Data processing method and device, electronic equipment and storage medium

By constructing a collection of user trajectory sequences and combining weather, time and space factors, the long-term and real-time access preferences of users are modeled, which solves the problem of inaccurate recommendations in the existing technology and achieves more efficient user interest data recommendations.

CN119025739APending Publication Date: 2024-11-26RAJAX NETWORK &TECHNOLOGY (SHANGHAI) CO LTD
View PDF 0 Cites 1 Cited by

Patent Information

Application Number
CN202310586899.9
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2023-05-23
Publication Date
2024-11-26

AI Technical Summary

Technical Problem

The prior art lacks accuracy and efficiency in user recommendations, making it difficult to accurately understand the user's access preferences, resulting in poor recommendation results.

Method used

By obtaining user access check-in data, a collection of user trajectory sequences is constructed, and combined with weather, time and space factors, the user's long-term and real-time access preferences are modeled, and the user's long-term and real-time access preferences are generated, and the user's long-term and real-time access preferences are then used to create a user's long-term and real-time access preferences, and then the interest data candidates are selected. Sets to predict probability and recommend.

Benefits of technology

It improves the accuracy and efficiency of recommendation results, can better extract user historical trajectory information, combine long-term and real-time preferences to predict, and provide more accurate user interest data recommendations.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119025739A_ABST
    Figure CN119025739A_ABST
Patent Text Reader

Abstract

The invention provides a data processing method and device, electronic equipment and a storage medium, and relates to the technical field of Internet. The method comprises the following steps: acquiring access sign-in data of a user, constructing a sequence set of access sign-in of the user according to the access sign-in data, and determining a historical sequence and a current sequence of the user according to the sequence set; based on the historical sequence and the current sequence, fusing factors of weather, time and space, modeling the long-term access preference of the user to obtain a long-term access preference representation of the user; modeling the user real-time access preference based on the current sequence to obtain a user real-time access preference representation; according to the method, the user long-term access preference representation and the user real-time access preference representation are combined, probability prediction is carried out on the interest data candidate set, interest data recommendation is carried out according to the prediction result, and the recommendation result is accurate and efficient.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of Internet technology, and in particular to a data processing method and device, an electronic device, and a storage medium. Background Art

[0002] With the development of technology and the improvement of people's living standards, people can obtain a variety of products or services in their daily life or work. At present, when recommending products or services to users, the recommendations are not accurate and efficient enough, so this technical problem needs to be solved urgently. Summary of the invention

[0003] In view of the above problems, the present application is proposed to provide a data processing method and device, electronic device and storage medium that overcome the above problems or at least partially solve the above problems. The technical solution is as follows:

[0004] In a first aspect, a data processing method is provided, comprising:

[0005] Acquire the user's access sign-in data, construct a set of user's access sign-in trajectory sequences according to the access sign-in data, and determine the user's historical trajectory sequence and current trajectory sequence according to the trajectory sequence set;

[0006] Based on the historical trajectory sequence and the current trajectory sequence, the user's long-term access preference is modeled by integrating weather, time and space factors to obtain a representation of the user's long-term access preference;

[0007] Modeling the user's real-time access preference based on the current trajectory sequence to obtain a representation of the user's real-time access preference;

[0008] Combining the user's long-term access preference representation and the user's real-time access preference representation, a probability prediction is performed on the interest data candidate set, and the interest data is recommended based on the prediction result.

[0009] In a possible implementation, each trajectory sequence in the trajectory sequence set is composed of a plurality of visit check-in points in the order of user visit time, wherein each visit check-in point is extracted from the visit check-in data.

[0010] In a possible implementation, constructing a set of user visit sign-in trajectory sequences according to the visit sign-in data, and determining a historical trajectory sequence and a current trajectory sequence of the user according to the trajectory sequence set includes:

[0011] Extracting each visit check-in point according to the visit check-in data, and sequentially determining a plurality of visit check-in points of each trajectory sequence according to a preset division period;

[0012] Using the multiple access check-in points of each determined trajectory sequence, a set of trajectory sequences of user access check-in is constructed;

[0013] Based on the preset division period and the order of each trajectory sequence in the trajectory sequence set, a historical trajectory sequence and a current trajectory sequence of the user are determined.

[0014] In a possible implementation, based on the historical trajectory sequence and the current trajectory sequence, factors of weather, time and space are integrated to model the user's long-term access preference, and obtain the user's long-term access preference representation, including:

[0015] The similarity weights of the historical trajectory sequence and the current trajectory sequence are calculated respectively according to the three dimensions of weather, time and space, and then the factors of weather, time and space are aggregated and represented according to the similarity weights to obtain the user's long-term access preference representation.

[0016] In a possible implementation, similarity weights of the historical trajectory sequence and the current trajectory sequence are calculated respectively according to the three dimensions of weather, time and space, and then weather, time and space factors are aggregated and represented according to the similarity weights to obtain a user's long-term access preference representation, including:

[0017] Encoding each visit check-in point in the historical trajectory sequence and the current trajectory sequence to obtain an encoded historical trajectory sequence and an encoded current trajectory sequence;

[0018] Calculate the weather and time similarity aggregation of the encoded historical trajectory sequence and the encoded current trajectory sequence;

[0019] Calculate the spatial distance similarity aggregation of the encoded historical trajectory sequence and the encoded current trajectory sequence;

[0020] The weather and time similarity aggregation of the encoded historical trajectory sequence and the encoded current trajectory sequence, as well as the spatial distance similarity aggregation of the encoded historical trajectory sequence and the encoded current trajectory sequence, are combined to obtain the user's long-term access preference representation.

[0021] In a possible implementation, the calculating of weather and time similarity aggregation of the encoded historical trajectory sequence and the encoded current trajectory sequence includes:

[0022] For each visit check-in point in the encoded historical trajectory sequence, calculate the similarity with the weather and time conditions corresponding to the current moment in the encoded current trajectory sequence, and obtain the similarity weight of each visit check-in point in the encoded historical trajectory sequence, and then aggregate to obtain the vector representation of the encoded historical trajectory sequence;

[0023] For the encoded current trajectory sequence, the direct averaging method is used instead of the weather and time similarity method. The similarity weights of each visit check-in point in the encoded current trajectory sequence are the same, and then the vector representation of the encoded current trajectory sequence is obtained by aggregation.

[0024] In a possible implementation, the calculating of spatial distance similarity aggregation of the encoded historical trajectory sequence and the encoded current trajectory sequence includes:

[0025] The longitude and latitude of each visit check-in point in the encoded historical trajectory sequence are averaged to obtain the center coordinates of the encoded historical trajectory sequence;

[0026] The spatial distance similarity aggregation of the encoded historical trajectory sequence and the encoded current trajectory sequence is obtained by weighted aggregation of the latitude and longitude position of the previous moment of the current moment in the encoded current trajectory sequence and the distance of the center coordinates of the encoded historical trajectory sequence.

[0027] In a possible implementation, modeling the user's real-time access preference based on the current trajectory sequence to obtain the user's real-time access preference representation includes:

[0028] For the encoded current trajectory sequence, the long short-term memory network model is used to learn the short-term temporal relationship. The visit check-in point at the previous moment of the current moment in the encoded current trajectory sequence is used as the real-time access preference vector to obtain the user's real-time access preference representation.

[0029] In a possible implementation, combining the user's long-term access preference representation and the user's real-time access preference representation, probabilistically predicting the interest data candidate set, and recommending the interest data according to the prediction result, includes:

[0030] The distribution probability of the interest data candidate set is calculated by combining the user's long-term access preference representation and the user's real-time access preference representation, and the interest data with probability meeting the preset conditions is selected for recommendation.

[0031] In a second aspect, a data processing device is provided, comprising:

[0032] a determination unit, configured to obtain the user's visit sign-in data, construct a set of user visit sign-in trajectory sequences according to the visit sign-in data, and determine the user's historical trajectory sequence and current trajectory sequence according to the trajectory sequence set;

[0033] A long-term access preference modeling unit, configured to model the user's long-term access preference based on the historical trajectory sequence and the current trajectory sequence, integrating weather, time and space factors, and obtaining a user's long-term access preference representation;

[0034] A real-time access preference modeling unit, used to model the user's real-time access preference based on the current trajectory sequence to obtain a real-time access preference representation of the user;

[0035] The recommendation unit is used to combine the user's long-term access preference representation and the user's real-time access preference representation to perform probability prediction on the interest data candidate set, and recommend the interest data according to the prediction result.

[0036] In a possible implementation, each trajectory sequence in the trajectory sequence set is composed of a plurality of visit check-in points in the order of user visit time, wherein each visit check-in point is extracted from the visit check-in data.

[0037] In a possible implementation manner, the determining unit is further configured to:

[0038] Extracting each visit check-in point according to the visit check-in data, and sequentially determining a plurality of visit check-in points of each trajectory sequence according to a preset division period;

[0039] Using the multiple access check-in points of each determined trajectory sequence, a set of trajectory sequences of user access check-in is constructed;

[0040] Based on the preset division period and the order of each trajectory sequence in the trajectory sequence set, a historical trajectory sequence and a current trajectory sequence of the user are determined.

[0041] In a possible implementation, the long-term access preference modeling unit is further used to:

[0042] The similarity weights of the historical trajectory sequence and the current trajectory sequence are calculated respectively according to the three dimensions of weather, time and space, and then the factors of weather, time and space are aggregated and represented according to the similarity weights to obtain the user's long-term access preference representation.

[0043] In a possible implementation, the long-term access preference modeling unit is further used to:

[0044] Encoding each visit check-in point in the historical trajectory sequence and the current trajectory sequence to obtain an encoded historical trajectory sequence and an encoded current trajectory sequence;

[0045] Calculate the weather and time similarity aggregation of the encoded historical trajectory sequence and the encoded current trajectory sequence;

[0046] Calculate the spatial distance similarity aggregation of the encoded historical trajectory sequence and the encoded current trajectory sequence;

[0047] The weather and time similarity aggregation of the encoded historical trajectory sequence and the encoded current trajectory sequence, as well as the spatial distance similarity aggregation of the encoded historical trajectory sequence and the encoded current trajectory sequence, are combined to obtain the user's long-term access preference representation.

[0048] In a possible implementation, the long-term access preference modeling unit is further used to:

[0049] For each visit check-in point in the encoded historical trajectory sequence, calculate the similarity with the weather and time conditions corresponding to the current moment in the encoded current trajectory sequence, and obtain the similarity weight of each visit check-in point in the encoded historical trajectory sequence, and then aggregate to obtain the vector representation of the encoded historical trajectory sequence;

[0050] For the encoded current trajectory sequence, the direct averaging method is used instead of the weather and time similarity method. The similarity weights of each visit check-in point in the encoded current trajectory sequence are the same, and then the vector representation of the encoded current trajectory sequence is obtained by aggregation.

[0051] In a possible implementation, the long-term access preference modeling unit is further used to:

[0052] The longitude and latitude of each visit check-in point in the encoded historical trajectory sequence are averaged to obtain the center coordinates of the encoded historical trajectory sequence;

[0053] The spatial distance similarity aggregation of the encoded historical trajectory sequence and the encoded current trajectory sequence is obtained by weighted aggregation of the latitude and longitude position of the previous moment of the current moment in the encoded current trajectory sequence and the distance of the center coordinates of the encoded historical trajectory sequence.

[0054] In a possible implementation, the real-time access preference modeling unit is further used to:

[0055] For the encoded current trajectory sequence, the long short-term memory network model is used to learn the short-term temporal relationship. The visit check-in point at the previous moment of the current moment in the encoded current trajectory sequence is used as the real-time access preference vector to obtain the user's real-time access preference representation.

[0056] In a possible implementation, the recommendation unit is further configured to:

[0057] The distribution probability of the interest data candidate set is calculated by combining the user's long-term access preference representation and the user's real-time access preference representation, and the interest data with probability meeting the preset conditions is selected for recommendation.

[0058] According to a third aspect, an electronic device is provided, comprising a processor and a memory, wherein a computer program is stored in the memory, and the processor is configured to run the computer program to execute any one of the data processing methods described above.

[0059] According to a fourth aspect, a storage medium is provided, wherein the storage medium stores a computer program, wherein the computer program is configured to execute any of the above-mentioned data processing methods when run.

[0060] By means of the above technical scheme, the data processing method and device, electronic device and storage medium provided by the embodiments of the present application can obtain the user's access check-in data, construct a trajectory sequence set of the user's access check-in according to the access check-in data, and determine the user's historical trajectory sequence and current trajectory sequence according to the trajectory sequence set; based on the historical trajectory sequence and the current trajectory sequence, the factors of weather, time and space are integrated to model the user's long-term access preferences and obtain the user's long-term access preference representation; based on the current trajectory sequence, the user's real-time access preferences are modeled to obtain the user's real-time access preference representation; combined with the user's long-term access preference representation and the user's real-time access preference representation, the interest data candidate set is probabilistically predicted, and the interest data is recommended based on the prediction results. It can be seen that the embodiments of the present application can simultaneously consider the user's historical trajectory sequence and the current trajectory sequence, and perform similarity aggregation representation of the user's historical trajectory information through weather, time and space information, so as to better extract the user's historical access trajectory information and obtain the user's long-term access preference representation; and based on the current trajectory sequence, the user's real-time access preference is modeled to obtain the user's real-time access preference representation, and then the user's long-term access preference representation and the user's real-time access preference representation are combined to perform probability prediction on the candidate set of interest data, and recommend interest data based on the prediction results, so that the recommendation results are more accurate and efficient. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in describing the embodiments of the present application are briefly introduced below.

[0062] Figure 1 A flow chart showing a data processing method provided by an embodiment of the present application is shown;

[0063] Figure 2 A flow chart showing a data processing method provided by another embodiment of the present application is shown;

[0064] Figure 3 A structural diagram of a data processing device provided in an embodiment of the present application is shown;

[0065] Figure 4 A structural diagram of an electronic device provided in an embodiment of the present application is shown. DETAILED DESCRIPTION

[0066] The exemplary embodiments of the present application will be described in more detail below with reference to the accompanying drawings. Although the exemplary embodiments of the present application are shown in the accompanying drawings, it should be understood that the present application can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided in order to enable a more thorough understanding of the present application and to fully convey the scope of the present application to those skilled in the art.

[0067] It should be noted that the terms "first", "second", etc. 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 sequence. It should be understood that such use is 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. In addition, the term "including" and its variants are to be interpreted as open-ended terms meaning "including but not limited to".

[0068] In order to solve the above technical problems, the present application embodiment provides a data processing method, such as Figure 1 As shown, the data processing method may include the following steps S101 to S104:

[0069] Step S101, obtaining the user's visit sign-in data, constructing a set of user visit sign-in trajectory sequences according to the visit sign-in data, and determining the user's historical trajectory sequence and current trajectory sequence according to the trajectory sequence set;

[0070] Step S102, based on the historical trajectory sequence and the current trajectory sequence, integrating weather, time and space factors, modeling the user's long-term access preference, and obtaining the user's long-term access preference representation;

[0071] Step S103, modeling the user's real-time access preference based on the current trajectory sequence to obtain a representation of the user's real-time access preference;

[0072] Step S104 , combining the user's long-term access preference representation and the user's real-time access preference representation, probabilistically predicting the interest data candidate set, and recommending the interest data based on the prediction result.

[0073] The embodiments of the present application can simultaneously consider the user's historical trajectory sequence and the current trajectory sequence, and perform similarity aggregation representation of the user's historical trajectory information through weather, time and space information, so as to better extract the user's historical access trajectory information and obtain the user's long-term access preference representation; and based on the current trajectory sequence, the user's real-time access preference is modeled to obtain the user's real-time access preference representation, and then combine the user's long-term access preference representation and the user's real-time access preference representation to perform probability prediction on the interest data candidate set, and recommend the interest data based on the prediction results, so that the recommendation results are more accurate and efficient.

[0074] In the embodiment of the present application, a possible implementation method is provided. Each trajectory sequence in the trajectory sequence set in step S101 above can be composed of multiple access check-in points in the order of user access time, wherein each access check-in point is extracted from the access check-in data. The access check-in data here may include user identification, access check-in point identification, access time, geographic location information of the access check-in point, etc., which is not limited in the present embodiment.

[0075] In the embodiment of the present application, a possible implementation method is provided. In the above step S101, a set of trajectory sequences of user visit sign-in is constructed according to the visit sign-in data, and the historical trajectory sequence and current trajectory sequence of the user are determined according to the trajectory sequence set. Specifically, the following steps A1 to A3 may be included:

[0076] Step A1, extracting each visit check-in point according to the visit check-in data, and sequentially determining a plurality of visit check-in points of each trajectory sequence according to a preset division period.

[0077] In this step, the preset division period may be 24 hours or 48 hours, etc., which may be determined according to actual needs, and this embodiment does not impose any limitation on this.

[0078] Step A2: construct a set of user visit check-in trajectory sequences using the multiple visit check-in points of each determined trajectory sequence.

[0079] In this step, each trajectory sequence may be composed of multiple trajectory sequences, for example, the multiple trajectory sequences are trajectory sequence 1, trajectory sequence 2, trajectory sequence 3, ... trajectory sequence n, where n is a positive integer, and a trajectory sequence set of user visit check-in may be constructed.

[0080] Step A3: determining the user's historical trajectory sequence and current trajectory sequence based on the preset division period and the order of each trajectory sequence in the trajectory sequence set.

[0081] In this step, taking the example of the above multiple trajectory sequences as an example, based on the preset division period and the order of each trajectory sequence in the trajectory sequence set is trajectory sequence 1, trajectory sequence 2, trajectory sequence 3...trajectory sequence n, then trajectory sequence 1, trajectory sequence 2, trajectory sequence 3...trajectory sequence n-1 can be used as the user's historical trajectory sequence, and trajectory sequence n can be used as the user's current trajectory sequence.

[0082] In this embodiment, a trajectory sequence set of user access check-in can be constructed according to the access check-in data, and the user's historical trajectory sequence and current trajectory sequence can be determined according to the trajectory sequence set, so as to facilitate the subsequent modeling of the user's long-term access preference and the user's real-time access preference, and the modeling is efficient and accurate.

[0083] The present application provides a possible implementation method. In the above step S102, based on the historical trajectory sequence and the current trajectory sequence, the factors of weather, time and space are integrated to model the user's long-term access preference, and obtain the user's long-term access preference representation, which may specifically include the following steps B1:

[0084] Step B1, calculate the similarity weights between the historical trajectory sequence and the current trajectory sequence according to the three dimensions of weather, time and space, and then aggregate the weather, time and space factors according to the similarity weights to obtain the user's long-term access preference representation.

[0085] This embodiment can calculate the similarity weights of the historical trajectory sequence and the current trajectory sequence according to the three dimensions of weather, time and space, and then aggregate the weather, time and space factors according to the similarity weights to obtain the user's long-term access preference representation, which can better reflect the user's long-term access preference and improve the accuracy of subsequent predictions.

[0086] In the embodiment of the present application, a possible implementation method is provided. Step B1 calculates the similarity weights of the historical trajectory sequence and the current trajectory sequence respectively according to the three dimensions of weather, time and space, and then aggregates the weather, time and space factors according to the similarity weights to obtain the user's long-term access preference representation, which may specifically include the following steps B11 to B14:

[0087] Step B11, encoding each visit check-in point in the historical trajectory sequence and the current trajectory sequence to obtain an encoded historical trajectory sequence and an encoded current trajectory sequence;

[0088] Step B12, calculating the weather and time similarity aggregation of the encoded historical trajectory sequence and the encoded current trajectory sequence;

[0089] Step B13, calculating the spatial distance similarity aggregation of the encoded historical trajectory sequence and the encoded current trajectory sequence;

[0090] Step B14, combining the weather and time similarity aggregation of the encoded historical trajectory sequence and the encoded current trajectory sequence, and the spatial distance similarity aggregation of the encoded historical trajectory sequence and the encoded current trajectory sequence, to obtain the user's long-term access preference representation.

[0091] This embodiment first encodes each visit check-in point in the historical trajectory sequence and the current trajectory sequence to obtain the encoded historical trajectory sequence and the encoded current trajectory sequence, and then calculates the weather and time similarity aggregation of the encoded historical trajectory sequence and the encoded current trajectory sequence, and calculates the spatial distance similarity aggregation of the encoded historical trajectory sequence and the encoded current trajectory sequence. Then, the weather and time similarity aggregation of the encoded historical trajectory sequence and the encoded current trajectory sequence, as well as the spatial distance similarity aggregation of the encoded historical trajectory sequence and the encoded current trajectory sequence are combined to obtain the user's long-term access preference representation, which can better reflect the user's long-term access preference and improve the accuracy of subsequent predictions.

[0092] The present application provides a possible implementation method. The above step B12 calculates the weather and time similarity aggregation of the encoded historical trajectory sequence and the encoded current trajectory sequence, which may specifically include the following steps B121 and B122:

[0093] Step B121, for each visit check-in point in the encoded historical trajectory sequence, calculate the similarity between the weather and time conditions corresponding to the current moment in the encoded current trajectory sequence, and obtain the similarity weight of each visit check-in point in the encoded historical trajectory sequence, and then aggregate to obtain the vector representation of the encoded historical trajectory sequence;

[0094] Step B122, for the encoded current trajectory sequence, a direct average method is used instead of the weather and time similarity method, and the similarity weights of the visit check-in points in the encoded current trajectory sequence are the same, and then the vector representation of the encoded current trajectory sequence is obtained by aggregation.

[0095] According to the above steps B121 and B122, this embodiment can accurately and efficiently calculate the weather and time similarity aggregation of the encoded historical trajectory sequence and the encoded current trajectory sequence.

[0096] The present application provides a possible implementation method. The above step B13 calculates the spatial distance similarity aggregation of the encoded historical trajectory sequence and the encoded current trajectory sequence, which may specifically include the following steps B131 and B132:

[0097] Step B131, averaging the longitude and latitude of each visit check-in point in the encoded historical trajectory sequence to obtain the center coordinates of the encoded historical trajectory sequence;

[0098] Step B132, weighted aggregation is performed by the distance between the latitude and longitude position of the previous moment of the current moment in the encoded current trajectory sequence and the center coordinates of the encoded historical trajectory sequence to obtain the spatial distance similarity aggregation of the encoded historical trajectory sequence and the encoded current trajectory sequence.

[0099] According to the above steps B131 and B132, this embodiment can accurately and efficiently calculate the spatial distance similarity aggregation of the encoded historical trajectory sequence and the encoded current trajectory sequence.

[0100] The present application provides a possible implementation method. In the above step S103, the user's real-time access preference is modeled based on the current trajectory sequence to obtain the user's real-time access preference representation, which may specifically include the following steps C1:

[0101] Step C1: for the encoded current trajectory sequence, the encoded current trajectory sequence is processed by a long short-term memory network model to learn the short-term temporal relationship, and the visit check-in point at the previous moment of the current moment in the encoded current trajectory sequence is used as the real-time access preference vector to obtain the user's real-time access preference representation.

[0102] In this step, for example, if the current moment in the current trajectory sequence is t, then the previous moment of the current moment is t-1. The encoded current trajectory sequence can be processed by a long short-term memory network model to learn short-term temporal relationships. The visit check-in point at the previous moment of the current moment in the encoded current trajectory sequence is used as the real-time access preference vector to obtain the user's real-time access preference representation, which can better reflect the user's real-time access preference or short-term access preference and improve the accuracy of subsequent predictions.

[0103] In the embodiment of the present application, a possible implementation method is provided. In the above step S104, the user's long-term access preference representation and the user's real-time access preference representation are combined to perform probability prediction on the interest data candidate set, and the interest data is recommended according to the prediction result. Specifically, the following steps D1 may be included:

[0104] Step D1, combining the user's long-term access preference representation and the user's real-time access preference representation, calculates the distribution probability of the interest data candidate set, and selects the interest data whose probability meets the preset conditions for recommendation.

[0105] In this step, the candidate set of interest data may be a candidate set consisting of multiple interest faces, and any of the multiple interest faces may be composed of one or more places of interest. If a certain interest face is composed of one place of interest, the probability of the place of interest may be calculated by combining the user's long-term access preference representation and the user's real-time access preference representation as the probability of the interest face; if a certain interest face is composed of multiple places of interest, the probabilities of multiple places of interest may be calculated by combining the user's long-term access preference representation and the user's real-time access preference representation, and then the comprehensive probability of multiple places of interest is obtained according to the probabilities of multiple places of interest as the probability of the interest face.

[0106] The above introduces Figure 1 There are multiple implementation methods for each link of the illustrated embodiment. The data processing method of the embodiment of the present application will be further explained below through specific embodiments.

[0107] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with the relevant laws, regulations and standards of the relevant countries and regions, and provide corresponding operation entrances for users to choose to authorize or refuse.

[0108] This specific embodiment proposes a data processing method based on spatiotemporal weather interest prediction, which considers the user's historical trajectory sequence and current trajectory sequence at the same time, and uses spatiotemporal weather information to perform similarity aggregation expression on the user's historical access preference information, so as to better extract the user's historical trajectory information, such as Figure 2 As shown, the method mainly includes four steps: user trajectory definition, long-term access preference modeling, real-time or short-term access preference modeling, and interest surface prediction. The following will introduce these four steps in detail.

[0109] 1. User trajectory definition

[0110] Here L = {l1, l2, ..., l |L|} represents a set of visit check-in points. All visit check-in points involved in this embodiment belong to this set. |L| represents the number of elements in the set L. The visit check-in point here can be a POI (Point of Interest). For each user, a user visit check-in trajectory sequence set S = {S1, S2, ..., S n}, where n is a positive integer, the visit check-in data may include user identification, visit check-in point identification, visit time, geographic location information of the visit check-in point, etc., which is not limited in this embodiment.

[0111] In the trajectory sequence set S, each trajectory S m ∈S, is composed of multiple access check-in points in the order of user access time, that is, m is a positive integer between 1 and n. m | represents the set S m The number of elements in .

[0112] Given a trajectory sequence set S, the problem of interest face prediction can be described as a historical trajectory sequence {S1, S2, ..., S n-1} and the current trajectory sequence S n ={l1,l2,...,l t-1}Predict the user's likely interest pages at time t, where l t-1 It is the user’s most recent visit check-in point at time t-1.

[0113] 2. Long-term access preference modeling

[0114] The historical traces of user access check-ins hide a lot of user access preference information, so the user's long-term access preference can be modeled based on the historical trace sequence. The intuitive idea of ​​long-term access preference modeling is to extract the information that best matches the current scene from the historical trace information. Therefore, this solution calculates the similarity weights of the historical trace sequence and the current trace sequence based on the three dimensions of weather, time, and space, and finally aggregates them according to the weights.

[0115] a) Encoding: LSTM (Long Short Term Memory) encoding can be used here. LSTM is a time recurrent neural network.

[0116] For the trajectory sequence set S = {S1, S2, ..., S n}, a layer of LSTM can be selected to encode each visit check-in point in each trajectory sequence.

[0117] By encoding, the historical trajectory sequence The encoding is represented as Here h is a positive integer between 1 and n-1. h | represents the set S h The number of elements in the current trajectory sequence S n ={l1,l2,...,l t-1} encoding is represented as {h1, h2, ..., h t-1}.

[0118] b) Weather-time similarity aggregation. In order to extract trajectory information that is more relevant to the historical trajectory sequence and the current trajectory sequence, weather-time similarity aggregation is first performed on the historical trajectory sequence, that is, it focuses more on expressing visit check-in trajectories that are similar in weather conditions and time periods.

[0119] First, the time can be divided into 12 time slices = 6 time periods on weekdays + 6 time periods on weekends. For example, the 6 time periods on weekdays can be 6:00-8:00 in the morning, 8:00-11:00 in the morning, 11:00-14:00 at noon, 14:00-17:00 in the afternoon, 17:00-22:00 in the evening, and 22:00-6:00 tomorrow. The weather can be divided into 8 states, such as light rain, heavy rain, cloudy, overcast, etc. It should be noted that the examples here are only illustrative and do not limit this embodiment.

[0120] For the visit check-in point in trajectory sequence i and the visit check-in point in trajectory sequence j, the time slice similarity τ of these two visit check-in points time,i,j Similarity with weather τ weather,i,j , the similarity can be calculated using the following formula:

[0121]

[0122] Among them, H i represents the set of user's access check-in points in a certain time slice or weather in the i trajectory sequence where the access check-in point is located, H j It represents the set of user's visit check-in points in a certain time slice or weather in the j trajectory sequence where the visit check-in point is located. The final weather time total similarity formula is expressed as:

[0123] τ i,j =τ weather,i,j * τ time,i,j

[0124] For the historical trajectory sequence S h ={l1,l2,...,l |Sh|}, calculate the similarity ω with the weather conditions at the current time c of the current trajectory sequence k , k ranges from 1 to |S h | is a positive integer, and finally aggregated to obtain the historical trajectory sequence S h The vector representation of is as follows:

[0125]

[0126]

[0127] Here τ c,k =τweather,c,k *τ time,c,k ,and then Hc represents the current trajectory sequence of c at the current time, which is the set of the user's visit check-in points under the corresponding weather at the current time c; H k Represents the historical trajectory sequence S of the visit check-in point k h The set of user visit check-in points under the weather corresponding to the current time c. For example, if the weather corresponding to the current time c is light rain, then Hc represents the set of user visit check-in points under the weather corresponding to light rain in the current trajectory sequence, and H k Represents the historical trajectory sequence S of the visit check-in point k h The middle is the collection of the visit check-in points of Xiaoyu's users.

[0128] further, H c represents the current trajectory sequence of the current time c, and the set of user's access check-in points in the time slice corresponding to the current time c; H k Represents the historical trajectory sequence S of the visit check-in point k h The set of user access check-in points in the time slice corresponding to the current moment c. For example, the time slice corresponding to the current moment c is 11:00-14:00 on weekdays, then H c In the current trajectory sequence, it represents the set of user check-in points between 11:00 and 14:00 on weekdays. k Indicates the historical trajectory sequence S where the visit check-in point k is located h It is a collection of users' access check-in points from 11:00 to 14:00 on weekdays.

[0129] For the current trajectory sequence S n , we can take the direct average method instead of the weather time aggregation method. Here, we consider that each recently visited check-in point in the current trajectory sequence can reflect the user's recent visit preference, so s n The calculation of is as follows:

[0130]

[0131] c) Spatial distance aggregation: In step b), the trajectory sequence set S = {S1, S2, ..., S n The temporal weather aggregation of {s1, s2, ..., s n}After that, spatial aggregation is used to obtain a unified vector representation of the trajectory sequence.

[0132] First, for each historical trajectory sequence S h ∈{S1,S2,...,S n-1} to obtain the center coordinates of each trajectory by averaging the longitude and latitude of the track:

[0133]

[0134]

[0135] Here is the l and lat l In order to model the spatial location relationship of long-term visit preferences, an aggregation method based on location distance is adopted here. By considering the visit check-in point l at time t-1 t-1 The latitude and longitude positions of the historical trajectory sequence are weighted and aggregated with the distances of the center coordinates of each historical trajectory sequence. The trajectory with a smaller spatial distance has a greater impact on the prediction of the point of interest or the surface of interest. Therefore, the final long-term access preference is expressed as As shown below:

[0136]

[0137] Where W h is the training parameter vector matrix, d n,h Indicates the visit check-in point l t-1 and the historical trajectory sequence S h The spatial distance

[0138] 3. Short-term access preference modeling

[0139] Short-term access preference mainly considers the current trajectory sequence S n ={l1,l2,...,l t-1}, considering the preference at time t-1, LSTM is still used to process the current trajectory sequence to learn the short-term temporal relationship, and the visit check-in point at time t-1 is represented by h t-1 is the short-term access preference vector.

[0140] 4. Interest prediction

[0141] After obtaining the user's long-term access preference representation and short-term access preference representation in the above steps 2 and 3 respectively, the following formula is used to calculate the distribution probability of the candidate set of places of interest:

[0142]

[0143] in, It represents the concatenation symbol of long-term access preference representation and short-term access preference representation. is the training matrix parameter mapped to all candidate places of interest. Therefore, the place of interest with the greatest possibility at the next time t is the place of interest with the highest probability, and then the interest surface can be determined based on the place of interest. If a certain interest surface is composed of one place of interest, the probability of the place of interest can be calculated as the probability of the interest surface by combining the user's long-term access preference representation and the user's real-time access preference representation; if a certain interest surface is composed of multiple places of interest, the probabilities of multiple places of interest can be calculated by combining the user's long-term access preference representation and the user's real-time access preference representation, and then the comprehensive probability of multiple places of interest can be obtained based on the probabilities of multiple places of interest as the probability of the interest surface.

[0144] The log-likelihood loss function is as follows:

[0145]

[0146] Where N represents the total number of training samples, p k Represents the predicted probability corresponding to the kth training sample.

[0147] This embodiment is based on a data processing method for spatiotemporal weather interest prediction, which takes into account the user's historical trajectory sequence and current trajectory sequence at the same time, and uses spatiotemporal weather information to perform similarity aggregation expression on the user's historical access preference information, which can better extract the user's historical trajectory information, and then accurately and efficiently predict and recommend places of interest and interest surfaces.

[0148] It should be noted that the size of the sequence number of each step in the above embodiment does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiment of the present application. In practical applications, all the above possible implementation methods can be combined in any way to form possible embodiments of the present application, which will not be described one by one here.

[0149] Based on the data processing methods provided in the above embodiments and based on the same inventive concept, an embodiment of the present application also provides a data processing device.

[0150] Figure 3 is a structural diagram of a data processing device provided in an embodiment of the present application. Figure 3 As shown, the data processing device may specifically include a determination unit 310 , a long-term access preference modeling unit 320 , a real-time access preference modeling unit 330 and a recommendation unit 340 .

[0151] The determination unit 310 is used to obtain the user's visit sign-in data, construct a track sequence set of the user's visit sign-in according to the visit sign-in data, and determine the user's historical track sequence and current track sequence according to the track sequence set;

[0152] A long-term access preference modeling unit 320 is used to model the user's long-term access preference based on the historical trajectory sequence and the current trajectory sequence, integrating weather, time and space factors, and obtain a user's long-term access preference representation;

[0153] A real-time access preference modeling unit 330, configured to model the user's real-time access preference based on the current trajectory sequence to obtain a representation of the user's real-time access preference;

[0154] The recommendation unit 340 is used to perform probability prediction on the interest data candidate set in combination with the user's long-term access preference representation and the user's real-time access preference representation, and recommend the interest data according to the prediction result.

[0155] In an embodiment of the present application, a possible implementation method is provided, where each trajectory sequence in the trajectory sequence set is composed of multiple visit check-in points in the order of user visit time, wherein each visit check-in point is extracted from the visit check-in data.

[0156] A possible implementation manner is provided in an embodiment of the present application, wherein the determining unit 310 is further configured to:

[0157] Extracting each visit check-in point according to the visit check-in data, and sequentially determining a plurality of visit check-in points of each trajectory sequence according to a preset division period;

[0158] Using the multiple access check-in points of each determined trajectory sequence, a set of trajectory sequences of user access check-in is constructed;

[0159] Based on the preset division period and the order of each trajectory sequence in the trajectory sequence set, a historical trajectory sequence and a current trajectory sequence of the user are determined.

[0160] A possible implementation method is provided in an embodiment of the present application, wherein the long-term access preference modeling unit 320 is further configured to:

[0161] The similarity weights of the historical trajectory sequence and the current trajectory sequence are calculated respectively according to the three dimensions of weather, time and space, and then the factors of weather, time and space are aggregated and represented according to the similarity weights to obtain the user's long-term access preference representation.

[0162] A possible implementation method is provided in an embodiment of the present application, wherein the long-term access preference modeling unit 320 is further configured to:

[0163] Encoding each visit check-in point in the historical trajectory sequence and the current trajectory sequence to obtain an encoded historical trajectory sequence and an encoded current trajectory sequence;

[0164] Calculate the weather and time similarity aggregation of the encoded historical trajectory sequence and the encoded current trajectory sequence;

[0165] Calculate the spatial distance similarity aggregation of the encoded historical trajectory sequence and the encoded current trajectory sequence;

[0166] The weather and time similarity aggregation of the encoded historical trajectory sequence and the encoded current trajectory sequence, as well as the spatial distance similarity aggregation of the encoded historical trajectory sequence and the encoded current trajectory sequence, are combined to obtain the user's long-term access preference representation.

[0167] A possible implementation method is provided in an embodiment of the present application, wherein the long-term access preference modeling unit 320 is further configured to:

[0168] For each visit check-in point in the encoded historical trajectory sequence, calculate the similarity with the weather and time conditions corresponding to the current moment in the encoded current trajectory sequence, and obtain the similarity weight of each visit check-in point in the encoded historical trajectory sequence, and then aggregate to obtain the vector representation of the encoded historical trajectory sequence;

[0169] For the encoded current trajectory sequence, the direct averaging method is used instead of the weather and time similarity method. The similarity weights of each visit check-in point in the encoded current trajectory sequence are the same, and then the vector representation of the encoded current trajectory sequence is obtained by aggregation.

[0170] A possible implementation method is provided in an embodiment of the present application, wherein the long-term access preference modeling unit 320 is further configured to:

[0171] The longitude and latitude of each visit check-in point in the encoded historical trajectory sequence are averaged to obtain the center coordinates of the encoded historical trajectory sequence;

[0172] The spatial distance similarity aggregation of the encoded historical trajectory sequence and the encoded current trajectory sequence is obtained by weighted aggregation of the latitude and longitude position of the previous moment of the current moment in the encoded current trajectory sequence and the distance of the center coordinates of the encoded historical trajectory sequence.

[0173] A possible implementation method is provided in an embodiment of the present application, wherein the real-time access preference modeling unit 330 is further configured to:

[0174] For the encoded current trajectory sequence, the long short-term memory network model is used to learn the short-term temporal relationship. The visit check-in point at the previous moment of the current moment in the encoded current trajectory sequence is used as the real-time access preference vector to obtain the user's real-time access preference representation.

[0175] A possible implementation method is provided in an embodiment of the present application, wherein the recommendation unit 340 is further configured to:

[0176] The distribution probability of the interest data candidate set is calculated by combining the user's long-term access preference representation and the user's real-time access preference representation, and the interest data with probability meeting the preset conditions is selected for recommendation.

[0177] Based on the same inventive concept, an embodiment of the present application also provides an electronic device, including a processor and a memory, wherein a computer program is stored in the memory, and the processor is configured to run the computer program to execute the data processing method of any one of the above embodiments.

[0178] In an exemplary embodiment, an electronic device is provided, such as Figure 4 As shown, Figure 4 The electronic device 400 shown includes: a processor 401 and a memory 403. The processor 401 and the memory 403 are connected, such as through a bus 402. Optionally, the electronic device 400 may also include a transceiver 404. It should be noted that in actual applications, the transceiver 404 is not limited to one, and the structure of the electronic device 400 does not constitute a limitation on the embodiments of the present application.

[0179] Processor 401 may be a CPU (Central Processing Unit), a general-purpose processor, a DSP (Digital Signal Processor), an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array) or other programmable logic devices, transistor logic devices, hardware components or any combination thereof. It may implement or execute various exemplary logic blocks, modules and circuits described in conjunction with the disclosure of this application. Processor 401 may also be a combination that implements computing functions, such as a combination of one or more microprocessors, a combination of a DSP and a microprocessor, etc.

[0180] The bus 402 may include a path to transmit information between the above components. The bus 402 may be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus. The bus 402 may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 4Only one thick line is used in the diagram, but this does not mean that there is only one bus or only one type of bus.

[0181] The memory 403 can be a ROM (Read Only Memory) or other types of static storage devices that can store static information and instructions, a RAM (Random Access Memory) or other types of dynamic storage devices that can store information and instructions, or an EEPROM (Electrically Erasable Programmable Read Only Memory), a CD-ROM (Compact Disc Read Only Memory) or other optical disk storage, optical disk storage (including compressed optical disk, laser disk, optical disk, digital versatile disk, Blu-ray disk, etc.), a magnetic disk storage medium or other magnetic storage device, or any other medium that can be used to carry or store the desired program code in the form of instructions or data structures and can be accessed by a computer, but is not limited to these.

[0182] The memory 403 is used to store the computer program code for executing the solution of the present application, and the execution is controlled by the processor 401. The processor 401 is used to execute the computer program code stored in the memory 403 to implement the contents shown in the above method embodiment.

[0183] Among them, electronic devices include but are not limited to: mobile terminals such as mobile phones, laptops, digital broadcast receivers, PDAs (personal digital assistants), PADs (tablet computers), PMPs (portable multimedia players), vehicle-mounted terminals (such as vehicle-mounted navigation terminals), etc., and fixed terminals such as digital TVs, desktop computers, etc. Figure 4 The electronic device shown is merely an example and should not bring any limitation to the functions and scope of use of the embodiments of the present application.

[0184] Based on the same inventive concept, an embodiment of the present application further provides a storage medium, in which a computer program is stored, wherein the computer program is configured to execute the data processing method of any one of the above embodiments when running.

[0185] Those skilled in the art can clearly understand that the specific working processes of the systems, devices, and modules described above can refer to the corresponding processes in the aforementioned method embodiments, and for the sake of brevity, they are not further described here.

[0186] Those skilled in the art can understand that the technical solution of the present application can be essentially or all or 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, which includes a number of program instructions, so that an electronic device (such as a personal computer, a server, or a network device, etc.) executes all or part of the steps of the method described in each embodiment of the present application when running the program instructions. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk, etc., various media that can store program codes.

[0187] Alternatively, all or part of the steps of implementing the aforementioned method embodiments may be accomplished by hardware associated with program instructions (such as electronic devices such as personal computers, servers, or network devices), and the program instructions may be stored in a computer-readable storage medium. When the program instructions are executed by a processor of an electronic device, the electronic device executes all or part of the steps of the methods described in the embodiments of the present application.

[0188] The above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that within the spirit and principles of the present application, they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not deviate from the protection scope of the present application.

Claims

1. A data processing method, characterized in that: include: Acquire the user's access sign-in data, construct a set of user's access sign-in trajectory sequences according to the access sign-in data, and determine the user's historical trajectory sequence and current trajectory sequence according to the trajectory sequence set; Based on the historical trajectory sequence and the current trajectory sequence, the user's long-term access preference is modeled by integrating weather, time and space factors to obtain a representation of the user's long-term access preference; Modeling the user's real-time access preference based on the current trajectory sequence to obtain a representation of the user's real-time access preference; Combining the user's long-term access preference representation and the user's real-time access preference representation, a probability prediction is performed on the interest data candidate set, and the interest data is recommended based on the prediction result.

2. The method according to claim 1, characterized in that Each trajectory sequence in the trajectory sequence set is composed of a plurality of visit check-in points in the order of user visit time, wherein each visit check-in point is extracted from the visit check-in data.

3. The method according to claim 2, characterized in that Constructing a user's access sign-in trajectory sequence set according to the access sign-in data, and determining the user's historical trajectory sequence and current trajectory sequence according to the trajectory sequence set, including: Extracting each visit check-in point according to the visit check-in data, and sequentially determining a plurality of visit check-in points of each trajectory sequence according to a preset division period; Using the multiple access check-in points of each determined trajectory sequence, a set of trajectory sequences of user access check-in is constructed; Based on the preset division period and the order of each trajectory sequence in the trajectory sequence set, a historical trajectory sequence and a current trajectory sequence of the user are determined.

4. The method according to any one of claims 1 to 3, characterized in that Based on the historical trajectory sequence and the current trajectory sequence, the factors of weather, time and space are integrated to model the user's long-term access preference, and obtain the user's long-term access preference representation, including: The similarity weights of the historical trajectory sequence and the current trajectory sequence are calculated respectively according to the three dimensions of weather, time and space, and then the factors of weather, time and space are aggregated and represented according to the similarity weights to obtain the user's long-term access preference representation.

5. The method according to claim 4, characterized in that The similarity weights of the historical trajectory sequence and the current trajectory sequence are calculated respectively according to the three dimensions of weather, time and space, and then the weather, time and space factors are aggregated and represented according to the similarity weights to obtain the user's long-term access preference representation, including: Encoding each visit check-in point in the historical trajectory sequence and the current trajectory sequence to obtain an encoded historical trajectory sequence and an encoded current trajectory sequence; Calculate the weather and time similarity aggregation of the encoded historical trajectory sequence and the encoded current trajectory sequence; Calculate the spatial distance similarity aggregation of the encoded historical trajectory sequence and the encoded current trajectory sequence; The weather and time similarity aggregation of the encoded historical trajectory sequence and the encoded current trajectory sequence, as well as the spatial distance similarity aggregation of the encoded historical trajectory sequence and the encoded current trajectory sequence, are combined to obtain the user's long-term access preference representation.

6. The method according to claim 5, characterized in that The calculating of weather and time similarity aggregation of the encoded historical trajectory sequence and the encoded current trajectory sequence includes: For each visit check-in point in the encoded historical trajectory sequence, calculate the similarity with the weather and time conditions corresponding to the current moment in the encoded current trajectory sequence, and obtain the similarity weight of each visit check-in point in the encoded historical trajectory sequence, and then aggregate to obtain the vector representation of the encoded historical trajectory sequence; For the encoded current trajectory sequence, the direct averaging method is used instead of the weather and time similarity method. The similarity weights of each visit check-in point in the encoded current trajectory sequence are the same, and then the vector representation of the encoded current trajectory sequence is obtained by aggregation.

7. The method according to claim 5, characterized in that The calculating of spatial distance similarity aggregation of the encoded historical trajectory sequence and the encoded current trajectory sequence includes: The longitude and latitude of each visit check-in point in the encoded historical trajectory sequence are averaged to obtain the center coordinates of the encoded historical trajectory sequence; The spatial distance similarity aggregation of the encoded historical trajectory sequence and the encoded current trajectory sequence is obtained by weighted aggregation of the latitude and longitude position of the previous moment of the current moment in the encoded current trajectory sequence and the distance of the center coordinates of the encoded historical trajectory sequence.

8. A data processing device, characterized in that: include: a determination unit, configured to obtain the user's access sign-in data, construct a set of user's access sign-in trajectory sequences according to the access sign-in data, and determine the user's historical trajectory sequence and current trajectory sequence according to the trajectory sequence set; A long-term access preference modeling unit, configured to model the user's long-term access preference based on the historical trajectory sequence and the current trajectory sequence, integrating weather, time and space factors, and obtaining a user's long-term access preference representation; A real-time access preference modeling unit, used to model the user's real-time access preference based on the current trajectory sequence to obtain a real-time access preference representation of the user; The recommendation unit is used to combine the user's long-term access preference representation and the user's real-time access preference representation to perform probability prediction on the interest data candidate set, and recommend the interest data according to the prediction result.

9. An electronic device, characterized in that: The method comprises a processor and a memory, wherein a computer program is stored in the memory, and the processor is configured to run the computer program to execute the data processing method according to any one of claims 1 to 7.

10. A storage medium, characterized in that: The storage medium stores a computer program, wherein the computer program is configured to execute the data processing method according to any one of claims 1 to 7 when running.

Citation Information

Cited By

  • Data cache prefetching method and device, storage medium and program product

    CN121579809A