Rider identification method and device
By screening the key features and the application of random forest models in the rider's feature collection, the problem of low efficiency and insufficient accuracy of traditional rider identity authentication methods has been solved, and more efficient and accurate rider recognition has been achieved.
Patent Information
- Application Number
- CN202411318003.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-20
- Publication Date
- 2025-01-28
AI Technical Summary
Traditional rider identity authentication methods are inefficient and have low accuracy, and there are security risks such as identity impersonation.
Based on information value from the rider's feature collection, a key feature set is screened, the key feature data sets for users to be recognized are obtained, and rider recognition is used with random forest models.
It improved the efficiency and accuracy of riding identification, and reduced hidden safety hazards such as identity.
Smart Images
Figure CN119377804A_ABST
Abstract
Description
Technical Field
[0001] The present disclosure relates to the field of Internet technology, and in particular to a rider identification method and device. Background Art
[0002] With the rapid development of the food delivery industry, the safety and management of riders have become important issues. Traditional methods of rider identity authentication usually use manual review or photo comparison, which are not only inefficient, but also have low accuracy in rider identification and pose security risks such as identity fraud. Summary of the invention
[0003] The present disclosure aims to solve one of the technical problems in the related art at least to some extent.
[0004] To this end, a first object of the present disclosure is to propose a rider identification method to improve the efficiency and accuracy of rider identification.
[0005] A second objective of the present disclosure is to provide a rider identification device.
[0006] A third objective of the present disclosure is to provide an electronic device.
[0007] A fourth objective of the present disclosure is to provide a computer-readable storage medium.
[0008] A fifth object of the present disclosure is to provide a computer program product.
[0009] To achieve the above-mentioned purpose, the first embodiment of the present disclosure proposes a rider identification method, including:
[0010] A key feature subset is selected from the rider feature set based on information value, wherein the information value corresponding to any key feature in the key feature subset is greater than an information value threshold;
[0011] Acquire a key feature data set corresponding to the user to be identified, wherein the key feature data in the key feature data set corresponds one-to-one to the key feature;
[0012] Rider identification is performed according to the key feature data set to obtain a rider identification result corresponding to the user to be identified.
[0013] Optionally, before filtering out a key feature subset from the rider feature set based on information value, the method further includes:
[0014] Obtain a user sample set and a user feature data set corresponding to any user sample in the user sample set, wherein the user feature data in the user feature data set corresponds one-to-one to the rider features in the rider feature set;
[0015] The user sample set is labeled to obtain a labeled user sample set, wherein if the user sample meets the rider user requirement, the user sample is labeled as a rider positive sample; if the user sample does not meet the rider user requirement, the user sample is labeled as a rider negative sample;
[0016] According to the user feature data set and the labeled user sample set, the information value corresponding to any rider feature in the rider feature set is determined.
[0017] Optionally, obtaining a user feature data set corresponding to any user sample in the user sample set includes:
[0018] Obtaining basic user information corresponding to any user sample in the user sample set from the subsystem warehouse;
[0019] Obtaining user network signaling information corresponding to any one of the user samples from a network signaling data warehouse, wherein the user network signaling information includes communication information, location information, and rider-related application information;
[0020] The user basic information and the user network signaling information are associated to form a user feature set wide table, and a user feature data set corresponding to any user sample is obtained.
[0021] Optionally, the labeling the user sample set includes:
[0022] Obtaining virtual number call information corresponding to any user sample in the user sample set;
[0023] If the virtual number call information indicates that the call frequency between any user sample and the virtual number of the target label is greater than the call frequency threshold, it is determined that any user sample meets the rider user requirement, and any user sample is marked as a rider positive sample, wherein the target label includes a food delivery label and a takeaway label;
[0024] If the virtual number call information indicates that the call frequency between any user sample and the virtual number of the target label is not greater than the call frequency threshold, it is determined that any user sample does not meet the rider user requirement, and any user sample is marked as a rider negative sample.
[0025] Optionally, determining the information value corresponding to any rider feature in the rider feature set according to the user feature data set and the labeled user sample set includes:
[0026] For any rider feature in the rider feature set, the labeled user sample set is binned according to the user feature data set to obtain a plurality of labeled user sample subsets;
[0027] Determine the weight of evidence corresponding to the labeled user sample subset according to the number of positive samples of the riders and the number of negative samples of the riders corresponding to the labeled user sample subset;
[0028] According to the evidence weights corresponding to the multiple labeled user sample subsets, the information value corresponding to any rider feature is determined.
[0029] Optionally, performing rider identification according to the key feature data set to obtain a rider identification result corresponding to the user to be identified includes:
[0030] Obtain key feature sample data set;
[0031] Performing random forest model fitting according to the key feature sample data set to obtain a rider identification model that meets the rider identification requirements;
[0032] The key feature data set is input into the rider identification model to perform rider identification, and a rider identification result corresponding to the user to be identified is obtained.
[0033] Optionally, performing rider identification according to the key feature data set to obtain a rider identification result corresponding to the user to be identified includes:
[0034] Perform rider identification according to the key feature data set to obtain a rider prediction probability corresponding to the user to be identified;
[0035] If the rider prediction probability is greater than the prediction probability threshold, extracting the historical rider recognition result of the user to be recognized;
[0036] If the historical rider identification result meets the rider verification requirements, the rider identification result corresponding to the user to be identified is determined to be a rider.
[0037] Optionally, the key feature subset includes at least one of the following key features:
[0038] Usage behavior characteristics corresponding to rider-related applications, where the usage behavior characteristics include the number of applications, the number of times the applications are used, and the total traffic used by the applications;
[0039] The number of cells whose dwell time is greater than the dwell time threshold;
[0040] Total number of calls;
[0041] Total number of virtual number segments for calls.
[0042] Optionally, the rider-related application includes at least one of the following:
[0043] Order-taking applications;
[0044] Battery replacement applications;
[0045] Navigation app.
[0046] To achieve the above-mentioned purpose, a second aspect of the present disclosure provides a rider identification device, comprising:
[0047] A feature screening unit, used for screening out a key feature subset from the rider feature set based on information value, wherein the information value corresponding to any key feature in the key feature subset is greater than an information value threshold;
[0048] A data collection unit, used to obtain a key feature data set corresponding to the user to be identified, wherein the key feature data in the key feature data set corresponds to the key feature one by one;
[0049] The rider identification unit is used to perform rider identification according to the key feature data set to obtain a rider identification result corresponding to the user to be identified.
[0050] To achieve the above-mentioned purpose, a third aspect of the present disclosure provides an electronic device, including: a processor, and a memory communicatively connected to the processor;
[0051] Memory stores computer-executable instructions;
[0052] The processor executes the computer-executable instructions stored in the memory to implement the method shown in any one of the first aspects above.
[0053] To achieve the above-mentioned objectives, an embodiment of the fourth aspect of the present disclosure proposes a computer-readable storage medium, in which computer-executable instructions are stored. When the computer-executable instructions are executed by a processor, they are used to implement the method shown in any one of the above-mentioned first aspects.
[0054] To achieve the above-mentioned objective, a fifth aspect of the present disclosure provides a computer program product, including a computer program, which, when executed by a processor, implements the method shown in any one of the above-mentioned first aspects.
[0055] In summary, the method, apparatus, device and storage medium provided by the present disclosure can introduce more predictive features in the rider identification process by using information value to screen out key features that affect the identification of riders, thereby improving the efficiency and accuracy of rider identification.
[0056] Additional aspects and advantages of the present disclosure will be given in part in the following description and in part will be obvious from the following description or learned through practice of the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS
[0057] The above and / or additional aspects and advantages of the present disclosure will become apparent and easily understood from the following description of the embodiments in conjunction with the accompanying drawings, in which:
[0058] Figure 1 A schematic diagram of a flow chart of a rider identification method provided by an embodiment of the present disclosure;
[0059] Figure 2 A schematic flow chart of a rider identification method provided by another embodiment of the present disclosure;
[0060] Figure 3 A schematic diagram of an interaction of rider communication provided by an embodiment of the present disclosure;
[0061] Figure 4 A schematic diagram of the structure of a rider identification device provided in an embodiment of the present disclosure. DETAILED DESCRIPTION
[0062] Embodiments of the present disclosure are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present disclosure, and should not be construed as limiting the present disclosure.
[0063] The present disclosure is described in detail below with reference to specific embodiments.
[0064] In the first embodiment, if Figure 1 As shown, Figure 1 The present invention provides a flow chart of a method for identifying a rider provided by an embodiment of the present invention. The method can be implemented by a computer program and can be run on a device for identifying a rider. The computer program can be integrated into an application or run as an independent tool application.
[0065] Among them, the rider identification device can be an electronic device with a rider identification function.
[0066] For example, the rider identification method includes the following steps:
[0067] S101, selecting a key feature subset from the rider feature set based on information value, wherein the information value corresponding to any key feature in the key feature subset is greater than an information value threshold;
[0068] According to some embodiments, information value (IV) is used to indicate the contribution of a feature to target prediction, that is, the predictive ability of the feature. Generally speaking, the higher the IV value, the stronger the predictive ability of the feature and the higher the information contribution.
[0069] In some embodiments, a rider feature refers to a feature related to a rider. A rider feature set refers to a set formed by gathering at least one rider feature.
[0070] According to some embodiments, a key feature refers to a rider feature whose information value is greater than an information value threshold.
[0071] In some embodiments, the information value threshold does not specifically refer to a fixed threshold for a certain sound. The information value threshold can be determined, for example, according to an actual application scenario.
[0072] S102, obtaining a key feature data set corresponding to the user to be identified, wherein the key feature data in the key feature data set correspond to the key features one by one;
[0073] According to some embodiments, the user to be identified refers to a user for whom rider identification is required.
[0074] In some embodiments, the key feature data refers to feature data corresponding to the key feature.
[0075] S103, performing rider identification according to the key feature data set to obtain a rider identification result corresponding to the user to be identified.
[0076] According to some embodiments, the rider identification result is used to indicate whether the user to be identified is a rider.
[0077] In summary, the method provided in this embodiment can introduce more predictive features in the rider identification process by using information value to screen out key features that affect the identification of riders, thereby improving the efficiency and accuracy of rider identification.
[0078] This embodiment also provides another rider identification method, which can be executed by an electronic device.
[0079] For example, the rider identification method may include the following steps:
[0080] S201, obtaining a user sample set and a user feature data set corresponding to any user sample in the user sample set;
[0081] According to some embodiments, the user feature data in the user feature data set corresponds one-to-one to the rider features in the rider feature set.
[0082] In some embodiments, Figure 2The following is a flow chart of a rider identification method provided by an embodiment of the present disclosure. Figure 2 As shown, the user basic information corresponding to any user sample in the user sample set can be obtained from the subsystem warehouse; the user network signaling information corresponding to any user sample can be obtained from the network signaling data warehouse; the user basic information and the user network signaling information are associated to form a user feature set wide table to obtain a user feature data set corresponding to any user sample.
[0083] In some embodiments, user network signaling information includes but is not limited to communication information, location information, rider-related application information, etc. Therefore, by using the user Internet signaling deep recognition results, excluding mass applications, and introducing the relevant application data used by riders, more predictive features can be obtained, thereby improving the accuracy of rider identification.
[0084] Among them, through location information, the rider’s location and movement trajectory can be identified, network demand and traffic distribution can be better understood, and network resource configuration can be optimized.
[0085] It should be noted that riders can use order-taking software to receive delivery orders. During the delivery process, they can ride electric bicycles for delivery and use battery-swapping software to replace batteries to ensure power supply. In addition, riders can also use map navigation to plan the best delivery route to ensure that items are delivered to the destination on time. These tools and technologies enable riders to perform delivery work more efficiently. The riders' behavior is reflected in their online behavior by merging the applications they use into order-taking applications, battery-swapping applications, and navigation applications. Order-taking applications are rider-specific applications, and battery-swapping applications and navigation applications are rider-assistive software.
[0086] That is, rider-related applications may include at least one of the following:
[0087] Order-taking applications;
[0088] Battery replacement applications;
[0089] Navigation app.
[0090] In some embodiments, in order to further improve the rider feature set, the rider-related application information may also include dimensional information such as the number of applications, the number of times the applications are used, the duration of application use, and the total flow of application use. Furthermore, combined with the location information, the rider's working hours, idle hours, delivery time, route, efficiency and other work data can also be distinguished, which can provide data support and decision-making reference for users.
[0091] In some embodiments, by analyzing the rider's related application information and communication information, call packages, data packages, long-endurance terminals and other services that are more in line with their communication characteristics can be recommended.
[0092] S202, labeling the user sample set to obtain a labeled user sample set;
[0093] According to some embodiments, in the process of labeling the user sample set, if the user sample meets the rider user requirements, the user sample can be labeled as a rider positive sample; if the user sample does not meet the rider user requirements, the user sample can be labeled as a rider negative sample.
[0094] In some embodiments, the rider user requirement refers to the requirement used when identifying whether the user corresponding to the user sample is a rider.
[0095] It should be noted that Figure 3 This is a schematic diagram of a rider communication interaction provided by an embodiment of the present disclosure. Figure 3 As shown, the relevant food delivery platforms can provide virtual numbers as intermediate numbers for riders and customers to ensure that both parties can contact each other safely and accurately. Riders and customers can only see virtual numbers but not each other's real numbers. The use of such virtual numbers can protect the privacy and information security of riders and customers, and can also avoid disputes caused by telephone harassment or information leakage. In addition, the virtual number can be tagged and recorded on the number tagging platform, and the number platform's label only records the label of the virtual number.
[0096] That is to say, the virtual number call information corresponding to any user sample in the user sample set can be obtained; if the virtual number call information indicates that the call frequency between any user sample and the virtual number of the target label is greater than the call frequency threshold, then it is determined that any user sample meets the rider user requirements, and any user sample is marked as a rider positive sample, where the target label includes a food delivery label and a takeaway label; if the virtual number call information indicates that the call frequency between any user sample and the virtual number of the target label is not greater than the call frequency threshold, then it is determined that any user sample does not meet the rider user requirements, and any user sample is marked as a rider negative sample. Therefore, the accuracy of sample labeling can be improved.
[0097] In some embodiments, the call frequency threshold does not specifically refer to a fixed threshold. The call frequency threshold can be adjusted according to actual application scenarios, for example.
[0098] According to some embodiments, the user sample set may also be acquired based on a virtual number, that is, the user sample set may be obtained by acquiring the call status with the virtual number of the target tag.
[0099] Take a scenario as an example, Figure 2As shown in the figure, we can crawl the label data of virtual numbers, extract label numbers such as "food delivery" to establish number groups, filter the user numbers of rider order-taking applications that are frequently used based on mobile user online behavior information, extract the user numbers that frequently talk to the number group and mark them as rider positive samples, select some numbers of other professional users such as office groups and mark them as rider negative samples, and merge the rider positive samples and rider negative samples into a labeled user sample set. Therefore, a large number of positive samples can be obtained, and the samples are objectively accurate.
[0100] S203, determining the information value corresponding to any rider feature in the rider feature set according to the user feature data set and the labeled user sample set;
[0101] According to some embodiments, for any rider feature in the rider feature set, the labeled user sample set can be binned according to the user feature data set to obtain multiple labeled user sample subsets; based on the number of rider positive samples and the number of rider negative samples corresponding to the labeled user sample subsets, the evidence weight corresponding to the labeled user sample subsets can be determined; based on the evidence weights corresponding to multiple labeled user sample subsets, the information value corresponding to any rider feature can be determined.
[0102] In some embodiments, when binning the labeled user sample set according to the user feature data set, monotonic binning can be used for binning. Monotonic binning is a binning technique used for scorecard development. In order to improve the stability of the prediction results, the model often needs to bin features (also called discretization or grouping). Monotonic binning attempts to convert numerical variables into categorical variables by creating bins that have a monotonic relationship with the target.
[0103] For example, for continuous variables, you can use: equal-interval binning, equal-frequency binning, and custom intervals; for discrete variables, if there are too many bins, you can merge the bins.
[0104] According to some embodiments, the weight of evidence (WOE) represents the relationship between a predictable variable and a binary variable, and is used to measure the predictive power of the independent variable relative to the dependent variable. Since it evolved from the field of credit scoring, it is often described as an indicator for measuring the separation of good samples and bad samples.
[0105] In some embodiments, WOE may be determined according to the following formula:
[0106]
[0107] Among them, Bad T Indicates the number of negative rider samples in the labeled user sample set, Good TIndicates the number of positive samples of riders in the labeled user sample set, Bad i represents the number of negative rider samples in the labeled user sample subset i, Good i Represents the number of positive rider samples in the labeled user sample subset i.
[0108] in, represents the proportion of marginal negative samples in the labeled user sample subset i, Represents the proportion of marginal positive samples in the labeled user sample subset i.
[0109] Taking a scenario as an example, for continuous variables, the user feature data can be split into 10 parts (or smaller parts, depending on the distribution), the number of positive samples of riders and the number of negative samples of riders in each group (box) are calculated, the marginal negative sample ratio and the marginal positive sample ratio in each group (box) are calculated, and the WOE is calculated by obtaining the natural divisor of the marginal negative sample ratio and the marginal positive sample ratio.
[0110] According to some embodiments, the IV corresponding to any rider feature may be determined according to the following formula:
[0111]
[0112] Among them, n is the number of user sample subsets after annotation. It can be seen that the calculation of IV value is the weighted sum of WOE value, which can be used to measure the strength of the relationship between the variable represented by WOE and the binary variable.
[0113] S204, selecting a key feature subset from the rider feature set based on information value;
[0114] According to some embodiments, the correspondence between information value and prediction ability can be divided into the following levels:
[0115] If the information value is less than 0.02, the predictive ability of the feature is useless for prediction;
[0116] If 0.02<information value<0.1, the predictive ability of the feature is a weak predictor variable;
[0117] If 0.1<information value<0.3, the predictive ability of the feature is a medium predictor;
[0118] If 0.3<information value<0.5, the predictive ability of the feature is a strong predictor;
[0119] If the information value is > 0.5, the predictive ability of the feature is questionable or too good to be true.
[0120] In some embodiments, by measuring the importance of variables in the rider feature set based on the IV value, and then screening out key features, resources can be effectively saved in subsequent model training and prediction. At the same time, when the feature needs to be expanded later, the IV value of the feature can also be automatically calculated, and the feature can be automatically updated according to the prediction ability threshold.
[0121] Taking a scenario as an example, there are 19 rider features in the rider feature set, of which the information value of 12 rider features is greater than 0.3, then these 12 rider features can be used as key features.
[0122] According to some embodiments, the key feature subset selected from the rider feature set based on information value may include at least one of the following key features:
[0123] Usage behavior characteristics of rider-related applications, including the number of applications, the number of times applications are used, and the total traffic volume of applications;
[0124] The number of cells whose dwell time is greater than the dwell time threshold;
[0125] Total number of calls;
[0126] Total number of virtual number segments for calls.
[0127] In some embodiments, the dwell time threshold does not specifically refer to a fixed threshold. The threshold can be determined by its corresponding information value, and the threshold can be, for example, 5 minutes.
[0128] In some embodiments, the usage behavior characteristics corresponding to the rider-related applications can also be usage behavior characteristics within a preset time period. For example, the number of battery replacement applications within a week, the number of times the battery replacement applications are used within a week, and the total flow rate used by the battery replacement applications within a week can be selected according to their corresponding IV values.
[0129] In some embodiments, based on the effect of IV value on the model's predictive ability, features that are useless for prediction are eliminated, and the retained key features are used in model training and predictive identification. Among them, 9 features that are more characteristic of riders are introduced: the number of order-taking applications, the number of times the order-taking applications are used, the total traffic used by the order-taking applications, the number of battery swapping applications, the number of times the battery swapping applications are used, the total traffic used by the battery swapping applications, the number of navigation applications, the number of times the navigation applications are used, and the total traffic used by navigation applications. This can effectively save resources in subsequent model training and prediction, and can improve the accuracy of rider identification.
[0130] S205, obtaining a key feature data set corresponding to the user to be identified;
[0131] According to some embodiments, the method for obtaining the key feature data set may refer to the user feature data set, which will not be repeated here.
[0132] S206, obtaining a key feature sample data set;
[0133] According to some embodiments, the key feature sample data set may be obtained from the user feature data set, that is, feature data corresponding to the key feature is obtained from the user feature data set to obtain the key feature sample data set.
[0134] S207, performing random forest model fitting according to the key feature sample data set to obtain a rider identification model that meets the rider identification requirements;
[0135] According to some embodiments, Random Forest (RF) refers to a classifier that uses multiple decision trees to train and predict samples. It contains classifiers of multiple decision trees, and the categories of its outputs are determined by the mode of the categories output by individual trees. Random Forest is a flexible and easy-to-use machine learning algorithm that can get good results in most cases even without hyperparameter tuning. Random Forest is also one of the most commonly used algorithms because it is simple and can be used for both classification and regression. Random Forest integrates all classification voting results and designates the category with the most votes as the final output, which is the simplest idea of bagging.
[0136] In some embodiments, Figure 2 As shown, when fitting a random forest model based on a set of key feature sample data, the following steps can be taken:
[0137] S2071, splitting the key feature sample data set according to a preset ratio to obtain a training set and a validation set;
[0138] S2072, randomly selecting a key feature subset from the key feature set, randomly selecting a sample record subset from the training set, and constructing a decision tree model based on the key feature subset and the sample record subset;
[0139] S2073, repeating step S2072, constructing a combination of several decision tree models to form an initial rider identification model, generating rules for identifying riders, and the inference results output by the initial rider identification model are obtained by synthesizing all decision tree models;
[0140] S2074, during the model training phase, the random forest model results are persisted and saved. The model training effect is judged by viewing the confusion matrix of the validation set. When the precision / recall results meet expectations, the model training phase is completed. Otherwise, the model is adjusted and retrained. The initial rider recognition model calculates the output result after receiving the input data. The output result is a two-dimensional vector, corresponding to the probability of being predicted as a rider and whether it is a rider (1 represents a rider, 0 represents a non-rider).
[0141] In some embodiments, the preset ratio does not specifically refer to a fixed ratio. The preset ratio can be adjusted according to the actual application scenario. For example, the fixed ratio can be 7:3, 70% as the training set and 30% as the verification set.
[0142] In some embodiments, when constructing a decision tree model based on a key feature subset and a sample record subset, tree growth may be performed based on the residual minimization principle until the iteration condition is met.
[0143] In some embodiments, the confusion matrix is a situation analysis table that summarizes the prediction results of the classification model in machine learning, and summarizes the records in the data set in a matrix form according to the two criteria of the real category and the category predicted by the classification model. The rows of the matrix represent the real values, the columns of the matrix represent the predicted values, and the matrix representation is as follows:
[0144]
[0145] Among them, the meanings of some symbols defined in the classification evaluation indicators are as follows:
[0146] TP (True Positive): predicts the positive class as the positive class number. If the actual number is 0, the prediction is also 0.
[0147] FN (False Negative): predict the positive class as the negative class number, the actual number is 0, and the prediction is 1;
[0148] FP (False Positive): predict the negative class as the positive class number, the actual number is 1, and the prediction is 0;
[0149] TN (True Negative): predicts the negative class as the negative class number, the actual value is 1, and the prediction is also 1;
[0150] Accuracy
[0151]
[0152] Precision:
[0153]
[0154] Recall rate:
[0155]
[0156] S208, inputting the key feature data set into the rider identification model to perform rider identification, and obtaining a rider identification result corresponding to the user to be identified;
[0157] According to some embodiments, the rider identification model can calculate the output result after receiving the key feature data set, and the output result is a two-dimensional vector, which corresponds to the key feature and its corresponding predicted probability of being a rider and whether the rider is a rider; then, these key features and the probability value of the user being a rider can be spliced to obtain the rider identification result.
[0158] In some embodiments, the identified rider data can also be extracted to evaluate the model accuracy. When the verified model accuracy reaches the expected value, the final training model file is saved. Otherwise, the misidentified rider data is checked and the model is analyzed and adjusted.
[0159] According to some embodiments, after rider identification is performed based on the key feature data set and the rider prediction probability corresponding to the user to be identified is obtained, if the rider prediction probability is greater than the prediction probability threshold, the historical rider identification result of the user to be identified can be extracted; if the historical rider identification result meets the rider verification requirement, the rider identification result corresponding to the user to be identified is determined to be the rider. Therefore, since when identifying a rider with a preset behavior type, the rider within the target time period is identified, and the identification is performed based on at least two target identification factors, the identification accuracy can be improved, and the accuracy of rider identification can be further improved.
[0160] Take a scenario as an example. When the rider identification result output by the rider identification model is a daily granularity prediction result, if the rider prediction probability of the user to be identified is > 85%, the historical rider identification results of the last 7 days can be extracted. If the number of days in which the user to be identified is a rider in the historical rider identification results is not less than 5 days, it can be determined that the user to be identified is a rider. Therefore, the set of rider numbers identified on a daily basis can be included in the communication whitelist, so that users will not miss or reject calls from rider users, and rider user numbers will not be mistakenly shut down, thereby protecting the rights and interests of rider users.
[0161] It should be noted that it is also possible to extract some identified riders and the numbers they actively call from the multiple rider identification results obtained after rider identification, crawl the number labels on the number marking platform, and obtain the number of "food delivery" labels called by the identified riders. This method can be used to verify the accuracy of rider identification.
[0162] In summary, the method provided in this embodiment analyzes and compares the rider's mobile signaling through feature extraction and machine learning algorithms based on mobile signaling, so as to achieve rapid and accurate identification of the rider, thereby improving the efficiency and accuracy of identity authentication. Secondly, not only the overall characteristics of the target are collected, but also the target identity can be given more accurately through the identification of multiple additional target information, with better robustness and adaptability, and strong applicability. In addition, by accurately identifying the rider, the company can better manage and monitor the rider's behavior and prevent criminals from impersonating others to commit illegal acts. This can include preventing identity impersonation, protecting customer privacy, etc., thereby enhancing communication security, reducing security risks, avoiding security risks such as identity impersonation, and improving the safety of the rider.
[0163] By using information value to screen out key features that affect rider identification, more predictive features can be introduced into the rider identification process, which can improve the efficiency and accuracy of rider identification.
[0164] In order to implement the above embodiments, the present disclosure also proposes a rider identification device.
[0165] like Figure 4 As shown, the rider identification device 400 comprises:
[0166] The feature screening unit 401 is used to screen out a key feature subset from the rider feature set based on information value, wherein the information value corresponding to any key feature in the key feature subset is greater than an information value threshold;
[0167] The data collection unit 402 is used to obtain a key feature data set corresponding to the user to be identified, wherein the key feature data in the key feature data set corresponds to the key feature one by one;
[0168] The rider identification unit 403 is used to perform rider identification according to the key feature data set to obtain a rider identification result corresponding to the user to be identified.
[0169] Optionally, before filtering out the key feature subset from the rider feature set based on the information value, the feature screening unit 401 is further used to:
[0170] Obtain a user sample set and a user feature data set corresponding to any user sample in the user sample set, wherein the user feature data in the user feature data set corresponds one-to-one to the rider features in the rider feature set;
[0171] The user sample set is labeled to obtain a labeled user sample set, wherein if the user sample meets the rider user requirements, the user sample is labeled as a rider positive sample; if the user sample does not meet the rider user requirements, the user sample is labeled as a rider negative sample;
[0172] Based on the user feature data set and the labeled user sample set, determine the information value corresponding to any rider feature in the rider feature set.
[0173] Optionally, when the feature screening unit 401 is used to obtain a user feature data set corresponding to any user sample in the user sample set, it is specifically used to:
[0174] Obtain basic user information corresponding to any user sample in the user sample set from the subsystem warehouse;
[0175] Obtain user network signaling information corresponding to any user sample from the network signaling data warehouse, wherein the user network signaling information includes communication information, location information, and rider-related application information;
[0176] The user basic information and the user network signaling information are associated to form a user feature set wide table, and a user feature data set corresponding to any user sample is obtained.
[0177] Optionally, when the feature screening unit 401 is used to label the user sample set, it is specifically used to:
[0178] Obtain call information of a virtual number corresponding to any user sample in the user sample set;
[0179] If the virtual number call information indicates that the call frequency between any user sample and the virtual number of the target label is greater than the call frequency threshold, it is determined that any user sample meets the rider user requirements, and any user sample is marked as a rider positive sample, where the target label includes a food delivery label and a takeaway label;
[0180] If the virtual number call information indicates that the call frequency between any user sample and the virtual number of the target label is not greater than the call frequency threshold, it is determined that any user sample does not meet the rider user requirements, and any user sample is marked as a rider negative sample.
[0181] Optionally, the feature screening unit 401 is used to determine the information value corresponding to any rider feature in the rider feature set according to the user feature data set and the labeled user sample set, specifically to:
[0182] For any rider feature in the rider feature set, the labeled user sample set is binned according to the user feature data set to obtain multiple labeled user sample subsets;
[0183] Determine the weight of evidence corresponding to the labeled user sample subset according to the number of positive samples and negative samples of the riders corresponding to the labeled user sample subset;
[0184] According to the evidence weights corresponding to multiple labeled user sample subsets, the information value corresponding to any rider feature is determined.
[0185] Optionally, the rider identification unit 403 is used to perform rider identification according to the key feature data set, and when obtaining the rider identification result corresponding to the user to be identified, it is specifically used to:
[0186] Obtain key feature sample data set;
[0187] A random forest model is fitted based on the key feature sample data set to obtain a rider identification model that meets the rider identification requirements;
[0188] The key feature data set is input into the rider identification model to perform rider identification and obtain the rider identification result corresponding to the user to be identified.
[0189] Optionally, the rider identification unit 403 is used to perform rider identification according to the key feature data set, and when obtaining the rider identification result corresponding to the user to be identified, it is specifically used to:
[0190] Perform rider identification based on the key feature data set to obtain the rider prediction probability corresponding to the user to be identified;
[0191] If the rider prediction probability is greater than the prediction probability threshold, the historical rider recognition results of the user to be recognized are extracted;
[0192] If the historical rider identification results meet the rider verification requirements, the rider identification result corresponding to the user to be identified is determined to be the rider.
[0193] Optionally, the key feature subset includes at least one of the following key features:
[0194] Usage behavior characteristics of rider-related applications, including the number of applications, the number of times applications are used, and the total traffic volume of applications;
[0195] The number of cells whose dwell time is greater than the dwell time threshold;
[0196] Total number of calls;
[0197] Total number of virtual number segments for calls.
[0198] Optionally, the rider-related application includes at least one of the following:
[0199] Order-taking applications;
[0200] Battery replacement applications;
[0201] Navigation app.
[0202] It should be noted that the aforementioned explanation of the rider identification method embodiment is also applicable to the rider identification device of this embodiment and will not be repeated here.
[0203] In summary, the device provided by the embodiment of the present disclosure can introduce more predictive features in the rider identification process by using information value to screen out key features that affect the identification of riders, thereby improving the efficiency and accuracy of rider identification.
[0204] In order to implement the above embodiments, the present disclosure also proposes an electronic device, including: a processor, and a memory communicatively connected to the processor; the memory stores computer-executable instructions; the processor executes the computer-executable instructions stored in the memory to implement the method provided by the above embodiments.
[0205] In order to implement the above embodiments, the present disclosure further proposes a computer-readable storage medium, in which computer-executable instructions are stored. When the computer-executable instructions are executed by a processor, they are used to implement the methods provided by the above embodiments.
[0206] In order to implement the above embodiments, the present disclosure also proposes a computer program product, including a computer program, which implements the methods provided by the above embodiments when executed by a processor.
[0207] The collection, storage, use, processing, transmission, provision and disclosure of user personal information involved in this disclosure shall comply with the relevant laws and regulations and shall not violate public order and good morals.
[0208] It should be noted that personal information from users should be collected for legitimate and reasonable purposes and should not be shared or sold outside of these legitimate uses. In addition, such collection / sharing should be carried out after receiving the user's informed consent, including but not limited to notifying the user to read the user agreement / user notice and sign the agreement / authorization including authorization of relevant user information before the user uses the function. In addition, any necessary steps should be taken to protect and safeguard access to such personal information data and ensure that others who have access to personal information data comply with its privacy policy and procedures.
[0209] The present disclosure is expected to provide implementation schemes for users to selectively block the use or access of personal information data. That is, the present disclosure is expected to provide hardware and / or applications to prevent or block access to such personal information data. Once the personal information data is no longer needed, risks can be minimized by limiting data collection and deleting the data. In addition, when applicable, such personal information is de-identified to protect the privacy of users.
[0210] In the description of the aforementioned embodiments, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present disclosure. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, unless they are contradictory.
[0211] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of the features. In the description of the present disclosure, "plurality" means at least two, such as two, three, etc., unless otherwise clearly and specifically defined.
[0212] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, segment or portion of code that includes one or more executable instructions for implementing the steps of a custom logical function or process, and the scope of the preferred embodiments of the present disclosure includes additional implementations in which functions may not be performed in the order shown or discussed, including performing functions in a substantially simultaneous manner or in reverse order depending on the functions involved, which should be understood by technicians in the technical field to which the embodiments of the present disclosure belong.
[0213] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute the instructions), or in combination with these instruction execution systems, devices or apparatuses. For the purpose of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in combination with these instruction execution systems, devices or apparatuses. More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk box (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), a fiber optic device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium and then editing, interpreting or otherwise processing in a suitable manner if necessary, and then stored in a computer memory.
[0214] It should be understood that the various parts of the present disclosure can be implemented with hardware, applications, firmware, or a combination thereof. In the above-mentioned embodiments, multiple steps or methods can be implemented with applications or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented with hardware, as in another embodiment, it can be implemented with any one of the following technologies known in the art or a combination thereof: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.
[0215] A person skilled in the art may understand that all or part of the steps in the above-mentioned embodiment method may be completed by instructing related hardware through a program, and the program may be stored in a computer-readable storage medium, which, when executed, includes one or a combination of the steps of the method embodiment.
[0216] In addition, each functional unit in each embodiment of the present disclosure may be integrated into a processing module, or each unit may exist physically separately, or two or more units may be integrated into one module. The above-mentioned integrated module may be implemented in the form of hardware or in the form of an application function module. If the integrated module is implemented in the form of an application function module and sold or used as an independent product, it may also be stored in a computer-readable storage medium.
[0217] The storage medium mentioned above may be a read-only memory, a disk or an optical disk, etc. Although the embodiments of the present disclosure have been shown and described above, it can be understood that the above embodiments are exemplary and cannot be understood as limitations of the present disclosure. A person of ordinary skill in the art may change, modify, replace and modify the above embodiments within the scope of the present disclosure.
Claims
1. A rider identification method, characterized in that: include: A key feature subset is selected from the rider feature set based on information value, wherein the information value corresponding to any key feature in the key feature subset is greater than an information value threshold; Acquire a key feature data set corresponding to the user to be identified, wherein the key feature data in the key feature data set corresponds one-to-one to the key feature; Rider identification is performed according to the key feature data set to obtain a rider identification result corresponding to the user to be identified.
2. The method according to claim 1, characterized in that Before selecting the key feature subset from the rider feature set based on information value, the method further includes: Obtain a user sample set and a user feature data set corresponding to any user sample in the user sample set, wherein the user feature data in the user feature data set corresponds one-to-one to the rider features in the rider feature set; The user sample set is labeled to obtain a labeled user sample set, wherein if the user sample meets the rider user requirement, the user sample is labeled as a rider positive sample; if the user sample does not meet the rider user requirement, the user sample is labeled as a rider negative sample; According to the user feature data set and the labeled user sample set, the information value corresponding to any rider feature in the rider feature set is determined.
3. The method according to claim 2, characterized in that Obtaining a user feature data set corresponding to any user sample in the user sample set includes: Obtaining basic user information corresponding to any user sample in the user sample set from the subsystem warehouse; Obtaining user network signaling information corresponding to any one of the user samples from a network signaling data warehouse, wherein the user network signaling information includes communication information, location information, and rider-related application information; The user basic information and the user network signaling information are associated to form a user feature set wide table, and a user feature data set corresponding to any user sample is obtained.
4. The method according to claim 2, characterized in that: The marking of the user sample set includes: Obtaining virtual number call information corresponding to any user sample in the user sample set; If the virtual number call information indicates that the call frequency between any user sample and the virtual number of the target label is greater than the call frequency threshold, it is determined that any user sample meets the rider user requirement, and any user sample is marked as a rider positive sample, wherein the target label includes a food delivery label and a takeaway label; If the virtual number call information indicates that the call frequency between any user sample and the virtual number of the target label is not greater than the call frequency threshold, it is determined that any user sample does not meet the rider user requirement, and any user sample is marked as a rider negative sample.
5. The method according to claim 2, characterized in that: Determining the information value corresponding to any rider feature in the rider feature set according to the user feature data set and the labeled user sample set includes: For any rider feature in the rider feature set, the labeled user sample set is binned according to the user feature data set to obtain a plurality of labeled user sample subsets; Determine the weight of evidence corresponding to the labeled user sample subset according to the number of positive samples of the riders and the number of negative samples of the riders corresponding to the labeled user sample subset; According to the evidence weights corresponding to the multiple labeled user sample subsets, the information value corresponding to any rider feature is determined.
6. The method according to claim 1, characterized in that The rider identification is performed according to the key feature data set to obtain the rider identification result corresponding to the user to be identified, including: Obtain key feature sample data set; Performing random forest model fitting according to the key feature sample data set to obtain a rider identification model that meets the rider identification requirements; The key feature data set is input into the rider identification model to perform rider identification, and a rider identification result corresponding to the user to be identified is obtained.
7. The method according to claim 1, characterized in that The rider identification is performed according to the key feature data set to obtain the rider identification result corresponding to the user to be identified, including: Perform rider identification according to the key feature data set to obtain a rider prediction probability corresponding to the user to be identified; If the rider prediction probability is greater than the prediction probability threshold, extracting the historical rider recognition result of the user to be recognized; If the historical rider identification result meets the rider verification requirements, the rider identification result corresponding to the user to be identified is determined to be a rider.
8. The method according to claim 1, characterized in that The key feature subset includes at least one of the following key features: Usage behavior characteristics corresponding to rider-related applications, where the usage behavior characteristics include the number of applications, the number of times the applications are used, and the total traffic used by the applications; The number of cells whose dwell time is greater than the dwell time threshold; Total number of calls; Total number of virtual numbers for calls.
9. The method according to claim 8, characterized in that The rider-related applications include at least one of the following: Order-taking applications; Battery replacement applications; Navigation app.
10. A rider identification device, characterized in that: include: A feature screening unit, used for screening out a key feature subset from the rider feature set based on information value, wherein the information value corresponding to any key feature in the key feature subset is greater than an information value threshold; A data collection unit, used to obtain a key feature data set corresponding to the user to be identified, wherein the key feature data in the key feature data set corresponds to the key feature one by one; The rider identification unit is used to perform rider identification according to the key feature data set to obtain a rider identification result corresponding to the user to be identified.