A method and system for determining user intention target
Through grid construction and clustering calculation methods, a prediction model of user intention targets is established, which solves the problems of low application coverage and poor prediction accuracy in the existing technology, and improves the prediction accuracy of user intention targets.
Patent Information
- Application Number
- CN202111421625.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2021-11-26
- Publication Date
- 2024-12-27
- Estimated Expiration
- 2041-11-26
AI Technical Summary
The prior art has low application coverage and poor prediction accuracy when predicting user intention targets, especially due to the deviation caused by the user's ignoring the difference in submission of demand information or historical behavioral information acquisition.
By using the grid construction method, multiple nodes are set, each node is determined jointly by the predicted probability representation value of the first intention target and the predicted probability representation value of the second intention target, and obtain the user's historical behavior information for clustering calculations, and establish A prediction model of user intention targets and predict the current user intention targets based on this model.
It improves the prediction accuracy of user intention goals, reduces prediction bias caused by the differences in user behavior information, and enhances the robustness of the prediction model.
Smart Images
Figure CN114119090B_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of computer technology, and in particular to a method and system for determining a user's intended target. Background Art
[0002] With the development of computer networks, more and more business platforms are hosted in computer networks to provide users with various services. In order to accurately provide users with the services they need, business platforms need to predict the services they need. When predicting the services required by users, the following two methods can usually be used.
[0003] The first method is to predict user services based on user demand information obtained from the user's client. This method can directly obtain user demand preferences, but because it requires user cooperation and participation, and many users will ignore submitting demand information, this method has a low application coverage rate. At the same time, there is also the possibility that users submit demand information at will, which reduces the accuracy of user service prediction.
[0004] The second method is to collect statistics on the historical behavior information of users, and according to the proportion of the historical behavior information of users corresponding to different user services, the user service corresponding to the historical behavior information with a high proportion is used as the predicted user service. However, due to the differences in the acquisition of historical behavior information of users corresponding to different user services, the amount of historical behavior information of users for different user services is inaccurate. If the user service is predicted based on this calculated proportion, the prediction accuracy of the user service will be reduced.
[0005] To give a specific example, suppose that a real estate agency platform predicts a user's intended target, and the user's intended target is the intention to buy a new house or the intention to buy a second-hand house, and pushes new house information or second-hand house information to the user according to the predicted user's intended target. When the real estate agency platform predicts the user's intended target, two methods are used: the first method is to obtain the house purchase preference demand information from the user's client and predict the user's intended target, but since most users ignore submitting the house purchase preference demand information, the application coverage of this method is low, and there are users who submit the house purchase preference demand information at will, which affects the prediction accuracy of the user's intended target. The second method is to characterize the user's demand preference by the proportion of the number of online and offline historical behavior information of the user, and take the user's intended target corresponding to the online and offline historical behavior information with a high proportion as the predicted user's intended target. However, there is a significant difference in the number of online and offline historical behavior information corresponding to the user's second-hand house purchase intention and new house purchase intention, respectively. Usually, the new house purchase intention is significantly less than the second-hand house purchase intention, so the predicted user intention target has a large deviation. Summary of the invention
[0006] In view of this, an embodiment of the present application provides a method and system for determining a user's intended target, which can improve the accuracy of predicting the user's intended target.
[0007] The embodiment of the present application is implemented as follows:
[0008] A method for determining a user's intended target, the method comprising:
[0009] Setting a plurality of nodes, each node is jointly determined by a predicted probability representation value of a first intended target and a predicted probability representation value of a second intended target, and a grid including all of the nodes covers a predicted probability representation value range of the first intended target and a predicted probability representation value range of the second intended target;
[0010] Obtaining historical behavior information of multiple users, calculating the predicted probability representation value of each user's intended target, and classifying each user into a corresponding node according to the obtained predicted probability representation value of the first intended target and the predicted probability representation value of the second intended target of each user;
[0011] Obtaining the behavior information of all users in each node in a set future time period, performing clustering calculation within the grid, and obtaining the user intention target of each node;
[0012] Receive the behavior information of the current user, calculate the predicted probability representation value of the first intention target and the predicted probability representation value of the second intention target of the current user, determine the corresponding node, and use the user intention target of the corresponding node as the intention target of the current user.
[0013] Preferably, the calculation of the predicted probability representation value of each user's intended target includes:
[0014] For each user, the historical behavior information related to the first intention target is input into the set first intention target neural network model to obtain the predicted probability representation value of the first intention target of the user; the historical behavior information related to the second intention target is input into the set second intention target neural network model to obtain the predicted probability representation value of the second intention target of the user.
[0015] Preferably, the first intended target neural network model is a neural network score model of the first intended target, and the predicted probability representation value of the first intended target of the user is a score value of the first intended target of the user;
[0016] The second intended target neural network model is a neural network score model of the second intended target, and the predicted probability representation value of the second intended target of the user is the score value of the second intended target of the user.
[0017] Preferably, the user intention target of the node obtained by calculation includes:
[0018] For each node, based on the behavior information of all users in the node during the set time period, calculate the difference between the probability value of the user implementing the first intention goal and the probability value of the user implementing the second intention goal, and the ratio of the number of users implementing the first intention goal to the number of users implementing the overall intention goal;
[0019] The difference between the probability value of the user implementing the first intention goal and the probability value of the user implementing the second intention goal of all nodes in the grid, as well as the proportion of the number of users implementing the first intention goal to the number of users implementing the overall intention goal, are normalized to the maximum-minimum value, and clustered using the K-means clustering algorithm, where the K value is 3, to obtain the clustering result;
[0020] The obtained clustering results are used as the user intention targets of each node in the grid.
[0021] Preferably, before performing the maximum-minimum normalization process, the method further includes:
[0022] A plurality of historical time points are set, and at each historical time point, based on the behavior information of all users in the node in the set time period, the difference between the probability value of the users at the historical time point implementing the first intention goal and the probability value of implementing the second intention goal is calculated, as well as the ratio of the number of users implementing the first intention goal to the number of users implementing the overall intention goal;
[0023] The differences between the probability values of the users implementing the first intention goal and the probability values of the users implementing the second intention goal at multiple historical time points, and the proportion of the number of users implementing the first intention goal to the number of users implementing the overall intention goal, are averaged and used as the differences between the probability values of the users implementing the first intention goal and the probability values of the users implementing the second intention goal, and the proportion of the number of users implementing the first intention goal to the number of users implementing the overall intention goal.
[0024] Preferably, the set time period is the next 14 days;
[0025] The multiple historical time points are historical time points for 30 consecutive days.
[0026] Preferably, before taking the obtained clustering results as the user intention targets of each node in the grid, the method further includes:
[0027] The obtained clustering results are denoised.
[0028] Preferably, the denoising process of the obtained clustering results includes:
[0029] The clustering result of each node in the grid is denoised using an eight-neighborhood approach.
[0030] A non-transitory computer-readable storage medium storing instructions, characterized in that when the instructions are executed by a processor, the processor executes the steps of any of the above-mentioned methods for determining a user's intended target.
[0031] A computer program product includes a computer program or an instruction, wherein when the computer program or the instruction is executed by a processor, the steps of any one of the above-mentioned methods for determining a user's intended target are implemented.
[0032] As seen above, the embodiment of the present application sets up multiple nodes, each node is jointly determined by the predicted probability representation value of the first intention target and the predicted probability representation value of the second intention target, including a grid of all the nodes covering the predicted probability representation value range of the first intention target and the predicted probability representation value range of the second intention target; historical behavior information of multiple users is obtained, and the predicted probability representation value of the intention target of each user is calculated, and each user is classified into a corresponding node according to the obtained predicted probability representation value of the first intention target and the predicted probability representation value of the second intention target of each user; the behavior information of all users in each node for a set future time period is obtained, and clustering calculations are performed within the grid to obtain the user intention target of each node; the behavior information of the current user is received, the predicted probability representation value of the first intention target and the predicted probability representation value of the second intention target of the current user are calculated, the corresponding node is determined, and the user intention target of the corresponding node is used as the intention target of the current user. In this way, the embodiment of the present application adopts a grid construction method to accurately model the user's intended target prediction based on the user's historical behavior information, and predicts the current user's intended target based on the established model, thereby improving the accuracy of predicting the user's intended target. BRIEF DESCRIPTION OF THE DRAWINGS
[0033] Figure 1 A flow chart of a method for determining a user's intended target provided in an embodiment of the present application;
[0034] Figure 2 A schematic diagram of the clustering results of the grid provided in the embodiment of the present application;
[0035] Figure 3 Schematic diagram of the eight neighborhoods of the node provided for this application;
[0036] Figure 4 A schematic diagram of the clustering result of the denoised grid provided in this application;
[0037] Figure 5 A schematic diagram of the system structure for determining a user's intended target provided in an embodiment of the present application;
[0038] Figure 6 A schematic diagram of an electronic device provided according to another embodiment of the present application. DETAILED DESCRIPTION
[0039] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.
[0040] The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequential order. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein, for example. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.
[0041] The technical solution of the present application is described in detail with specific embodiments below. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described in detail in some embodiments.
[0042] From the background technology, it can be seen that when predicting the user's intended target, the first method is to determine it directly based on the obtained user preference demand information, and the second method is to determine it based on the obtained user historical behavior information. When the first method is used, it is difficult to obtain user preference demand information, so it is difficult to predict the user's intended target; the second method is used because it directly uses the user's historical behavior information for prediction, and does not analyze the difference in the amount of historical behavior information obtained from different users due to different acquisition methods, which leads to inaccurate prediction of the user's intended target. In order to solve the above problems, the present application adopts a network construction method, accurately models the user's intended target prediction based on the user's historical behavior information, and predicts the current user's intended target based on the established model, thereby improving the accuracy of predicting the user's intended target.
[0043] Figure 1The method flow chart for determining the user's intended target provided in the embodiment of the present application includes the following specific steps:
[0044] Step 101, setting a plurality of nodes, each node is jointly determined by a predicted probability representation value of a first intended target and a predicted probability representation value of a second intended target, and a grid including all the nodes covers a predicted probability representation value range of the first intended target and a predicted probability representation value range of the second intended target;
[0045] Step 102: Obtain historical behavior information of multiple users, calculate the predicted probability representation value of each user's intended target, and classify each user into a corresponding node according to the obtained predicted probability representation value of the first intended target and the predicted probability representation value of the second intended target of each user;
[0046] Step 103: Obtain the behavior information of all users in each node in a set future time period, perform clustering calculation in the grid, and obtain the user intention target of each node;
[0047] Step 104, receive the behavior information of the current user, calculate the predicted probability representation value of the first intention target and the predicted probability representation value of the second intention target of the current user, determine the corresponding node, and use the user intention target of the corresponding node as the intention target of the current user.
[0048] In step 101 of the above method, a planar grid is constructed, which is divided into a plurality of nodes, each of which is determined jointly by the predicted probability representation value of one target and the predicted probability representation value of another target, that is, the coordinate value of each node is set to (x, y), where x represents the predicted probability representation value of the first intended target, and y represents the probability predicted representation value of the second intended target. The grid covers the predicted probability representation value range of the first intended target and the predicted probability representation value range of the second intended target.
[0049] Here, as needed, the grid can be set to a grid including 40*40 nodes, and the number of nodes in the grid is 1600. The score value is used as the predicted probability representation value, that is, the predicted probability representation value of the first intended target of the user is the score value of the first intended target of the user; the predicted probability representation value of the second intended target of the user is the score value of the second intended target of the user. The predicted probability representation value range of the first intended target and the predicted probability representation value range of the second intended target are set to 2 to 10 points respectively, and 0.2 points are used as the node division unit to construct a grid including 40*40 nodes.
[0050] In this case, the types of intended targets ultimately predicted by the present application include three types, one is the first intended target, another is the second intended target, and the third is uncertain whether it is the first intended target or the second intended target.
[0051] Step 102 of the above method is to divide the multiple users into groups based on the historical behavior information of the multiple users, and classify the divided groups into corresponding nodes. Specifically, the calculation of the predicted probability representation value of each user's intended target includes:
[0052] For each user, the historical behavior information related to the first intention target is input into the set first intention target neural network model to obtain the predicted probability representation value of the first intention target of the user; the historical behavior information related to the second intention target is input into the set second intention target neural network model to obtain the predicted probability representation value of the second intention target of the user.
[0053] Here, the first intention neural network model and the second intention neural network model have the same structure, which are composed of a deep neural network (DNN) and a long short-term memory network (LSTM), which process the non-time-related historical behavior information and the time-related historical behavior information in the user's historical behavior information respectively, and then summarize the processing results to obtain the predicted probability value, and then map them according to the corresponding relationship between the predicted probability value and the predicted probability representation value to obtain the corresponding predicted probability representation value.
[0054] In step 102 of the above method, the calculation of the predicted probability representation value of each user's intended target includes:
[0055] For each node, based on the behavior information of all users in the node during the set time period, calculate the difference between the probability value of the user implementing the first intention goal and the probability value of the user implementing the second intention goal, and the ratio of the number of users implementing the first intention goal to the number of users implementing the overall intention goal;
[0056] The difference between the probability value of the user implementing the first intention goal and the probability value of the user implementing the second intention goal of all nodes in the grid, as well as the proportion of the number of users implementing the first intention goal to the number of users implementing the overall intention goal, are normalized to the maximum-minimum value, and clustered using the K-means clustering algorithm, where the K value is 3, to obtain the clustering result;
[0057] The obtained clustering results are used as the user intention targets of each node in the grid.
[0058] Here, the difference between the probability value of the user implementing the first intention goal and the probability value of the user implementing the second intention goal is expressed as:
[0059] The probability difference of users implementing their intended goals = the probability value of users implementing their first intended goal - the probability value of users implementing their second intended goal. The proportion of users implementing their first intended goal to the number of users implementing their overall intended goals is expressed as:
[0060]
[0061] The probability difference of users implementing the intended goal and the proportion of users implementing the first intended goal for each node are obtained for a historical time point. In order to obtain accurate data and ensure the accuracy of the subsequent prediction modeling grid, the embodiment of the present invention will select multiple historical time points, perform the above calculations on the node, summarize the calculated data and take the average value as the final data result.
[0062] Specifically, before the maximum-minimum normalization process is performed, the method further includes:
[0063] A plurality of historical time points are set, and at each historical time point, based on the behavior information of all users in the node in the set time period, the difference between the probability value of the users at the historical time point implementing the first intention goal and the probability value of implementing the second intention goal is calculated, as well as the ratio of the number of users implementing the first intention goal to the number of users implementing the overall intention goal;
[0064] The differences between the probability values of the users implementing the first intention goal and the probability values of the users implementing the second intention goal at multiple historical time points, and the proportion of the number of users implementing the first intention goal to the number of users implementing the overall intention goal, are averaged and used as the differences between the probability values of the users implementing the first intention goal and the probability values of the users implementing the second intention goal, and the proportion of the number of users implementing the first intention goal to the number of users implementing the overall intention goal.
[0065] Here, the set time period is the next 14 days; the multiple historical time points are historical time points for 30 consecutive days.
[0066] In the embodiment of the present application, the K-means clustering algorithm is used for clustering, the K value is 3, and the clustering results obtained include:
[0067] a. Randomly select K values from the grid as the centroids of the K subsets;
[0068] b. Calculate the distances from the unselected values to the centroids of the K subsets respectively, assign the unselected values to the nearest subset according to the distance, and reconstruct the K subsets;
[0069] c. Reconstruct K subsets, and recalculate the centroid of each of the reconstructed K subsets based on the reconstructed K subsets;
[0070] d. Calculate the distances from the selected values in the grid (not the values used as the centroid) to the centroids of the K subsets respectively, and assign the unselected values to the nearest subsets according to the distances. Return to step c and continue until the reconstructed K subsets do not change.
[0071] The values in the grid include the probability difference of each node's users implementing the intended goal and the proportion of users who implement the first intended goal. K is 3. The distances from the selected values in the network to the centroids of the K subsets are calculated using the Euclidean distance method:
[0072]
[0073] After calculation, the clustering result obtained is that the intended target of each node in the grid is: the first intended target, the second intended target, or it is uncertain whether it is the first intended target or the second intended target.
[0074] In the above method, in order to more accurately predict the user's intended target, before the obtained clustering results are used as the user's intended target for each node in the grid, it also includes: denoising the obtained clustering results. Various methods can be used for denoising, and the best implementation method is: denoising the clustering results of each node in the grid using an eight-neighborhood method. Specifically, the clustering results of each node in the grid are traversed to determine whether the number of nodes with the most identical clustering results among the eight adjacent nodes of each node exceeds half of the number of nodes in the node area. If so and the clustering result of the node is not the same as the most identical clustering results, the clustering result of the node is set to the most identical clustering result.
[0075] After setting up the grid according to the above method, the process of predicting the current user's intended target as described in step 104 of the above method can be performed accordingly. After accurately predicting the user's intended target, the business platform can accurately push information related to the user's intended target to the user based on the user's intended target, so as to facilitate the user's selection.
[0076] After setting up the grid according to the above method, the effect of the set user's intention target model can also be evaluated, that is, according to the process of step 104, the behavior information of the test user is processed, the user's intention target is predicted, and compared with the actual user's intention target to determine its effect.
[0077] In the above scheme, the user's behavior information is related to the user's intended target. For intended targets in different fields, the user's behavior information to be obtained is completely different. For example, assuming that the real estate agency platform predicts the user's intended target, and the user's intended target is the intention to buy a new house or the intention to buy a second-hand house, then the user's behavior information includes online behavior information and offline behavior information. The online behavior information includes user activity information, browsing information, search information, attention information, sharing information and comment information, etc. The offline behavior information includes user follow-up information, user property viewing information and user interview information, etc.
[0078] In the above scheme, the grid is set for the prediction of two intended targets.
[0079] The following uses the example of a real estate agency platform predicting a user's intended target, where the user's intended target is the intention to purchase a new house or the intention to purchase a second-hand house, to explain this application in detail.
[0080] The first step is to build the grid
[0081] First, obtain the intention score of all users to purchase second-hand houses and the score of purchasing new houses (the score range is 2 to 10), and use 0.2 as the score interval to construct a grid with 40*40 nodes. Each node in the grid represents the user group that meets the corresponding intention score of purchasing second-hand houses and the score of purchasing new houses. After the construction is completed, calculate the user of each node in the set future time period, such as the second-hand house transaction and new house transaction in the next 14 days, so as to obtain the transaction rate difference and transaction proportion of each node. In order to make the results more accurate, set multiple historical time points, such as 30 days. For each historical time point, each node obtains the transaction rate difference and transaction proportion respectively. After averaging, the transaction rate difference and transaction proportion of each node are obtained.
[0082] Specifically, the whole process includes:
[0083] The first sub-step: for a historical time point, the historical behavior information of all users is obtained, and the historical behavior information involving the intention to purchase second-hand houses is input into the neural network score model of the intention to purchase second-hand houses to obtain the score value of the intention to purchase second-hand houses; the historical behavior information involving the intention to purchase new houses is input into the neural network score model of the intention to purchase new houses to obtain the score value of the intention to purchase new houses; the score value ranges from 2 to 10, and the higher the score, the stronger the purchase intention;
[0084] In this step, the two neural network score models have the same structure, consisting of DNN and LSTM, which process the online historical behavior information and offline historical behavior information respectively, summarize the processing results to obtain the intention probability value, and map it to the score value according to the mapping between the set intention probability value and the score value;
[0085] In the second sub-step, a grid with 40*40 nodes is constructed with 0.2 as the score interval. Each node in the grid represents a user group that meets the corresponding second-hand house purchase intention score and new house purchase score. After the construction is completed, the second-hand house transaction and new house transaction of each node in the next 14 days are calculated, and the transaction rate difference deal_rate_gap and transaction ratio deal_ratio of each node are obtained:
[0086] deal_rate_gap=deal_rate ershou -deal_rate xinfang
[0087]
[0088] Among them, deal_rate ershou and deal_rate xinfang The transaction rates of second-hand houses and new houses of users in this node, respectively, deal_cnt ers and deal_cnt xinfang They are the number of second-hand house transactions and the number of new house transactions of the users in this node respectively;
[0089] In the third sub-step, use the 30-day historical time points to calculate each node of the grid according to the second sub-step, take the average value, and obtain the deal rate difference deal_rate_gap and deal ratio deal_ratio contained in each node.
[0090] The second step is to cluster the nodes in the grid.
[0091] The values of each node in the grid are normalized and clustered using the K-means clustering algorithm.
[0092] First, the deal rate difference deal_rate_gap and deal ratio deal_ratio contained in each node in the grid are normalized using the minimum_maximum (Min_Max) method;
[0093] Secondly, cluster the nodes in the normalized grid, including:
[0094] 1) Randomly select K values from the grid as the centroids of the K subsets, where K is set to 3;
[0095] 2) Calculate the distances from the unselected values to the centroids of the K subsets respectively, assign the unselected values to the nearest subset according to the distance, and reconstruct the K subsets;
[0096] 3) reconstructing K subsets, and recalculating the centroid of each of the reconstructed K subsets based on the reconstructed K subsets;
[0097] 4) Calculate the distances from the selected values in the grid (not the values used as the centroid) to the centroids of the K subsets respectively, and assign the unselected values to the nearest subsets according to the distances. Return to step 3) and continue until the reconstructed K subsets do not change.
[0098] Here, K is set to 3, that is, the clustering result is to cluster the intention targets of the nodes in the grid into three categories: intention to buy second-hand houses, intention to buy new houses, and fuzzy intention to buy. The distances from the selected values in the network to the centroids of the K subsets are calculated using the Euclidean distance method:
[0099]
[0100] The clustering results obtained are as follows Figure 2 As shown, Figure 2 A schematic diagram of the clustering results of the grid provided in the embodiment of the present application. Among them, the ershou_cust_score of the horizontal axis and the nh_cust_score of the vertical axis are the intention score values for purchasing a second-hand house and the intention score values for purchasing a new house, respectively. The correspondence between the three intention target categories can be obtained by calculating the deal_ratio of each node transaction ratio of each category. The lowest mean is the intention to purchase a new house, followed by the fuzzy intention to purchase, and then the intention to purchase a second-hand house, which are respectively represented in the figure as the gray area above, the black area in the middle and lower part, and the gray area on the right.
[0101] The third step is to denoise the clustering results.
[0102] The clustering results are denoised using the eight-neighborhood method, the clustering results of abnormal nodes are corrected, and the final grid of nodes with intended targets is obtained.
[0103] Since the grid division granularity is relatively fine, the clustering results obtained by each node are affected by some noise. The clustering results of each node in the grid are denoised using an eight-neighborhood method. Figure 3 A schematic diagram of the eight neighborhoods of a node provided for this application, as shown in the figure, the node areas in eight directions adjacent to the node are the eight neighborhoods of the node.
[0104] The denoising process includes: traversing the clustering results of each node in the grid, determining whether the number of nodes with the most identical clustering results among the eight adjacent nodes of each node exceeds half of the number of nodes in the node area; if so and the clustering result of the node is not the same as the most identical clustering results, then setting the clustering result of the node to the most identical clustering results.
[0105] The clustering results are as follows: Figure 4 As shown, Figure 4 A schematic diagram of the clustering results of the denoised grid provided in this application.
[0106] The fourth step is to evaluate the obtained prediction model
[0107] The intention scores for purchasing second-hand houses and the intention scores for purchasing new houses of all users on the test date can be mapped to the grid of the prediction model to obtain corresponding nodes. The intention targets of the corresponding nodes are compared with the actual intention targets of the users to obtain evaluation results. The actual intention targets of the users are the transaction results of the users in the next 14 days, including: a. the transaction rate of second-hand houses and the transaction rate of new houses; b. the ratio of the transaction volume of second-hand houses to the overall transaction volume; c. the difference between the transaction rate of second-hand houses and the transaction rate of new houses.
[0108] The two indicators b and c can be used to determine the modeling effect of the prediction model. If the effect is not good, the prediction model can be readjusted.
[0109] It can be seen that the method provided by the present application provides a good solution for the modeling problem of predicting the intention target of users on real estate intermediary platforms or other business platforms. When the method of the present application is applied to the field of house buying and selling, the prediction bias caused by the acquisition of historical behavior data of users' intention to buy second-hand houses and new houses is avoided, making the effect of the established prediction model more robust.
[0110] Figure 5 A schematic diagram of the system structure for determining a user's intended target provided in an embodiment of the present application, the system comprises: a grid construction unit, a clustering unit and a prediction unit, wherein:
[0111] A grid construction unit is used to set a plurality of nodes, each node is jointly determined by the predicted probability representation value of the first intended target and the predicted probability representation value of the second intended target, and the grid including all the nodes covers the predicted probability representation value range of the first intended target and the predicted probability representation value range of the second intended target; obtain historical behavior information of a plurality of users, calculate the predicted probability representation value of the intended target of each user, and classify each user into a corresponding node according to the obtained predicted probability representation value of the first intended target and the predicted probability representation value of the second intended target of each user;
[0112] A clustering unit is used to obtain the behavior information of all users in each node in a set future time period, perform clustering calculations in the grid, obtain the user intention target of each node, and store the corresponding nodes in the grid construction unit;
[0113] The prediction unit is used to receive the behavior information of the current user, calculate the predicted probability representation value of the first intention target and the predicted probability representation value of the second intention target of the current user, determine the corresponding node, and use the user intention target of the corresponding node as the intention target of the current user.
[0114] In another embodiment of the present application, a non-transitory computer-readable storage medium is provided, wherein the non-transitory computer-readable storage medium stores instructions, which, when executed by a processor, cause the processor to execute a method for determining a user's intended target in the aforementioned embodiment. Figure 6 A schematic diagram of an electronic device provided by another embodiment of the present application. Figure 6 As shown, another embodiment of the present application further provides an electronic device, which may include a processor 601, wherein the processor 601 is used to execute the steps of the above method for determining a user's intended target. Figure 6 It can also be seen that the electronic device provided in the above embodiment also includes a non-transitory computer-readable storage medium 602, on which a computer program is stored. When the computer program is executed by the processor 601, the above method steps for determining the user's intended target are executed.
[0115] Specifically, the non-transitory computer-readable storage medium 602 can be a general storage medium, such as a mobile disk, a hard disk, FLASH, a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), or a portable compact disk read-only memory (CD-ROM), etc. When the computer program on the non-transitory computer-readable storage medium 602 is executed by the processor 601, it can trigger the processor 601 to execute the various steps of the above-mentioned method for determining the user's intended target.
[0116] In practical applications, the non-transitory computer-readable storage medium 602 may be included in the device / apparatus / system described in the above embodiment, or may exist independently without being assembled into the device / apparatus / system. The above computer-readable storage medium carries one or more programs, and when the above one or more programs are executed, each step in the above method for determining a user's intended target can be executed.
[0117] Yet another embodiment of the present application provides a computer program product, including a computer program or instructions, which, when executed by a processor, implements each step of the above-mentioned method for determining a user's intended target.
[0118] The flow chart and block diagram in the accompanying drawings of the present application show the possible architecture, function and operation of the system, method and computer program product according to the various embodiments disclosed in the present application. In this regard, each box in the flow chart or block diagram can represent a module, a program segment or a part of a code, and the above-mentioned module, program segment or a part of a code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some implementations as replacements, the functions marked in the box can also occur in the order of the standards in different figures. For example, the boxes represented by two connections can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram or flow chart, and the combination of the boxes in the block diagram or flow chart can be implemented with a dedicated hardware-based system that performs the specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.
[0119] Those skilled in the art will appreciate that the features described in the various embodiments and / or claims of the present disclosure may be combined and / or combined in a variety of ways, even if such combinations or combinations are not explicitly described in the present application. In particular, without departing from the spirit and teachings of the present application, the features described in the various embodiments and / or claims of the present application may be combined and / or combined in a variety of ways, and all of these combinations and / or combinations fall within the scope disclosed in the present application.
[0120] Specific embodiments are used herein to illustrate the principles and implementation methods of the present application. The description of the above embodiments is only used to help understand the method and core ideas of the present application, and is not used to limit the present application. For those skilled in the art, changes can be made in the specific implementation methods and application scopes according to the ideas, spirits and principles of the present application, and any modifications, equivalent substitutions, improvements, etc. made by them should be included in the scope of protection of the present application.
Claims
1. A method for determining a user's intended target, characterized in that: The method comprises: Setting a plurality of nodes, each node is jointly determined by a predicted probability representation value of a first intended target and a predicted probability representation value of a second intended target, and a grid including all of the nodes covers a predicted probability representation value range of the first intended target and a predicted probability representation value range of the second intended target; Obtaining historical behavior information of multiple users, calculating the predicted probability representation value of each user's intended target, and classifying each user into a corresponding node according to the obtained predicted probability representation value of the first intended target and the predicted probability representation value of the second intended target of each user; Obtaining the intended goals implemented by all users in each node in a set future time period, performing clustering calculations within the grid, and obtaining the intended goals of users in each node; Receive the behavior information of the current user, calculate the predicted probability representation value of the first intention target and the predicted probability representation value of the second intention target of the current user, determine the corresponding node, and use the user intention target of the corresponding node as the intention target of the current user; The performing of clustering calculation within the grid to obtain the user intention target of each node includes: For each node, based on the intended goals implemented by all users in the node in the set future time period, calculate the difference between the probability value of the user implementing the first intended goal and the probability value of the user implementing the second intended goal, and the ratio of the number of users implementing the first intended goal to the number of users implementing the overall intended goal; wherein the number of users implementing the overall intended goal is the sum of the number of users implementing the first intended goal and the number of users implementing the second intended goal; The difference between the probability value of the user implementing the first intention goal and the probability value of the user implementing the second intention goal of each node in the grid, and the proportion of the number of users implementing the first intention goal to the number of users implementing the overall intention goal, are normalized to obtain a normalized result; based on the normalized result, the K-means clustering algorithm is used to cluster the nodes in the grid, with K being 3, to obtain a clustering result; The obtained clustering results are used as the user intention targets of each node in the grid; The calculation of the predicted probability representation value of each user's intended target includes: For each user, the historical behavior information related to the first intention target is input into the set first intention target neural network model to obtain the predicted probability representation value of the first intention target of the user; the historical behavior information related to the second intention target is input into the set second intention target neural network model to obtain the predicted probability representation value of the second intention target of the user.
2. The method according to claim 1, characterized in that The first intended target neural network model is a neural network score model of the first intended target, and the predicted probability representation value of the first intended target of the user is a score value of the first intended target of the user; The second intended target neural network model is a neural network score model of the second intended target, and the predicted probability representation value of the second intended target of the user is the score value of the second intended target of the user.
3. The method according to claim 1, characterized in that Before the maximum-minimum normalization process is performed, the method further includes: Set multiple historical time points, and at each historical time point, calculate the difference between the probability value of the user implementing the first intention goal and the probability value of the user implementing the second intention goal at the historical time point according to the intention goals implemented by all users in each node in the set future time period, and the ratio of the number of users implementing the first intention goal to the number of users implementing the overall intention goals; For each node, the difference between the probability values of the user implementing the first intention goal and the probability values of the user implementing the second intention goal at multiple historical time points, and the proportion of the number of users implementing the first intention goal to the number of users implementing the overall intention goal, are averaged and used as the difference between the probability value of the user implementing the first intention goal and the probability value of the user implementing the second intention goal, and the proportion of the number of users implementing the first intention goal to the number of users implementing the overall intention goal, calculated for the node.
4. The method according to claim 3, characterized in that The set future time period is the next 14 days; the multiple historical time points are historical time points for 30 consecutive days.
5. The method according to claim 1, characterized in that Before taking the obtained clustering results as the user intention targets of each node in the grid, the method further includes: performing denoising processing on the obtained clustering results.
6. The method according to claim 5, characterized in that The denoising process for the obtained clustering result includes: denoising the clustering result of each node in the grid by using an eight-neighborhood method.
7. A non-transitory computer-readable storage medium storing instructions, characterized in that: When the instructions are executed by a processor, the processor is caused to perform the method for determining a user's intended target according to any one of claims 1 to 6.
8. A computer program product comprising a computer program or instructions, characterized in that When the computer program or instruction is executed by a processor, the method for determining a user's intended target according to any one of claims 1 to 6 is implemented.