute characteristics, access characteristics and travel characteristics respectively to complete the screening of user characteristics; the homomorphic encryption algorithm is used in the embodiment of the present invention, but other encryption algorithms can also be used, and the embodiment of the present invention does not make specific limitations.

[0075] Feature fusion step: after feature screening, the user features are fused to form a user data set, which includes user features and user risk labels.

[0076] As an optional implementation, the feature screening step includes:

[0077] Label encryption step: Encrypt the risk label of each Internet ticket purchasing user using a homomorphic encryption algorithm;

[0078] IV value calculation steps: calculate the IV value for each user group based on the user characteristics, sort and filter the IV values, and output the minimum data set of valid feature data.

[0079] S103, model building step: using the user training data set to build a risk user identification model based on differential privacy;

[0080] The model building step S103 further includes:

[0081] The random forest algorithm is used to build the model, and the differential privacy algorithm is combined to improve the security of the model. The risk user identification model uses the CART decision tree as a single decision tree of the random forest, selects the split value of the continuity attribute according to the quantile, uses the Laplace mechanism to add noise to the sample set and leaf nodes, and selects the best split attribute through the exponential mechanism. The evaluation function of the exponential mechanism selects the Gini index.

[0082] S104, risky user identification step: input the ticket purchasing user to be identified into the risky user identification model, and output the user's risk prediction result.

[0083] Embodiment 2

[0084] See also Figure 2 , Figure 2 This is a flow chart of another method for identifying risky users in a railway ticketing system according to an embodiment of the present invention. Figure 2 The method for identifying risky users in a railway ticketing system can be applied to an embodiment of the present application that provides a system for identifying risky users in a railway ticketing system, and can also be applied to electronic devices, which is not limited by the embodiment of the present invention.

[0085] The overall process of the method provided in the embodiment of the present application is as follows Figure 2 As shown in the figure, it is divided into three parts: secure construction of user basic features, feature screening based on homomorphic encryption, and differential privacy classification model construction. The basic features of users are enriched based on third-party data, and then features with strong classification capabilities are screened using IV technology. Finally, the multi-party data features are fused and input into the random forest algorithm protected by differential privacy to train the risk user identification model. Figure 3 As shown:

[0086] S201. Basic feature construction steps

[0087] The user feature construction is mainly divided into three parts: attribute features, access features, and travel features.

[0088] The user attribute characteristics and access characteristics of ticket purchasing come from the railway Internet ticketing system, and the travel characteristics come from the travel service system. Different systems are managed by different departments, and there are risks in sharing data between different departments.

[0089] Attribute features: User name features, age distribution features, location features, etc. can all be used as attribute features. User names are composed of uppercase and lowercase letters, numbers, and characters, and different pattern combination features can be extracted.

[0090] Access features: mainly divided into three parts: device environment information used by users when accessing, such as IP, WiFi, etc.; content features of user access, such as various business requests such as user login and ticket query, combined with time dimension and device information, can form risk user identification features of different granularity ranges; a series of user requests will form an access sequence, and then extract sequence-related features.

[0091] Travel characteristics: Train tickets are only one part of a passenger's journey. Currently, the railway Internet ticketing system also provides passengers with insurance, hotel, car-hailing and other services. The use of various services can also be used as a travel characteristic of passengers.

[0092] Third-party data is mainly used to supplement the information corresponding to the user's access characteristics, such as the network attributes of the IP address used by the user when accessing, the region to which it belongs, the type corresponding to the user's mobile phone number, and other information. Third-party data is mainly used to enrich the user's access characteristics.

[0093] The user's access characteristics and travel characteristics can expose the user's location and travel characteristics. In order to strengthen data security protection, the departure city, arrival city, and hotel will be encoded and anonymized to prevent direct access to the user's travel information in the event of a data leak.

[0094] S202. Feature screening steps based on homomorphic encryption

[0095] The classification and discrimination capabilities of variables are screened based on the IV (Information Value) information value of each variable. First, the calculation of the IV value is based on WOE (weight of Evidence). The user label is a label data set, and WOE encoding is performed for a variable. The features formed by the user include continuous features and discrete features. Discrete features are binned by feature values. Continuous feature attribute values ​​with a number less than 50 are directly grouped according to the feature value. Feature values ​​greater than 50 are divided into intervals according to quantiles and divided into 50 interval groups. After the variables are grouped, for the i-th group, the calculation formula of WOE is as follows:

[0096]

[0097] Where: P y1 is the proportion of risky users to all risky users, P y0 The proportion of normal users to all normal users; B is a risk user, B i is the number of risky users corresponding to the i-th group, G is the number of normal users, Gi is the number of normal users corresponding to the i-th group.

[0098] The positive or negative value of the current group WOE is determined by the ratio of risky users and normal users in the current group and the ratio of risky users and normal users in the whole group. When the ratio of the current group is smaller than the overall ratio of the sample, the WOE is negative, which has a negative impact; when the ratio of the current group is larger than the overall ratio, the WOE is positive, which is used to judge whether the individual will have a positive impact; when the ratio of the current group is equal to the overall ratio, the WOE is 0, so the value range of WOE is all real numbers. WOE does not take into account the proportion of samples in the group to the overall sample. If the WOE value of a group is very high, but the number of samples accounts for a very low proportion of the overall sample, the ability to predict the variable as a whole will decrease. Therefore, we also need to calculate IV.

[0099] The IV value is used to measure the predictive power of the independent variable. The calculation of IV is based on WOE, which is the weighted sum of WOE. The IV value calculation of group i is:

[0100]

[0101] The overall IV corresponding to this variable is:

[0102]

[0103] IV is multiplied by a weight based on WOE This ensures that the result of each grouping of the variable is non-negative, and can avoid the situation where the number of samples in a group is very small, but B i / G i The situation where the WOE is very large.

[0104] The model builder cannot synchronize the labels of the sample data set to the feature data provider to avoid leakage of the user's risk identification. Therefore, the feature screening of the data provider's ticketing system and travel service system needs to be completed using homomorphic encryption technology.

[0105] In the sample data set, user behavior features are extracted based on user attribute features, access features, and travel features. The user labels are first identified: risky users are 1, normal users are 0, and then the labels are encrypted. In the ciphertext state, the feature screening of the railway Internet ticketing system and travel service system is completed. The overall calculation process is as follows: Figure 4 As shown:

[0106] Step 1: Label y i After encryption, it is sent to the data provider synchronously, and {us i ,encry(y i ),Encry(1-y i )}; Among them, usi represents user i, encry(y i ) represents the positive label y i Encrypted value; Encry(1-y i ) represents the negative label y i The encrypted value of

[0107] Step 2: Each group is calculated according to the following formula {Encode(user_set i ),sum(Encry(y i )),sum(Encry(1-y i ))}, the labels corresponding to each sample in the group are summed and then sent to the label owner; where user_set i represents each user group i;

[0108] Step 3: On the label owner side, decrypt and calculate the number of positive and negative samples corresponding to each group. The formula is as follows:

[0109] npos i =Decry(sum(Encry(y i )))

[0110] nneg i =Decry(sum(Encry(1-y i )))

[0111] Step 4: The tag owner calculates the IV value of the variable.

[0112]

[0113]

[0114]

[0115] Among them, npos i Indicates the number of positive samples corresponding to each group i; nneg i Indicates the number of negative samples corresponding to each group i, pos total Represents the sum of the number of positive samples corresponding to all groups; neg total represents the sum of the number of negative samples corresponding to all groups; woe i represents the weight of evidence for the ith group.

[0116] Step 5: The label owner calculates the IV values ​​for the characteristic variables of the railway Internet ticketing system and the travel service system, and synchronizes the calculated results to the data provider. The railway Internet ticketing system and the travel service system sort the IV values ​​of all variables and provide the top 50% of the features to the label owner. This ensures that the data provider outputs the minimum set of valid feature data.

[0117] After completing the feature screening, the label owner will merge the features provided by different data providers to form a data set with complete features and labels, which will serve as the training set and test set for the risk user identification model.

[0118] S203. Steps for building a classification model based on differential privacy

[0119] The risk user identification model uses the random forest algorithm and combines differential privacy to improve the security of the algorithm. The model uses the CART decision tree as a single decision tree of the random forest, selects the split value of the continuous attribute according to the quantile, uses the Laplace mechanism to add noise to the sample set and leaf nodes, selects the best split attribute through the exponential mechanism, and selects the Gini index as the evaluation function of the exponential mechanism. The samples and attributes used to build the decision tree are randomly extracted from the data set and the attribute set respectively. This randomness can train multiple different classifiers and improve the applicability of the model. The prediction results of the model are voted, and the category with the most prediction results among multiple classifiers is selected as the final risk label.

[0120] The algorithm process is as follows:

[0121] Input: training dataset D, number of attributes F, privacy protection budget B, number of decision trees T, maximum depth of decision tree d max

[0122] Output: Random forest that satisfies B-differential privacy protection

[0123] Stop condition: If the sample set labels are consistent or the data set is smaller than the minimum split sample or greater than the maximum depth or the Gini index after splitting is less than or equal to the minimum Gini index.

[0124] Step 1 distributes the privacy protection budget equally to each tree ε' = B / T.

[0125] Step 2: Evenly distribute the privacy protection budget allocated to each tree to each layer of the decision tree, and then evenly divide the privacy protection budget allocated to each node into 2 parts, ε = ε' / (2*(d max +1)), half is used for dataset sample counting, and half is used for dataset partitioning (attribute selection / leaf node counting).

[0126] Step 3 for t = 1 to T

[0127] (1) Random selection | D t | samples and split attributes as training data set D t ;

[0128] (2) Add Laplace noise to the training set, |D t |noisy=|D t |+Lap(1 / ε), if the stopping condition is met, stop building the tree and return the leaf node label with Laplace noise added, otherwise execute (3);

[0129] (3) Traverse the value of each attribute separately. For discrete variables, directly use the corresponding Value value. For continuous variables, traverse the quantiles for variables exceeding a certain threshold, and calculate the optimal split value for each attribute separately;

[0130] (4) According to the exponential mechanism, the following probabilities are used

[0131]

[0132] To select the best split attribute. Where A is the attribute set, q(D,A) represents the Gini index, and D is the data set;

[0133] (5) Divide the data set according to the best attribute. If the left or right data set after division meets the stopping condition, stop building the tree and return the leaf node label with Laplace noise added. Otherwise, divide the data set according to the split node, recursively build the left and right subtrees, and execute (2)-(5).

[0134] Step 4 returns T decision trees that satisfy differential privacy B protection.

[0135] Based on the constructed differential privacy classification model, risky users are identified. The process is as follows:

[0136] (1) For each user, calculate the risk result predicted by each tree.

[0137] (2) Select the label with the highest mode in the tree prediction results as the user’s final risk label.

[0138] The risky user identification model combined with differential privacy protection can protect users' statistical information and further improve the security of model training and use.

[0139] Embodiment 3

[0140] The embodiment of the present application provides a risk user identification system for a railway ticketing system, which adopts the risk user identification method for a railway ticketing system as described above, such as Figure 5 As shown, the system includes:

[0141] Data providing module 101: used to provide multi-source sample data of the railway ticketing system, construct characteristics of ticket purchasing users based on the multi-source sample data and generate risk labels of ticket purchasing users, wherein the multi-source sample data includes: data on attribute characteristics, access characteristics and travel characteristics of ticket purchasing users;

[0142] The label possessing module 102 is used to encrypt the private information of the user label based on the risk label of the ticket purchasing user by using an encryption method, and then send it to the data providing module, and use a feature screening algorithm to screen the features of the ticket purchasing user, and fuse the screened features of the ticket purchasing user to generate a user training data set;

[0143] Model building module 103: used to build a risk user identification model based on differential privacy using a user training data set;

[0144] Risky user identification module 104: used to input the ticket purchasing user to be identified into the risky user identification model and output the risk prediction result of the user.

[0145] As an optional implementation, the data providing module 101 further includes:

[0146] Third-party data providing module: used to supplement the characteristics of ticket purchasing users with third-party data characteristics. User characteristics include: basic characteristics of ticket purchasing users and third-party data characteristics;

[0147] Among them, the label ownership module encrypts the label and sends it to the data providing module synchronously; the data providing module sums the labels corresponding to each sample in the group and sends them to the label ownership module; the label ownership module decrypts and calculates the number of positive and negative samples corresponding to each user group; the label ownership module calculates the IV value of the characteristic variables of the data providing module, and synchronizes the calculated results to the data providing module. After the data providing module sorts the IV values ​​of all variables, it provides the minimum set of valid characteristic data to the label ownership module.

[0148] Embodiment 4

[0149] The present application embodiment provides an electronic device, such as Figure 6 As shown, the device includes: a memory storing executable program code; a processor coupled to the memory; the processor calls the executable program code stored in the memory to execute the steps of the railway risk user identification method as described above.

[0150] Embodiment 5

[0151] An embodiment of the present application provides a computer storage medium, which stores computer instructions. When the computer instructions are called, they are used to execute the steps of the railway risk user identification method as described above.

[0152] In addition, combined Figure 1 The railway risk user identification method described in the embodiment of the present application can be implemented by a computer device. Figure 6 Schematic diagram of the hardware structure of a computer device according to an embodiment of the present application.

[0153] In some of these embodiments, the computer device may further include a communication interface 83 and a bus 80. Figure 6 As shown, the processor 81, the memory 82, and the communication interface 83 are connected via a bus 80 and communicate with each other.

[0154] Specifically, the processor 81 may include a central processing unit (CPU), or an application specific integrated circuit (ASIC), or may be configured to implement one or more integrated circuits of the embodiments of the present application.

[0155] The memory 82 may be used to store or cache various data files required for processing and / or communication, as well as possible computer program instructions executed by the processor 81 .

[0156] The processor 81 implements any one of the railway risk user identification methods in the above embodiments by reading and executing computer program instructions stored in the memory 82 .

[0157] The technical features of the above-described embodiments may be arbitrarily combined. To make the description concise, not all possible combinations of the technical features in the above-described embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0158] The above-mentioned embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the patent of the present application shall be subject to the attached claims.

Claims

1. A risk user identification method for a railway ticketing system, characterized in that: The method comprises: Feature construction step: constructing the features of ticket purchasing users and generating risk labels for ticket purchasing users based on the multi-source sample data of the railway ticketing system, wherein the multi-source sample data includes: data on the attribute features, access features and travel features of ticket purchasing users; Feature screening and fusion step: based on the risk label of the ticket purchasing user, after encrypting the privacy information of the risk label by using an encryption method, the features of the ticket purchasing user are screened by using a feature screening algorithm, and the screened features of the ticket purchasing user are fused to generate a user training data set; Model building step: using the user training data set to build a risk user identification model based on differential privacy; Risky user identification step: input the ticket purchasing user to be identified into the risky user identification model, and output the user's risk prediction result.

2. The risk user identification method for the railway ticketing system according to claim 1 is characterized in that: The feature construction step also includes: Third-party data feature construction step: third-party data features are used to supplement the ticket purchasing user features, and the ticket purchasing user features include: basic features of the ticket purchasing user and third-party data features.

3. The risk user identification method for a railway ticketing system according to claim 1 or 2 is characterized in that: The feature construction step further comprises: Attribute feature construction step: extract risk user identification features based on the identity information registered by the ticket purchasing user; Access feature construction step: extract risk user identification features based on at least one of the device environment information used by the ticket purchasing user for access, the content features accessed by the ticket purchasing user, and the time features of the ticket purchasing user's access; Travel feature construction step: extracting the risk user identification features based on the travel-related behavioral features of the ticket purchasing user.

4. The risk user identification method for a railway ticketing system according to claim 1 or 2 is characterized in that: The feature screening and fusion steps include: Feature screening step: After encrypting the risk label of the ticket purchasing user using a homomorphic encryption algorithm, the information value IV of the attribute feature, access feature and travel feature is calculated respectively to complete the screening of the ticket purchasing user features; Feature fusion step: fusing the ticket purchasing user features after feature screening to form the user data set, which includes: the user features and the user risk labels.

5. The risk user identification method for the railway ticketing system according to claim 4 is characterized in that: The feature screening step comprises: Label encryption step: Encrypt the risk label of each ticket purchasing user using a homomorphic encryption algorithm; IV value calculation step: IV value calculation is performed for the user characteristics of each ticket purchasing user group, and the IV values ​​are sorted and screened respectively to output the minimum data set of valid characteristic data.

6. The risk user identification method for the railway ticketing system according to claim 1 is characterized in that: The model building step also includes: The risk identification model is constructed using the random forest algorithm, and the security of the model is improved by combining the differential privacy algorithm; wherein, the risk user identification model uses the CART decision tree as a single decision tree of the random forest, selects the split value of the continuity attribute according to the quantile, uses the Laplace mechanism to add noise to the sample set and leaf nodes, and selects the best split attribute through the exponential mechanism, and the evaluation function of the exponential mechanism selects the Gini index.

7. A risk user identification system for a railway ticketing system, using the risk user identification method for a railway ticketing system as claimed in any one of claims 1 to 6, characterized in that: The system comprises: Data providing module: used to provide multi-source sample data of the railway ticketing system, construct the characteristics of ticket purchasing users based on the multi-source sample data and generate risk labels for ticket purchasing users, wherein the multi-source sample data includes: data on attribute characteristics, access characteristics and travel characteristics of ticket purchasing users; Label possessing module: used to encrypt the privacy information of the user label based on the risk label of the ticket purchasing user by using an encryption method, and then send it to the data providing module, and use a feature screening algorithm to screen the features of the ticket purchasing user, and fuse the screened features of the ticket purchasing user to generate a user training data set; Model building module: used to build a risk user identification model based on differential privacy using the user training data set; Risky user identification module: used to input the ticket purchasing user to be identified into the risky user identification model and output the user's risk prediction result.

8. The risk user identification system for railway ticketing system according to claim 7 is characterized in that: The data providing module also includes: The third-party data providing module is used to supplement the ticket purchasing user characteristics with third-party data characteristics, and the user characteristics include: basic characteristics of the ticket purchasing user and third-party data characteristics.

9. The risk user identification system for railway ticketing system according to claim 7 or 8, characterized in that: The tag owning module encrypts the tag and sends it to the data providing module synchronously; The data providing module calculates the sum of the labels corresponding to each sample in the group and sends it to the label owning module; The label owning module decrypts and calculates the number of positive and negative samples corresponding to each user group; The label owning module calculates IV values ​​for the characteristic variables of the data providing module and synchronizes the calculated results to the data providing module. The data providing module sorts the IV values ​​of all variables and provides the minimum set of valid characteristic data to the label owning module.

10. An electronic device, characterized in that: The device includes: a memory storing executable program code; a processor coupled to the memory; the processor calls the executable program code stored in the memory to execute the risk user identification method for the railway ticketing system as described in any one of claims 1-6.

11. A computer storage medium, characterized in that: The computer storage medium stores computer instructions, which, when called, are used to execute the risk user identification method for a railway ticketing system as described in any one of claims 1-6.

Citation Information

Patent Citations

  • stochastic forest algorithm oriented to differential privacy protection

    CN109284626A

  • A risk identification method and device

    CN109544166A

  • Object processing method, device and equipment based on privacy protection

    CN112364367A

  • Business travel operation platform implementation method and system based on big data

    CN113516500A

  • Black product user identification method and system, electronic equipment and storage medium

    CN113902039A