Vehicle-cargo matching method, device, storage medium and electronic device integrating vehicle owner preferences

Abstract

The present invention provides a vehicle-cargo matching method, device, storage medium and electronic device that integrates the preferences of vehicle owners, and relates to the field of vehicle-cargo matching technology. The present invention makes full use of the historical behavior sequence information of vehicle owners. Historical behavior is a reflection of the preferences of vehicle owners. Some preferences are static or change slowly over time, and some preferences are dynamic and change in real time. They show the preferences of vehicle owners from different angles, so combining these two different types of preference forms can better reflect the personalized preferences of vehicle owners. On this basis, the present invention proposes a new vehicle-cargo matching method idea, namely "prediction + decision-making", which effectively mines the long-term and short-term preferences of vehicle owners by establishing a deep learning model, improves the accuracy of prediction, and then integrates the results of preference prediction into the construction process of the vehicle-cargo matching decision model, effectively improving the efficiency of vehicle-cargo matching, reducing matching costs, and improving vehicle owner satisfaction and stickiness, so it is more valuable in practical applications.

Description

Technical Field

[0001] The present invention relates to the technical field of vehicle-cargo matching, and in particular to a vehicle-cargo matching method, device, storage medium and electronic device that integrate vehicle owner preferences. Background Art

[0002] The development of social economy has accelerated the circulation of goods, services and information, thereby expanding the society's demand for logistics and transportation, development and innovation. With the development of online freight platforms, road freight business has exploded. It is extremely important to help car owners find the source of goods they are interested in among the massive amount of information. The problem of vehicle-cargo matching is a key issue in road freight logistics.

[0003] Existing approaches to vehicle-cargo matching include establishing a matching index system based on vehicle and cargo information, recommending vehicle and cargo information based on the matching results, and then achieving vehicle-cargo matching. Other approaches derive a vehicle owner's cargo preference feature vector from a cargo feature vector associated with the vehicle owner's behavior, and then calculate the matching degree between the vehicle owner and the cargo source based on this feature vector.

[0004] These methods don't fully utilize the historical interaction sequences of vehicle owners. Since vehicle owners' preferences are constantly evolving, capturing their dynamic changes over time can enhance their expressive power and improve prediction accuracy. Furthermore, incorporating personalized preferences into vehicle-cargo matching decision models can help improve transportation resource utilization, reduce idle distance, and lower transportation costs. Summary of the Invention

[0005] (1) Technical problems solved

[0006] In response to the shortcomings of the existing technology, the present invention provides a vehicle-cargo matching method, device, storage medium and electronic device that integrates vehicle owner preferences, solving the problems that the existing technology does not fully consider the dynamic changes of vehicle owner preferences over time, the recommended cargo sources are not accurate, and the transportation resource utilization efficiency is low.

[0007] (2) Technical solution

[0008] To achieve the above objectives, the present invention is implemented through the following technical solutions:

[0009] In a first aspect, a vehicle-cargo matching method integrating vehicle owner preferences is provided, the method comprising:

[0010] Obtain vehicle-cargo related data; the vehicle-cargo related data includes: vehicle owner data, historical cargo source data, vehicle owner historical interaction behavior sequence data, and cargo source data to be matched; and the vehicle owner historical interaction behavior sequence data is the cargo source data that the vehicle owner actually transported and gave positive reviews on the online freight platform in chronological order;

[0011] Predicting the vehicle owner's comprehensive preference for matching cargo sources based on vehicle-cargo related data; the comprehensive preference is the probability of the vehicle owner interacting with the cargo source to be matched;

[0012] Obtain a vehicle-cargo matching decision model that considers comprehensive preferences, vehicle loading rate, and matching costs;

[0013] The optimal solution of the vehicle-cargo matching decision model is obtained as the vehicle-cargo matching result.

[0014] Furthermore, the prediction of the vehicle owner's comprehensive preference for matching cargo sources based on vehicle-cargo related data includes:

[0015] Predict the comprehensive preferences of car owners for matching cargo sources based on their short-term and long-term preferences.

[0016] Furthermore, the prediction of the comprehensive preference of the car owner for matching goods based on the car owner's short-term preference and long-term preference includes:

[0017] Obtain the trained short-term preference representation module, long-term preference representation module, weight coefficients and trained multi-layer perceptron;

[0018] The trained short-term preference representation module and long-term preference representation module are used to obtain the car owner's short-term preference vector and long-term preference vector for matching cargo sources. The short-term preference vector and long-term preference vector are weighted and fused based on the weight coefficient to obtain a comprehensive preference vector.

[0019] The comprehensive preference vector and the cargo source vector are used as the input of the trained multi-layer perceptron to obtain the comprehensive preference of the car owner for matching cargo sources.

[0020] Furthermore, the vehicle-cargo matching decision model includes:

[0021] Objective function:

[0022] maxZ=w1P+w2LR+w3(1-MC)

[0023] in,

[0024] w1 represents the emphasis the platform places on the comprehensive preferences of car owners;

[0025] w2 represents the platform's preference for vehicle loading rate;

[0026] w3 indicates the platform's emphasis on reducing matching costs;

[0027] P, LR, and MC represent the comprehensive preference score rate, loading rate, and matching cost, respectively, w1, w2, w3∈[0,1], w1+w2+w3=1;

[0028]

[0029]

[0030]

[0031] pr vc ∈[0,1] represents the comprehensive preference of the car owner v for the cargo source c;

[0032] x vc is the decision variable for matching the car owner v with the cargo source c; and

[0033]

[0034] Then we get the vehicle-cargo matching decision variable matrix:

[0035]

[0036] m and n are the number of car owners to be matched and the number of goods to be matched respectively;

[0037] d vc is the distance between the current location of the vehicle owner v and the starting point of the cargo source c;

[0038] l v ,w v ,b v The length, load capacity and volume of the vehicle owned by the vehicle owner v;

[0039] l c ,w c ,b c is the length, weight and volume of the source c;

[0040] D is the maximum matching distance given;

[0041] Constraints:

[0042] Condition 1: Each car owner is matched with at most one source of goods, expressed as:

[0043]

[0044] Condition 2: Each source of goods matches at most one vehicle owner, expressed as:

[0045]

[0046] Condition 3: The length of the cargo source matched by each vehicle owner does not exceed the length of the vehicle, expressed as:

[0047]

[0048] Condition 4: The weight of the cargo matched by each vehicle owner does not exceed the vehicle load, expressed as:

[0049]

[0050] Condition 5: The volume of cargo matched by each vehicle owner does not exceed the vehicle capacity, expressed as:

[0051]

[0052] Condition 6: Each vehicle owner’s distance to the starting point of the matching cargo source does not exceed the maximum matching distance, expressed as:

[0053]

[0054] Among them, V and C are the set of car owners and the set of cargo sources respectively.

[0055] Furthermore, an ant colony optimization algorithm is used to obtain the optimal solution of the vehicle-cargo matching decision model.

[0056] Furthermore, the LSTM model is trained based on vehicle-cargo related data to obtain a short-term preference representation module.

[0057] Furthermore, long-term preferences are determined based on the similarity between the car owner's historical interaction behavior sequence data and the source of goods to be matched.

[0058] In a second aspect, a vehicle-cargo matching device that integrates vehicle owner preferences is provided, the device comprising:

[0059] The vehicle-cargo data acquisition module is used to acquire vehicle-cargo data; the vehicle-cargo data includes: vehicle owner data, historical cargo data, vehicle owner historical interaction behavior sequence data, and cargo data to be matched; the vehicle owner historical interaction behavior sequence data is the cargo data of the vehicle owner who actually transported and gave positive reviews on the online freight platform in chronological order;

[0060] A comprehensive preference acquisition module is used to predict the vehicle owner's comprehensive preference for matching cargo sources based on vehicle-cargo related data; the comprehensive preference is the probability of the vehicle owner interacting with the cargo source to be matched;

[0061] The vehicle-cargo matching decision model acquisition module is used to obtain the vehicle-cargo matching decision model that considers comprehensive preferences, vehicle loading rate, and matching cost;

[0062] The decision model solving module is used to obtain the optimal solution of the vehicle-cargo matching decision model as the vehicle-cargo matching result.

[0063] In a third aspect, a computer-readable storage medium is provided, which stores a computer program for vehicle-cargo matching, wherein the computer program enables a computer to execute the above-mentioned vehicle-cargo matching method that integrates vehicle owner preferences.

[0064] In a fourth aspect, an electronic device is provided, including:

[0065] one or more processors;

[0066] Memory; and

[0067] One or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, the programs including a method for executing the above-mentioned vehicle-cargo matching method integrating vehicle owner preferences.

[0068] (3) Beneficial effects

[0069] The present invention provides a vehicle-cargo matching method, device, storage medium, and electronic device that integrates vehicle owner preferences. Compared with existing technologies, it has the following advantages:

[0070] 1) The present invention makes full use of the owner's historical behavior sequence information. Historical behavior is a reflection of the owner's preferences. Some preferences are static or change slowly over time, while some preferences are dynamic and change in real time. They show the owner's preferences from different perspectives, so combining these two different types of preference forms can better reflect the owner's personalized preferences. On this basis, the present invention proposes a new vehicle-cargo matching method idea, namely "prediction + decision-making". By establishing a deep learning model to effectively explore the owner's long-term and short-term preferences, improve the accuracy of the prediction, and then integrate the results of the preference prediction into the construction process of the vehicle-cargo matching decision model, it effectively improves the efficiency of vehicle-cargo matching, reduces matching costs, and enhances the owner's satisfaction and stickiness. Therefore, it is more valuable in practical applications.

[0071] 2) This invention uses a deep learning model to model owner preferences and normalizes the time intervals between behavior sequences, enhancing the model's perception of time and enabling more accurate modeling of owners' short-term interests. Furthermore, the introduction of an attention mechanism adaptively calculates the owner's preference representation vector, effectively filtering out historical behaviors that are most relevant to predicted supply, and further identifying the importance of long-term and short-term preferences for current behavior prediction, thereby improving the accuracy of preference prediction.

[0072] 3) This invention obtains vehicle-cargo data, mines the vehicle owner's personalized preferences based on this data, sets the loading rate and matching cost based on basic vehicle and cargo information, and establishes a vehicle-cargo matching decision model based on the predicted personalized preferences, loading rate, and matching cost. Constraints are set to optimize overall efficiency, and the optimal solution for the vehicle-cargo matching model is obtained using an ant colony optimization algorithm, which is the optimal vehicle-cargo matching solution. Compared to traditional methods that only consider the degree of matching, this invention can more accurately consider the vehicle owner's personalized preferences, improve vehicle owner satisfaction, increase vehicle owner stickiness to the platform, improve overall resource utilization, reduce costs, and improve efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0073] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0074] Figure 1 This is a flow chart of vehicle-cargo matching according to an embodiment of the present invention;

[0075] Figure 2 Flowchart for predicting comprehensive preference in an embodiment of the present invention. DETAILED DESCRIPTION

[0076] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention are clearly and completely described. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. All other embodiments derived by persons of ordinary skill in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0077] The embodiments of the present application provide a vehicle-cargo matching method, device, storage medium, and electronic device that integrate vehicle owner preferences, thereby solving the problems in the prior art of insufficiently considering the dynamic changes of vehicle owner preferences over time, low accuracy of recommended cargo sources, and low efficiency in transportation resource utilization.

[0078] In order to better understand the above technical solution, the above technical solution will be described in detail below with reference to the accompanying drawings and specific implementation methods.

[0079] Example 1:

[0080] like Figure 1 As shown, the present invention provides a vehicle-cargo matching method integrating vehicle owner preferences, the method being executed by a computer and comprising:

[0081] Obtain vehicle-cargo related data; the vehicle-cargo related data includes: vehicle owner data, historical cargo source data, vehicle owner historical interaction behavior sequence data, and cargo source data to be matched; and the vehicle owner historical interaction behavior sequence data is the cargo source data that the vehicle owner actually transported and gave positive reviews on the online freight platform in chronological order;

[0082] Predicting the vehicle owner's comprehensive preference for matching cargo sources based on vehicle-cargo related data; the comprehensive preference is the probability of the vehicle owner interacting with the cargo source to be matched;

[0083] Obtain a vehicle-cargo matching decision model that considers comprehensive preferences, vehicle loading rate, and matching costs;

[0084] The optimal solution of the vehicle-cargo matching decision model is obtained as the vehicle-cargo matching result.

[0085] The beneficial effects of this embodiment are:

[0086] 1) The present invention makes full use of the historical behavior sequence information of the car owner. Historical behavior is a reflection of the user's preferences. Some preferences are static or change slowly over time, while some preferences are dynamic and change in real time. They show the user's preferences from different angles, so combining these two different types of preference forms can better reflect the car owner's personalized preferences. On this basis, the present invention proposes a new vehicle-cargo matching method idea, namely "prediction + decision-making". By establishing a deep learning model to effectively explore the user's long-term and short-term preferences, improve the accuracy of the prediction, and then integrate the results of the preference prediction into the construction process of the vehicle-cargo matching decision model, it effectively improves the efficiency of vehicle-cargo matching, reduces matching costs, and enhances user satisfaction and stickiness. Therefore, it is more valuable in practical applications.

[0087] The implementation process of the embodiment of the present invention is described in detail below:

[0088] Typically, in online freight platforms, shippers publish cargo information through their app and wait for calls from vehicle owners. Vehicle owners then view cargo information in the information hall and, upon finding cargo they find interesting, contact the shipper by phone. Once both parties reach an agreement, a transaction can be concluded. The proposed vehicle-cargo matching method primarily identifies and recommends cargo that the vehicle owner is interested in, based on the vast amount of data in the information hall. This method facilitates transactions and achieves a perfect match between vehicle and cargo.

[0089] The owner's historical interaction behavior sequence data refers to the source data of goods that the owner actually transported and gave positive reviews, with a score of 1. During model training, negative samples without interaction are constructed, and their score is 0. Sequence data refers to behavioral data with timestamps, arranged in chronological order.

[0090] S1. Obtain vehicle and cargo related data.

[0091] In specific implementation, vehicle and cargo related data include:

[0092] Owner data, represented as V = {v1, v2, ..., v m}; Specifically, the owner v k Contains information such as the owner's ID, gender, age group, home address, vehicle length, load, and capacity;

[0093] The source data to be matched is represented as C = {c1,c2,...,c n};

[0094] Historical supply data, expressed as HC = {hc1,hc2,...,hcp};

[0095] Specific, to be matched source c k And historical sources hc k All contain information such as cargo source ID, starting point, destination, cargo type, length, weight, volume, freight, loading and unloading type, etc.

[0096] The historical interaction behavior sequence of the car owner v∈V is expressed as:

[0097]

[0098] in, Indicates that the owner v is at time About the source of goods The interactive behavior is the data of the goods that are actually transported and given good reviews. Such goods are positive samples and have a score of 1.

[0099] And B v The elements in are arranged in chronological order.

[0100] S2. Predicting the vehicle owner's comprehensive preference for matching cargo sources based on vehicle-cargo related data, where the comprehensive preference is the probability of the vehicle owner interacting with the cargo source.

[0101] In specific implementation, this embodiment does not specifically limit the method for predicting the comprehensive preference of the vehicle owner for matching cargo sources.

[0102] For example, the method (DIEN model) in the literature (Zhou G, Mou N, Fan Y, et al. Deep interest evolution network for click-through rate prediction [C] / / Proceedings of the AAAI conference on artificial intelligence. 2019, 33(01): 5941-5948) can be used to predict the comprehensive preferences of car owners for matching sources. After the time interval is normalized, the improvements to the LSTM model can also be replaced with improvements to the GRU and bidirectional GRU models.

[0103] The following is only one feasible method to obtain comprehensive preferences.

[0104] like Figure 2 As shown, in order to better describe the preferences of car owners, this embodiment comprehensively considers the impact of car owners' long-term preferences and short-term preferences on comprehensive preferences. Long-term preferences are static or change slowly over time, while short-term preferences are dynamic and change in real time. The specific steps are as follows:

[0105] S201. Obtain a trained short-term preference representation module, a long-term preference representation module, a weight coefficient, and a trained multi-layer perceptron.

[0106] The short-term preference representation module uses the LSTM model to input the owner's historical interaction behavior sequence data and the source of goods to be predicted, and output the owner's short-term preference vector.

[0107] And the specific equation of the LSTM model is as follows:

[0108] i k =σ s (x k W i +h k-1 U i +b i )

[0109] f k =σ s (x k W f +h k-1 U f +b f )

[0110] o k =σ s (x k W o +h k-1 U o +b o )

[0111]

[0112]

[0113] in,

[0114] W * , U * ∈R D×D D, is the dimension of the embedding layer and the hidden layer;

[0115] i k 、f k 、o k They are input gate, forget gate, and output gate respectively;

[0116] x k is the input of the kth source;

[0117] b * is the bias vector;

[0118] c k It is the cell state;

[0119] h k is the kth hidden state;

[0120] σ s is the sigmoid activation function;

[0121] σ t is the tanh activation function;

[0122] It is the element-wise multiplication operation.

[0123] Considering the non-uniform time intervals in the historical interaction behavior sequence, in order to normalize the time interval and enhance the time sensitivity of the model, a time-aware feature is introduced. Characterize the relative time interval of historical interaction behavior sequences and introduce time-aware features Characterization k State and state to be predicted t p The absolute time interval, is the average time interval of all historical behaviors of the owner v, which is expressed as follows:

[0124]

[0125]

[0126] in,

[0127] W r 、W s ∈R D ;

[0128] σ t is the tanh activation function.

[0129] The above time-aware features are converted into a dense vector through the embedding layer:

[0130]

[0131]

[0132] in,

[0133] W *r 、W *s ∈R D×D ;

[0134] σ s is the sigmoid activation function.

[0135] Based on this, k 、c k Updated to:

[0136]

[0137]

[0138] To highlight the influence of cargo sources in different historical behaviors on the cargo sources to be matched and adaptively represent the interests of car owners, an attention mechanism is introduced. The attention score can be calculated as follows:

[0139]

[0140] in,

[0141]

[0142] x j Embedding vector representing the source input;

[0143] e tg Represents the embedding vector of the source to be matched, when x k With e tg The more relevant, the attention score a k The larger the value, the higher the source vector. The source vector is obtained by densifying the sparse source features of the input through the embedding layer. Embedding represents the general operation of processing the data input to the model.

[0144] Use attention scores to adjust cell state and hidden state:

[0145]

[0146]

[0147] The last hidden state is selected as the short-term preference representation of the owner v, as follows:

[0148]

[0149] The long-term preference representation module is used to input the owner's historical interaction behavior data and the data of the cargo source to be predicted, and output the owner's adaptive long-term preference vector. The owner's long-term preference is adaptively weighted based on the owner's historical interaction cargo source and the cargo source to be predicted:

[0150]

[0151] in, Indicates the similarity between the historical interaction sources of vehicle owner v and the source to be matched. The calculation method is as follows:

[0152]

[0153] in, x j Represents the embedding vector of the source input, xtg Represents the embedding vector of the source to be matched, x j 、x tg It is obtained by densifying the input sparse source vector through the embedding layer.

[0154] For the multi-layer perceptron (MLP), the activation function uses the sigmoid function. The fused comprehensive preference vector and the supply vector are concatenated as input, and the output is the comprehensive preference prediction result.

[0155] And the negative log-likelihood function shown below is used as the loss function:

[0156]

[0157] in,

[0158] N represents the number of historical behavior data of all car owners as training sets,

[0159] In this embodiment, the historical interaction behavior sequence information of the car owner is divided into a training set, a validation set, and a test set in a ratio of 8:1:1 according to the chronological order.

[0160] y j =1 means that the owner of the vehicle has interacted with the source of goods in the training set, y j =0 means the owner has no interaction with the source of goods. It represents the comprehensive preference prediction result, that is, the probability that the car owner interacts with the source of goods.

[0161] Regularization terms are added to prevent overfitting of the model, and the optimization goal is to minimize the loss function and the regularization term:

[0162]

[0163] Where w is a set of trainable parameters, and λ>0. The maximum number of iterations, epoch_num, is set to 100. The objective function J is trained using backpropagation and gradient descent. In this embodiment, the gradient descent method uses the Adam optimization algorithm with a learning rate η = 1e-4, exponential decay rates β1 = 0.9, and β2 = 0.999. When the number of iterations reaches epoch_num, training is stopped to minimize J.

[0164] S202. Utilize the trained short-term preference representation module and long-term preference representation module to obtain the car owner's short-term preference vector and long-term preference vector for matching cargo sources, and perform weighted fusion on the short-term preference vector and the long-term preference vector based on a weight coefficient to obtain the car owner's comprehensive preference vector for matching cargo sources.

[0165] In the specific implementation, the short-term preference representation module and the long-term preference representation module trained above are used to obtain the short-term preference representation vector of the car owner v for the matching cargo source c. and long-term preference representation vector

[0166] Considering that long-term and short-term preferences have different degrees of influence on the driver's behavior, this embodiment adaptively learns the weight coefficient a of the two and then weights the long-term and short-term preferences together. The calculation method is as follows:

[0167]

[0168]

[0169] in,

[0170] They are the connection between the long-term preference vector of car owner v, the short-term preference vector of car owner v, and the context. The context mainly includes the time interval and location of the car owner's historical interaction behavior. After the preference vector is trained, it does not change with the source of goods c to be matched.

[0171] σ s is the sigmoid activation function;

[0172] W a ∈R D×D is a trainable parameter;

[0173] b a is the bias vector.

[0174] The fused preference vector and supply vector are used as the input of the trained multi-layer perceptron, and the comprehensive preference prediction result pr is obtained through the multi-layer perceptron vc :

[0175] pr vc =MLP([p v ,x tg ]

[0176] Among them, MLP is a multi-layer perceptron, and the activation function uses the sigmoid function. v ,e tg ] is the concatenation of the owner's preference vector and the cargo source vector, thus obtaining the owner's comprehensive preference prediction set for matching cargo sources:

[0177]

[0178] Among them, pr vc ∈[0,1] represents the comprehensive preference of the car owner v for the cargo source c.

[0179] S3. Obtain a vehicle-cargo matching decision model that considers comprehensive preferences, vehicle loading rate, and matching costs.

[0180] In specific implementation, the vehicle-cargo matching decision model includes an objective function and constraints, among which:

[0181] Objective function:

[0182] maxZ=w1P+w2LR+w3(1-MC)

[0183] in,

[0184] w1 represents the emphasis the platform places on the comprehensive preferences of car owners;

[0185] w2 represents the platform's preference for vehicle loading rate;

[0186] w3 indicates the platform's emphasis on reducing matching costs;

[0187] P, LR, and MC represent the comprehensive preference score rate, loading rate, and matching cost, respectively, w1, w2, w3∈[0,1], w1+w2+w3=1;

[0188]

[0189]

[0190]

[0191] pr vc ∈[0,1] represents the comprehensive preference of the car owner v for the cargo source c;

[0192] x vc is the decision variable for matching the car owner v with the cargo source c; and

[0193]

[0194] Then the vehicle-cargo matching decision variable matrix is ​​formed:

[0195]

[0196] m and n are the number of car owners to be matched and the number of goods to be matched respectively;

[0197] d vc is the distance between the current location of the vehicle owner v and the starting point of the cargo source c;

[0198] l v ,w v ,b v The length, load capacity and volume of the vehicle owned by the vehicle owner v;

[0199] l c,w c ,b c is the length, weight and volume of the source c;

[0200] D is the maximum matching distance given;

[0201] Constraints:

[0202] Condition 1: Each car owner is matched with at most one source of goods, expressed as:

[0203]

[0204] Condition 2: Each source of goods matches at most one vehicle owner, expressed as:

[0205]

[0206] Condition 3: The length of the cargo source matched by each vehicle owner does not exceed the length of the vehicle, expressed as:

[0207]

[0208] Condition 4: The weight of the cargo matched by each vehicle owner does not exceed the vehicle load, expressed as:

[0209]

[0210] Condition 5: The volume of cargo matched by each vehicle owner does not exceed the vehicle capacity, expressed as:

[0211]

[0212] Condition 6: Each vehicle owner’s distance to the starting point of the matching cargo source does not exceed the maximum matching distance, expressed as:

[0213]

[0214] Among them, V and C are the set of car owners and the set of cargo sources respectively.

[0215] S4. Using a solver, an optimal solution of the vehicle-cargo matching decision model is obtained as a vehicle-cargo matching result.

[0216] In specific implementation, there is no limitation on the solution method of the vehicle-cargo matching decision model, and the vehicle-cargo matching decision model can also be solved based on the improved ant colony algorithm.

[0217] For example, Ling H, Fu Y, Hua M, et al. An Adaptive Parameter Controlled AntColony Optimization Approach for Peer-to-Peer Vehicle and Cargo Matching[J].

[0218] The following only gives a feasible method to solve the vehicle-cargo matching decision model.

[0219] Specifically, the existing ant colony optimization algorithm can be used to obtain the optimal solution of the model, and the steps are as follows:

[0220] a1. Consider a vehicle-cargo matching solution as a path from the starting point to the end point of an ant. Therefore, the process of the algorithm seeking the optimal vehicle-cargo matching solution is the process of the ant seeking the optimal path.

[0221] a2. Initialization parameters, including: ant colony size m, number of cities n, pheromone concentration matrix τ, heuristic information matrix η, pheromone importance factor α, heuristic function importance factor β, pheromone volatility factor ρ, optimal value Z (0) , the optimal solution X (0) , maximum number of iterations Max_iter.

[0222] a3. For each ant, randomly select a city as its starting point based on:

[0223]

[0224] Calculate the transition probability between the ants and the remaining cities and make a selection, allow k is a collection of cities that the ants have not yet reached. As the ants visit again and again, allow k The number of elements in the table continues to decrease until all cities have been visited.

[0225] a4. Record the access path length and route of each ant, and promptly update the optimal access path length and corresponding route in the current iteration, while also recording the changes in pheromone concentration on the global path.

[0226] a5. If the current number of iterations has not reached the preset maximum number of iterations, the number of iterations iter = iter + 1, and the process returns to a3; otherwise, the iteration ends, the optimal access route is output, and the optimal solution for vehicle-cargo matching is obtained.

[0227] Example 2:

[0228] A vehicle-cargo matching device integrating vehicle owner preferences, the device comprising:

[0229] The vehicle-cargo data acquisition module is used to acquire vehicle-cargo data; the vehicle-cargo data includes: vehicle owner data, historical cargo data, vehicle owner historical interaction behavior sequence data, and cargo data to be matched; the vehicle owner historical interaction behavior sequence data is the cargo data of the vehicle owner who actually transported and gave positive reviews on the online freight platform in chronological order;

[0230] A comprehensive preference acquisition module is used to predict the vehicle owner's comprehensive preference for matching cargo sources based on vehicle-cargo related data; the comprehensive preference is the probability of the vehicle owner interacting with the cargo source to be matched;

[0231] The vehicle-cargo matching decision model acquisition module is used to obtain the vehicle-cargo matching decision model that considers comprehensive preferences, vehicle loading rate, and matching cost;

[0232] The decision model solving module is used to obtain the optimal solution of the vehicle-cargo matching decision model as the vehicle-cargo matching result.

[0233] It is understandable that the vehicle-cargo matching device that integrates vehicle owner preferences provided in an embodiment of the present invention corresponds to the above-mentioned vehicle-cargo matching method that integrates vehicle owner preferences. The explanation, examples, beneficial effects, etc. of the relevant contents can refer to the corresponding contents in the vehicle-cargo matching method that integrates vehicle owner preferences, and will not be repeated here.

[0234] Example 3:

[0235] A computer-readable storage medium stores a computer program for vehicle-cargo matching, wherein the computer program causes a computer to execute the following steps:

[0236] Obtain vehicle-cargo related data; the vehicle-cargo related data includes: vehicle owner data, historical cargo source data, vehicle owner historical interaction behavior sequence data, and cargo source data to be matched; and the vehicle owner historical interaction behavior sequence data is the cargo source data that the vehicle owner actually transported and gave positive reviews on the online freight platform in chronological order;

[0237] Predicting the vehicle owner's comprehensive preference for matching cargo sources based on vehicle-cargo related data; the comprehensive preference is the probability of the vehicle owner interacting with the cargo source to be matched;

[0238] Obtain a vehicle-cargo matching decision model that considers comprehensive preferences, vehicle loading rate, and matching costs;

[0239] The optimal solution of the vehicle-cargo matching decision model is obtained as the vehicle-cargo matching result.

[0240] Example 4:

[0241] An electronic device, comprising:

[0242] one or more processors;

[0243] Memory; and

[0244] One or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, the programs comprising steps for performing the following steps:

[0245] Obtain vehicle-cargo related data; the vehicle-cargo related data includes: vehicle owner data, historical cargo source data, vehicle owner historical interaction behavior sequence data, and cargo source data to be matched; and the vehicle owner historical interaction behavior sequence data is the cargo source data that the vehicle owner actually transported and gave positive reviews on the online freight platform in chronological order;

[0246] Predicting the vehicle owner's comprehensive preference for matching cargo sources based on vehicle-cargo related data; the comprehensive preference is the probability of the vehicle owner interacting with the cargo source to be matched;

[0247] Obtain a vehicle-cargo matching decision model that considers comprehensive preferences, vehicle loading rate, and matching costs;

[0248] The optimal solution of the vehicle-cargo matching decision model is obtained as the vehicle-cargo matching result.

[0249] In summary, compared with the prior art, the present invention has the following beneficial effects:

[0250] 1) The present invention makes full use of the owner's historical behavior sequence information. Historical behavior is a reflection of the owner's preferences. Some preferences are static or change slowly over time, while some preferences are dynamic and change in real time. They show the owner's preferences from different perspectives, so combining these two different types of preference forms can better reflect the owner's personalized preferences. On this basis, the present invention proposes a new vehicle-cargo matching method idea, namely "prediction + decision-making". By establishing a deep learning model to effectively explore the owner's long-term and short-term preferences, improve the accuracy of the prediction, and then integrate the results of the preference prediction into the construction process of the vehicle-cargo matching decision model, it effectively improves the efficiency of vehicle-cargo matching, reduces matching costs, and enhances the owner's satisfaction and stickiness. Therefore, it is more valuable in practical applications.

[0251] 2) This invention uses a deep learning model to model owner preferences and normalizes the time intervals between behavior sequences, enhancing the model's perception of time and enabling more accurate modeling of owners' short-term interests. Furthermore, the introduction of an attention mechanism adaptively calculates the owner's preference representation vector, effectively filtering out historical behaviors that are most relevant to predicted supply, and further identifying the importance of long-term and short-term preferences for current behavior prediction, thereby improving the accuracy of preference prediction.

[0252] 3) This method obtains vehicle-cargo data, mines the vehicle owner's personalized preferences based on this data, sets the loading rate and matching cost based on basic vehicle and cargo information, and establishes a vehicle-cargo matching decision model based on the predicted personalized preferences, loading rate, and matching cost. Constraints are set to optimize overall efficiency, and the optimal solution for the vehicle-cargo matching model is obtained using an ant colony optimization algorithm, which is the optimal vehicle-cargo matching solution. Compared to traditional methods that only consider the degree of matching, this method can more accurately consider the vehicle owner's personalized preferences, improve vehicle owner satisfaction, increase vehicle owner stickiness to the platform, improve overall resource utilization, reduce costs, and improve efficiency.

[0253] It should be noted that, through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus the necessary general hardware platform. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, which can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., and includes a number of instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or certain parts of the embodiment. In this article, relational terms such as first and second are merely used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "comprise", "include" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device that includes a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. Without further constraints, an element defined by the phrase "comprises a..." does not preclude the existence of additional identical elements in the process, method, article or apparatus that includes the element.

[0254] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A vehicle-cargo matching method integrating vehicle owner preferences, characterized in that: The method includes: Obtain vehicle-cargo related data; the vehicle-cargo related data includes: vehicle owner data, historical cargo source data, vehicle owner historical interaction behavior sequence data, and cargo source data to be matched; and the vehicle owner historical interaction behavior sequence data is the cargo source data that the vehicle owner actually transported and gave positive reviews on the online freight platform in chronological order; Predicting the vehicle owner's comprehensive preference for matching cargo sources based on vehicle-cargo related data; the comprehensive preference is the probability of the vehicle owner interacting with the cargo source to be matched; Obtain a vehicle-cargo matching decision model that considers comprehensive preferences, vehicle loading rate, and matching costs; Obtaining the optimal solution of the vehicle-cargo matching decision model as the vehicle-cargo matching result; The vehicle-cargo matching decision model includes: Objective function: in, Indicates the degree to which the platform prioritizes the overall preferences of car owners; Indicates the platform's preference for vehicle loading rate; Indicates the platform's emphasis on reducing matching costs; 、 、 Represent the comprehensive preference score rate, loading rate, and matching cost respectively. ， ; Indicates the owner v About the source of goods c comprehensive preferences; For car owners v About the source of goods c matching decision variables; and Then we get the vehicle-cargo matching decision variable matrix: They are the number of car owners to be matched and the number of goods to be matched respectively; For car owners v Current Location and Supply c distance from the starting point; For car owners v The length, load capacity and volume of the owned vehicles; For the source of goods c length, weight and volume; D is the maximum matching distance given; Constraints: Condition 1: Each car owner is matched with at most one source of goods, expressed as: ， Condition 2: Each source of goods matches at most one vehicle owner, expressed as: ， Condition 3: The length of the cargo source matched by each vehicle owner does not exceed the length of the vehicle, expressed as: ， ， Condition 4: The weight of the cargo matched by each vehicle owner does not exceed the vehicle load, expressed as: ， ， Condition 5: The volume of cargo matched by each vehicle owner does not exceed the vehicle capacity, expressed as: ， ， Condition 6: Each vehicle owner’s distance to the starting point of the matching cargo source does not exceed the maximum matching distance, expressed as: ， ， in, They are respectively the car owner set and the cargo source set.

2. The vehicle-cargo matching method integrating vehicle owner preferences as claimed in claim 1, characterized in that: The prediction of the vehicle owner's comprehensive preference for matching cargo sources based on vehicle-cargo related data includes: Predict the comprehensive preferences of car owners for matching cargo sources based on their short-term and long-term preferences.

3. The vehicle-cargo matching method integrating vehicle owner preferences as claimed in claim 2, characterized in that: The prediction of the comprehensive preference of the car owner for matching cargo sources based on the car owner's short-term preference and long-term preference includes: Obtain the trained short-term preference representation module, long-term preference representation module, weight coefficients and trained multi-layer perceptron; The trained short-term preference representation module and long-term preference representation module are used to obtain the car owner's short-term preference vector and long-term preference vector for matching cargo sources. The short-term preference vector and long-term preference vector are weighted and fused based on the weight coefficient to obtain a comprehensive preference vector. The comprehensive preference vector and the cargo source vector are used as the input of the trained multi-layer perceptron to obtain the comprehensive preference of the car owner for matching cargo sources.

4. The vehicle-cargo matching method integrating vehicle owner preferences as claimed in claim 1, characterized in that: The ant colony optimization algorithm is used to obtain the optimal solution of the vehicle-cargo matching decision model.

5. The vehicle-cargo matching method integrating vehicle owner preferences as claimed in claim 1, characterized in that: The LSTM model is trained based on vehicle-cargo related data to obtain a short-term preference representation module.

6. The vehicle-cargo matching method integrating vehicle owner preferences as claimed in claim 1, characterized in that: Determine long-term preferences based on the similarity between the car owner's historical interaction behavior sequence data and the source of goods to be matched.

7. A vehicle-cargo matching device that integrates vehicle owner preferences, characterized in that: The device includes: The vehicle-cargo data acquisition module is used to acquire vehicle-cargo data; the vehicle-cargo data includes: vehicle owner data, historical cargo data, vehicle owner historical interaction behavior sequence data, and cargo data to be matched; the vehicle owner historical interaction behavior sequence data is the cargo data of the vehicle owner who actually transported and gave positive reviews on the online freight platform in chronological order; A comprehensive preference acquisition module is used to predict the vehicle owner's comprehensive preference for matching cargo sources based on vehicle-cargo related data; the comprehensive preference is the probability of the vehicle owner interacting with the cargo source to be matched; The vehicle-cargo matching decision model acquisition module is used to obtain the vehicle-cargo matching decision model that considers comprehensive preferences, vehicle loading rate, and matching cost; A decision model solving module, used to obtain the optimal solution of the vehicle-cargo matching decision model as the vehicle-cargo matching result; The vehicle-cargo matching decision model includes: Objective function: in, Indicates the degree to which the platform prioritizes the overall preferences of car owners; Indicates the platform's preference for vehicle loading rate; Indicates the platform's emphasis on reducing matching costs; 、 、 Represent the comprehensive preference score rate, loading rate, and matching cost respectively. ， ; Indicates the owner v About the source of goods c comprehensive preferences; For car owners v About the source of goods c matching decision variables; and Then we get the vehicle-cargo matching decision variable matrix: They are the number of car owners to be matched and the number of goods to be matched respectively; For car owners v Current Location and Supply c distance from the starting point; For car owners v The length, load capacity and volume of the owned vehicles; For the source of goods c length, weight and volume; D is the maximum matching distance given; Constraints: Condition 1: Each car owner is matched with at most one source of goods, expressed as: ， Condition 2: Each source of goods matches at most one vehicle owner, expressed as: ， Condition 3: The length of the cargo source matched by each vehicle owner does not exceed the length of the vehicle, expressed as: ， ， Condition 4: The weight of the cargo matched by each vehicle owner does not exceed the vehicle load, expressed as: ， ， Condition 5: The volume of cargo matched by each vehicle owner does not exceed the vehicle capacity, expressed as: ， ， Condition 6: Each vehicle owner’s distance to the starting point of the matching cargo source does not exceed the maximum matching distance, expressed as: ， ， in, They are respectively the car owner set and the cargo source set.

8. A computer-readable storage medium, characterized in that It stores a computer program for vehicle-cargo matching, wherein the computer program enables a computer to execute the vehicle-cargo matching method integrating vehicle owner preferences as described in any one of claims 1 to 6.

9. An electronic device, characterized in that: include: one or more processors; Memory; as well as One or more programs, wherein the one or more programs are stored in the memory and are configured to be executed by the one or more processors, the programs including a method for executing the vehicle-cargo matching method integrating vehicle owner preferences as described in any one of claims 1-6.

Images