Intelligent service matching and pushing method for multiple types of downstream users

Abstract

The invention relates to the technical field of instant logistics, and discloses a business intelligent matching and pushing method for multiple types of downstream users, and the method comprises the following steps: collecting historical order data and driver data, the order data comprising order time, a goods taking place, a goods delivery place and order completion time; the driver data comprises a driver position, a driver familiar area and a driver current state; order features are generated through historical order data, driver features are generated through driver data, the order features and the driver features are combined to generate comprehensive state features, the comprehensive state features are input into an order distribution model, the order distribution model outputs an order distribution strategy, and the order distribution strategy comprises a next order received by each driver; pushing a next order to each driver according to an order distribution strategy output by the order distribution model; the pushing method provided by the invention not only improves the overall efficiency of cargo distribution, but also improves the satisfaction of all participants, and reduces the operation cost at the same time.

Description

Technical Field

[0001] The present invention relates to the field of instant logistics, and more specifically, to a business intelligent matching and push method for multiple types of downstream users. Background Art

[0002] The delivery order of the freight company needs to be pushed to the truck driver through the freight platform. The driver who receives the order will deliver the goods to the designated receiving place. The order distribution methods are mainly divided into two categories: grabbing orders and dispatching orders. The grabbing order mode relies on the personal experience of the truck driver, and is prone to picking and blindly grabbing orders. The delivery efficiency is highly random. Therefore, most freight platforms have begun to choose the dispatching order mode;

[0003] The dispatching process of the freight platform mainly considers the following four factors: the distance between the driver and the pickup point, the distance between the delivery address and the delivery journey, the time when the freight company places the order and the expected delivery time. By calculating these four factors, the optimal delivery efficiency is selected and the order is dispatched to the nearest freight driver;

[0004] For a driver, the delivery efficiency is highest when working in a familiar work area. When dispatching orders through the above-mentioned dispatching process, it is easy for the driver to leave his familiar work area, which will undoubtedly reduce the overall delivery efficiency. Currently, a return order that allows the freight driver to quickly return to the familiar area is given to the freight driver who leaves the familiar area, so that he can return to the familiar area as soon as possible after receiving the return order. However, due to the uncertainty of the distribution of return orders and the difficulty in predicting the regional order volume, the allocation of return orders becomes complicated and difficult to accurately allocate. Summary of the invention

[0005] The present invention provides a business intelligent matching and push method for multiple types of downstream users, which solves the technical problems raised in the above background technology.

[0006] The present invention provides a business intelligent matching and pushing method for multiple types of downstream users, comprising the following steps:

[0007] Step 101, collecting historical order data and driver data, the order data includes order time, pickup location, delivery location, order completion time, the driver data includes driver location, driver familiar area and driver current status, the driver current status includes delivery status and non-delivery status;

[0008] Step 102, generating order features through historical order data, generating driver features through driver data, combining order features and driver features to generate comprehensive state features, and inputting them into an order allocation model, the order allocation model outputs an order allocation strategy, and the order allocation strategy includes the next order received by each driver;

[0009] Step 103: Push the next order to each driver according to the order allocation strategy output by the order allocation model.

[0010] In a preferred embodiment, the calculation formula for the combination of order characteristics and driver characteristics is:

[0011] X=CONCAT(D d , Q s )

[0012] Among them, X represents the comprehensive state characteristics, D d represents the order characteristics, Q s represents driver features, and CONCAT represents vector concatenation.

[0013] In a preferred embodiment, the order allocation model includes an input layer, a policy network, and a value network;

[0014] The strategy network includes a first hidden layer, a second hidden layer and an output layer. The input layer inputs the comprehensive state features into the first hidden layer, the first hidden layer generates the first hidden features which are input into the second hidden layer, the second hidden layer generates the second hidden features which are input into the output layer, and the output layer outputs the probability value of each optional order allocation strategy, and selects the order in the order allocation strategy corresponding to the maximum probability as the next order received by the driver.

[0015] In a preferred embodiment, the calculation formula of the first hidden layer is:

[0016] h1=σ(W1·X+b1)

[0017] Among them, h1 represents the first hidden feature, σ represents the ReLU activation function, W1 represents the first hidden weight matrix, X represents the comprehensive state feature, and b1 represents the bias vector of the first hidden layer;

[0018] The calculation formula for the second hidden layer is:

[0019] h2=σ(W2·h1+b2)

[0020] Wherein, h2 represents the second hidden feature, σ represents the ReLU activation function, W2 represents the second hidden weight matrix, h1 represents the first hidden feature, and b2 represents the bias vector of the second hidden layer;

[0021] The calculation formula of the output layer is:

[0022] p=softmax(W p h2+b p )

[0023] Among them, p represents the probability of an order in the order allocation strategy, softmax represents the softmax function, and W prepresents the policy network weight matrix of the output layer, h2 represents the second hidden feature, b p Represents the bias vector of the output layer.

[0024] In a preferred embodiment, the value network includes a value processing layer and a value generation layer. The input layer inputs the comprehensive state characteristics into the value processing layer, the value processing layer generates a value vector, the value vector is input into the value generation layer, and the value generation layer outputs the estimated value of the current state.

[0025] In a preferred embodiment, the calculation formula of the value processing layer is:

[0026] v=σ(W v1 ·X+b v1 )

[0027] Where v represents the value vector, σ represents the ReLU activation function, and W v1 represents the weight matrix of the value processing layer, X represents the comprehensive state feature, b v1 Represents the bias parameters of the value processing layer;

[0028] The calculation formula of the value generation layer is:

[0029] V(s)=W v2 ·v+b v2

[0030] Among them, V(s) represents the estimated value of the state in state s, W v2 represents the weight matrix of the value generation layer, v represents the value vector, b v2 Represents the bias parameter of the value generation layer.

[0031] In a preferred embodiment, the state space includes order data and driver data;

[0032] Define the action space: Generate the action space by combining the next orders that each driver can accept;

[0033] The orders include return orders and non-return orders;

[0034] It should be noted that each order is regarded as a vector, which points from the merchant point to the user point, and the vector represents the order vector;

[0035] Definition of return order: define a return vector from the driver's current position to the center of the driver's familiar area, and use the upper and lower tangent vectors of the driver's current position and the boundary of the driver's familiar area as the upper and lower limit thresholds of the return vector. If the angle between the vector of any order and the return vector is within the upper and lower limit thresholds of the return vector, then the order is a return order, otherwise it is a non-return order.

[0036] One action includes the next order received by each driver.

[0037] In a preferred embodiment, the feedback reward is calculated after the driver completes the cargo delivery task pushed by the platform for one day. The feedback reward calculation formula is as follows:

[0038] r=ω1R eff +ω2R sat +ω3R cust +ω4R back

[0039] Among them, r represents the total feedback reward value obtained, R eff Represents the delivery efficiency reward value, R sat represents the driver satisfaction reward value, R cust Represents the satisfaction reward value of the receiving user, R back represents the return time reward value, ω1, ω2, ω3, and ω4 represent the first, second, third, and fourth reward parameters respectively;

[0040] The calculation formula for the delivery efficiency bonus value is:

[0041]

[0042] Among them, R eff represents the distribution efficiency reward value, T delivery Indicates the total time from order receipt to delivery completion;

[0043] The calculation formula for the driver satisfaction reward value is:

[0044]

[0045] Among them, R sat represents the driver satisfaction reward value, F order Indicates the frequency of drivers accepting orders, T work represents the working hours of the driver, I in represents the income level of the driver, and α represents the adjustment factor;

[0046] The calculation formula for the recipient user satisfaction reward value is:

[0047]

[0048] Among them, R cust Indicates the recipient user satisfaction reward value, Indicates the feedback rating of the recipient user;

[0049] The calculation formula for the return time bonus value is:

[0050]

[0051] Among them, R back Represents the return time reward value, ρ and τ are both positive numbers, and ρ>τ.

[0052] A business intelligent matching system for multiple types of downstream users, including the following modules:

[0053] Data collection module, used to collect historical order data and driver data;

[0054] A data preprocessing module is used to preprocess the collected data;

[0055] The feature generation module is used to generate order features and driver features and combine them into comprehensive status features;

[0056] The order allocation model module is used to input comprehensive state characteristics and output order allocation strategies;

[0057] The action and feedback module pushes the next order to each driver based on the output order allocation strategy.

[0058] A storage medium stores non-transitory computer-readable instructions. When the non-transitory computer-readable instructions are executed by a computer, the steps in the above-mentioned business intelligent matching and pushing method for multiple types of downstream users can be executed.

[0059] The beneficial effects of the present invention are:

[0060] 1. The push method of the present invention not only improves the overall efficiency of cargo distribution, but also improves the satisfaction of all participants, while reducing operating costs. It is a comprehensive optimization solution;

[0061] 2. The present invention can more efficiently assign orders to suitable drivers through an intelligent matching and push method, thereby reducing delivery time and improving delivery efficiency. It uses the return order priority principle to encourage drivers to return to areas they are familiar with, which can reduce the average delivery time of future orders. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] Figure 1 It is a flow chart of the business intelligent matching and pushing method for multiple types of downstream users of the present invention.

[0063] Figure 2 It is a block diagram of the business intelligent matching system for multiple types of downstream users of the present invention.

[0064] In the figure: 100, data acquisition module; 200, data preprocessing module; 300, feature generation module; 400, order allocation model module; 500, action and feedback module. DETAILED DESCRIPTION

[0065] The subject matter described herein will now be discussed with reference to example embodiments. It should be understood that the discussion of these embodiments is only to enable those skilled in the art to better understand and implement the subject matter described herein, and the functions and arrangements of the elements discussed may be changed without departing from the scope of protection of the contents of this specification. Each example may omit, replace or add various processes or components as needed. In addition, the features described relative to some examples may also be combined in other examples.

[0066] like Figure 1 As shown, a business intelligent matching and pushing method for multiple types of downstream users includes the following steps:

[0067] Step 101, collecting historical order data and driver data, the order data includes order time, pickup location, delivery location, order completion time, the driver data includes driver location, driver familiar area and driver current status, the driver current status includes delivery status and non-delivery status;

[0068] The acquisition of historical order data can be obtained from the background storage data of the cargo platform;

[0069] Driver data information can also be obtained from the backend storage data of the cargo platform;

[0070] The driver's location can be represented by real-time GPS coordinates. The dynamically updated order list in the cargo platform includes the location and time information of the orders to be delivered in the order data. The driver's familiar area can be based on the driver's historical delivery records. Clustering algorithms (such as K-means) are used to determine the driver's main activity area and generate the driver's familiar area. The driver's current status includes whether there are any ongoing orders and whether the driver is in the familiar area. The time information in the order data and driver data includes the current time, predicted pickup and delivery times, and any time window restrictions.

[0071] It should be noted that the collected historical order data and driver data are preprocessed, and the preprocessing methods include:

[0072] Outlier detection and processing, removing or correcting obviously erroneous data points, such as unreasonable coordinates, negative delivery times, etc.

[0073] Data standardization: normalize or standardize numerical features to reduce the impact of scale differences;

[0074] It is important to further clarify that new features can also be constructed, such as distance calculation, speed estimation, and definition of familiar areas (based on the areas where the driver has made the most deliveries in the past).

[0075] Step 102, generating order features through historical order data, generating driver features through driver data, combining order features and driver features to generate comprehensive state features, and inputting them into an order allocation model, the order allocation model outputs an order allocation strategy, and the order allocation strategy includes the next order received by each driver;

[0076] The calculation formula for combining order characteristics and driver characteristics is:

[0077] X=CONCAT(D d , Q s )

[0078] Among them, X represents the comprehensive state characteristics, D d represents the order characteristics, Q s represents driver features, CONCAT represents vector concatenation;

[0079] The order allocation model includes input layer, strategy network and value network;

[0080] The strategy network includes a first hidden layer, a second hidden layer and an output layer. The input layer inputs the comprehensive state feature into the first hidden layer. The first hidden layer generates a first hidden feature which is input into the second hidden layer. The second hidden layer generates a second hidden feature which is input into the output layer. The output layer outputs the probability value of each optional order allocation strategy, and selects the order in the order allocation strategy corresponding to the maximum probability as the next order received by the driver.

[0081] The calculation formula for the first hidden layer is:

[0082] h1=σ(W1·X+b1)

[0083] Among them, h1 represents the first hidden feature, σ represents the ReLU activation function, W1 represents the first hidden weight matrix, X represents the comprehensive state feature, and b1 represents the bias vector of the first hidden layer;

[0084] The calculation formula for the second hidden layer is:

[0085] h2=σ(W2·h1+b2)

[0086] Wherein, h2 represents the second hidden feature, σ represents the ReLU activation function, W2 represents the second hidden weight matrix, h1 represents the first hidden feature, and b2 represents the bias vector of the second hidden layer;

[0087] The calculation formula of the output layer is:

[0088] p=softmax(W p h2+b p )

[0089] Among them, p represents the probability of an order in the order allocation strategy, softmax represents the softmax function, and W p represents the policy network weight matrix of the output layer, h2 represents the second hidden feature, b p Represents the bias vector of the output layer;

[0090] The value network includes a value processing layer and a value generation layer. The input layer inputs the comprehensive state features into the value processing layer, the value processing layer generates a value vector, the value vector is input into the value generation layer, and the value generation layer outputs the estimated value of the current state.

[0091] The calculation formula of the value processing layer is:

[0092] v=σ(W v1 ·X+b v1 )

[0093] Where v represents the value vector, σ represents the ReLU activation function, and W v1 represents the weight matrix of the value processing layer, X represents the comprehensive state feature, b v1 Represents the bias parameters of the value processing layer;

[0094] The calculation formula of the value generation layer is:

[0095] V(s)=W v2 ·v+b v2

[0096] Among them, V(s) represents the estimated value of the state in state s, W v2 represents the weight matrix of the value generation layer, v represents the value vector, b v2 Represents the bias parameters of the value generation layer;

[0097] The state space includes order data and driver data;

[0098] Define the action space: Generate the action space by combining the next orders that each driver can accept;

[0099] The orders include return orders and non-return orders;

[0100] It should be noted that each order is regarded as a vector, which points from the merchant point to the user point, and the vector represents the order vector;

[0101] Definition of a return order: A return vector is defined, pointing from the driver's current position to the center of the driver's familiar area, and the upper and lower tangent vectors based on the driver's current position and the boundary of the driver's familiar area are used as the upper and lower limit threshold ranges of the return vector. If the angle between the vector of any order and the return vector is within the upper and lower limit threshold ranges of the return vector, it means that the order is a return order, otherwise it is a non-return order. That is, for any order, the angle between the order vector and the return vector of the order can be calculated to determine whether it is a return order.

[0102] One action includes the next order received by each driver;

[0103] It should be noted that the next order assigned to the driver is based on the principle of return order priority, unless non-return orders can significantly improve overall efficiency.

[0104] The feedback reward is calculated after the driver completes the one-day cargo delivery task pushed by the platform. The feedback reward calculation formula is as follows:

[0105] r=ω1R eff +ω2R sat +ω3R cust +ω4R back

[0106] Among them, r represents the total feedback reward value obtained, R eff Represents the delivery efficiency reward value, R sat represents the driver satisfaction reward value, R cust Represents the satisfaction reward value of the receiving user, R back represents the return time reward value, ω1, ω2, ω3, and ω4 represent the first, second, third, and fourth reward parameters respectively;

[0107] The calculation formula for the delivery efficiency bonus value is:

[0108]

[0109] Among them, R eff represents the distribution efficiency reward value, T delivery Indicates the total time from order receipt to delivery completion;

[0110] The calculation formula for the driver satisfaction reward value is:

[0111]

[0112] Among them, R sat represents the driver satisfaction reward value, F order Indicates the frequency of drivers accepting orders, T work represents the working hours of the driver, I in represents the income level of the driver, and α represents the adjustment factor;

[0113] The calculation formula for the recipient user satisfaction reward value is:

[0114]

[0115] Among them, R cust Indicates the recipient user satisfaction reward value, Indicates the feedback rating of the recipient user;

[0116] The calculation formula for the return time bonus value is:

[0117]

[0118] Among them, R back represents the return time bonus value, ρ and τ are both positive numbers, and ρ>τ;

[0119] The purpose of designing the reward function of the present invention is to balance the conflicts among delivery efficiency, driver satisfaction, recipient user satisfaction and return time, with the goal of maximizing the number of orders completed per unit time and encouraging drivers to return to familiar areas, so as to reduce the average delivery time of future orders;

[0120] The training of the order allocation model consists of the following steps:

[0121] Step 1: Randomly initialize the parameters of the policy network and the value network;

[0122] Step 2: reset the environment and obtain the first state;

[0123] Step 3: Get the feedback r based on the interaction between the action and the environment and the output of the current policy network t , the next state s t+1 and whether the termination condition done is triggered, the termination condition is that all drivers return to the corresponding driver familiar area;

[0124] Experience t ,a t ,r t ,s t+1 , done is stored in the experience pool, an experience is a four-tuple s t ,a t ,r t ,s t+1 or t ,a t ,r t ,done;

[0125] Step 4: Update the strategy network and value network;

[0126] The method for updating the policy network comprises the following steps:

[0127] 1. Randomly sample a certain number of experience samples from the experience replay pool to form a batch dataset, and calculate the advantage function of the experience samples;

[0128] δ t = r t + γV(s t+1 ) - V(s t )

[0129]

[0130] where r t represents the immediate reward obtained at time step t, V(s t ) represents the value function estimate of state s t , δ t represents the TD error, represents the n-step Monte Carlo return, represents the estimated value of the advantage function at time step t, γ represents the discount factor, and λ represents the truncation coefficient;

[0131] 2. Use the current policy network to calculate the action probability π(a t | s t ; θ), use the previously updated policy network to calculate the action probability π(a t | s t ; θ oid ), and calculate the ratio

[0132] 3. Use the clip objective function of PPO to limit the update step size;

[0133] The calculation formula of the policy loss function is:

[0134] L CLIP (θ) = E t [min(r t (θ)A t , clip(r t (θ), 1 - ∈, 1 + ∈)A t )]

[0135] where A t represents the estimate of the advantage function, and ∈ represents the clip parameter;

[0136] 4. Calculate the gradient of the policy loss function with respect to the policy network parameter θ, and use the Adam optimizer to update the parameter;

[0137] The method for updating the value network includes the following steps:

[0138] 1. The policy loss function of the value network is:

[0139] Lv(φ) = E t [(V(s t |φ) - G t ) 2

[0140] where G t represents the actual return, and V(s t |φ) represents the prediction of the value network;

[0141] 2. Calculate the gradient of the loss function of the value network with respect to the parameter φ, and update the parameter through the Adam optimizer;

[0142] Step Five, update the policy network and the value network every fixed number of network update times so that their parameters are the same as those of the current order allocation model;

[0143] Step Six, terminate the steps after the policy network and the value network converge or the number of update times reaches the set value.

[0144] Step 103, push the next order to each driver according to the order allocation policy output by the order allocation model.

[0145] As Figure 2 shown, a business intelligent matching system for multiple types of downstream users includes the following modules:

[0146] Data acquisition module 100, used to acquire historical order data and driver data;

[0147] Data preprocessing module 200, used to preprocess the acquired data;

[0148] Feature generation module 300, used to generate order features and driver features, and combine them into comprehensive state features;

[0149] Order allocation model module 400, used to input the comprehensive state features and output the order allocation policy;

[0150] Action and feedback module 500, according to the output order allocation policy, push the next order to each driver.

[0151] A storage medium stores non - transient computer - readable instructions, which can execute the steps in the above - mentioned business intelligent matching and pushing method for multiple types of downstream users when executed by a computer.

[0152] ​The above describes an embodiment of the present embodiment, but the present embodiment is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the guidance of the present embodiment, ordinary technicians in this field can also make many forms, all of which are within the protection of the present embodiment.

Claims

1. A business intelligent matching and pushing method for multiple types of downstream users, characterized in that: The following steps are involved: Step 101, collecting historical order data and driver data, the order data includes order time, pickup location, delivery location, order completion time, the driver data includes driver location, driver familiar area and driver current status, the driver current status includes delivery status and non-delivery status; Step 102, generating order features through historical order data, generating driver features through driver data, combining order features and driver features to generate comprehensive state features, and inputting them into an order allocation model, the order allocation model outputs an order allocation strategy, and the order allocation strategy includes the next order received by each driver; Step 103: Push the next order to each driver according to the order allocation strategy output by the order allocation model.

2. According to claim 1, a method for intelligent matching and pushing services for multiple types of downstream users is characterized in that: The calculation formula for the combination of order characteristics and driver characteristics is: X=CONCAT(D d ,Q s ) Among them, X represents the comprehensive state characteristics, D d represents the order characteristics, Q s represents driver features, and CONCAT represents vector concatenation.

3. The method for intelligent business matching and pushing for multiple types of downstream users according to claim 2 is characterized in that: The order allocation model includes input layer, strategy network and value network; The strategy network includes a first hidden layer, a second hidden layer and an output layer. The input layer inputs the comprehensive state features into the first hidden layer, the first hidden layer generates the first hidden features which are input into the second hidden layer, the second hidden layer generates the second hidden features which are input into the output layer, and the output layer outputs the probability value of each optional order allocation strategy, and selects the order in the order allocation strategy corresponding to the maximum probability as the next order received by the driver.

4. The method for intelligent business matching and pushing for multiple types of downstream users according to claim 3 is characterized in that: The calculation formula for the first hidden layer is: h1=σ(W1·X+b1) Among them, h1 represents the first hidden feature, σ represents the ReLU activation function, W1 represents the first hidden weight matrix, X represents the comprehensive state feature, and b1 represents the bias vector of the first hidden layer; The calculation formula for the second hidden layer is: h2=σ(W2·h1+b2) Wherein, h2 represents the second hidden feature, σ represents the ReLU activation function, W2 represents the second hidden weight matrix, h1 represents the first hidden feature, and b2 represents the bias vector of the second hidden layer; The calculation formula of the output layer is: p=softmax(W p h2+b p ) Among them, p represents the probability of an order in the order allocation strategy, softmax represents the softmax function, and W p represents the policy network weight matrix of the output layer, h2 represents the second hidden feature, b p Represents the bias vector of the output layer.

5. The method for intelligent business matching and pushing for multiple types of downstream users according to claim 4 is characterized in that: The value network includes a value processing layer and a value generation layer. The input layer inputs the comprehensive state features into the value processing layer, the value processing layer generates a value vector, the value vector is input into the value generation layer, and the value generation layer outputs the estimated value of the current state.

6. The method for intelligent business matching and pushing for multiple types of downstream users according to claim 5, characterized in that: The calculation formula of the value processing layer is: v=σ(W v1 ·X+b v1 ) Where v represents the value vector, σ represents the ReLU activation function, and W v1 represents the weight matrix of the value processing layer, X represents the comprehensive state feature, b v1 Represents the bias parameters of the value processing layer; The calculation formula of the value generation layer is: V(s)=W v2 ·v+b v2 Among them, V(s) represents the estimated value of the state in state s, W v2 represents the weight matrix of the value generation layer, v represents the value vector, b v2 Represents the bias parameter of the value generation layer.

7. The method for intelligent business matching and pushing for multiple types of downstream users according to claim 6, characterized in that: The state space includes order data and driver data; Define the action space: Generate the action space by combining the next orders that each driver can accept; The orders include return orders and non-return orders; It should be noted that each order is regarded as a vector, which points from the merchant point to the user point, and the vector represents the order vector; Definition of return order: define a return vector from the driver's current position to the center of the driver's familiar area, and use the upper and lower tangent vectors of the driver's current position and the boundary of the driver's familiar area as the upper and lower limit thresholds of the return vector. If the angle between the vector of any order and the return vector is within the upper and lower limit thresholds of the return vector, then the order is a return order, otherwise it is a non-return order. One action includes the next order received by each driver.

8. The method for intelligent business matching and pushing for multiple types of downstream users according to claim 7, characterized in that: The feedback reward is calculated after the driver completes the one-day cargo delivery task pushed by the platform. The feedback reward calculation formula is as follows: r=ω1R eff +ω2R sat +ω3R cust +ω4R back Among them, r represents the total feedback reward value obtained, R eff Represents the delivery efficiency reward value, R sat represents the driver satisfaction reward value, R cust Represents the satisfaction reward value of the receiving user, R back represents the return time reward value, ω1, ω2, ω3, and ω4 represent the first, second, third, and fourth reward parameters respectively; The calculation formula for the delivery efficiency bonus value is: Among them, R eff represents the distribution efficiency reward value, T delivery Indicates the total time from order receipt to delivery completion; The calculation formula for the driver satisfaction reward value is: Among them, R sat represents the driver satisfaction reward value, F order Indicates the frequency of drivers accepting orders, T work represents the working hours of the driver, I in represents the income level of the driver, and α represents the adjustment factor; The calculation formula for the recipient user satisfaction reward value is: Among them, R cust Indicates the reward value of the recipient’s satisfaction. Indicates the feedback rating of the recipient user; The calculation formula for the return time bonus value is: Among them, R back Represents the return time reward value, ρ and τ are both positive numbers, and ρ>τ.

9. A business intelligent matching system for multiple types of downstream users, characterized in that: Includes the following modules: Data collection module, used to collect historical order data and driver data; A data preprocessing module is used to preprocess the collected data; The feature generation module is used to generate order features and driver features and combine them into comprehensive status features; The order allocation model module is used to input comprehensive state characteristics and output order allocation strategies; The action and feedback module pushes the next order to each driver based on the output order allocation strategy.

10. A storage medium storing non-transitory computer-readable instructions, which, when executed by a computer, can execute the steps in the business intelligent matching and pushing method for multiple types of downstream users as described in claim 8.

Images