Hotel room price prediction method and device, equipment, storage medium and program product

Abstract

The invention provides a hotel room price prediction method and device, equipment, a storage medium and a program product, and relates to the technical field of computers, and the method comprises the steps: determining the regional association degree between regional nodes according to the inter-regional feature factors between the regional nodes; the regional association degree represents the population flow density between the regions; determining an edge weight coefficient of a GCN (Graphics Convolutional Network) model according to the regional correlation degree; and inputting the node feature factors of each region node into the GCN model, and obtaining a room price prediction result of the hotel in each region node output by the GCN model in combination with the edge weight coefficient. According to the method, the edge weight coefficient of the GCN model is improved through the regional correlation degree, and the spatial regional topological correlation characteristics between the regional nodes, the regional characteristic factors of the region where the method is located and the hotel characteristic factors can be considered more comprehensively, so that the pricing of the hotel room price is more observable, objective and scientific.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to a hotel room price prediction method, device, equipment, storage medium and program product. Background Art

[0002] With the rise of tourism and the increasing convenience of transportation, the hotel industry has also developed rapidly. At the same time, the competition in the hotel industry is also particularly fierce, especially in the formulation of hotel room rate strategies, which will have a key impact on hotel customer acquisition, profit and revenue. Therefore, it is very important to formulate a reasonable pricing strategy.

[0003] At present, hotel pricing methods mainly include cost-oriented method and market-oriented method. However, the above hotel pricing methods lack scientificity. Summary of the invention

[0004] The present invention provides a hotel room price prediction method, device, equipment, storage medium and program product, which are used to solve the defect that the hotel pricing method in the prior art lacks scientificity and realizes the pricing of hotel room prices to be more observable, objective and scientific.

[0005] The present invention provides a hotel room price prediction method, comprising the following steps: Determining the regional correlation between the regional nodes according to the regional characteristic factors between the regional nodes; the regional correlation represents the population flow density between the regions; According to the regional association degree, determining the edge weight coefficient of the graph convolutional network (GCN) model; The node characteristic factors of each of the regional nodes are input into the GCN model to obtain the room price prediction results of the hotels in each of the regional nodes output by the GCN model.

[0006] According to a hotel room price prediction method provided by the present invention, determining the edge weight coefficient of the graph convolutional network (GCN) model according to the regional association degree includes: Constructing a correlation matrix according to the regional correlation between the regional nodes; The normalized incidence matrix is ​​summed with the identity matrix to obtain the target matrix; Sum each row of the target matrix to obtain an angle matrix; The matrix elements of the diagonal matrix are used as edge weight coefficients of the GCN model.

[0007] According to a hotel room price prediction method provided by the present invention, the iterative output of the GCN model is implemented based on the following method: In each layer of GCN, the features of neighboring nodes are aggregated; Perform linear transformation on the aggregated features, and use activation function to perform nonlinear mapping on the linearly transformed features to update the node features of the current node and obtain the updated feature vector; The updated feature vector is passed to the next layer, and the node feature update operation is repeated until the set number of layers is reached.

[0008] According to a hotel room price prediction method provided by the present invention, the regional correlation between the regional nodes is determined based on the inter-regional characteristic factors between the regional nodes, including: Inputting the inter-regional characteristic factors between the regional nodes into the extreme gradient boosting tree XGB regression model to obtain the regional correlation between the regional nodes output by the XGB regression model; The inter-region characteristic factors include at least one of the straight-line distance between regional centers, traffic information between regions, and regional facility information within associated regions.

[0009] According to a hotel room price prediction method provided by the present invention, the XGB regression model includes multiple decision trees, and the objective function of the XGB regression model is determined based on the following method: Determine the objective function of the XGB regression model according to the number of the regional nodes, the actual value of each of the regional nodes, the predicted values ​​of the first t decision trees for the regional nodes, and the model complexity of the t decision trees; The actual value represents the actual regional correlation of the regional node; and the predicted value represents the regional correlation predicted by the regional node.

[0010] According to a hotel room price prediction method provided by the present invention, the regional nodes are determined based on the following method: Divide the area by street, and obtain multiple street areas; The street area is used as the area node.

[0011] The present invention also provides a hotel price prediction device, comprising the following modules: A regional correlation determination module is used to determine the regional correlation between the regional nodes according to the regional characteristic factors between the regional nodes; the regional correlation represents the population flow density between the regions; An edge weight coefficient determination module, used to determine the edge weight coefficient of the graph convolutional network (GCN) model according to the region association degree; The hotel price prediction module is used to input the node characteristic factors of each of the regional nodes into the GCN model to obtain the price prediction results of the hotels in each of the regional nodes output by the GCN model.

[0012] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, any of the above-mentioned hotel room price prediction methods is implemented.

[0013] The present invention also provides a non-transitory computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the method for predicting hotel room prices as described above is implemented.

[0014] The present invention also provides a computer program product, comprising a computer program, wherein when the computer program is executed by a processor, the computer program implements any of the above-mentioned hotel room price prediction methods.

[0015] The hotel price prediction method, device, equipment, storage medium and program product provided by the present invention determine the regional correlation between the nodes of each region according to the regional characteristic factors between the nodes of each region; the regional correlation represents the population flow density between regions; according to the regional correlation, the edge weight coefficient of the graph convolution network GCN model is determined; the node characteristic factors of each regional node are input into the GCN model to obtain the prediction results of the hotel prices in each regional node output by the GCN model. The present invention improves the edge weight coefficient of the GCN model through the regional correlation, and can more comprehensively consider the spatial regional topological correlation characteristics between each regional node, the regional characteristic factors of the region where the node is located, and the characteristic factors of the hotel itself, so that the pricing of hotel prices is more observable, objective and scientific. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions in the present invention or 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 some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0017] Figure 1 It is a flow chart of the hotel room price prediction method provided by the present invention.

[0018] Figure 2 It is a schematic diagram of the degree of correlation between regions provided by the present invention.

[0019] Figure 3 It is a flow chart of a hotel price strategy recommendation method based on XGB improved GCN edge weights and considering spatial regional topological correlation provided by the present invention.

[0020] Figure 4 It is a structural schematic diagram of the hotel room price prediction device provided by the present invention.

[0021] Figure 5 It is a structural schematic diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION

[0022] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0023] In the relevant technologies, the cost-oriented method is a pricing method based on hotel costs, mainly including the one-thousandth pricing method, the cost-plus pricing method, the break-even pricing method, etc. Among them, the one-thousandth pricing method sets the hotel room price based on the house cost, labor cost, decoration cost and other costs invested before the hotel is built and operated; the cost-plus pricing method uses the expected target rate of return as the product pricing basis, and calculates the average room price by calculating the operating profit index; the break-even pricing method calculates the room price under the condition of determining the room demand and sales volume, so that the hotel can achieve a break-even state, or achieve the expected profit target.

[0024] The market-oriented method sets prices based on the supply and demand relationship and price competition relationship in the market, and mainly includes economic pricing models and pricing methods based on AI models. Among them, the economic pricing model is based on theories such as revenue management and game theory, and formulates scientific and reasonable room prices by analyzing the relationship between pricing and room inventory; the pricing room price based on the AI ​​model determines the relevant pricing model by studying and analyzing the relationship between the segmentation characteristics and prices of each hotel.

[0025] Hotel cost factors include labor, linens, rent, depreciation and other expenses. Cost factors are complex and diverse, and it is impossible to estimate and determine costs accurately, which can easily lead to high or low room rates. Furthermore, the cost-oriented pricing method is insensitive to changes in the market environment, and lacks consideration of market characteristic factors such as room inventory, star standards, number of surrounding hotels, etc. that affect hotel room prices, and the pricing lacks scientificity.

[0026] There are many and rich features that affect hotel prices. In addition to the hotel's own characteristics such as the number of hotel rooms, star standards, hardware facilities, the number of surrounding hotels, and surrounding transportation facilities, the spatial regional topological correlation characteristics between the target area where the hotel is located and other related areas, that is, the population flow density and correlation degree are also factors that need to be considered. This is related to the inter-regional topological characteristics such as the flow distance between regions, the degree of transportation convenience, and the correlation between business districts, medical treatment, and scenic spots between regions. In related technologies, most of the machine learning AI models based on market-oriented methods only consider the feature information related to the hotel itself, and lack consideration of the population flow density and flow correlation between the target area and other related areas, that is, lack of consideration of the correlation of spatial regional topological characteristics such as business districts, medical treatment, and scenic spots, and lack of certain observability and scientific statistical analysis.

[0027] Based on this, the present invention proposes a hotel room price prediction method.

[0028] Combine the following Figure 1-Figure 5 The present invention describes the hotel room price prediction method, device, equipment, storage medium and program product.

[0029] Figure 1 is a flow chart of the hotel room price prediction method provided by the present invention, such as Figure 1 As shown, the method includes the following: Step 101: Determine the regional correlation between the regional nodes according to the inter-regional characteristic factors between the regional nodes.

[0030] In the embodiment of the present invention, the regional node can be understood as a node in a graph convolutional neural network (GCN), wherein the regional node is determined by dividing the region by street, obtaining multiple street regions, and then using the street regions as regional nodes. For example, a city is divided into regions by street, and each street region is defined as a node in GCN. Each regional node in the city is connected to its neighboring regional nodes through edges to form a graph structure of GCN.

[0031] In the graph convolutional neural network GCN, each region node is connected to other neighboring region nodes in the network through edges, and then a new node feature input vector is obtained through aggregation calculation. Figure 2 As shown in the figure, assuming that the predicted hotel is in area A, area A is taken as the target area, and other areas B, C, D, and E are taken as associated areas. The inter-regional correlation degree (i.e., regional correlation degree) between the target area A and the associated areas can be expressed as , , , The degree of association between regions is the population flow density from target region A to other associated regions, that is, it represents the population flow density between regions.

[0032] The degree of inter-regional correlation between the target area and other related areas is affected by inter-regional characteristic factors such as the straight-line distance between regional centers, inter-regional traffic information, and regional facility information within the related areas. Among them, inter-regional traffic information includes the number of inter-regional direct subway routes, the average number of inter-regional direct subway stations, the number of inter-regional direct bus routes, the average number of inter-regional direct bus stations, and the time required for direct subways and buses; regional facility information within the related areas includes the number of office buildings within the related areas, the number of scenic spots within the related areas, the number of hospitals within the related areas, and the number of business districts within the related areas (such as the number of commercial centers or shopping areas).

[0033] In the embodiment of the present invention, based on the eXtreme Gradient Boosting (XGBoost) regression model, namely the XGB regression model, the regional association is predicted according to the inter-regional characteristic factors. For example, the inter-regional characteristic factors between the regional nodes are used as the feature input vector The XGB regression model predicts the regional correlation between nodes in each region based on the input inter-regional characteristic factors. .

[0034] The XGB regression model consists of several decision trees, where the steps for building a decision tree are as follows: (1) Confirm the objective function: ; in, is the objective function, is the number of regional nodes, For sample The actual value of (i.e. the actual regional association of the regional node), For the front Decision trees are used together to analyze the samples The predicted value of (i.e., the regional association degree predicted by the regional node), where the sample refers to the regional node, For the feature input vector, for and The loss function uses square error for calculation; is a regular term, representing the model complexity of the t-th decision tree: ; in, is the depth of the current decision tree, is the node value of the leaf node, and These are all hyperparameters used to control the severity of the penalty and prevent overfitting.

[0035] (2) Minimization of objective function: Based on the decision tree modeling principle, the training goal of the model is to minimize the objective function in the above formula. According to Taylor expansion, the objective function can be simplified to: ; in, is the first-order derivative of the loss function; is the second-order derivative of the loss function.

[0036] at this time, Can be defined as , For sample The leaf node where is the predicted value of the leaf node. The objective function can be finally simplified as: ; in, For leaf nodes The cumulative sum of the first-order derivatives of the included samples; For leaf nodes The cumulative sum of the second-order derivatives of the included samples; To fall in The sample set of leaf nodes.

[0037] XGB is constructed to minimize the standard function , so take the derivative of the above formula and set it to 0, we can get: ; in, The final predicted value for the leaf node.

[0038] (3) Confirm the XGB model loop iteration cutoff condition: The XGB tree will continuously loop through step (2) according to the characteristic factors, split and generate new nodes, and the depth of the decision tree will also increase. At the same time, XGB is composed of multiple decision trees, and the current decision tree will further perform residual fitting based on the prediction results of the previous decision tree. Therefore, it is necessary to set the corresponding maximum depth of the decision tree and the number of decision trees to prevent the model from overfitting. For example, set the maximum depth of the decision tree Maxdepth=5 and the number of decision trees Maxnum=6.

[0039] After the XGB regression model is trained, the inter-regional characteristic factors between the regional nodes are input into the XGB regression model to obtain the regional correlation between the regional nodes output by the XGB regression model. For example, the inter-regional characteristic factors between region A and region B are input into the XGB regression model to obtain the regional correlation between region A and region B output by the XGB regression model. and .

[0040] Step 102: Determine the edge weight coefficient of the graph convolutional network (GCN) model according to the region association degree.

[0041] The edge weight coefficient of the GCN model is improved according to the regional correlation predicted by the XGB regression model. For example, the angle matrix is ​​obtained by normalizing the regional correlation matrix and adding it to the unit matrix as the edge weight coefficient of the GCN model.

[0042] Step 103, input the node characteristic factors of each of the regional nodes into the GCN model to obtain the room price prediction results of the hotels in each of the regional nodes output by the GCN model.

[0043] In an embodiment of the present invention, the node feature vector is a characteristic factor affecting the room price pricing in the target area where the hotel is located, that is, the node characteristic factors, including at least one of the number of business district complexes, the number of hospitals, the number of scenic spots, the number of office buildings, the number of hotels, the number of transportation routes (such as the number of subway routes and bus routes within 1km), the distance between the hotel and the boarding station (such as a railway station, a high-speed railway station), the number of hotel rooms, the hotel star rating, the number of parking spaces, and hotel food information (such as hotel breakfast conditions).

[0044] After determining the edge weight coefficient of the GCN model according to the regional correlation, the node characteristic factors of each regional node are input into the GCN model to obtain the room price prediction results of the hotels in each regional node output by the GCN model.

[0045] The hotel room price prediction method proposed in the embodiment of the present invention determines the regional correlation between the regional nodes according to the regional characteristic factors between the regional nodes; the regional correlation represents the population flow density between the regions; according to the regional correlation, the edge weight coefficient of the graph convolution network GCN model is determined; the node characteristic factors of each regional node are input into the GCN model to obtain the room price prediction result of the hotel in each regional node output by the GCN model. The present invention improves the edge weight coefficient of the GCN model through the regional correlation, which can more comprehensively consider the spatial regional topological correlation characteristics between each regional node, the regional characteristic factors of the region where the node is located, and the characteristic factors of the hotel itself, so that the pricing of hotel room prices is more observable, objective and scientific.

[0046] Based on the above embodiment, determining the edge weight coefficient of the graph convolutional network GCN model according to the region association degree includes: Step 1020, constructing a correlation matrix according to the regional correlation degrees between the regional nodes; Step 1021, summing the normalized incidence matrix and the identity matrix to obtain a target matrix; Step 1022, sum each row of the target matrix to obtain an angle matrix; Step 1023: Use the matrix elements of the diagonal matrix as edge weight coefficients of the GCN model.

[0047] After obtaining the regional correlation between the nodes in each region predicted by the XGB regression model, Figure 3 Taking the five regions A, B, C, D, and E in the figure as an example, a correlation matrix is ​​constructed based on the regional correlation between the nodes in each region. D , as shown in the following table:

[0048] In the correlation matrix D In the above, only the correlation degree between the target area and other related areas is considered, while the node characteristic factors of the area itself, that is, the node feature vector, are not taken into account. Therefore, in the embodiment of the present invention, the node feature vectors of each area need to be integrated to obtain the final diagonal matrix , the integration method is as follows: First, the incidence matrix D Normalize to get the normalized ,Right now: ; After normalization, we get the correlation matrix for:

[0049] Furthermore, the normalized correlation matrix With the identity matrix Sum and get the target matrix : + ; Target Matrix for:

[0050] Furthermore, the target matrix Sum each row of to form a diagonal matrix, which is the diagonal matrix of the final output graph: .

[0051] In one embodiment, the iterative output of the GCN model is implemented in the following manner: in each layer of GCN, the features of neighbor nodes are aggregated; the aggregated features are linearly transformed, and the activation function is used to perform nonlinear mapping on the linearly transformed features to update the node features of the current node to obtain an updated feature vector; the updated feature vector is passed to the next layer, and the node feature update operation is repeated until the set number of layers is reached. For example, the iterative output process of the GCN model is as follows: 1) Feature aggregation: In each layer of GCN, the node features are updated by aggregating the features of neighboring nodes. Specifically, the feature vector of each node is weighted averaged with the feature vectors of its neighboring nodes; 2) Linear transformation: Apply a linear transformation to the aggregated features, such as through weight matrix multiplication, to further extract node features; 3) Activation function: Apply activation function (such as ReLU) to perform nonlinear mapping on the features after linear transformation to increase the expressive power of the model; 4) Inter-layer transfer: The updated feature vector is transferred to the next layer, and the above process is repeated until the set number of layers or convergence condition is reached; 5) Output layer: After the last layer of GCN, the output layer (such as softmax) generates the final prediction result or node classification.

[0052] According to the iterative optimization principle of graph convolutional networks, the calculation of graph networks is an iterative process that continuously considers neighbor and self information. Each iteration is a feature reorganization, and the features of the next layer are the graph convolution of the features of the previous layer. Therefore, in the embodiment of the present invention, the feature iteration process is: ; in, For the The feature matrix of the layer, that is, the node feature matrix, where the feature matrix of the 0th layer It is composed of the feature vectors of each node. For example, the characteristic factors that affect the housing price strategy in a certain node area are as follows:

[0053] The characteristic factors of this node area constitute the characteristic vector, which is one of the rows of the characteristic matrix. Finally, the vectors of all node areas are gathered to form the final characteristic matrix. .

[0054] make , then after introducing the network weight, a two-layer GCN network That is: ; in, is the weight of each 0 layer in the network, is the weight of each layer in the network, In order to reorganize the feature matrix output by the first layer according to the structure of the graph, is a non-linear activation function.

[0055] In addition to considering the characteristic attributes of the hotel itself and the regional node in which it is located, the embodiment of the present invention also considers the correlation characteristics such as commerce, transportation, etc. between the regional node and other regional nodes in the city, which will also affect the price strategy of the hotel in the region, such as the straight-line distance between regional centers, the number of direct subway routes between regions, the average number of direct subway stations between regions, and the number of office buildings, hospitals, and scenic spots in the associated regions as the correlation influences between regional nodes, making the statistical analysis more observable and objective.

[0056] In order to further analyze and illustrate the hotel room price prediction method proposed in the present invention, refer to the following examples.

[0057] The embodiment of the present invention specifically proposes a hotel price strategy recommendation method based on XGB to improve GCN edge weights and consider spatial regional topological correlation. Figure 3 As shown, the embodiment of the present invention is based on the GCN graph convolutional neural network, which divides the city into regions based on streets, and each street region is used as a node in the graph network. The regional attributes such as the business district, medical treatment, scenic spots, CBD office buildings, and the number of hotels in the target area where the hotel is predicted, as well as the hotel's own attributes such as the predicted number of hotel rooms, hardware facilities, and hotel star rating are used as node feature factors. At the same time, the embodiment of the present invention introduces the idea of ​​spatial regional topological association, considering that the population flow density from the target area where the hotel is located to other associated areas (i.e., the degree of regional association) is also a major factor affecting hotel room prices. Therefore, the straight-line distance between regions, the number of direct subways, the number of office buildings in the associated area, and other inter-regional feature factors are used as inputs, and then the degree of regional association is predicted through the XGB regression algorithm.

[0058] Furthermore, this method improves the GCN edge weights based on the predicted degree of correlation between regions, normalizes the correlation matrix of the degree of correlation between regions, and sums it with the unit matrix to obtain the correlation angle matrix as the GCN edge weight coefficient, and finally establishes a GCN graph prediction model to output the recommended room price strategy for hotels in each area of ​​the city.

[0059] refer to Figure 3 The implementation steps of the hotel price strategy recommendation method considering spatial regional topological correlation based on XGB improved GCN edge weights include: (1) Input node characteristic factors such as the straight-line distance between regional centers, the number of direct subway lines and the average number of stations between regions, the number of direct bus routes and the average number of stations between regions, the number of office buildings in the associated regions, and the number of scenic spots in the associated regions into the XGB regression model; (2) The XGB regression model outputs the inter-regional population mobility density (i.e., regional correlation); (3) Based on the regional correlation angle matrix constructed by the regional correlation degree, the edge weight of GCN is reconstructed; (4) Input node characteristic factors such as the number of commercial complexes, the number of hospitals, the number of scenic spots, the number of CBD office buildings, the number of other hotels, the number of subway routes within 1 km, the number of bus routes within 1 km, the number of hotel rooms, the hotel star rating, the number of parking spaces, and the hotel breakfast conditions into the GCN model; (5) The GCN model outputs the recommended housing price strategy for each region. The housing price strategy output by the GCN model includes the hotel housing price prediction results, which are used to reasonably price the hotel room prices.

[0060] The embodiment of the present invention more comprehensively considers the characteristic factors affecting the hotel price strategy. In addition to considering the hotel's own attributes such as the number of predicted hotel rooms, hardware facilities, and hotel star rating, it also divides the regional nodes into streets based on the GCN graph network characteristics, and considers the regional attributes such as the number of business districts, medical treatment, scenic spots, CBD office buildings, and other hotels in the regional nodes. On the other hand, the spatial regional topological association characteristics that affect the hotel price strategy are integrated and increased, and the population flow density between regions is predicted by XGB regression to map the degree of association between regions, and to improve the GCN model edge weight. In the above manner, in addition to considering the characteristic attributes of the predicted hotel itself and the regional nodes, the association characteristics such as trade, transportation, etc. between the regional nodes and other regional nodes in the city will also affect the price strategy of the hotel in the region, such as the straight-line distance between regional centers, the number of direct subway routes between regions, the average number of direct subway stations between regions, and the number of office buildings, hospitals, and scenic spots in the associated regions are also considered as the association influence between regional nodes, making the statistical analysis more observable and objective.

[0061] The hotel price prediction device provided by the present invention is described below. The hotel price prediction device described below and the hotel price prediction method described above can be referenced to each other.

[0062] refer to Figure 4 The hotel price prediction device provided by the present invention includes: a regional association determination module 401, an edge weight coefficient determination module 402 and a hotel price prediction module 403.

[0063] The regional correlation determination module 401 is used to determine the regional correlation between the regional nodes according to the regional characteristic factors between the regional nodes; the regional correlation represents the population flow density between the regions; An edge weight coefficient determination module 402 is used to determine the edge weight coefficient of the graph convolutional network GCN model according to the region association degree; The hotel room price prediction module 403 is used to input the node characteristic factors of each of the regional nodes into the GCN model to obtain the room price prediction results of the hotels in each of the regional nodes output by the GCN model.

[0064] The hotel room price prediction device proposed in the embodiment of the present invention determines the regional correlation between the regional nodes according to the regional characteristic factors between the regional nodes; the regional correlation represents the population flow density between the regions; the edge weight coefficient of the graph convolution network GCN model is determined according to the regional correlation; the node characteristic factors of each regional node are input into the GCN model to obtain the room price prediction result of the hotel in each regional node output by the GCN model. The present invention improves the edge weight coefficient of the GCN model through the regional correlation, which can more comprehensively consider the spatial regional topological correlation characteristics between each regional node, the regional characteristic factors of the region where the node is located, and the characteristic factors of the hotel itself, so that the pricing of hotel room prices is more observable, objective and scientific.

[0065] In one embodiment, the edge weight coefficient determination module 402 is specifically used to: Constructing a correlation matrix according to the regional correlation between the regional nodes; The normalized incidence matrix is ​​summed with the identity matrix to obtain the target matrix; Sum each row of the target matrix to obtain an angle matrix; The matrix elements of the diagonal matrix are used as edge weight coefficients of the GCN model.

[0066] In one embodiment, the iterative output of the GCN model is implemented based on the following method: In each layer of GCN, the features of neighboring nodes are aggregated; Perform linear transformation on the aggregated features, and use activation function to perform nonlinear mapping on the linearly transformed features to update the node features of the current node and obtain the updated feature vector; The updated feature vector is passed to the next layer, and the node feature update operation is repeated until the set number of layers is reached.

[0067] In one embodiment, the region association determination module 401 is specifically configured to: Inputting the inter-regional characteristic factors between the regional nodes into the extreme gradient boosting tree XGB regression model to obtain the regional correlation between the regional nodes output by the XGB regression model; The inter-region characteristic factors include at least one of the straight-line distance between regional centers, traffic information between regions, and regional facility information within associated regions.

[0068] In one embodiment, the XGB regression model includes multiple decision trees, and the objective function of the XGB regression model is determined based on the following method: Determine the objective function of the XGB regression model according to the number of the regional nodes, the actual value of each of the regional nodes, the predicted values ​​of the first t decision trees for the regional nodes, and the model complexity of the t decision trees; The actual value represents the actual regional correlation of the regional node; and the predicted value represents the regional correlation predicted by the regional node.

[0069] In one embodiment, the regional nodes are determined based on: Divide the area by street, and obtain multiple street areas; The street area is used as the area node.

[0070] Figure 5 An example of a physical structure diagram of an electronic device is shown in FIG. Figure 5 As shown, the electronic device may include: a processor 510, a communication interface 520, a memory 530 and a communication bus 540, wherein the processor 510, the communication interface 520 and the memory 530 communicate with each other through the communication bus 540. The processor 510 may call the logic instructions in the memory 530 to execute the hotel room price prediction method, which includes: determining the regional correlation between each of the regional nodes according to the regional characteristic factors between the regional nodes; the regional correlation represents the population flow density between the regions; determining the edge weight coefficient of the graph convolution network GCN model according to the regional correlation; inputting the node characteristic factors of each of the regional nodes into the GCN model, and obtaining the room price prediction results of the hotels in each of the regional nodes output by the GCN model.

[0071] In addition, the logic instructions in the above-mentioned memory 530 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0072] On the other hand, the present invention also provides a computer program product, which includes a computer program. The computer program can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the hotel room price prediction method provided by the above-mentioned methods, which includes: determining the regional correlation between each of the regional nodes based on the regional characteristic factors between the regional nodes; the regional correlation represents the population flow density between the regions; determining the edge weight coefficient of the graph convolutional network GCN model based on the regional correlation; inputting the node characteristic factors of each of the regional nodes into the GCN model, and obtaining the room price prediction results of the hotels in each of the regional nodes output by the GCN model.

[0073] On the other hand, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to execute the hotel room price prediction method provided by the above-mentioned methods, the method comprising: determining the regional correlation between each of the regional nodes based on the regional characteristic factors between the regional nodes; the regional correlation characterizes the population flow density between the regions; determining the edge weight coefficient of the graph convolutional network GCN model based on the regional correlation; inputting the node characteristic factors of each of the regional nodes into the GCN model, and obtaining the room price prediction results of the hotels in each of the regional nodes output by the GCN model.

[0074] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0075] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, can also be implemented by hardware. 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, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for 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 some parts of the embodiments.

[0076] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. 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 embodiments of the present invention.

Claims

1. A hotel room price prediction method, characterized in that: include: Determine the regional correlation between the regional nodes according to the regional characteristic factors between the regional nodes; the regional correlation represents the population flow density between the regions; According to the regional association degree, determining the edge weight coefficient of the graph convolutional network (GCN) model; The node characteristic factors of each of the regional nodes are input into the GCN model to obtain the room price prediction results of the hotels in each of the regional nodes output by the GCN model.

2. The hotel room price prediction method according to claim 1, characterized in that: Determining the edge weight coefficient of the graph convolutional network (GCN) model according to the region association degree includes: Constructing a correlation matrix according to the regional correlation between the regional nodes; The normalized incidence matrix is ​​summed with the identity matrix to obtain the target matrix; Sum each row of the target matrix to obtain an angle matrix; The matrix elements of the diagonal matrix are used as edge weight coefficients of the GCN model.

3. The hotel room price prediction method according to claim 1, characterized in that: The iterative output of the GCN model is achieved in the following way: In each layer of GCN, the features of neighboring nodes are aggregated; Perform linear transformation on the aggregated features, and use activation function to perform nonlinear mapping on the linearly transformed features to update the node features of the current node and obtain the updated feature vector; The updated feature vector is passed to the next layer, and the node feature update operation is repeated until the set number of layers is reached.

4. The hotel room price prediction method according to claim 1, characterized in that: Determining the regional association degree between the regional nodes according to the inter-regional characteristic factors between the regional nodes includes: Inputting the inter-regional characteristic factors between the regional nodes into the extreme gradient boosting tree XGB regression model to obtain the regional correlation between the regional nodes output by the XGB regression model; The inter-region characteristic factors include at least one of the straight-line distance between regional centers, traffic information between regions, and regional facility information within associated regions.

5. The hotel room price prediction method according to claim 4, characterized in that: The XGB regression model includes multiple decision trees, and the objective function of the XGB regression model is determined based on the following method: Determine the objective function of the XGB regression model according to the number of the regional nodes, the actual value of each of the regional nodes, the predicted values ​​of the first t decision trees for the regional nodes, and the model complexity of the t decision trees; The actual value represents the actual regional correlation of the regional node; and the predicted value represents the regional correlation predicted by the regional node.

6. The hotel room price prediction method according to claim 1, characterized in that: The regional nodes are determined based on the following method: Divide the area by street, and obtain multiple street areas; The street area is used as the area node.

7. A hotel room price prediction device, characterized in that: include: A regional correlation determination module, used to determine the regional correlation between the regional nodes according to the inter-regional characteristic factors between the regional nodes; The regional correlation represents the population flow density between regions; An edge weight coefficient determination module, used to determine the edge weight coefficient of the graph convolutional network (GCN) model according to the region association degree; The hotel price prediction module is used to input the node characteristic factors of each of the regional nodes into the GCN model to obtain the price prediction results of the hotels in each of the regional nodes output by the GCN model.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the hotel room price prediction method according to any one of claims 1 to 6 is implemented.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the hotel room price prediction method according to any one of claims 1 to 6 is implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the hotel room price prediction method according to any one of claims 1 to 6 is implemented.

Images