A user relationship identification method based on a car-hailing history

Abstract

The application discloses a user relationship identification method based on a history of taking a car, and comprises the following steps: S1, collecting user historical data of taking a car; S2, automatically marking the type of geographic information; S3, calculating the confidence address of each user; and S4, determining the relationship between users according to the confidence address. The application provides a user relationship identification method based on a history of taking a car to solve the problem of difficult identification of brushing, the method collects passenger historical data of taking a car and driver historical data of taking a car to go home, and adopts the method of manually marking residences and offices to identify the neighbor and colleague relationship between users, can match the driver and the passenger in taking an order, and can improve the risk control threshold of other associated users according to the brushing user found; in summary, the method can mine the relationship between users, solve the problem of associated identification of brushing, improve the accuracy of risk control identification, reduce asset loss, invest limited activity resources into real users, and thus improve the activity effect.

Description

Technical Field

[0001] This invention relates to the fields of ride-hailing, carpooling, and taxi services, and particularly to a method for identifying user relationships based on ride-hailing history. Background Technology

[0002] In ride-hailing, taxi, and carpooling services, platforms often offer substantial coupons to passengers and high rewards to drivers, incentivizing either to engage in fraudulent order placement to increase earnings. Identifying relationships between users, such as neighbors or colleagues, can more accurately detect fraudulent activity. If one person engages in fraudulent order placement, the likelihood of their colleagues or neighbors doing so increases. Similarly, if many people in a neighborhood engage in fraudulent order placement, the probability of others in that neighborhood doing so is also very high. Therefore, identifying user relationships can significantly improve the accuracy of fraudulent order detection.

[0003] Based on this, the present invention discloses a user relationship identification method based on ride-hailing history to solve the user relationship identification problem, thereby assisting ride-hailing platforms in determining the risk of a user engaging in fraudulent activities based on the determined relationship. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to overcome the defects of the prior art and provide a user relationship identification method based on ride-hailing history.

[0005] This invention provides the following technical solution:

[0006] This invention provides a user relationship identification method based on ride-hailing history, comprising the following steps:

[0007] S1. Collect users' historical ride-hailing data:

[0008] (1) Collect historical ride-hailing data set T1 of passenger users within 1 year, specifically including user ID, origin geographical name, and destination geographical name;

[0009] (2) Collect data set T2 of home-trip orders set by drivers within one year, including user ID, origin geographical name, and destination geographical name;

[0010] (3) After merging T1 and T2, remove duplicates from all different starting point geographical names or ending point geographical names and store them in the set Q, p∈Q, where p is the geographical name of the starting point or ending point.

[0011] S2. Automatically label geographic information by type:

[0012] For each p, p∈Q, the geographic type is marked by calling the map points in the map software. If the geographic name is a residential area, the type is marked as 1, indicating "residence". If the geographic name is an office area, the type is marked as 2, indicating "office". All other areas, such as hospitals, are marked as 0.

[0013] The function for finding the geographic type from the geographic name is denoted as f(p), and the value range of f(p) is 0, 1, 2;

[0014] S3. Calculate the confidence address for each user:

[0015] For each user x, their different geographical names are summarized as (p,t); p is the user's origin or destination address, t is the number of times the address appears, the set of user x's residential addresses is denoted as C(x), (p,t)∈C(x), f(p)=1, and each different address appears only once in this set; the set of user x's office addresses is denoted as D(x), (p,t)∈D(x), f(p)=2, and each different address appears only once in this set;

[0016] The maximum residential confidence address H(x) = p1 of user x must satisfy:

[0017] (1)(p1,t1)∈C(x);

[0018] (2)

[0019] (3) t1≥α, where α is the preset confidence coefficient, let α=3;

[0020] If C(x) has no p1 that satisfies the condition, then there is no maximum residential address, and in this case, p1 = NIL is recorded.

[0021] The maximum office confidence address W(x) = p2 for user x must satisfy:

[0022] (1)(p2,t2)∈D(x);

[0023] (2)

[0024] (3) t2≥α, where α is the preset confidence coefficient, let α=3;

[0025] If D(x) does not have a p2 that satisfies the condition, then there is no maximum office address, and in this case, p2 = NIL is recorded.

[0026] S4. Determine the relationship between users based on their trusted addresses:

[0027] For users x and y, if x and y satisfy H(x) = H(y) and H(x) ≠ NIL, then users x and y are neighbors; otherwise, it does not mean they are not neighbors, but there is currently insufficient evidence to prove that x and y are neighbors. If x and y satisfy W(x) = W(y) and W(x) ≠ NIL, then users x and y are colleagues; otherwise, it does not mean they are not colleagues, but there is currently insufficient evidence to prove that x and y are colleagues. Both neighbor and colleague relationships exist simultaneously.

[0028] S5. Raise the risk control threshold based on relationships with neighbors and colleagues:

[0029] (a) Order risk control, where x and y are the passenger and driver respectively:

[0030] (a.1) For colleague relationships, an isolation strategy is adopted. If x and y have a colleague relationship, and x is a passenger and y is a driver, y can be matched with other passengers and x can be matched with other drivers, but the system will not match x and y into a single order, and x and y are not visible to each other.

[0031] (a.2) For neighbor relationships, the first-order isolation strategy is adopted. If x and y are neighbors, and x is a passenger and y is a driver, x and y can only be matched for the first order. For the second order and subsequent orders, x and y cannot be matched. x and y are not visible to each other.

[0032] (b) Discount risk control, if x is a passenger:

[0033] For x, a level-limiting strategy is adopted. Let R(H(x)) be the number of passengers identified in the ride-hailing risk control system as neighbors or colleagues who have engaged in fraudulent transactions. If R(H(x)) ≥ β, where β is a preset risk threshold, then high-value and high-discount coupons will be restricted from being issued to passengers in that area.

[0034] (c) Order dispatch and reward restrictions, if y is a driver:

[0035] For y, a downgrade strategy is adopted. Let S(H(y)) be the number of drivers identified in the ride-hailing risk control system as having engaged in fraudulent transactions (neighbors or colleagues). If S(H(x)) ≥ r, where r is a preset risk threshold, then the restrictions on drivers in this area are as follows:

[0036] (1) The dispatch priority is lower than that of normal drivers;

[0037] (2) Restrict the acceptance of high-value orders;

[0038] (3) Participation in high-reward activities is prohibited;

[0039] (4) Add the driver to the pool of people who are suspected of fraudulent orders, lower the threshold for identifying the driver as a fraudulent order driver, and make it easier for the driver to be identified as a fraudulent order driver.

[0040] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0041] This invention provides a user relationship identification method based on ride-hailing history to solve the problem of difficulty in identifying fraudulent orders. The method summarizes passenger ride-hailing history data and driver home trip history data, and uses manual labeling of residences and offices to identify neighbor and colleague relationships between users. Based on these relationships, it can match drivers and passengers for ride-hailing orders, and raise the risk control threshold for other related users based on the identified fraudulent order users.

[0042] In summary, this invention, based on the characteristics of the travel industry, can uncover the relationships between users, solve the problem of identifying associations in fraudulent transactions, improve the accuracy of risk control identification, reduce asset losses, and invest limited activity resources in genuine users, thereby improving the effectiveness of the activity. Attached Figure Description

[0043] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0044] Figure 1 This is a schematic diagram of the process steps of the present invention. Detailed Implementation

[0045] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are for illustration and explanation only and are not intended to limit the present invention. All identical reference numerals in the drawings refer to the same components.

[0046] Example 1

[0047] like Figure 1 This invention provides a user relationship identification method based on ride-hailing history, comprising the following steps:

[0048] S1. Collect users' historical ride-hailing data:

[0049] (1) Collect historical ride-hailing data set T1 of passenger users within 1 year, specifically including user ID, origin geographical name, and destination geographical name;

[0050] (2) Collect data set T2 of home-trip orders set by drivers within one year, including user ID, origin geographical name, and destination geographical name;

[0051] (3) After merging T1 and T2, remove duplicates from all different starting point geographical names or ending point geographical names and store them in the set Q, p∈Q, where p is the geographical name of the starting point or ending point.

[0052] S2. Automatically label geographic information by type:

[0053] For each p, p∈Q, the geographic type is marked by calling the map points in the map software. If the geographic name is a residential area, the type is marked as 1, indicating "residence". If the geographic name is an office area, the type is marked as 2, indicating "office". All other areas, such as hospitals, are marked as 0.

[0054] The function for finding the geographic type from the geographic name is denoted as f(p), and the value range of f(p) is 0, 1, 2;

[0055] S3. Calculate the confidence address for each user:

[0056] For each user x, their different geographical names are summarized as (p,t); p is the user's origin or destination address, t is the number of times the address appears, the set of user x's residential addresses is denoted as C(x), (p,t)∈C(x), f(p)=1, and each different address appears only once in this set; the set of user x's office addresses is denoted as D(x), (p,t)∈D(x), f(p)=2, and each different address appears only once in this set;

[0057] The maximum residential confidence address H(x) = p1 of user x must satisfy:

[0058] (1)(p1,t1)∈C(x);

[0059] (2)

[0060] (3) t1≥α, where α is the preset confidence coefficient, let α=3;

[0061] If C(x) has no p1 that satisfies the condition, then there is no maximum residential address, and in this case, p1 = NIL is recorded.

[0062] The maximum office confidence address W(x) = p2 for user x must satisfy:

[0063] (1)(p2,t2)∈D(x);

[0064] (2)

[0065] (3) t2≥α, where α is the preset confidence coefficient, let α=3;

[0066] If D(x) does not have a p2 that satisfies the condition, then there is no maximum office address, and in this case, p2 = NIL is recorded.

[0067] S4. Determine the relationship between users based on their trusted addresses:

[0068] For users x and y, if x and y satisfy H(x) = H(y) and H(x) ≠ NIL, then users x and y are neighbors; otherwise, it does not mean they are not neighbors, but there is currently insufficient evidence to prove that x and y are neighbors. If x and y satisfy W(x) = W(y) and W(x) ≠ NIL, then users x and y are colleagues; otherwise, it does not mean they are not colleagues, but there is currently insufficient evidence to prove that x and y are colleagues. Both neighbor and colleague relationships exist simultaneously.

[0069] S5. Raise the risk control threshold based on relationships with neighbors and colleagues:

[0070] (a) Order risk control, where x and y are the passenger and driver respectively:

[0071] (a.1) For colleague relationships, an isolation strategy is adopted. If x and y have a colleague relationship, and x is a passenger and y is a driver, y can be matched with other passengers and x can be matched with other drivers, but the system will not match x and y into a single order, and x and y are not visible to each other.

[0072] (a.2) For neighbor relationships, the first-order isolation strategy is adopted. If x and y are neighbors, and x is a passenger and y is a driver, x and y can only be matched for the first order. For the second order and subsequent orders, x and y cannot be matched. x and y are not visible to each other.

[0073] (b) Discount risk control, if x is a passenger:

[0074] For x, a level-limiting strategy is adopted. Let R(H(x)) be the number of passengers identified in the ride-hailing risk control system as neighbors or colleagues who have engaged in fraudulent transactions. If R(H(x)) ≥ β, where β is a preset risk threshold, then high-value and high-discount coupons will be restricted from being issued to passengers in that area.

[0075] (c) Order dispatch and reward restrictions, if y is a driver:

[0076] For y, a downgrade strategy is adopted. Let S(H(y)) be the number of drivers identified in the ride-hailing risk control system as having engaged in fraudulent transactions (neighbors or colleagues). If S(H(x)) ≥ r, where r is a preset risk threshold, then the restrictions on drivers in this area are as follows:

[0077] (1) The dispatch priority is lower than that of normal drivers;

[0078] (2) Restrict the acceptance of high-value orders;

[0079] (3) Participation in high-reward activities is prohibited;

[0080] (4) Add the driver to the pool of people who are suspected of fraudulent orders, lower the threshold for identifying the driver as a fraudulent order driver, and make it easier for the driver to be identified as a fraudulent order driver.

[0081] This invention provides a user relationship identification method based on ride-hailing history to solve the problem of difficulty in identifying fraudulent orders. The method summarizes passenger ride-hailing history data and driver home trip history data, and uses manual labeling of residences and offices to identify neighbor and colleague relationships between users. Based on this relationship, it can match drivers and passengers for ride-hailing orders (as shown in step S5), and raise the risk control threshold for other related users based on the identified fraudulent order users.

[0082] In summary, this invention, based on the characteristics of the travel industry, can uncover the relationships between users, solve the problem of identifying associations in fraudulent transactions, improve the accuracy of risk control identification, reduce asset losses, and invest limited activity resources in genuine users, thereby improving the effectiveness of the activity.

[0083] Finally, it should be noted that the above descriptions are merely preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for identifying user relationships based on ride-hailing history, characterized in that, Includes the following steps: S1. Collect users' historical ride-hailing data: (1) Collect historical ride-hailing data set T1 of passenger users within 1 year, specifically including user ID, origin geographical name, and destination geographical name; (2) Collect data set T2 of home-trip orders set by drivers within one year, including user ID, origin geographical name, and destination geographical name; (3) After merging T1 and T2, remove duplicates from all different starting point geographical names or ending point geographical names and store them in the set Q, p∈Q, where p is the geographical name of the starting point or ending point. S2. Automatically label geographic information by type: For each p, p∈Q, the geographic type is marked by calling the map points in the map software. If the geographic name is a residential area, the type is marked as 1, indicating "residence". If the geographic name is an office area, the type is marked as 2, indicating "office". All other areas are marked as 0. The function for finding the geographic type from the geographic name is denoted as f(p), and the value range of f(p) is 0, 1, 2; S3. Calculate the confidence address for each user: For each user x, their different geographical names are summarized as (p,t); p is the user's origin or destination address, t is the number of times the address appears, the set of user x's residential addresses is denoted as C(x), (p,t)∈C(x), f(p)=1, and each different address appears only once in this set; the set of user x's office addresses is denoted as D(x), (p,t)∈D(x), f(p)=2, and each different address appears only once in this set; The maximum residential confidence address H(x) = p1 of user x must satisfy: (1)(p1,t1)∈C(x); (2) (3) t1≥α, where α is the preset confidence coefficient, let α=3; If C(x) has no p1 that satisfies the condition, then there is no maximum residential address, and in this case, p1 = NIL is recorded. The maximum office confidence address W(x) = p2 for user x must satisfy: (1)(p2,t2)∈D(x); (2) (3) t2≥α, where α is the preset confidence coefficient, let α=3; If D(x) does not have a p2 that satisfies the condition, then there is no maximum office address, and in this case, p2 = NIL is recorded. S4. Determine the relationship between users based on their trusted addresses: For users x and y, if x and y satisfy H(x) = H(y) and H(x) ≠ NIL, then users x and y are neighbors; otherwise, it does not mean they are not neighbors, but there is currently insufficient evidence to prove that x and y are neighbors. If x and y satisfy W(x) = W(y) and W(x) ≠ NIL, then users x and y are colleagues; otherwise, it does not mean they are not colleagues, but there is currently insufficient evidence to prove that x and y are colleagues. Both neighbor and colleague relationships exist simultaneously. S5. Raise the risk control threshold based on relationships with neighbors and colleagues: (a) Order risk control, where x and y are the passenger and driver respectively: (a.1) For colleague relationships, an isolation strategy is adopted. If x and y have a colleague relationship, and x is a passenger and y is a driver, y can be matched with other passengers and x can be matched with other drivers, but the system will not match x and y into a single order, and x and y are not visible to each other. (a.2) For the neighbor relationship, the first isolation strategy is adopted, if x and y have a neighbor relationship, and x is a passenger and y is a driver, x and y can only match the first single, the second single and the subsequent x and y cannot match the single, and x and y cannot see each other; (b) Preferential strength risk control, if x is a passenger: For x, the limit level strategy is adopted, R(H(x)) is the number of passengers of the identified single brushing neighbors or colleagues in the taxi risk control system, if R(H(x))≥β, β is a preset risk threshold, then the passengers in the region are limited to issue high-value and high-discount coupons; (c) Order dispatching and reward restriction, if y is a driver: For y, the demotion strategy is adopted, S(H(y)) is the number of drivers of the identified single brushing neighbors or colleagues in the taxi risk control system, if S(H(x))≥r, r is a preset risk threshold, then the drivers in the region are restricted to: (1) The order dispatching priority is lower than that of normal drivers; (2) Limiting to take high-value orders; (3) Cannot participate in high-reward activities; (4) Put into the single brushing observation object pool, reduce the threshold of the single brushing identification degree of the driver, and it is easier to be judged as a single brushing driver.

Images