CN108829766B - Interest point recommendation method, system, equipment and computer readable storage medium - Google Patents

Interest point recommendation method, system, equipment and computer readable storage medium Download PDF

Info

Publication number
CN108829766B
CN108829766B CN201810529606.2A CN201810529606A CN108829766B CN 108829766 B CN108829766 B CN 108829766B CN 201810529606 A CN201810529606 A CN 201810529606A CN 108829766 B CN108829766 B CN 108829766B
Authority
CN
China
Prior art keywords
interest
gate
representing
time
lstm model
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN201810529606.2A
Other languages
Chinese (zh)
Other versions
CN108829766A (en
Inventor
赵朋朋
周晓方
许佳捷
崔志明
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Suzhou University
Original Assignee
Suzhou University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Suzhou University filed Critical Suzhou University
Priority to CN201810529606.2A priority Critical patent/CN108829766B/en
Publication of CN108829766A publication Critical patent/CN108829766A/en
Application granted granted Critical
Publication of CN108829766B publication Critical patent/CN108829766B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/049—Temporal neural networks, e.g. delay elements, oscillating neurons or pulsed inputs
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q50/00—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
    • G06Q50/01—Social networking

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Health & Medical Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • Business, Economics & Management (AREA)
  • General Physics & Mathematics (AREA)
  • Computing Systems (AREA)
  • Computational Linguistics (AREA)
  • Artificial Intelligence (AREA)
  • Data Mining & Analysis (AREA)
  • Molecular Biology (AREA)
  • Biophysics (AREA)
  • General Engineering & Computer Science (AREA)
  • Biomedical Technology (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Evolutionary Computation (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Economics (AREA)
  • Human Resources & Organizations (AREA)
  • Marketing (AREA)
  • Primary Health Care (AREA)
  • Strategic Management (AREA)
  • Tourism & Hospitality (AREA)
  • General Business, Economics & Management (AREA)
  • Devices For Checking Fares Or Tickets At Control Points (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)

Abstract

The invention discloses a method, a system, equipment and a computer readable storage medium for recommending points of interest, wherein the method comprises the following steps: acquiring time information and position information of a group of interest points; inputting time information and position information of the check-in sequence to a pre-trained LSTM model; obtaining the operation result of each interest point output by the LSTM model; determining a target interest point based on the operation result; the memory unit, the output gate and the hidden state of the LSTM model are controlled by a time gate and a distance gate, wherein the time gate represents the time interval between every two adjacent interest points of check-in, and the distance gate represents the distance interval between every two adjacent interest points of check-in. The interest point recommendation method, the system, the equipment and the computer readable storage medium disclosed by the invention solve the technical problem of improving the prediction accuracy of the next interest point of the current prediction user to a certain extent.

Description

Interest point recommendation method, system, equipment and computer readable storage medium
Technical Field
The present invention relates to the field of neural network technology, and more particularly, to a method, system, device, and computer-readable storage medium for point of interest recommendation.
Background
In recent years, the rapid growth of geographic location-based social networking services such as Foursquare, Facebook Places, Yelp, etc., has attracted many users to share their locations and experiences and have accumulated large amounts of geotagged data. The geotag data provides a good opportunity for knowing the moving behavior of the user, for example, which Point of Interest (POI) the user wants to go next can be predicted according to the geotag data.
One conventional method for predicting a user's interest point is to predict a user's next interest point based on ST-RNN (Spatial Temporal recursive Neural Networks).
However, the spatial and temporal relationships of the adjacent check-in records cannot be well modeled by the ST-RNN, and the next interest point of the user can be predicted only according to the recently visited interest point of the user, so that the accuracy of the predicted interest point of the ST-RNN is poor.
In summary, how to improve the prediction accuracy of the existing method for predicting the user interest point is a problem to be solved urgently by those skilled in the art.
Disclosure of Invention
The invention aims to provide an interest point recommendation method, which can solve the technical problem of improving the prediction accuracy of the existing method for predicting user interest points to a certain extent. The invention also provides a point of interest recommendation system, equipment and a computer readable storage medium.
In order to achieve the above purpose, the invention provides the following technical scheme:
a point of interest recommendation method, comprising:
acquiring time information and position information of a group of interest points;
inputting the time information and the position information to a pre-trained LSTM model;
obtaining the operation result of each interest point output by the LSTM model;
determining a target interest point based on the operation result;
the memory unit, the output gate and the hidden state of the LSTM model are controlled by a time gate and a distance gate, wherein the time gate represents the time interval between every two adjacent checked-in interest points, and the distance gate represents the distance interval between every two adjacent checked-in interest points.
Preferably, the update equation of the LSTM model trained in advance includes:
it＝σ(Wi[ht-1,xt]+bi)；
ft＝σ(Wf[ht-1,xt]+bf)；
Figure BDA0001676890040000021
T1t＝σ(xtWxt1+σ(ΔttWt1)+bt1),s.t.Wxt1≤0；
T2t＝σ(xtWxt2+σ(ΔttWt2)+bt2)；
D1t＝σ(xtWxd1+σ(ΔdtWd1)+bd1),s.t.Wxd1≤0；
D2t＝σ(xtWxd2+σ(ΔdtWd2)+bd2)；
Figure BDA0001676890040000022
Figure BDA0001676890040000023
ot＝σ(Wo[ht-1,xt]+ΔttWto+ΔdtWdo+bo)；
Figure BDA0001676890040000024
wherein itAn input gate representing a time t; w represents the corresponding weight; h represents the hidden state; x is the number oftRepresenting the time information and the position information; b represents the corresponding bias; sigma represents mapping values to a sigmoid layer between (0,1), 1 represents complete retention, and 0 represents complete forgetting; f. oftDenotes the t-thA forgetting gate at a moment;
Figure BDA0001676890040000025
representing a state of a short term memory cell of the memory cells; c. CtRepresenting a state of a long-term memory cell of the memory cells; t is1tRepresenting a first time gate for controlling the time interval effect of the short term interest; t is2tRepresenting a second time gate for controlling the time interval effect of long-term interest; d1tRepresenting a first range gate for controlling the range-interval effect of short-term interest; d2tRepresenting a second range gate for controlling the range-interval effect of long-term interest; Δ t represents the time interval between the interest points of two adjacent check-ins; Δ d represents the distance interval between the interest points of two adjacent check-ins; otRepresenting the output gate.
Preferably, the update equation of the LSTM model trained in advance includes:
it＝σ(Wi[ht-1,xt]+bi)；
Figure BDA0001676890040000031
T1t＝σ(xtWxt1+σ(ΔttWt1)+bt1),s.t.Wxt1≤0；
T2t＝σ(xtWxt2+σ(ΔttWt2)+bt2)；
D1t＝σ(xtWxd1+σ(ΔdtWd1)+bd1),s.t.Wxd1≤0；
D2t＝σ(xtWxd2+σ(ΔdtWd2)+bd2)；
Figure BDA0001676890040000032
Figure BDA0001676890040000033
ot＝σ(Wo[ht-1,xt]+ΔttWto+ΔdtWdo+bo)；
Figure BDA0001676890040000034
wherein itAn input gate representing a time t; w represents the corresponding weight; h represents the hidden state; x is the number oftRepresenting the time information and the position information; b represents the corresponding bias; sigma represents mapping values to a sigmoid layer between (0,1), 1 represents complete retention, and 0 represents complete forgetting;
Figure BDA0001676890040000035
representing a state of a short term memory cell of the memory cells; c. CtRepresenting a state of a long-term memory cell of the memory cells; t is1tRepresenting a first time gate for controlling the time interval effect of the short term interest; t is2tRepresenting a second time gate for controlling the time interval effect of long-term interest; d1tRepresenting a first range gate for controlling the range-interval effect of short-term interest; d2tRepresenting a second range gate for controlling the range-interval effect of long-term interest; Δ t represents the time interval between the interest points of two adjacent check-ins; Δ d represents the distance interval between the interest points of two adjacent check-ins; otRepresenting the output gate.
Preferably, the pre-training of the LSTM model includes:
constructing an initial LSTM model;
acquiring a training interest point sample and a testing interest point sample;
inputting the training interest point samples to the initial LSTM model;
judging whether the output interest point of the initial LSTM model is consistent with the test interest point sample, if so, determining that the initial LSTM model is the LSTM model trained in advance, if not, adjusting the parameters of the initial LSTM model, and returning to the step of inputting the training interest point sample to the initial LSTM model;
wherein when Wxt1When > 0, set Wxt10; when W isxd1When > 0, set Wxd1＝0。
Preferably, the determining the target interest point based on the operation result includes:
performing softmax operation on the operation result to obtain the probability distribution of each interest point;
determining the target point of interest based on the probability distribution.
Preferably, the determining the target interest point based on the probability distribution includes:
and determining the first N interest points with the maximum probability distribution values as the target interest points.
Preferably, after determining the target interest point based on the operation result, the method further includes:
and sending the target interest point to a user terminal.
A point of interest recommendation system comprising:
the first acquisition module is used for acquiring time information and position information of sign-in of a group of interest points;
the input module is used for inputting the time information and the position information to a pre-trained LSTM model;
the second acquisition module is used for acquiring the operation result of each interest point output by the LSTM model;
the determining module is used for determining a target interest point based on the operation result;
the memory unit, the output gate and the hidden state of the LSTM model are controlled by a time gate and a distance gate, wherein the time gate represents the time interval between every two adjacent checked-in interest points, and the distance gate represents the distance interval between every two adjacent checked-in interest points.
A point of interest recommendation device, comprising:
a memory for storing a computer program;
a processor for implementing the steps of the point of interest recommendation method as described in any of the above when executing the computer program.
A computer-readable storage medium having stored thereon a computer program which, when being executed by a processor, carries out the steps of the point of interest recommendation method according to any one of the preceding claims.
The invention provides an interest point recommendation method, which predicts a target interest point according to time information and position information of sign-in of a group of interest points by means of a pre-trained LSTM model, wherein a memory unit, an output gate and a hidden state of the pre-trained LSTM model all comprise a time gate and a distance gate, the time gate represents a time interval between every two adjacent sign-in interest points, the distance gate represents a distance interval between every two adjacent sign-in interest points, and when the sign-in time of the interest points in a sign-in sequence is earlier and farther than the current sign-in time, the interest points reflect long-term interest of a user in time and distance; when the distance between the user and the current check-in time is closer, the short-term interest of the user in time and distance can be reflected better. Namely, the LSTM model applied by the invention considers the influence of long-term and short-term time and distance on the target interest point, which completely accords with the influence of long-term interest when the user selects the preferred interest point in real life, and meanwhile, the short-term interest makes the preference of the next interest point of the user more definite. Compared with the prior art that the ST-RNN can only predict the next interest point of the user according to the recently visited interest points of the user, the prediction accuracy is higher. In summary, the interest point recommendation method provided by the invention solves the technical problem of how to improve the prediction accuracy of the existing prediction user interest point to a certain extent. The interest point recommendation system, the interest point recommendation equipment and the computer readable storage medium provided by the invention also solve the corresponding technical problems.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to the provided drawings without creative efforts.
Fig. 1 is a flowchart of a point of interest recommendation method according to an embodiment of the present invention;
fig. 2 is a schematic diagram of a first structure of a pre-trained LSTM model in a point of interest recommendation method according to an embodiment of the present invention;
fig. 3 is a second structural diagram of an LSTM model trained in advance in the method for recommending a point of interest according to the embodiment of the present invention;
fig. 4 is a schematic structural diagram of a point of interest recommendation system according to an embodiment of the present invention;
fig. 5 is a schematic structural diagram of a point of interest recommendation device according to an embodiment of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
The action execution subject of each step in the method for recommending a point of interest provided by the embodiment of the present invention may be a system for recommending a point of interest provided by the embodiment of the present invention, and the system may be embedded in a computer, a server, and the like. For convenience of description, the action execution subject of each step in the method for recommending a point of interest provided by the embodiment of the present invention is set as a point of interest recommendation system, referred to as a recommendation system for short, provided by the embodiment of the present invention.
Referring to fig. 1, fig. 1 is a flowchart illustrating a point of interest recommendation method according to an embodiment of the present invention.
The interest point recommendation method provided by the embodiment of the invention can comprise the following steps:
step S101: and acquiring a group of time information and position information of the check-in of the interest points.
The interest points are recorded information of users after checking in on the social network service based on the geographic position, and generally comprise time information and position information. In practical application, the recommendation system can read the check-in sequence of each user interest point from the social network service, including time information and position information, and also can receive externally input time information and position information of a group of interest points transmitted by an uploading port of the recommendation system. The number of the points of interest in the set of points of interest may be determined according to actual needs, and may be all the points of interest that a certain user checked in for a period of time, it should be noted that what the recommendation system predicts may be the next possible point of interest for the user, which may be all the points of interest recorded by the social networking service for a period of time.
Step S102: inputting time information and position information to a pre-trained LSTM (Long Short-Term Memory network) model, wherein a Memory unit, an output gate and a hidden state of the LSTM model are controlled by a time gate and a distance gate, the time gate represents the time interval between every two adjacent signed interest points, and the distance gate represents the distance interval between every two adjacent signed interest points.
After acquiring a group of time information and distance information of interest point check-in, the recommendation system can input the acquired time information and distance information into a pre-trained LSTM model, and the LSTM model outputs an operation result related to each interest point according to the input time information and distance information, wherein the interest points can be a group of vectors related to each interest point, and the like. Because the memory unit, the output gate and the hidden state of the pre-trained LSTM model are all influenced by the control of a time gate and a distance gate, the time gate represents the time interval between every two adjacent checked-in interest points, the distance gate represents the distance interval between every two adjacent checked-in interest points, and when the check-in time of the interest points in the check-in sequence is far away from the current check-in time, the interest points reflect the long-term interest of the user in time and distance; when the distance between the user and the current check-in time is closer, the short-term interest of the user in time and distance can be reflected better. Namely, the LSTM model applied by the invention considers the influence of long-term and short-term time and distance on the target interest point, which completely accords with the influence of long-term interest when the user selects the preferred interest point in real life, and meanwhile, the short-term interest makes the preference of the next interest point of the user more definite.
Step S103: and acquiring the operation result of each interest point output by the LSTM model.
Step S104: and determining the target interest point based on the operation result.
After the recommendation system obtains the operation result, the target interest point can be determined according to the operation result, and specifically, one or some most prominent interest points in the operation result can be determined as the target interest point. Because the memory unit, the output gate and the hidden state of the pre-trained LSTM model all comprise a time gate and a distance gate, the time gate represents the time interval between every two adjacent checked-in interest points, and the distance gate represents the distance interval between every two adjacent checked-in interest points, when the check-in time of the interest points in the check-in sequence is far away from the current check-in time, the interest points reflect the long-term interest of the user in time and distance; when the distance between the user and the current check-in time is closer, the short-term interest of the user in time and distance can be reflected better. Namely, the LSTM model applied by the invention considers the influence of long-term and short-term time and distance on the target interest point, which completely accords with the influence of long-term interest when the user selects the preferred interest point in real life, and meanwhile, the short-term interest makes the preference of the next interest point of the user more definite. The target interest point is also referred to herein as the next interest point to be visited by the user.
The invention provides an interest point recommendation method, which predicts a target interest point according to time information and position information of sign-in of a group of interest points by means of a pre-trained LSTM model, wherein a memory unit, an output gate and a hidden state of the pre-trained LSTM model all comprise a time gate and a distance gate, the time gate represents a time interval between every two adjacent sign-in interest points, the distance gate represents a distance interval between every two adjacent sign-in interest points, and when the sign-in time of the interest points in a sign-in sequence is earlier and farther than the current sign-in time, the interest points reflect long-term interest of a user in time and distance; when the distance between the user and the current check-in time is closer, the short-term interest of the user in time and distance can be reflected better. Namely, the LSTM model applied by the invention considers the influence of long-term and short-term time and distance on the target interest point, which completely accords with the influence of long-term interest when the user selects the preferred interest point in real life, and meanwhile, the short-term interest makes the preference of the next interest point of the user more definite. Compared with the prior art that the ST-RNN can only predict the next interest point of the user according to the recently visited interest points of the user, the prediction accuracy is higher. In summary, the interest point recommendation method provided by the invention solves the technical problem of how to improve the prediction accuracy of the existing prediction user interest point to a certain extent.
In the method for recommending a point of interest provided in an embodiment of the present invention, the updating equation of the pre-trained LSTM model may include:
it＝σ(Wi[ht-1,xt]+bi)；
ft＝σ(Wf[ht-1,xt]+bf)；
Figure BDA0001676890040000081
T1t＝σ(xtWxt1+σ(ΔttWt1)+bt1),s.t.Wxt1≤0；
T2t＝σ(xtWxt2+σ(ΔttWt2)+bt2)；
D1t＝σ(xtWxd1+σ(ΔdtWd1)+bd1),s.t.Wxd1≤0；
D2t＝σ(xtWxd2+σ(ΔdtWd2)+bd2)；
Figure BDA0001676890040000082
Figure BDA0001676890040000083
ot＝σ(Wo[ht-1,xt]+ΔttWto+ΔdtWdo+bo)；
Figure BDA0001676890040000084
wherein itAn input gate representing a time t; w denotes the corresponding weight, WiRepresents the weight of the input gate, WfWeight representing forgetting gate, WcRepresenting the weight, W, of the memory celloRepresents the weight of the output gate, WtoWeight, W, of output gate when time information is tdoIndicating distance information asdThe weight of the time output gate;hindicates a hidden state, ht-1Indicates the hidden state at time (t-1), htRepresenting a hidden state at time t; x is the number oftRepresenting interest point feature vectors at t moment in the check-in sequence; b denotes the corresponding offset, biIndicating the offset of the input gate, bfIndicating the offset of a forgetting gate, bcRepresenting the bias of the memory cell, boIndicating the offset of the output gate; sigma represents mapping values to a sigmoid layer between (0,1), 1 represents complete retention, and 0 represents complete forgetting; f. oftA forgetting gate showing the t-th time;
Figure BDA0001676890040000085
representing a state of a short term memory cell of the memory cells; c. CtRepresenting the state of a long-term memory cell of said memory cells, ct-1Indicating the state of the long-term memory cell at time (t-1); t is1tAnd T2tAre the same and all represent time gates at the same time, except T1tShowing a first time gate for controlling the time interval influence of short-term interest, T2tRepresenting a second time gate for controlling the time interval effect of long-term interest; wxt1、Wt1、Wxt2、Wt2、Wxd1、Wd1、Wxd2、Wd2All represent the weight at the t-th moment; d1tRepresenting a first range gate for controlling the range-interval effect of short-term interest; d2tRepresenting a second range gate for controlling the range-interval effect of long-term interest; Δ ttRepresenting the time interval between two adjacent interest points checked in at the time t; Δ dtRepresenting the distance interval between two adjacent interest points checked in at the time t; otAn output gate is shown. Referring to fig. 2, fig. 2 is a first structural diagram of an LSTM model trained in advance in a method for recommending a point of interest according to an embodiment of the present invention.
In practical applications, in order to reduce the number of parameters and improve efficiency, the input gate and the forgetting gate in the previous embodiment may be coupled, and in the method for recommending a point of interest provided in the embodiment of the present invention, the updating equation of the pre-trained LSTM model may include:
it＝σ(Wi[ht-1,xt]+bi)；
Figure BDA0001676890040000091
T1t＝σ(xtWxt1+σ(ΔttWt1)+bt1),s.t.Wxt1≤0；
T2t＝σ(xtWxt2+σ(ΔttWt2)+bt2)；
D1t＝σ(xtWxd1+σ(ΔdtWd1)+bd1),s.t.Wxd1≤0；
D2t＝σ(xtWxd2+σ(ΔdtWd2)+bd2)；
Figure BDA0001676890040000092
Figure BDA0001676890040000093
ot＝σ(Wo[ht-1,xt]+ΔttWto+ΔdtWdo+bo)；
Figure BDA0001676890040000094
wherein itAn input gate representing a time t; w denotes the corresponding weight, WiRepresents the weight of the input gate, WcRepresenting the weight, W, of the memory celloRepresents the weight of the output gate, WtoWeight, W, of output gate when time information is tdoIndicating distance information asdThe weight of the time output gate;hindicates a hidden state, ht-1To represent(t-1)Hidden state of the moment, htRepresenting a hidden state at time t; x is the number oftRepresenting time information and location information;brepresenting the corresponding offset, biIndicating the offset of the input gate, bcRepresenting the bias of the memory cell, boIndicating the offset of the output gate; sigma represents mapping values to a sigmoid layer between (0,1), 1 represents complete retention, and 0 represents complete forgetting;
Figure BDA0001676890040000095
represents the aboveA state of a short term memory cell of the memory cells; c. CtRepresenting the state of a long-term memory cell of said memory cells, ct-1Indicating the state of the long-term memory cell at time (t-1); t is1tAnd T2tAre the same and all represent time gates at the same time, except T1tShowing a first time gate for controlling the time interval influence of short-term interest, T2tRepresents a second time gate for controlling the time interval influence of long-term interest, Wxt1、Wt1、Wxt2、Wt2、Wxd1、Wd1、Wxd2、Wd2All represent the weight at the t-th moment; d1tRepresenting a first range gate for controlling the range-interval effect of short-term interest; d2tRepresenting a second range gate for controlling the range-interval effect of long-term interest; Δ ttRepresenting the time interval between two adjacent interest points checked in at the time t; Δ dtRepresenting the distance interval between two adjacent interest points checked in at the time t; otAn output gate is shown. Referring to fig. 3, fig. 3 is a schematic diagram illustrating a second structure of a previously trained LSTM model in a point of interest recommendation method according to an embodiment of the present invention.
In the method for recommending the point of interest provided by the embodiment of the invention, the pre-training of the LSTM model may include the following steps:
constructing an initial LSTM model;
acquiring a training interest point sample and a testing interest point sample;
inputting training interest point samples to an initial LSTM model;
judging whether the output interest point of the initial LSTM model is consistent with the test interest point sample, if so, determining the initial LSTM model to be a pre-trained LSTM model, otherwise, adjusting the parameters of the initial LSTM model, and returning to the step of inputting the training interest point sample to the initial LSTM model;
wherein when Wxt1When > 0, set Wxt10; when W isxd1When > 0, set Wxd1＝0。
The training sample of interest sample as referred to herein includes time information and distance information of the point of interest check-in. In practical application, in order to adapt to the next interest point recommendation, in the process of inputting the training interest point sample to the initial LSTM model, the training interest point sample may be converted into the following form:
Figure BDA0001676890040000101
the transformed training point of interest samples are then input into the initial LSTM model, at which time vt uCorresponding to xt，d(lt+1,lt) Corresponding to Δ dtD (,) represents a function that computes the distance of two geographical locations,
Figure BDA0001676890040000102
corresponding to Δ tt. In addition, in the process of adjusting parameters, the initial LSTM model outputs htAnd vt uA probability distribution of all interest points is calculated, and then based on the output sum vt+1 uThe gradient steps are taken to optimize the loss, specifically, a variant of Adam's random gradient descent can be used to optimize the parameters in the initial LSTM model, Adam adapting the learning rate of each parameter by making smaller updates to frequent parameters and larger updates to infrequent parameters.
In the method for recommending an interest point provided in the embodiment of the present invention, the target interest point is determined based on the operation result, which may specifically be:
performing softmax operation on the operation result to obtain the probability distribution of each interest point;
a target point of interest is determined based on the probability distribution.
In practical application, after obtaining the operation result, the recommendation system may perform softmax operation on the operation result to obtain a probability distribution of each interest point, where the probability distribution is the probability that the user goes to each interest point after the current interest point sequence, and then determine the target interest point according to the probability distribution of each interest point.
In the method for recommending an interest point provided in the embodiment of the present invention, the target interest point is determined based on probability distribution, which may specifically be:
and determining the first N interest points with the maximum probability distribution values as target interest points.
In practical applications, the first N interest points with the largest probability distribution value may be determined as target interest points, and the value of N may be determined according to actual needs, such as 1, 5, 10, and the like. Furthermore, the determined N points of interest may be presented in the form of a list or the like.
In the method for recommending an interest point provided in an embodiment of the present invention, after determining a target interest point based on an operation result, the method may further include:
and sending the target interest point to the user terminal.
In practical application, after determining the target interest point, the recommendation system may further send the target interest point to a user terminal for a user to make a decision, where the user terminal includes a mobile phone, a computer, a mailbox, and the like of the user.
The invention also provides an interest point recommendation system which has the corresponding effect of the interest point recommendation method provided by the embodiment of the invention. Referring to fig. 4, fig. 4 is a schematic structural diagram of a point of interest recommendation system according to an embodiment of the present invention.
The interest point recommendation system provided by the embodiment of the invention can comprise:
the first acquisition module 101 is configured to acquire time information and location information of a group of points of interest;
an input module 102, configured to input time information and location information to a pre-trained LSTM model;
a second obtaining module 103, configured to obtain an operation result of each interest point output by the LSTM model;
a determining module 104, configured to determine a target interest point based on the operation result;
the memory unit, the output gate and the hidden state of the LSTM model all comprise a time gate and a distance gate, wherein the time gate represents the time interval between every two adjacent interest points of check-in, and the distance gate represents the distance interval between every two adjacent interest points of check-in.
For more specific working processes of the modules, reference may be made to corresponding contents disclosed in the foregoing embodiments, and details are not described herein again.
Referring to fig. 5, fig. 5 is a schematic structural diagram of an interest point recommendation device according to an embodiment of the present invention. The point of interest recommendation device comprises a memory 201 and a processor 202, and the processor 202 implements the following steps when executing the computer program stored in the memory 201:
acquiring time information and position information of a group of interest points;
inputting time information and position information to a pre-trained LSTM model;
obtaining the operation result of each interest point output by the LSTM model;
determining a target interest point based on the operation result;
the memory unit, the output gate and the hidden state of the LSTM model are controlled by a time gate and a distance gate, wherein the time gate represents the time interval between every two adjacent interest points of check-in, and the distance gate represents the distance interval between every two adjacent interest points of check-in.
In this embodiment, when the processor 202 executes the computer subprogram stored in the memory 201, the following steps may be specifically implemented: the update equation for the previously trained LSTM model is set as:
it＝σ(Wi[ht-1,xt]+bi)；
ft＝σ(Wf[ht-1,xt]+bf)；
Figure BDA0001676890040000121
T1t＝σ(xtWxt1+σ(ΔttWt1)+bt1),s.t.Wxt1≤0；
T2t＝σ(xtWxt2+σ(ΔttWt2)+bt2)；
D1t＝σ(xtWxd1+σ(ΔdtWd1)+bd1),s.t.Wxd1≤0；
D2t＝σ(xtWxd2+σ(ΔdtWd2)+bd2)；
Figure BDA0001676890040000122
Figure BDA0001676890040000125
ot＝σ(Wo[ht-1,xt]+ΔttWto+ΔdtWdo+bo)；
Figure BDA0001676890040000123
wherein itAn input gate representing a time t; w denotes the corresponding weight, WiRepresents the weight of the input gate, WfWeight representing forgetting gate, WcRepresenting the weight, W, of the memory celloRepresents the weight of the output gate, WtoWeight, W, of output gate when time information is tdoRepresenting the weight of the output gate when the distance information is d; h denotes a hidden state, ht-1Indicates the hidden state at time (t-1), htRepresenting a hidden state at time t; x is the number oftRepresenting interest point feature vectors at t moment in the check-in sequence; b denotes the corresponding offset, biIndicating the offset of the input gate, bfIndicating the offset of a forgetting gate, bcRepresenting the bias of the memory cell, boIndicating the offset of the output gate; sigma represents mapping values to a sigmoid layer between (0,1), 1 represents complete retention, and 0 represents complete forgetting; f. oftA forgetting gate showing the t-th time;
Figure BDA0001676890040000124
representing a state of a short term memory cell of the memory cells; c. CtRepresenting the state of a long-term memory cell of said memory cells, ct-1Indicating the state of the long-term memory cell at time (t-1); t is1tAnd T2tAre the same and all represent time gates at the same time, except T1tShowing a first time gate for controlling the time interval influence of short-term interest, T2tRepresenting a second time gate for controlling the time interval effect of long-term interest; wxt1、Wt1、Wxt2、Wt2、Wxd1、Wd1、Wxd2、Wd2All represent the weight at the t-th moment; d1tRepresenting a first range gate for controlling the range-interval effect of short-term interest; d2tRepresenting a second range gate for controlling the range-interval effect of long-term interest; Δ ttRepresenting the time interval between two adjacent interest points checked in at the time t; Δ dtRepresenting the distance interval between two adjacent interest points checked in at the time t; otAn output gate is shown.
In this embodiment, when the processor 202 executes the computer subprogram stored in the memory 201, the following steps may be specifically implemented: the update equation for the previously trained LSTM model is set as:
it＝σ(Wi[ht-1,xt]+bi)；
Figure BDA0001676890040000131
T1t＝σ(xtWxt1+σ(ΔttWt1)+bt1),s.t.Wxt1≤0；
T2t＝σ(xtWxt2+σ(ΔttWt2)+bt2)；
D1t＝σ(xtWxd1+σ(ΔdtWd1)+bd1),s.t.Wxd1≤0；
D2t＝σ(xtWxd2+σ(ΔdtWd2)+bd2)；
Figure BDA0001676890040000132
Figure BDA0001676890040000133
ot＝σ(Wo[ht-1,xt]+ΔttWto+ΔdtWdo+bo)；
Figure BDA0001676890040000134
wherein itAn input gate representing a time t; w denotes the corresponding weight, WiRepresents the weight of the input gate, WcRepresenting the weight, W, of the memory celloRepresents the weight of the output gate, WtoWeight, W, of output gate when time information is tdoRepresenting the weight of the output gate when the distance information is d; h denotes a hidden state, ht-1Indicates the hidden state at time (t-1), htRepresenting a hidden state at time t; x is the number oftRepresenting time information and location information; b denotes the corresponding offset, biIndicating the offset of the input gate, bcRepresenting the bias of the memory cell, boIndicating the offset of the output gate; sigma represents mapping values to a sigmoid layer between (0,1), 1 represents complete retention, and 0 represents complete forgetting;
Figure BDA0001676890040000135
representing a state of a short term memory cell of the memory cells; c. CtRepresenting the state of a long-term memory cell of said memory cells, ct-1Indicating the state of the long-term memory cell at time (t-1); t is1tAnd T2tAre the same in nature and are all shown in the tableShowing time gates at the same time, only T1tShowing a first time gate for controlling the time interval influence of short-term interest, T2tRepresents a second time gate for controlling the time interval influence of long-term interest, Wxt1、Wt1、Wxt2、Wt2、Wxd1、Wd1、Wxd2、Wd2All represent the weight at the t-th moment; d1tRepresenting a first range gate for controlling the range-interval effect of short-term interest; d2tRepresenting a second range gate for controlling the range-interval effect of long-term interest; Δ ttRepresenting the time interval between two adjacent interest points checked in at the time t; Δ dtRepresenting the distance interval between two adjacent interest points checked in at the time t; otAn output gate is shown.
In this embodiment, when the processor 202 executes the computer subprogram stored in the memory 201, the following steps may be specifically implemented: constructing an initial LSTM model; acquiring a training interest point sample and a testing interest point sample; inputting training interest point samples to an initial LSTM model; judging whether the output interest point of the initial LSTM model is consistent with the test interest point sample, if so, determining the initial LSTM model to be a pre-trained LSTM model, otherwise, adjusting the parameters of the initial LSTM model, and returning to the step of inputting the training interest point sample to the initial LSTM model; wherein when Wxt1When > 0, set Wxt10; when W isxd1When > 0, set Wxd1＝0。
In this embodiment, when the processor 202 executes the computer subprogram stored in the memory 201, the following steps may be specifically implemented: performing softmax operation on the operation result to obtain the probability distribution of each interest point; a target point of interest is determined based on the probability distribution.
In this embodiment, when the processor 202 executes the computer subprogram stored in the memory 201, the following steps may be specifically implemented: and determining the first N interest points with the maximum probability distribution values as target interest points.
In this embodiment, when the processor 202 executes the computer subprogram stored in the memory 201, the following steps may be specifically implemented: and sending the target interest point to the user terminal.
The invention also discloses a computer readable storage medium, on which a computer program is stored, which, when being executed by a processor, implements the steps of the point of interest recommendation method as described in any of the above embodiments.
For a description of relevant parts in the point of interest recommendation system, the device and the computer-readable storage medium provided by the embodiments of the present invention, reference is made to detailed descriptions of corresponding parts in the point of interest recommendation method provided by the embodiments of the present invention, and details are not repeated herein. In addition, parts of the above technical solutions provided in the embodiments of the present invention that are consistent with the implementation principles of the corresponding technical solutions in the prior art are not described in detail, so as to avoid redundant description.
The previous description of the disclosed embodiments is provided to enable any person skilled in the art to make or use the present invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other embodiments without departing from the spirit or scope of the invention. Thus, the present invention is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims (8)

1. A method for recommending points of interest, comprising:
acquiring time information and position information of a group of interest points;
inputting the time information and the position information to a pre-trained LSTM model;
obtaining the operation result of each interest point output by the LSTM model;
determining a target interest point based on the operation result;
the memory unit, the output gate and the hidden state of the LSTM model are controlled by a time gate and a distance gate, wherein the time gate represents the time interval between every two adjacent interest points of check-in, and the distance gate represents the distance interval between every two adjacent interest points of check-in;
wherein the pre-trained update equation of the LSTM model comprises:
it＝σ(Wi[ht-1,xt]+bi)；
ft＝σ(Wf[ht-1,xt]+bf)；
Figure FDA0003208234050000011
T1t＝σ(xtWxt1+σ(ΔttWt1)+bt1),s.t.Wxt1≤0；
T2t＝σ(xtWxt2+σ(ΔttWt2)+bt2)；
D1t＝σ(xtWxd1+σ(ΔdtWd1)+bd1),s.t.Wxd1≤0；
D2t＝σ(xtWxd2+σ(ΔdtWd2)+bd2)；
Figure FDA0003208234050000012
Figure FDA0003208234050000013
ot＝σ(Wo[ht-1,xt]+ΔttWto+ΔdtWdo+bo)；
Figure FDA0003208234050000014
wherein itAn input gate representing a time t;w represents the corresponding weight; h represents the hidden state; x is the number oftRepresenting the time information and the position information; b represents the corresponding bias; sigma represents mapping values to a sigmoid layer between (0,1), 1 represents complete retention, and 0 represents complete forgetting; f. oftA forgetting gate showing the t-th time;
Figure FDA0003208234050000015
representing a state of a short term memory cell of the memory cells; c. CtRepresenting a state of a long-term memory cell of the memory cells; t is1tRepresenting a first time gate for controlling the time interval effect of the short term interest; t is2tRepresenting a second time gate for controlling the time interval effect of long-term interest; d1tRepresenting a first range gate for controlling the range-interval effect of short-term interest; d2tRepresenting a second range gate for controlling the range-interval effect of long-term interest; Δ t represents the time interval between the interest points of two adjacent check-ins; Δ d represents the distance interval between the interest points of two adjacent check-ins; otRepresenting the output gate;
or, the updating equation of the LSTM model trained in advance includes:
it＝σ(Wi[ht-1,xt]+bi)；
Figure FDA0003208234050000021
T1t＝σ(xtWxt1+σ(ΔttWt1)+bt1),s.t.Wxt1≤0；
T2t＝σ(xtWxt2+σ(ΔttWt2)+bt2)；
D1t＝σ(xtWxd1+σ(ΔdtWd1)+bd1),s.t.Wxd1≤0；
D2t＝σ(xtWxd2+σ(ΔdtWd2)+bd2)；
Figure FDA0003208234050000022
Figure FDA0003208234050000023
ot＝σ(Wo[ht-1,xt]+ΔttWto+ΔdtWdo+bo)；
Figure FDA0003208234050000024
wherein itAn input gate representing a time t; w represents the corresponding weight; h represents the hidden state; x is the number oftRepresenting the time information and the position information; b represents the corresponding bias; sigma represents mapping values to a sigmoid layer between (0,1), 1 represents complete retention, and 0 represents complete forgetting;
Figure FDA0003208234050000025
representing a state of a short term memory cell of the memory cells; c. CtRepresenting a state of a long-term memory cell of the memory cells; t is1tRepresenting a first time gate for controlling the time interval effect of the short term interest; t is2tRepresenting a second time gate for controlling the time interval effect of long-term interest; d1tRepresenting a first range gate for controlling the range-interval effect of short-term interest; d2tRepresenting a second range gate for controlling the range-interval effect of long-term interest; Δ t represents the time interval between the interest points of two adjacent check-ins; Δ d represents the distance interval between the interest points of two adjacent check-ins; otRepresenting the output gate.
2. The method of claim 1, wherein pre-training the LSTM model comprises:
constructing an initial LSTM model;
acquiring a training interest point sample and a testing interest point sample;
inputting the training interest point samples to the initial LSTM model;
judging whether the output interest point of the initial LSTM model is consistent with the test interest point sample, if so, determining that the initial LSTM model is the LSTM model trained in advance, if not, adjusting the parameters of the initial LSTM model, and returning to the step of inputting the training interest point sample to the initial LSTM model;
wherein when Wxt1When > 0, set Wxt10; when W isxd1When > 0, set Wxd1＝0。
3. The method of claim 2, wherein determining the target point of interest based on the operation result comprises:
performing softmax operation on the operation result to obtain the probability distribution of each interest point;
determining the target point of interest based on the probability distribution.
4. The method of claim 3, wherein determining the target point of interest based on the probability distribution comprises:
and determining the first N interest points with the maximum probability distribution values as the target interest points.
5. The method of claim 1, wherein after determining the target point of interest based on the operation result, further comprising:
and sending the target interest point to a user terminal.
6. A point of interest recommendation system, comprising:
the first acquisition module is used for acquiring time information and position information of sign-in of a group of interest points;
the input module is used for inputting the time information and the position information to a pre-trained LSTM model;
the second acquisition module is used for acquiring the operation result of each interest point output by the LSTM model;
the determining module is used for determining a target interest point based on the operation result;
the memory unit, the output gate and the hidden state of the LSTM model are controlled by a time gate and a distance gate, wherein the time gate represents the time interval between every two adjacent interest points of check-in, and the distance gate represents the distance interval between every two adjacent interest points of check-in;
wherein the pre-trained update equation of the LSTM model comprises:
it＝σ(Wi[ht-1,xt]+bi)；
ft＝σ(Wf[ht-1,xt]+bf)；
Figure FDA0003208234050000031
T1t＝σ(xtWxt1+σ(ΔttWt1)+bt1),s.t.Wxt1≤0；
T2t＝σ(xtWxt2+σ(ΔttWt2)+bt2)；
D1t＝σ(xtWxd1+σ(ΔdtWd1)+bd1),s.t.Wxd1≤0；
D2t＝σ(xtWxd2+σ(ΔdtWd2)+bd2)；
Figure FDA0003208234050000032
Figure FDA0003208234050000033
ot＝σ(Wo[ht-1,xt]+ΔttWto+ΔdtWdo+bo)；
Figure FDA0003208234050000034
wherein itAn input gate representing a time t; w represents the corresponding weight; h represents the hidden state; x is the number oftRepresenting the time information and the position information; b represents the corresponding bias; sigma represents mapping values to a sigmoid layer between (0,1), 1 represents complete retention, and 0 represents complete forgetting; f. oftA forgetting gate showing the t-th time;
Figure FDA0003208234050000041
representing a state of a short term memory cell of the memory cells; c. CtRepresenting a state of a long-term memory cell of the memory cells; t is1tRepresenting a first time gate for controlling the time interval effect of the short term interest; t is2tRepresenting a second time gate for controlling the time interval effect of long-term interest; d1tRepresenting a first range gate for controlling the range-interval effect of short-term interest; d2tRepresenting a second range gate for controlling the range-interval effect of long-term interest; Δ t represents the time interval between the interest points of two adjacent check-ins; Δ d represents the distance interval between the interest points of two adjacent check-ins; otRepresenting the output gate;
or, the updating equation of the LSTM model trained in advance includes:
it＝σ(Wi[ht-1,xt]+bi)；
Figure FDA0003208234050000042
T1t＝σ(xtWxt1+σ(ΔttWt1)+bt1),s.t.Wxt1≤0；
T2t＝σ(xtWxt2+σ(ΔttWt2)+bt2)；
D1t＝σ(xtWxd1+σ(ΔdtWd1)+bd1),s.t.Wxd1≤0；
D2t＝σ(xtWxd2+σ(ΔdtWd2)+bd2)；
Figure FDA0003208234050000043
Figure FDA0003208234050000044
ot＝σ(Wo[ht-1,xt]+ΔttWto+ΔdtWdo+bo)；
Figure FDA0003208234050000045
wherein itAn input gate representing a time t; w represents the corresponding weight; h represents the hidden state; x is the number oftRepresenting the time information and the position information; b represents the corresponding bias; sigma represents mapping values to a sigmoid layer between (0,1), 1 represents complete retention, and 0 represents complete forgetting;
Figure FDA0003208234050000046
representing a state of a short term memory cell of the memory cells; c. CtRepresents the aboveThe state of the long term memory cell in the memory cell; t is1tRepresenting a first time gate for controlling the time interval effect of the short term interest; t is2tRepresenting a second time gate for controlling the time interval effect of long-term interest; d1tRepresenting a first range gate for controlling the range-interval effect of short-term interest; d2tRepresenting a second range gate for controlling the range-interval effect of long-term interest; Δ t represents the time interval between the interest points of two adjacent check-ins; Δ d represents the distance interval between the interest points of two adjacent check-ins; otRepresenting the output gate.
7. A point of interest recommendation device, comprising:
a memory for storing a computer program;
a processor for implementing the steps of the point of interest recommendation method according to any one of claims 1 to 5 when executing said computer program.
8. A computer-readable storage medium, characterized in that a computer program is stored on the computer-readable storage medium, which computer program, when being executed by a processor, carries out the steps of the point of interest recommendation method according to any one of claims 1 to 5.
CN201810529606.2A 2018-05-29 2018-05-29 Interest point recommendation method, system, equipment and computer readable storage medium Active CN108829766B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201810529606.2A CN108829766B (en) 2018-05-29 2018-05-29 Interest point recommendation method, system, equipment and computer readable storage medium

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201810529606.2A CN108829766B (en) 2018-05-29 2018-05-29 Interest point recommendation method, system, equipment and computer readable storage medium

Publications (2)

Publication Number Publication Date
CN108829766A CN108829766A (en) 2018-11-16
CN108829766B true CN108829766B (en) 2021-09-21

Family

ID=64146096

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201810529606.2A Active CN108829766B (en) 2018-05-29 2018-05-29 Interest point recommendation method, system, equipment and computer readable storage medium

Country Status (1)

Country Link
CN (1) CN108829766B (en)

Families Citing this family (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109461131A (en) * 2018-11-20 2019-03-12 中山大学深圳研究院 A kind of real-time deblurring method of intelligent inside rear-view mirror based on neural network algorithm
CN109800359B (en) * 2018-12-20 2021-08-17 北京百度网讯科技有限公司 Information recommendation processing method and device, electronic equipment and readable storage medium
CN110134885B (en) * 2019-05-22 2022-12-20 广东工业大学 Interest point recommendation method, device, equipment and computer storage medium
CN110289096B (en) * 2019-06-28 2021-12-07 电子科技大学 ICU (intensive Care Unit) intra-hospital mortality prediction method based on deep learning
CN111310583B (en) * 2020-01-19 2023-02-10 中国科学院重庆绿色智能技术研究院 Vehicle abnormal behavior identification method based on improved long-term and short-term memory network
CN112990430B (en) * 2021-02-08 2021-12-03 辽宁工业大学 Group division method and system based on long-time and short-time memory network
CN113111581B (en) * 2021-04-09 2022-03-11 重庆邮电大学 LSTM trajectory prediction method combining space-time factors and based on graph neural network
CN113590971B (en) * 2021-08-13 2023-11-07 浙江大学 Interest point recommendation method and system based on brain-like space-time perception characterization

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104281622A (en) * 2013-07-11 2015-01-14 华为技术有限公司 Information recommending method and information recommending device in social media
CN106169150A (en) * 2016-06-29 2016-11-30 广州市香港科大霍英东研究院 Item recommendation method based on user interest prediction and device
CN107230351A (en) * 2017-07-18 2017-10-03 福州大学 A kind of Short-time Traffic Flow Forecasting Methods based on deep learning
CN107562787A (en) * 2017-07-31 2018-01-09 北京三快在线科技有限公司 A kind of POI coding methods and device, POI recommend method, electronic equipment

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US10003924B2 (en) * 2016-08-10 2018-06-19 Yandex Europe Ag Method of and server for processing wireless device sensor data to generate an entity vector associated with a physical location

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104281622A (en) * 2013-07-11 2015-01-14 华为技术有限公司 Information recommending method and information recommending device in social media
CN106169150A (en) * 2016-06-29 2016-11-30 广州市香港科大霍英东研究院 Item recommendation method based on user interest prediction and device
CN107230351A (en) * 2017-07-18 2017-10-03 福州大学 A kind of Short-time Traffic Flow Forecasting Methods based on deep learning
CN107562787A (en) * 2017-07-31 2018-01-09 北京三快在线科技有限公司 A kind of POI coding methods and device, POI recommend method, electronic equipment

Non-Patent Citations (4)

* Cited by examiner, † Cited by third party
Title
What to Do Next:Modeling User Behaviors by Time-LSTM;Yu Zhu等;《Proceeding of the Twenty-Sixth International Joint Conference on Artificial Intelligence》;20170825;第3602-3608页 *
元搜索引擎中基于用户兴趣的个性化调度模型;徐科等;《清华大学学报(自然科学版)》;20051231;第1915-1919页 *
基于上下文感知和个性化度量嵌入的下一个兴趣点推荐;鲜学丰等;《计算机工程与科学》;20180430;第616-625页 *
基于用户签到行为的兴趣点推荐;任星怡;《计算机学报》;20170131;第28-49页 *

Also Published As

Publication number Publication date
CN108829766A (en) 2018-11-16

Similar Documents

Publication Publication Date Title
CN108829766B (en) Interest point recommendation method, system, equipment and computer readable storage medium
WO2021169577A1 (en) Wireless service traffic prediction method based on weighted federated learning
US20220092418A1 (en) Training method for air quality prediction model, prediction method and apparatus, device, program, and medium
US11507894B2 (en) System and method for ride order dispatching
CN109116299B (en) Fingerprint positioning method, terminal and computer readable storage medium
CN109948066B (en) Interest point recommendation method based on heterogeneous information network
US20140129501A1 (en) Method for predicting user operation and mobile terminal
CN110267292B (en) Cellular network flow prediction method based on three-dimensional convolutional neural network
EP2672781B1 (en) Method and mobile terminal for predicting user operation
JP2013205170A (en) Information processing device, information processing method, and program
KR20150009375A (en) Method and system for predicting power consumption
CN113852432A (en) RCS-GRU model-based spectrum prediction sensing method
CN111505740A (en) Weather prediction method, weather prediction device, computer equipment and storage medium
JP2013205171A (en) Information processing device, information processing method, and program
CN113822460A (en) Traffic flow prediction method and device, electronic equipment and storage medium
CN114637911A (en) Next interest point recommendation method of attention fusion perception network
CN109754135B (en) Credit behavior data processing method, apparatus, storage medium and computer device
US20230394554A1 (en) Method and internet of things system of charging pile recommendation for new energy vehicle in smart city
CN115510317A (en) Cloud-edge collaborative interest point recommendation method and device
CN115495661A (en) Self-adaptive interest point recommendation method based on long-term and short-term preference of user
CN116258923A (en) Image recognition model training method, device, computer equipment and storage medium
CN111556155B (en) Information pushing method and device
CN114492905A (en) Customer appeal rate prediction method and device based on multi-model fusion and computer equipment
CN110070371B (en) Data prediction model establishing method and equipment, storage medium and server thereof
CN112883292A (en) User behavior recommendation model establishment and position recommendation method based on spatio-temporal information

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant