A personalized route recommendation method based on price preference and user preference

Abstract

The application relates to a personalized route recommendation method based on price preference and user preference, which comprises the following steps: acquiring a road network structure diagram of a target region where a target user is located and a historical travel trajectory sequence of the target user in the target region; acquiring external weather characteristic information of the target region; extracting travel preference information of the target user according to the historical travel trajectory information of the target user; extracting price preference of the target user according to the Mahalanobis distance between road nodes in the target region; inputting the travel preference information of the target user, the external weather characteristic information of the target region and the travel preference information of the target user into a multi-head attention mechanism module after splicing to obtain comprehensive preference characteristics of the target user; and predicting an out-travel route of the target user according to the comprehensive preference characteristics of the target user, an out-travel starting point O and an out-travel ending point D and recommending the out-travel route to the target user.

Description

TECHNICAL FIELD

[0001] The application belongs to the field of intelligent traffic route recommendation, and particularly relates to a personalized route recommendation method based on price preference and user preference. BACKGROUND

[0002] With the increasing level of urbanization, the traffic problem is becoming increasingly serious. Personalized route recommendation (PRR) lays a foundation for the wide application of intelligent traffic. However, the complex traffic conditions bring new challenges to the personalized travel recommendation of users. Personalized route recommendation is conducive to improving traffic efficiency and has attracted more and more attention from scholars.

[0003] In recent years, domestic and foreign scholars have conducted extensive research on personalized route recommendation. At present, the research methods of personalized route recommendation can be divided into four categories: heuristic search algorithm method, using appropriate heuristic algorithm to obtain rapid response by effectively reducing the search space; machine learning method, describing the location correlation or space-time information in a principled way; deep learning method, using a learnable neural network to capture complex data features; and deep learning and heuristic search algorithm fusion method, using a neural network to automatically learn the cost function in the heuristic algorithm.

[0004] In intelligent traffic, for the research on personalized travel routes, scholars research historical traffic data to mine complex space-time features, and have achieved relatively rich research results. However, in the formal traffic network, some roads will often be congested due to various reasons, affecting the efficiency. Therefore, early travel route research often has prediction errors. Nowadays, with the development of artificial intelligence, deep learning technology appears in various researches, and deep learning has made major breakthroughs in various fields. The travel route prediction constructed by deep learning can not only fully analyze the space-time correlation of user historical trajectory data, but also consider the influence of external factors on the travel route.

[0005] In social networks, user preference and price preference have an important influence on topic propagation. However, the socialization of traffic networks and the personalized route recommendation of traffic routes based on social networks still have the following challenges:

[0006] 1. Social data is composed of user behavior sequences, and the capture of user interest points is a key factor for accurately predicting the direction of topic propagation. Traffic data is composed of card holes and vehicle trajectories, and the fusion of user preferences is a key factor for the achievement of personalized path recommendation. How to represent the socialization of traffic networks while integrating user preferences is a problem to be solved urgently.

[0007] 2. In social networks, node influence plays a key role in topic propagation cost. In traffic networks, the complex spatial characteristics of road networks have a significant impact on travel cost. How to socialize the representation of traffic networks while incorporating price preferences is a pressing problem.

[0008] 3. When recommending personalized travel routes for users, a trade-off between user preferences and price preferences must be made. Both preferences influence each other and jointly determine the final travel route. How to model the complex relationship between user preferences and price preferences while considering the impact of external traffic data characteristics on both preferences is a pressing problem. SUMMARY

[0009] To solve the problems in the background art, the present application provides a personalized route recommendation method based on price preference and user preference, comprising:

[0010] S1: Obtain the road network structure graph of the target area where the target user is located and the historical travel trajectory sequence of the target user in the target area through the data query API provided by the enterprise, and then preprocess the data;

[0011] S2: Obtain external weather feature information of the target area;

[0012] S3: Extract the travel preference information of the target user according to the historical travel trajectory information of the target user using the FC neural network model and the BERT model;

[0013] S4: Calculate the similarity between road nodes in the target area according to the Mahalanobis distance between road nodes, construct a road node similarity matrix, and create an incoming adjacency matrix and an outgoing adjacency matrix of the road network structure graph according to the road node similarity matrix;

[0014] S5: Extract the price preference of the target user using the adaptive GCN neural network according to the incoming adjacency matrix and the outgoing adjacency matrix of the road network structure graph:

[0015] S6: Concatenate the travel preference information of the target user, the external weather feature information of the target area, and the travel preference information of the user, and input them into a multi-head attention mechanism module to obtain comprehensive preference features of the user;

[0016] S7: Predict the travel route of the target user according to the comprehensive preference features of the target user, the travel origin O and the travel destination D, and recommend it to the target user.

[0017] Preferably, the road network structure graph G R comprises:

[0018] G R =(V,E)

[0019] Where V = {v1, v2, ..., v} N} represents the set of road nodes within the target area, v i Let E represent the i-th road node within the target area, and let E represent the edge set of the road network structure graph; the road network structure graph G... R The adjacency matrix is When node v i and node v j If they are directly connected geographically, then A i,j =1, otherwise A i,j =0, where N represents the number of road nodes within the target area.

[0020] Preferably, the external weather characteristic information includes:

[0021] Based on the varying impacts of weather on user travel, weather is categorized into 7 types, numbered sequentially from 1 to 7: rainy, cloudy, sunny, partly cloudy, foggy, snowy, and others. The external weather characteristic information EF is then represented as:

[0022] EF = (e 1t ,e 2t ,…,e Nt )

[0023] Among them, e it Let represent the external weather at the current time t for the i-th road node.

[0024] Preferably, the extraction of the target user's travel preference information includes:

[0025] The historical travel trajectory sequence of the target user is represented as p = {p1, p2, ..., p...} l ,…p L}, p l Represents the l-th travel trajectory of the target user; t lk This represents the k-th moment in the l-th travel trajectory of the target user; Indicates the target user at t lk The road node at the time;

[0026] S31: Construct the target user's location information matrix D based on the target user's historical travel trajectory sequence;

[0027]

[0028] Where, d ab ={0, 1}, d ab =1 indicates that the target user is at the b-th road node at time a, where a∈[t] 11 [,t], where t represents the current time; t11 represents the first time of the target user in the first travel trajectory, b e V, V represents the set of road nodes in the target area;

[0029] S32: input the position information matrix D of the target user into the FC neural network model to obtain the position information vector L of the target user;

[0030] L = FC(D)

[0031] S33: input the position information vector L of the target user into the BERT model to obtain the travel preference information of the target user;

[0032] UP = bert(L)

[0033] Wherein, UP represents the travel preference information of the target user.

[0034] Preferably, the similarity between the road nodes includes:

[0035]

[0036]

[0037] Wherein, represents the similarity between road nodes v i and v j , D(v i , v j ) represents the Mahalanobis distance between road nodes v i and v j , γ is a hyperparameter of , which is a scalar value, and M represents a symmetric semi-definite matrix.

[0038] Preferably, the incoming adjacency matrix and outgoing adjacency matrix of the road network structure graph include:

[0039] S41: after normalizing the road node similarity matrix D, the standardized adjacency matrix A' of the road network structure graph is obtained: A' = D -(1 / 2) GD -(1 / 2 )

[0040]

[0041] Wherein, D represents the degree matrix of the road network structure graph, and G represents the road node similarity matrix;

[0042] S42: create the incoming adjacency matrix and outgoing adjacency matrix of the road network structure graph according to the standardized adjacency matrix A' of the road network structure graph;

[0043]

[0044] ​

[0045] where σ is an activation function, G represents a road node similarity matrix, W out and W in denote learnable weight matrices, denotes an out-adjacency matrix of the road network structure graph, denotes an in-adjacency matrix of the road network structure graph.

[0046] Preferably, the price preference of the target user comprises:

[0047] PP = ReLU (Z out + Z in )

[0048]

[0049]

[0050]

[0051]

[0052]

[0053]

[0054] where F (L out , X, Γ) represents the spectrum of the updated Laplacian matrix L out , X represents a road node feature matrix of the road network structure graph, the initial feature of each road node in the road network structure graph is represented by the spatial coordinates of the road node, Γ represents a parameter for updating the Laplacian matrix Λ, K is the order of the convolution kernel and , k represents the convolution times of the function F(), U out is a matrix composed of eigenvectors of the Laplacian matrix L out , U in is a matrix composed of eigenvectors of the Laplacian matrix L in , Λ is a diagonal matrix of eigenvalues of the Laplacian matrix, L out is a normalized Laplacian matrix based on the out-degree of the road network, L in is a normalized Laplacian matrix based on the in-degree of the road network, I is an identity matrix, D out is an out-degree matrix, D in is an in-degree matrix.

[0055] Preferably, the predicted travel route of the target user comprises:

[0056] When the user travels to the road network structure graph GR v i , the estimated cost of user from road node v i to its neighbor nodes is calculated as:

[0057]

[0058]

[0059]

[0060]

[0061]

[0062] W G = P * A

[0063] where M represents the number of neighbor nodes of road node v i , P is the user's comprehensive preference, A is the estimated travel cost of road node v i to its neighbor nodes, g represents the cost from origin O to node , g A* (v i ) represents the estimated travel cost from origin O to road node vi, A is the total cost estimate of road node v i to neighbor nodes, A is the default cost estimate of road node v i to neighbor nodes, A is the travel cost estimate of road node v i to neighbor nodes combined with user preference, α, β are weight coefficient parameters, D i is the distance between road node v and its neighbor nodes, D i is the estimated travel time of road node v to its neighbor nodes, W G is the weight of user preference in road network, D i is the value of road node v and its neighbor nodes in weight matrix W G ;

[0064] Selecting a road node v i The neighbor node with the lowest estimated travel cost to its neighbor node is selected as the next travel road node of the user, and the above steps are repeated to predict the travel route P of the target user O->D .

[0065] The present application has at least the following beneficial effects

[0066] The present application extracts the travel preference and price preference of the user through the user trajectory sequence and the road network structure diagram, simultaneously, fuses the comprehensive preference characteristics obtained by the external weather feature information, and finally predicts the travel route of the target user according to the comprehensive preference characteristics of the target user, the travel starting point O and the travel terminal point D by using the A* algorithm. When the route is recommended, the route conforms to the user preference, reduces the travel cost of the user, improves the travel experience of the user, and can be applied to path recommendation, vehicle navigation, city planning and the like. BRIEF DESCRIPTION OF DRAWINGS

[0067] Figure 1 The present application is a method flowchart. DETAILED DESCRIPTION

[0068] The embodiments of the present application will be described in detail below with specific reference to the drawings. Those skilled in the art can easily understand other advantages and effects of the present application from the content disclosed in the specification. The present application can also be implemented or applied by different specific embodiments, and the details in the specification can be modified or changed based on different views and applications without departing from the spirit of the present application. It should be noted that the drawings provided in the following embodiments only illustrate the basic concept of the present application in a schematic manner, and the following embodiments and features in the embodiments can be combined with each other without conflict.

[0069] The drawings are only used for exemplary illustration, and the representation is only a schematic diagram, not a physical diagram, and cannot be understood as a limitation of the present application; in order to better illustrate the embodiments of the present application, some components in the drawings are omitted, enlarged or reduced, and do not represent the size of the actual product; it is understandable for those skilled in the art that some known structures and their descriptions in the drawings can be omitted.

[0070] The same or similar reference numerals in the drawings of the embodiments of the present application correspond to the same or similar components; in the description of the present application, it should be understood that if the orientations or positional relationships indicated by the terms "upper", "lower", "left", "right", "front", "back" and the like are based on the orientations or positional relationships shown in the drawings, they are only for the convenience of describing the present application and simplifying the description, and do not indicate or imply that the devices or elements referred to must have a particular orientation, be constructed and operated in a particular orientation, therefore the terms describing the positional relationship in the drawings are only used for exemplary illustration, and cannot be understood as a limitation on the present application, for those skilled in the art, the specific meanings of the above terms can be understood according to the specific circumstances.

[0071] Referring to Figure 1 , the present application provides a personalized route recommendation method based on price preference and user preference, comprising:

[0072] S1: obtaining the road network structure diagram of the target area where the target user is located and the historical travel trajectory sequence of the target user in the target area through the data query API provided by the enterprise, and then preprocessing the data;

[0073] Preferably, an embodiment of preprocessing data includes:

[0074] Simple data processing and cleaning, usually because the original data obtained has high repeatability and irrelevance, cannot be directly used for data analysis, through simple data cleaning, most unstructured data can be structured, so that abnormal values or null values no longer appear, and the accuracy of subsequent generated data is increased.

[0075] Preferably, the road network structure diagram G R includes:

[0076] G R =(V,E)

[0077] Wherein, V={v1,v2,…,v N} represents a set of road nodes in the target area, v i represents the i-th road node in the target area, E represents the edge set of the road network structure diagram; the adjacency matrix of the road network structure diagram G R is When the node v i and the node v j are directly connected in geographical space, A i,j =1, otherwise A i,j =0, N represents the number of road nodes in the target area;

[0078] S2: obtaining the external weather feature information EF of the target area:

[0079] Preferably, the external weather feature information includes:

[0080] According to the influence of weather on the user's trip, the weather is divided into 7 categories, and numbered in order of numbers 1-7, in turn, rain, overcast, sunny, cloudy, foggy, snowy and other, then the external weather feature information EF is represented as:

[0081] EF=(e 1t ,e 2t ,…,e Nt )

[0082] Wherein, e it represents the external weather of the i-th road node at the current time t.

[0083] S3: extracting the travel preference information of the target user according to the historical travel trajectory information of the target user by using the FC neural network model and the BERT model;

[0084] The historical travel trajectory sequence of the target user is represented as p={p1,p2,…,p l ,…p L}, p l represents the l-th travel trajectory of the target user; t lk represents the k-th time in the l-th travel trajectory of the target user; represents the road node where the target user is at t lk ;

[0085] The historical travel trajectory sequence of the target user is represented as p={p1,p2,…,p l ,…p L}, p l represents the l-th travel trajectory of the target user; t lk represents the k-th time in the l-th travel trajectory of the target user; represents the road node where the target user is at t lk ;

[0086] S31: constructing the position information matrix D of the target user according to the historical travel trajectory sequence of the target user;

[0087]

[0088] Wherein, d ab ={0,1},d ab =1 represents that the target user is at the b-th road node at the a-th time, a∈[t 11 ,t],t represents the current time; t 11represents the first time of the target user in the first travel trajectory, b is in V, and V represents a set of road nodes in the target area;

[0089] S32: input the position information matrix D of the target user into the FC neural network model to obtain the position information vector L of the target user;

[0090] L=FC(D)

[0091] S33: input the position information vector L of the target user into the BERT model to obtain the travel preference information of the target user;

[0092] UP=bert(L)

[0093] Wherein, UP represents the travel preference information of the target user.

[0094] The application constructs a traffic network socialized user travel preference mining method, and the application analogizes the social network and the traffic network, analogizes the "card mouth-topic propagation direction-topic propagation randomness" in the social network as the "card mouth-vehicle driving direction-road section selection randomness" in the traffic network, so that the traffic network socialization is realized. Meanwhile, by using the advantage of the Bert network in analyzing the user behavior sequence, the time correlation in the traffic network is mined, and then the user travel preference is captured, and the recommended travel route is improved.

[0095] S4: calculate the similarity between road nodes according to the Mahalanobis distance between road nodes in the target area, construct a road node similarity matrix, and create the incoming adjacency matrix and outgoing adjacency matrix of the road network structure graph according to the road node similarity matrix;

[0096] Preferably, the similarity between the road nodes comprises:

[0097]

[0098]

[0099] Wherein, represents the similarity between road nodes v i and v j , D(v i ,v j ) represents the Mahalanobis distance between road nodes v i and v j , γ is a hyperparameter of , which is a scalar value, M represents a symmetric positive semi-definite matrix, and (v i -v j ) represents the distance between the geographical coordinates of road nodes v i and v j .

[0100] Preferably, the incoming adjacency matrix and the outgoing adjacency matrix of the road network structure graph comprise:

[0101] S41: normalize the road node similarity matrix G to obtain a standardized adjacency matrix A' of the road network structure graph:

[0102] A' = D -(1 / 2) GD -(1 / 2)

[0103] wherein D represents a degree matrix of the road network structure graph, and G represents the road node similarity matrix;

[0104] S42: create the incoming adjacency matrix and the outgoing adjacency matrix of the road network structure graph according to the standardized adjacency matrix A' of the road network structure graph;

[0105]

[0106]

[0107] wherein σ is an activation function, G represents the road node similarity matrix, W out and W in represent learnable weight matrices, represents the outgoing adjacency matrix of the road network structure graph, represents the incoming adjacency matrix of the road network structure graph.

[0108] S5: extract the price preference of the target user by using the adaptive GCN neural network according to the incoming adjacency matrix and the outgoing adjacency matrix of the road network structure graph:

[0109] Preferably, the price preference of the target user comprises:

[0110] PP = ReLU (Z out + Z in )

[0111]

[0112]

[0113]

[0114]

[0115]

[0116]

[0117] wherein F (L out ​X, Γ) represents the spectrum of the updated Laplacian matrix L out , X represents the road node feature matrix of the road network structure graph, the initial feature of each road node in the road network structure graph is represented by the spatial coordinates of the road node, Γ represents the parameter for updating the Laplacian matrix Λ, K is the convolution kernel and the order of the function F() convolution times, U out is a matrix composed of the eigenvectors of the Laplacian matrix L out , U in is a matrix composed of the eigenvectors of the Laplacian matrix L in , Λ is the eigenvalue diagonal matrix of the Laplacian matrix, L out is a normalized Laplacian matrix based on the out-degree of the road network, L in is a normalized Laplacian matrix based on the in-degree of the road network, I is the unit matrix, D out is the out-degree matrix, D in is the in-degree matrix. The present application proposes a price preference mining method based on an adaptive graph convolution network. The present application analogizes the spatial correlation of road nodes as node influence, realizes the social representation of the traffic network. At the same time, by utilizing the advantages of adaptive graph convolution network in integrating user nodes and neighborhood node information, obtaining price preference, effectively mining the complex spatial correlation of the road network, and dynamically adjusting the recommended path cost.

[0118] S6: The travel preference information of the target user, the external weather feature information of the target area and the travel preference information of the user are spliced to input a multi-head attention mechanism module to obtain comprehensive preference features of the user:

[0119] P=MultiHead(Concat(UP,PP,EF))

[0120] Wherein, MultiHead represents a multi-head attention mechanism, and Concat represents feature splicing.

[0121] The present application proposes an attention mechanism-based mutual guidance learning mechanism. By utilizing the sensitivity of the multi-head attention mechanism to local features, external traffic data features are introduced, and an attention mechanism-based mutual guidance learning mechanism is proposed. The mechanism enables user preferences and price preferences to interact under different external features, learn from each other, enrich their respective semantics, and ultimately obtain comprehensive user travel preferences and price preferences, and optimize the travel route.

[0122] S7: According to the comprehensive preference features of the target user, the travel starting point O and the travel ending point D, the travel route of the target user is predicted and recommended to the target user.

[0123] Preferably, the step of predicting the travel route of the target user comprises:

[0124] When the user travels to the road network structure map G R Road node v i At that time, calculate the user's location from road node v i to its neighboring nodes The estimated cost is:

[0125]

[0126]

[0127]

[0128]

[0129]

[0130] W G =P*A

[0131] Where M represents road node v i The number of neighboring nodes, For road node v i to its neighboring nodes The estimated toll costs, Indicates the distance from the starting point O to the node. Cost, g A* (v i ) represents the distance from the starting point O to the road node v. i The estimated toll costs, For road node v i to neighboring nodes The estimated total cost of the nodes, For road node v i to neighboring nodes The default cost estimate To combine user preferences for road nodes v i to neighboring nodes The estimated cost of passage, To represent neighboring nodes The estimated cost to destination D, where α and β are weighting coefficient parameters. For road node v i to neighboring nodes Distance between geographic coordinates For road node v i to neighboring nodes The estimated travel time, W G The weight of user preferences in the road network, For road node v i with neighboring nodes In the weight matrix W G The value in;

[0132] Select road node v i The neighbor node with the lowest estimated travel cost to its neighbors is selected as the user's next path node. This process is repeated to predict the target user's travel route P. O-＞D .

[0133] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A personalized route recommendation method based on price preference and user preference, characterized in that, The method comprises the following steps: S1: obtaining a road network structure diagram of a target area where a target user is located and a historical travel trajectory sequence of the target user in the target area through a data query API provided by an enterprise, and then preprocessing the data; S2: obtaining external weather feature information of the target area; S3: extracting travel preference information of the target user according to the historical travel trajectory information of the target user by using an FC neural network model and a BERT model; S4: calculating the similarity between road nodes in the target area according to Mahalanobis distance between the road nodes, constructing a road node similarity matrix, and creating an incoming adjacency matrix and an outgoing adjacency matrix of the road network structure diagram according to the road node similarity matrix; S5: extracting price preference of the target user according to the incoming adjacency matrix and the outgoing adjacency matrix of the road network structure diagram by using an adaptive GCN neural network; S6: inputting the travel preference information of the target user, the external weather feature information of the target area and the travel preference information of the user into a multi-head attention mechanism module after splicing to obtain comprehensive preference features of the user; S7: predicting a travel route of the target user according to the comprehensive preference characteristics of the target user, the travel starting point and the travel ending point and recommending the travel route to the target user.

2. The personalized route recommendation method based on price preference and user preference according to claim 1, characterized in that, The road network structure diagram comprises: wherein, denotes a set of road nodes within the target region, denotes the i-th road node within the target region, denotes the i-th road node within the target region, denotes a set of edges of the road network structure graph; the adjacency matrix of the road network structure graph is , when node and node are directly connected in geographical space, , otherwise , N denotes the number of road nodes within the target region. 3.The personalized route recommendation method based on price preference and user preference of claim 1, wherein, The external weather feature information comprises: According to the influence of weather on the user's trip, the weather is divided into 7 categories, and numbered in order of numbers 1-7, in turn, rain, overcast, sunny, cloudy, fog, snow and other, then the external weather feature information is represented as: wherein, represents the external weather of the th road node at the current time point.

4. The personalized route recommendation method based on price preference and user preference according to claim 1, characterized in that, The extracting the travel preference information of the target user comprises: The target user's historical travel trajectory sequence is represented as follows , , The target user's first A travel itinerary; Indicates the target user in the 1st month. The first of the travel trajectories At that moment; Indicates the target user is The road node at the time; S31: Construct a location information matrix of the target user according to a historical travel trajectory sequence of the target user ; wherein, , denotes that the target user is at the th road node at the th time point, , denotes the current time point; denotes that the target user is at the 1th time point in the 1th travel trajectory, , denotes a set of road nodes within the target area; S32: Obtain the position information matrix of the target user Input the FC neural network model to obtain the position information vector L of the target user; S33: inputting a position information vector L of the target user into the BERT model to obtain the travel preference information of the target user; wherein, represents the travel preference information of the target user.

5. The personalized route recommendation method based on price preference and user preference according to claim 1, characterized in that, The similarity between the road nodes comprises: in, Represents road nodes and Similarity between them Represents road nodes and Mahalanobis distance between them yes The hyperparameter is a scalar value. This represents a symmetric positive semi-definite matrix.

6. The personalized route recommendation method based on price preference and user preference according to claim 1, characterized in that, The incoming adjacency matrix and the outgoing adjacency matrix of the road network structure diagram comprise: S41: Normalizing the road node similarity matrix obtaining a standardized adjacency matrix of the road network structure graph after normalization : wherein, denotes a degree matrix representing the road network structure graph, denotes a road node similarity matrix; S42: Standardized adjacency matrix according to the road network structure diagram Incoming and outgoing adjacency matrices of the road network structure diagram are created; wherein, is an activation function, denotes a road node similarity matrix, and denotes a learnable weight matrix, denotes an out-adjacency matrix of the road network structure graph, denotes an in-adjacency matrix of the road network structure graph.

7. The personalized route recommendation method based on price preference and user preference according to claim 1, characterized in that, The price preference of the target user comprises: wherein, denotes the spectrum of the updated Laplacian matrix denotes the road node feature matrix of the road network structure graph, the initial feature of each road node in the road network structure graph being represented by the spatial coordinates of the road node, denotes the parameters for updating the Laplacian matrix is the order of the convolution kernel and denotes the function the number of convolutions, is a matrix composed of the eigenvectors of the Laplacian matrix is a matrix composed of the eigenvectors of the Laplacian matrix is a diagonal matrix of the eigenvalues of the Laplacian matrix, is a normalized Laplacian matrix based on the out-degree of the road network, is a normalized Laplacian matrix based on the in-degree of the road network, is an identity matrix, is an out-degree matrix, is an in-degree matrix; denotes the out-adjacency matrix of the road network structure graph, denotes the in-adjacency matrix of the road network structure graph.​​​​​ 8.The personalized route recommendation method based on price preference and user preference of claim 1, wherein, The predicting the travel route of the target user comprises: When the user travels to a road node of the road network structure graph , the estimated cost of the user from the road node to its neighbor nodes is calculated as: in, Represents road nodes The number of neighboring nodes, Based on user preferences, Road nodes to its neighboring nodes The estimated toll costs, Indicates the starting point To the node The cost, Indicates the starting point To road node The estimated toll costs, Road nodes to neighboring nodes The estimated total cost of the nodes, Road nodes to neighboring nodes The default cost estimate To combine road nodes with user preferences to neighboring nodes The estimated cost of passage, To represent neighboring nodes To the finish line The estimated cost, , These are the weighting coefficient parameters. Road nodes to neighboring nodes Distance between geographic coordinates Road nodes to neighboring nodes The estimated travel time, The weight of user preferences in the road network, Road nodes with neighboring nodes In the weight matrix The value in; Road network structure diagram The adjacency matrix; Selecting a road node the neighbor node with the lowest estimated travel cost to its neighbor nodes as the next travel road node of the user, repeating the above steps to predict the travel route of the target user .

Images