CN116401460A - Method for improving information push accuracy, storage medium and electronic equipment - Google Patents
Method for improving information push accuracy, storage medium and electronic equipment Download PDFInfo
- Publication number
- CN116401460A CN116401460A CN202310444079.6A CN202310444079A CN116401460A CN 116401460 A CN116401460 A CN 116401460A CN 202310444079 A CN202310444079 A CN 202310444079A CN 116401460 A CN116401460 A CN 116401460A
- Authority
- CN
- China
- Prior art keywords
- information
- pushed
- sequence
- push
- piece
- 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.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 78
- 230000003993 interaction Effects 0.000 claims abstract description 21
- 238000012545 processing Methods 0.000 claims abstract description 16
- 230000000875 corresponding effect Effects 0.000 claims description 154
- 239000004743 Polypropylene Substances 0.000 claims description 9
- 238000004590 computer program Methods 0.000 claims description 8
- 238000012163 sequencing technique Methods 0.000 claims description 6
- 230000002596 correlated effect Effects 0.000 claims description 4
- 230000006870 function Effects 0.000 claims description 4
- -1 Polypropylene Polymers 0.000 claims description 3
- 229920001155 polypropylene Polymers 0.000 claims description 3
- 230000001960 triggered effect Effects 0.000 claims description 3
- 238000005457 optimization Methods 0.000 claims 1
- 230000001105 regulatory effect Effects 0.000 abstract 2
- 238000004364 calculation method Methods 0.000 description 7
- 230000006399 behavior Effects 0.000 description 6
- 238000013507 mapping Methods 0.000 description 4
- 238000004422 calculation algorithm Methods 0.000 description 3
- 230000003287 optical effect Effects 0.000 description 3
- 230000002452 interceptive effect Effects 0.000 description 2
- 239000013307 optical fiber Substances 0.000 description 2
- 230000000644 propagated effect Effects 0.000 description 2
- 238000003491 array Methods 0.000 description 1
- 238000010420 art technique Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000015572 biosynthetic process Effects 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 230000007423 decrease Effects 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 230000002093 peripheral effect Effects 0.000 description 1
- 238000003825 pressing Methods 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
- 239000004065 semiconductor Substances 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000002123 temporal effect Effects 0.000 description 1
- 239000002699 waste material Substances 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/953—Querying, e.g. by the use of web search engines
- G06F16/9535—Search customisation based on user profiles and personalisation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/958—Organisation or management of web site content, e.g. publishing, maintaining pages or automatic linking
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management
Abstract
The present invention relates to the field of data processing, and in particular, to a method for improving accuracy of information pushing, a storage medium, and an electronic device. The method includes clustering the plurality of adjustment reference information to generate a plurality of reference groups. And generating the matching degree of each adjustment reference information according to the interaction information corresponding to each adjustment reference information. And carrying out group matching processing on each piece of information to be adjusted, and generating group weights corresponding to each piece of information to be adjusted. And generating a target matching value corresponding to each piece of information to be adjusted according to the group weight and the initial matching value corresponding to each piece of information to be adjusted. And generating an updated push information sequence according to the target matching value corresponding to each piece of information to be adjusted. The sorting of the information to be regulated in the subsequent information pushing sequence can be regulated in time according to the preference degree of the user, so that the matching degree of the information to be pushed and the preference of the user is improved, and the accuracy of information pushing is improved more in time.
Description
Technical Field
The present invention relates to the field of data processing, and in particular, to a method for improving accuracy of information pushing, a storage medium, and an electronic device.
Background
Information pushing refers to a process that a server matches corresponding push content from a database according to identity attributes, interest attributes and the like of a user, and then presents the push content on a terminal of the user. In the related art, a server classifies users with the same or similar identity attribute and interest attribute into one category, namely, determines user categories for the users, and then pushes content interested by some users in one user category to other users in the user category as push content.
In the existing information pushing method, a deep autonomous learning model is used to calculate and determine matching information corresponding to each user from a large amount of information, and then pushing is performed. However, for some new users, the matching degree and accuracy of the pushed information are low due to the lack of the corresponding user characteristic values.
Disclosure of Invention
Aiming at the technical problems, the invention adopts the following technical scheme:
according to one aspect of the present invention, there is provided a method for improving accuracy of information push, the method comprising the steps of:
and generating a weight determination interval according to the starting time and the first target duration. The weight determination section comprises a plurality of adjustment reference information, wherein the adjustment reference information is initial push information displayed on a user interface.
Clustering is performed on the plurality of adjustment reference information to generate a plurality of reference groups.
Generating the matching degree H of each adjustment reference information according to the interaction information corresponding to each adjustment reference information 1 、H 2 、…、H h 、…、H U . Wherein H is h And (5) determining the matching degree of the h adjustment reference information in the interval for the weight. u is the total number of adjustment reference information in the weight determination section. h=1, 2, …, u. H h The following conditions are satisfied:
H h ＝M 1 *H h 1 +M 2 *H h 2 +M 3 *H h 3 。
wherein H is h 1 The total number of times the reference information is displayed on all user interfaces is adjusted for the h-th in the weight determination interval. H h 2 The total number of times the first tag is triggered at all user interfaces for the h-th adjustment reference information in the interval is determined for the weight. H h 3 The total number of times the text information is added to all user interfaces of the reference information is adjusted for the h-th in the weight determination section. M is M 1 、M 2 M and M 3 Respectively H h 1 、H h 2 H and H h 3 And (5) corresponding weight. M is M 1 <M 2 <M 3 。
And generating an adjustment interval according to the weight determination interval and the second target duration. The adjustment interval comprises a plurality of pieces of information to be adjusted, and the information to be adjusted is initial push information which is not displayed on the user interface.
And carrying out group matching processing on each piece of information to be adjusted, and generating group weights corresponding to each piece of information to be adjusted. The group weight is the average value of the matching degree of the reference group matched with the information to be adjusted. The group matching process is used for matching the information to be adjusted to the corresponding reference group.
And generating a target matching value corresponding to each piece of information to be adjusted by the group weight and the initial matching value corresponding to each piece of information to be adjusted. The target match value satisfies the following condition:
I q ＝PP q *CP q 。
wherein I is q And the target matching value corresponding to the q-th information to be adjusted. PP (Polypropylene) q The group weight of the q-th information to be adjusted. CP (control program) q Corresponding to the q-th information to be adjustedAn initial match value.
And ordering each piece of information to be adjusted according to the sequence from big to small according to the target matching value corresponding to each piece of information to be adjusted, and generating an updated push information sequence.
According to a second aspect of the present invention, there is provided a non-transitory computer readable storage medium storing a computer program which when executed by a processor implements a method of improving accuracy of information push as described above.
According to a third aspect of the present invention, there is provided an electronic device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, the processor implementing a method for improving accuracy of information push as described above when executing the computer program.
The invention has at least the following beneficial effects:
According to the method and the device, the weight determination section is generated according to the starting time and the first target time length, and the weight section is the time section where the user initially browses the push information. Through the interactive behavior characteristics generated by the browsing behavior of the user in a short time, the preference bias of the user can be rapidly captured. And the preference degree of the user for each type of information is represented by the average value of the matching degree of each reference group. Therefore, the sorting of the information to be adjusted in the subsequent information pushing sequence can be timely adjusted according to the preference degree, the matching degree of the information to be pushed and the preference of the user is improved, and the accuracy of information pushing is improved more timely.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for the description of the embodiments will be briefly described below, and it is apparent that the drawings in the following description are only some embodiments of the present invention, and other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
Fig. 1 is a flowchart of an information pushing method according to an embodiment of the present invention.
Fig. 2 is a flowchart of a multi-feature-based information pushing method according to an embodiment of the present invention.
Fig. 3 is a flowchart of a method for improving accuracy of information push according to an embodiment of the present invention.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to fall within the scope of the invention.
As an aspect of the present invention, as shown in fig. 1, there is provided an information push method including the steps of:
s100: and acquiring a feature library of the target user, wherein the feature library comprises a plurality of feature parameters for describing the target user.
The feature library of each target user, i.e., the user's electronic user representation, may be constructed using existing user feature acquisition methods. The included characteristic parameters can be gender characteristics, preference characteristics, location characteristics and the like of the user.
S200: according to the feature library of the target user, acquiring a plurality of pieces of information A to be pushed corresponding to the target user 1 、A 2 、…、A i 、…、A z . Wherein A is i ＝(A i 1 ，A i 2 ，…、A i a 、…、A i f(Ai) )，A i And the i information to be pushed is the i information to be pushed corresponding to the target user. And z is the total amount of information to be pushed corresponding to the target user. i=1, 2, …, z. A is that i a Is A i Corresponding a description parameter. f (Ai) is A i The total number of corresponding description parameters. a=1, 2, …, f (Ai). The plurality of description parameters include an interaction parameter, a location parameter, and a time parameter.
Specifically, the existing information content with higher correlation degree can be selected and matched according to the user characteristics. Of course, a matching relation table between the user characteristics and the information characteristics can be established according to the matching experience in each use scene, and a plurality of pieces of information to be pushed corresponding to each target user can be determined according to the matching relation table.
S300: generating a first correlation value B corresponding to each piece of information to be pushed according to the interaction parameter, the position parameter and the time parameter corresponding to each piece of information to be pushed 1 、B 2 、…、B i 、…、B z . Wherein B is i Is A i A corresponding first correlation value. B (B) i The following conditions are satisfied:
B i ＝W i 1 *C i 1 +W 2 *C i 2 +W 3 *C i 3 。
wherein W is i 1 Is C i 1 And corresponding interaction weight. W (W) i 1 And C i 1 Positive correlation. C (C) i 1 Is A i Corresponding interaction coefficient, C i 1 And A is a i The corresponding interaction parameters are positively correlated.
Specifically, the interaction parameters in the present example include the number of praise and comment of the information to be pushed, and the interaction coefficient is the sum of the number of praise and comment. And W can be set in advance according to actual conditions in specific use scenes i 1 And C i 1 Positive correlation between the two. As concrete, C i 1 ∈[0,8]W is then i 1 ＝8。C i 1 ∈[9,20]W is then i 1 ＝10。C i 1 ∈[21,80]W is then i 1 ＝12。C i 1 ∈[80,140]W is then i 1 ＝20。
W 2 Is a position weight, is a constant, e.g. W 2 ＝10。C i 2 Is A i Corresponding to the position coefficient. C (C) i 2 And A is a i Corresponding position parameter and position information contained in characteristic parameterThe degree of matching between the messages is positively correlated.
Specifically, A i The corresponding location parameter typically comprises at least one location information, such as may be contained in a location represented by the content in the push information and its corresponding tag. Such as a push message including both a text representing a trip and an image representing a scenic spot. The location parameters corresponding to the information include the departure location, destination location and sight spot location. In addition, the characteristic parameters corresponding to the user may also include the departure place and destination of a trip and at least one scenic spot position liked by the user.
Thereby, the position coefficient can be determined according to the degree of matching between the position parameter and the position information contained in the feature parameter. If only 1 position information is the same, C i 2 =10. With 2 identical positional information, C i 2 =15. With 3 identical positional information, C i 2 =20. With 4 identical positional information, C i 2 ＝40。W 2 The setting can be performed according to the actual situation in the specific use scene, for example, the setting can be 2, 3, 5 or the like.
W 3 Time weight, constant. For example W 3 ＝20。C i 3 Is A i Corresponding to the time coefficient. C (C) i 3 The following conditions are satisfied:
C i 3 ＝e^[-max(0，T i -1) 2 /k 1 ]. Where max () is a function taking the maximum value. T (T) i Is A i Corresponding duration of presence, k 1 For the adjustment factor, it is constant.
Preferably, k 1 ＝2*-(scale 2 2 x ln (decay)). Where scale is the decay scale factor, scale=7. decay is the attenuation coefficient, decay=0.01.
In the present embodiment, T i In days, if the duration of the presence is 12 hours, T i =0.5 according to the above C i 3 The calculation formula of (C) is a piecewise decay function, and according to the formula, when the existence duration is within one day, C i 3 ＝e 0 =1. The information released in the day can be regarded as fresher content, and C is maintained i 3 Is a constant value that does not decay. At the same time T i The size of 1 in-1 can be self-adjusting. Therefore, the algorithm formula can ensure that the content released in a certain period of time has higher identical pushing possibility, and further can avoid the condition that new content is discarded, so as to ensure the timeliness of the pushed information.
In addition, when T i >1, C i 3 Inversely proportional to the duration of the presence, i.e. the longer the duration of the presence, C i 3 The closer to 0, the less easily pushed, so that older content can be discarded in time. At the same time, when T i >1, C i 3 The method conforms to Gaussian distribution, so that the attenuation speed can be slow in the subsequent attenuation process, fast and slow. The characteristic is in accordance with the relation between the freshness of the pushed information and the pushing possibility, so that more accurate time dimension can be ensured to push the information.
S400: and sequencing each piece of information to be pushed according to the first correlation value corresponding to each piece of information to be pushed from big to small, and generating a first push information sequence.
The first correlation values calculated from the plurality of matching dimensions through the above S300 can more accurately reflect the matching degree between each piece of information to be pushed and the target user. Thus, the first push information sequence can be obtained after sorting from the big to the small.
S500: and performing coarse ordering processing on the first n pieces of information to be pushed in the first push information sequence by using a double-tower dssm model, and generating a second push information sequence.
Preferably, S500 includes:
s501: and acquiring a plurality of first characteristic parameters from the characteristic library of the target user, and generating a first target user characteristic set.
S502: and generating first similarity between the first n pieces of information to be pushed in the first push information sequence and the first target user feature set respectively through a double-tower dssm model.
S503: and according to the first similarity corresponding to each piece of information to be pushed, sequencing the first n pieces of information to be pushed in the first push information sequence according to the sequence from big to small, and generating a second push information sequence.
S600: and performing fine ordering processing on the first m pieces of information to be pushed in the second push information sequence by using the deepfm model to generate a target push information sequence. Wherein z is greater than or equal to n is greater than or equal to m. The correlation degree of the target push information sequence and the target user is larger than that of the second push information sequence and the target user.
Preferably, S600 includes:
and acquiring a plurality of second characteristic parameters from the characteristic library of the target user, and generating a second target user characteristic set. The second set of target user characteristics includes interaction characteristics between the target user and the information to be pushed.
And generating second similarity between the first m pieces of information to be pushed in the second push information sequence and the second target user feature set respectively through the deepfm model.
And according to the second similarity corresponding to each piece of information to be pushed, sequencing the first m pieces of information to be pushed in the second push information sequence according to the sequence from large to small, and generating a target push information sequence.
In the above step, the double-tower dssm model mainly maps the intrinsic characteristic information of the user and the intrinsic information of the information to be pushed into the same data space, and then obtains the first similarity.
In the deep fm model, due to the existence of the fm module, not only the inherent characteristic information of the user and the inherent information of the information to be pushed can be processed, but also the interaction characteristic between the user and the information to be pushed can be processed, so that the second similarity can be calculated more accurately. The coarse sorting and the fine sorting are sequentially carried out by respectively using the double-tower dssm model and the deep fm model, so that the information positioned at the front position is finer in processing and higher in correlation degree.
In the invention, firstly, corresponding multiple weights and coefficients are generated according to the matching relation between the characteristic parameters of the target user and the description parameters of the information to be pushed, and the corresponding multiple weights and coefficients can pass through B i ＝W i 1 *C i 1 +W 2 *C i 2 +W 3 *C i 3 And determining a first correlation value corresponding to each piece of information to be pushed. And ordering the information to be pushed based on the first correlation value. Since the first correlation value is calculated in a simpler manner, it can quickly determine a rough pushing sequence with less consumption of calculation resources.
And then, on the basis of the pushing sequence, performing coarse ordering processing and fine ordering processing in sequence by using an autonomous learning model again, and finally obtaining a target pushing information sequence. Because the coarse ordering process and the fine ordering process are both processed by selecting the data with the earlier ordering based on the previous ordering, the data processed in the model is also the data with higher correlation, thereby reducing the processing of the model on the data with lower correlation, and reducing the consumption of computing resources. In addition, as the computational complexity increases, the amount of data processed decreases, and eventually the finest processed data can be placed in the forward position of the push list. Therefore, even if the browsing time of the user is short, the content matching degree browsed by the user can be guaranteed to be higher, and the number of processed by using a complex algorithm in push information which is not browsed by the user can be reduced. And further, the waste of computing resources can be reduced while the matching precision is ensured.
As another possible embodiment of the present invention, the description parameters corresponding to the information to be pushed include real-time parameters.
In particular, the real-time parameters may be represented by the type of tag. The mapping relationship between the tag corresponding to the degree of urgency and the corresponding information content may be set in advance. Such as may be implemented by an autonomous learning model. Thus, after the push information is expected to be generated, the corresponding emergency label can be automatically matched according to the mapping relation.
After generating the first push information sequence, the method further comprises:
s410: and obtaining information to be pushed with the existence duration within a first time duration threshold t1 in the first push information sequence, and generating a third push information sequence.
t1 may be 1 day.
S420: and generating a first weighting coefficient corresponding to each piece of information to be pushed in the third push information sequence according to the real-time parameter corresponding to each piece of information to be pushed in the third push information sequence. D (D) b 1 The following conditions are satisfied:
D b 1 ＝1+k b *0.32. Wherein D is b 1 And the first weight coefficient corresponding to the b-th information to be pushed in the third pushing information sequence. k (k) b The real-time parameter k corresponding to the b-th information to be pushed in the third pushing information sequence b =0 or k b ＝1。
Correspondingly, as the information to be pushed all has the corresponding emergency label, the corresponding real-time parameters can be generated according to the emergency label. Such as: if the value corresponding to the emergency label of the information to be pushed is smaller than the emergency threshold, the corresponding real-time parameter is 0. If the value corresponding to the emergency label corresponding to the information to be pushed is larger than the emergency threshold, the corresponding real-time parameter is 1.
S430: and generating a second weight raising coefficient corresponding to each piece of information to be pushed in the third push information sequence according to the existence duration corresponding to each piece of information to be pushed in the third push information sequence and the times of user interface display. The second weighting coefficient satisfies the following condition:
D b 2 ＝max(α/(1+β*t b /t1+γ*x b /X) 0.5 ,1). Wherein α, β, γ are three super parameters, α=1.93, β=1.3, γ=1.4, respectively. D (D) b 2 And the second weight-raising coefficient corresponding to the b-th information to be pushed in the third pushing information sequence. t is t b And the existence duration corresponding to the b-th information to be pushed in the third pushing information sequence. X is x b For the third push information sequenceAnd the number of times of user interface display corresponding to the b-th information to be pushed in the column. X is the user interface display threshold. X=100.
Alpha, beta and gamma are super parameters larger than 0, and need to be adjusted by means of an online small flow experiment.
The main information to be pushed in the third push information sequence of this embodiment is some information with lower exposure amount just released. Specifically, the corresponding presence is less than 1 day long, and the number of user interface displays is less than the user interface display threshold. Typically, for just published information, to ensure that it is visible, it is provided a certain number of user interface displays over a period of time.
t b T1 and x b X is used to indicate the degree of freshness and the degree of exposure. The formula in the step shows that the supporting force of the newly released new information is maximum, and the supporting force decays along with time, namely the weight increasing coefficient is inversely related to the new and old degrees. Meanwhile, the more exposure is obtained by the information, the weaker the supporting force is, namely the weight increasing coefficient is inversely related to the exposure degree. Further, as can be seen from the above expression, even if the number of times the new information user interface is displayed is increased, the weighting coefficient is at least equal to 1, i.e., the new information user interface corresponds to natural distribution. Therefore, the new information pushing degree can be guaranteed not to be reduced by pressing.
S440: and generating a second correlation value corresponding to each piece of information to be pushed in the third push information sequence according to the first correlation value, the first weight-increasing coefficient and the second weight-increasing coefficient corresponding to each piece of information to be pushed in the third push information sequence. The second correlation value satisfies the following condition: d (D) b ＝max(1.2，D b 1 ，D b 2 )*D b xg . Wherein D is b And the second correlation value corresponding to the b-th information to be pushed in the third pushing information sequence. D (D) b xg And the first correlation value corresponding to the b-th information to be pushed in the third pushing information sequence.
S450: and ordering each piece of information to be pushed according to the sequence from big to small according to the second correlation value corresponding to each piece of information to be pushed in the third push information sequence, and generating a fourth push information sequence.
When fine-ranking is performed through the deep model, the model can be pushed according to interaction behaviors of a large number of users and information to be pushed. Thus, the longer the information release time is, the more interaction with the user is easy to generate, and the more easily the information is pushed. Correspondingly, since the release time of the new information is shorter, the interaction behavior generated by the user is fewer, and the new information is not easy to push. Thus, after the deep model processing in the above embodiment, some content with higher quality existing in the newly released information is easily missed.
In this embodiment, by setting the weighting coefficient, the information pushing degree of the new release, that is, the second correlation value, can be improved. In this case, the newly published information will have more opportunities to be viewed and browsed by the user. And the user directly judges the quality of the information content. Thus, high-quality content in the distributed information can be ensured, and the distributed information can be pushed and found more easily. And further, the situation that the hot spot information is missed can be avoided as much as possible.
As another possible embodiment of the present invention, after generating the first push information sequence, the method further comprises:
s460: obtaining duration t of coarse ordering process cp 。
S470: when t cl >And t2, taking the first push information sequence as a second push information sequence. Wherein t2 is a second duration threshold. t2=300 ms.
Preferably, after generating the second push information sequence, the method further comprises:
s480: acquiring duration t of the fine ordering process jp 。
S490: when t jp >And t3, taking the second push information sequence as a target push information sequence. Wherein t3 is a third duration threshold.
More preferably, t3=t2=300 ms.
Because the coarse arrangement and the fine arrangement performed by the autonomous learning model are required to be calculated more complicated, the processing time-out may be caused, and the situation that push information cannot be provided for the user in time may not be caused. Thus, in this embodiment, the coarse-ranking process and the fine-ranking process are limited by setting the time period threshold. When the processing is overtime, the first push information sequence can be used as a spam push sequence, so that the push information can be provided for the user more timely.
According to another aspect of the present invention, as shown in fig. 2, there is also provided a multi-feature based information pushing method, including the steps of:
s700: and acquiring an initial push information sequence corresponding to the target user. The initial push information sequence comprises a plurality of pieces of information to be pushed, and each piece of information to be pushed corresponds to one adjustment information set E 1 、E 2 、…、E c 、…、E v . Wherein E is c ＝(E c 1 ，E c 2 )，E c And the adjustment information set corresponding to the c-th information to be pushed in the initial push information sequence. v is the total amount of information to be pushed contained in the initial push information sequence. c=1, 2, …, v. E (E) c 1 And the sequence number is the sequence number corresponding to the c-th information to be pushed in the initial push information sequence. E (E) c 2 For E c A corresponding set of attributes. The attribute set comprises a tag attribute value, a topic attribute value and a position attribute value corresponding to the information to be pushed.
The initial push information sequence in this step may be a target push information sequence. The attribute set of the information to be pushed comprises a tag attribute value, a topic attribute value and a position attribute value corresponding to the information to be pushed. These attribute values may be obtained by prior art techniques, such as by machine learning models, to match corresponding tag attribute values, topic attribute values, and location attribute values according to information content. The location attribute values may also be obtained from GPS. The image information is exemplified, the corresponding label can be food, scenery, automobile, etc., the topic can be travel, tourism, etc., and the position can be the actual position and the departure place and destination position in the related formation. Then, a unique corresponding value is given to the mapping relation according to the mapping relation which is configured in advance.
S800: and generating adjustment parameters corresponding to each piece of information to be pushed according to the adjustment information set corresponding to each piece of information to be pushed. The adjustment parameters satisfy the following conditions:
wherein F is c And the adjustment parameter corresponding to the c-th information to be pushed in the initial push information sequence. J (J) d 1 For E c 2 The total number of occurrences before the d-th tag attribute value. f (de) is E c 2 The total number of categories of tag attribute values. J (J) f 2 For E c 2 The total number of occurrences before the f-th topic attribute value. f (fe) is E c 2 The total number of categories of topic attribute values. J (J) g 3 For E c 2 The total number of occurrences before the g-th position attribute value. f (ge) is E c 2 The total number of categories of the middle position attribute value. w (w) 1 、w 2 、w 3 W 4 The weights are respectively corresponding to the tag attribute value, the topic attribute value, the position attribute value and the arrangement sequence number.
Preferably, w 4 >max(w 1 ，w 2 ，w 3 )。
In the above formula, the weight corresponding to the arrangement sequence number is set to a larger value, so that the ordering mode in the initial push information sequence can be basically ensured not to be changed greatly, and only the content with higher similarity can be pulled apart. Therefore, the calculation result when the initial push information sequence is generated before can be reserved to the greatest extent, and the calculation amount can be saved to a certain extent.
Specifically, the following examples are given as illustrative examples:
w 1 ＝1.2。w 2 ＝1.2。w 3 ＝1.3。w 4 =1.7. The attribute sets of the first 4 pieces of information to be pushed in the initial push information sequence and the corresponding adjustment parameters are specifically shown in the following table 1:
table 1: detail table of first 4 pieces of information to be pushed in initial push information sequence
As can be seen from the calculation formula of the adjustment parameters and table 1, the more the same tag, topic or position appears, the larger the difference between the generated adjustment parameters, i.e. the larger the difference between the adjustment parameters between the information to be pushed with higher similarity, and the smaller the difference between the adjustment parameters between the information to be pushed with lower similarity. The adjusted ranks are 1, 4, 3, 2. Thereby reducing the probability of homogeneous content occurring together.
S900: and ordering each piece of information to be pushed according to the corresponding adjustment parameters of each piece of information to be pushed from small to large, and generating an optimized push information sequence.
In the invention, each piece of information to be pushed corresponds to an adjustment information set, and the adjustment information set comprises a plurality of description attribute values of the information to be pushed, such as which type of labels corresponds to the information to be pushed, which type of corresponding topics correspond to, and what type of attribute values correspond to the release position or the place represented in the information. The plurality of description attribute values can be used for approximately representing the corresponding information content to be pushed. When the adjustment parameters corresponding to each piece of information to be pushed are calculated, the difference of the adjustment parameters between pieces of information to be pushed with higher similarity is larger, and the difference of the adjustment parameters between pieces of information to be pushed with lower similarity is smaller. Therefore, when the information to be pushed with the similarity content is arranged at a position far away from the information to be pushed when the information to be pushed is arranged according to the adjustment parameters in the follow-up sequence, and therefore the homogeneity of the information can be reduced.
As a possible embodiment of the present invention, after generating the adjustment parameter corresponding to each piece of information to be pushed, the method further includes:
s810: when F c In (a) and (b)
When according to F c The previously generated adjustment parameters generate compensation parameters E c b 。
S820: according to w 4 、E c 1 E and E c b Generating an optimized parameter G corresponding to the c-th information to be pushed in the initial push information sequence c 。G c The following conditions are satisfied:
s830: will G c As the adjustment parameter corresponding to the c-th information to be pushed in the initial push information sequence.
Preferably, S810 includes: s811: will F c Average of values of corresponding parts of a plurality of adjustment parameters which have been generated before as compensation parameter E c b 。
The situation in table 1 is illustrated, where the 4 th information to be pushed is both the situation in this embodiment. Its corresponding compensation parameter E 4 b = (0+3.6+2.4+2.6+1.2×2+1.3×1)/3=4.1. Correspondingly, G 4 =1.7x4+4.1=10.9. The adjusted ranks are 1, 3, 4, 2.
Preferably, S810 includes: s812: will be combined with F c Adjacent values of the corresponding parts of the generated adjustment parameters as compensation parameters E c b 。
The situation in table 1 is illustrated, where the 4 th information to be pushed is both the situation in this embodiment. Its corresponding compensation parameter E 4 b = (0+1.2×2+1.3×1) =3.7. Correspondingly, G 4 =1.7x4+3.7=10.5. The adjusted ranks are 1, 3, 4, 2.
The compensation mode set in this embodiment is used when the information to be pushed
When F c Only=w 4 *E C 1 . Thereby F c The value of (c) becomes smaller if there is also some information to be pushed that differs significantly in its preceding position. But it corresponds to->
It is possible to cause the F corresponding to the previous information c Larger, and possibly larger than W 4 *E C 1 . Thus, F will be determined in the subsequent sorting c Only=w 4 *E C 1 The position of the push information of (c) is adjusted forward by a larger magnitude. Correspondingly, the situation can break the calculation result when the initial push information sequence is generated through complex calculation by the autonomous learning model before, and the correlation degree between the finally generated push information sequence and the user can be reduced. Therefore, on the premise of reducing the homogeneity of the information, the method can ensure that the finally generated push information sequence has higher correlation with the user as much as possible, and can also play a role in saving the calculated amount to a certain extent.
As a possible embodiment of the present invention, the optimized push information sequence comprises a plurality of push information sub-sequences, each push information sub-sequence comprising the same amount of information to be pushed.
After generating the optimized push information sequence, the method further comprises:
s910: if the B corresponding to the information to be pushed in the current push information subsequence max >Q1 or H max >Q2 or W max >And Q3, taking the information to be pushed in the current push information sub-sequence as information to be replaced. Wherein Q1 is a first homogeneity threshold. Q2 is a second homogeneity threshold. Q3 is a third homogeneity threshold. B (B) max And the maximum value of the repeated occurrence times of the tag attribute value of the information to be pushed is obtained. H max The maximum number of times the topic attribute value is repeated. W (W) max The maximum number of times of repeated appearance is the position attribute value.
S920: and carrying out position exchange on the information to be replaced and the target replacement information. The target replacement information is information to be pushed in a next push information sub-sequence adjacent to the current push information sub-sequence.
Since the amount of information placed on a display page is substantially fixed in actual use, the optimized push information sequence may be divided into a plurality of push information sub-sequences for display. And meanwhile, screening the homogeneous content by setting a threshold value, and exchanging the content with homogeneity with the content in the next display page. The scheme is more in line with the display characteristics of the terminal and the browsing habit of the user.
As another aspect of the present invention, as shown in fig. 3, a method for improving accuracy of information push is provided, where the method includes the following steps:
s10: and generating a weight determination interval according to the starting time and the first target duration. The weight determination section comprises a plurality of adjustment reference information, wherein the adjustment reference information is initial push information displayed on a user interface.
The start time may be the time when the user first receives push information.
S20: clustering is performed on the plurality of adjustment reference information to generate a plurality of reference groups.
Specifically, the clustering process may be performed using a K-Mean algorithm.
S30: generating the matching degree H of each adjustment reference information according to the interaction information corresponding to each adjustment reference information 1 、H 2 、…、H h 、…、H U . Wherein H is h And (5) determining the matching degree of the h adjustment reference information in the interval for the weight. u is the total number of adjustment reference information in the weight determination section. h=1, 2, …, u. H h The following conditions are satisfied:
H h ＝M 1 *H h 1 +M 2 *H h 2 +M 3 *H h 3 。
wherein H is h 1 The total number of times the reference information is displayed on all user interfaces is adjusted for the h-th in the weight determination interval. H h 2 Is the weightThe total number of times the h adjustment reference information in the interval is triggered to the first label in all user interfaces is determined. H h 3 The total number of times the text information is added to all user interfaces of the reference information is adjusted for the h-th in the weight determination section. M is M 1 、M 2 M and M 3 Respectively H h 1 、H h 2 H and H h 3 And (5) corresponding weight. M is M 1 <M 2 <M 3 。
Specifically, H h 1 、H h 2 H and H h 3 The browsing amount, the praise amount and the comment amount of the reference information can be respectively adjusted for the h, and the browsing amount, the praise amount and the comment amount can be acquired by setting a buried point.
S40: and generating an adjustment interval according to the weight determination interval and the second target duration. The adjustment interval comprises a plurality of pieces of information to be adjusted, and the information to be adjusted is initial push information which is not displayed on the user interface.
S50: and carrying out group matching processing on each piece of information to be adjusted, and generating group weights corresponding to each piece of information to be adjusted. The group weight is the average value of the matching degree of the reference group matched with the information to be adjusted. The group matching process is used for matching the information to be adjusted to the corresponding reference group.
S60: and generating a target matching value corresponding to each piece of information to be adjusted by the group weight and the initial matching value corresponding to each piece of information to be adjusted. The target match value satisfies the following condition:
I q ＝PP q *CP q 。
wherein I is q And the target matching value corresponding to the q-th information to be adjusted. PP (Polypropylene) q The group weight of the q-th information to be adjusted. CP (control program) q And the initial matching value corresponding to the q-th information to be adjusted.
S70: and ordering each piece of information to be adjusted according to the sequence from big to small according to the target matching value corresponding to each piece of information to be adjusted, and generating an updated push information sequence.
In this embodiment, a weight determination interval is generated according to the starting time and the first target duration, where the weight interval is a time interval in which the user initially browses the push information. Through the interactive behavior characteristics generated by the browsing behavior of the user in a short time, the preference bias of the user can be rapidly captured. And the preference degree of the user for each type of information is represented by the average value of the matching degree of each reference group. Therefore, the sorting of the information to be adjusted in the subsequent information pushing sequence can be timely adjusted according to the preference degree, the matching degree of the information to be pushed and the preference of the user is improved, and the accuracy of information pushing is improved more timely.
Embodiments of the present invention also provide a non-transitory computer readable storage medium that may be disposed in an electronic device to store at least one instruction or at least one program for implementing one of the methods embodiments, the at least one instruction or the at least one program being loaded and executed by the processor to implement the methods provided by the embodiments described above.
Embodiments of the present invention also provide an electronic device comprising a processor and the aforementioned non-transitory computer-readable storage medium.
Embodiments of the present invention also provide a computer program product comprising program code for causing an electronic device to carry out the steps of the method according to the various exemplary embodiments of the invention described in the present specification when the program product is run on the electronic device.
Furthermore, although the steps of the methods in the present disclosure are depicted in a particular order in the drawings, this does not require or imply that the steps must be performed in that particular order or that all illustrated steps be performed in order to achieve desirable results. Additionally or alternatively, certain steps may be omitted, multiple steps combined into one step to perform, and/or one step decomposed into multiple steps to perform, etc.
From the above description of embodiments, those skilled in the art will readily appreciate that the example embodiments described herein may be implemented in software, or may be implemented in software in combination with the necessary hardware. Thus, the technical solution according to the embodiments of the present disclosure may be embodied in the form of a software product, which may be stored in a non-volatile storage medium (may be a CD-ROM, a U-disk, a mobile hard disk, etc.) or on a network, including several instructions to cause a computing device (may be a personal computer, a server, a mobile terminal, or a network device, etc.) to perform the method according to the embodiments of the present disclosure.
In an exemplary embodiment of the present disclosure, an electronic device capable of implementing the above method is also provided.
Those skilled in the art will appreciate that the various aspects of the invention may be implemented as a system, method, or program product. Accordingly, aspects of the invention may be embodied in the following forms, namely: an entirely hardware embodiment, an entirely software embodiment (including firmware, micro-code, etc.) or an embodiment combining hardware and software aspects may be referred to herein as a "circuit," module "or" system.
An electronic device according to this embodiment of the invention. The electronic device is merely an example, and should not impose any limitations on the functionality and scope of use of embodiments of the present invention.
The electronic device is in the form of a general purpose computing device. Components of an electronic device may include, but are not limited to: the at least one processor, the at least one memory, and a bus connecting the various system components, including the memory and the processor.
Wherein the memory stores program code that is executable by the processor to cause the processor to perform steps according to various exemplary embodiments of the present invention described in the above section of the exemplary method of this specification.
The storage may include readable media in the form of volatile storage, such as random access memory (RCM) and/or cache memory, and may further include Read Only Memory (ROM).
The storage may also include a program/utility having a set (at least one) of program modules including, but not limited to: an operating system, one or more application programs, other program modules, and program data, each or some combination of which may include an implementation of a network environment.
The bus may be one or more of several types of bus structures including a memory bus or memory controller, a peripheral bus, an accelerated graphics port, a processor, or a local bus using any of a variety of bus architectures.
The electronic device may also communicate with one or more external devices (e.g., keyboard, pointing device, bluetooth device, etc.), with one or more devices that enable a user to interact with the electronic device, and/or with any device (e.g., router, modem, etc.) that enables the electronic device to communicate with one or more other computing devices. Such communication may be through an input/output (I/O) interface. And, the electronic device may also communicate with one or more networks such as a local area network (LCN), a wide area network (WCN), and/or a public network such as the internet via a network adapter. The network adapter communicates with other modules of the electronic device via a bus. It should be appreciated that although not shown, other hardware and/or software modules may be used in connection with an electronic device, including but not limited to: microcode, device drivers, redundant processors, external disk drive arrays, RCID systems, tape drives, data backup storage systems, and the like.
From the above description of embodiments, those skilled in the art will readily appreciate that the example embodiments described herein may be implemented in software, or may be implemented in software in combination with the necessary hardware. Thus, the technical solution according to the embodiments of the present disclosure may be embodied in the form of a software product, which may be stored in a non-volatile storage medium (may be a CD-ROM, a U-disk, a mobile hard disk, etc.) or on a network, including several instructions to cause a computing device (may be a personal computer, a server, a terminal device, or a network device, etc.) to perform the method according to the embodiments of the present disclosure.
In an exemplary embodiment of the present disclosure, a computer-readable storage medium having stored thereon a program product capable of implementing the method described above in the present specification is also provided. In some possible embodiments, the aspects of the invention may also be implemented in the form of a program product comprising program code for causing a terminal device to carry out the steps according to the various exemplary embodiments of the invention as described in the "exemplary method" section of this specification, when the program product is run on the terminal device.
The program product may employ any combination of one or more readable media. The readable medium may be a readable signal medium or a readable storage medium. The readable storage medium can be, for example, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or a combination of any of the foregoing. More specific examples (a non-exhaustive list) of the readable storage medium would include the following: an electrical connection having one or more wires, a portable disk, a hard disk, a random access memory (RCM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.
The computer readable signal medium may include a data signal propagated in baseband or as part of a carrier wave with readable program code embodied therein. Such a propagated data signal may take any of a variety of forms, including, but not limited to, electro-magnetic, optical, or any suitable combination of the foregoing. A readable signal medium may also be any readable medium that is not a readable storage medium and that can communicate, propagate, or transport a program for use by or in connection with an instruction execution system, apparatus, or device.
Program code embodied on a readable medium may be transmitted using any appropriate medium, including but not limited to wireless, wireline, optical fiber cable, RF, etc., or any suitable combination of the foregoing.
Program code for carrying out operations of the present invention may be written in any combination of one or more programming languages, including an object oriented programming language such as Java, C++ or the like and conventional procedural programming languages, such as the "C" programming language or similar programming languages. The program code may execute entirely on the user's computing device, partly on the user's device, as a stand-alone software package, partly on the user's computing device, partly on a remote computing device, or entirely on the remote computing device or server. In the case of remote computing devices, the remote computing device may be connected to the user computing device through any kind of network, including a local area network (LCN) or a wide area network (WCN), or may be connected to an external computing device (e.g., connected over the internet using an internet service provider).
Furthermore, the above-described drawings are only schematic illustrations of processes included in the method according to the exemplary embodiment of the present invention, and are not intended to be limiting. It will be readily appreciated that the processes shown in the above figures do not indicate or limit the temporal order of these processes. In addition, it is also readily understood that these processes may be performed synchronously or asynchronously, for example, among a plurality of modules.
It should be noted that although in the above detailed description several modules or units of a device for action execution are mentioned, such a division is not mandatory. Indeed, the features and functionality of two or more modules or units described above may be embodied in one module or unit in accordance with embodiments of the present disclosure. Conversely, the features and functions of one module or unit described above may be further divided into a plurality of modules or units to be embodied.
The present invention is not limited to the above embodiments, and any changes or substitutions that can be easily understood by those skilled in the art within the technical scope of the present invention are intended to be included in the scope of the present invention. Therefore, the protection scope of the invention is subject to the protection scope of the claims.
Claims (10)
1. The method for improving the accuracy of information push is characterized by comprising the following steps:
generating a weight determination interval according to the starting time and the first target duration; the weight determination section comprises a plurality of adjustment reference information, wherein the adjustment reference information is initial push information displayed on a user interface;
Clustering the plurality of adjustment reference information to generate a plurality of reference groups;
generating the matching degree H of each adjustment reference information according to the interaction information corresponding to each adjustment reference information 1 、H 2 、…、H h 、…、H U The method comprises the steps of carrying out a first treatment on the surface of the Wherein H is h Determining the matching degree of the h adjustment reference information in the interval for the weight; u is the total number of adjustment reference information in the weight determination section; h=1, 2, …, u; h h The following conditions are satisfied:
H h ＝M 1 *H h 1 +M 2 *H h 2 +M 3 *H h 3 ；
wherein H is h 1 Determining the total number of times of display of the reference information on all user interfaces for the h in the interval for the weight; h h 2 Determining the total number of times the h adjustment reference information in the interval is triggered to the first label in all user interfaces for the weight; h h 3 Determining the total number of times of adding text information to all user interfaces of the reference information for the h adjustment in the interval for the weight; m is M 1 、M 2 M and M 3 Respectively H h 1 、H h 2 H and H h 3 Corresponding weights; m is M 1 <M 2 <M 3 ；
Determining an interval and a second target duration according to the weight, and generating an adjustment interval; the adjustment interval comprises a plurality of pieces of information to be adjusted, wherein the information to be adjusted is initial push information which is not displayed on a user interface;
performing group matching processing on each piece of information to be adjusted to generate group weight corresponding to each piece of information to be adjusted; the group weight is the average value of the matching degree of the reference group matched with the information to be adjusted; the group matching process is used for matching the information to be adjusted to the corresponding reference group;
Generating a target matching value corresponding to each piece of information to be adjusted by the group weight and the initial matching value corresponding to each piece of information to be adjusted; the target matching value satisfies the following condition:
I q ＝PP q *CP q ；
wherein I is q The target matching value corresponding to the q-th information to be adjusted; PP (Polypropylene) q The group weight of the q-th information to be adjusted; CP (control program) q The initial matching value corresponding to the q-th information to be adjusted;
and ordering each piece of information to be adjusted according to the sequence from big to small according to the target matching value corresponding to each piece of information to be adjusted, and generating an updated push information sequence.
2. The method according to claim 1, wherein the method further comprises:
acquiring a feature library of a target user, wherein the feature library comprises a plurality of feature parameters for describing the target user;
according to the feature library of the target user, acquiring a plurality of pieces of information A to be pushed corresponding to the target user 1 、A 2 、…、A i 、…、A z The method comprises the steps of carrying out a first treatment on the surface of the Wherein A is i ＝(A i 1 ，A i 2 ，…、A i a 、…、A i f(Ai) )，A i The ith information to be pushed corresponding to the target user is obtained; z is the total amount of information to be pushed corresponding to the target user; i=1, 2, …, z; a is that i a Is A i A corresponding a description parameter; f (Ai) is A i The total number of corresponding description parameters; a=1, 2, …, f (Ai); the plurality of description parameters comprise interaction parameters, position parameters and time parameters;
Generating a first correlation value B corresponding to each piece of information to be pushed according to the interaction parameter, the position parameter and the time parameter corresponding to each piece of information to be pushed 1 、B 2 、…、B i 、…、B z The method comprises the steps of carrying out a first treatment on the surface of the Wherein B is i Is A i A corresponding first correlation value; b (B) i The following conditions are satisfied:
B i ＝W i 1 *C i 1 +W 2 *C i 2 +W 3 *C i 3 ；
wherein W is i 1 Is C i 1 Corresponding interaction weights; w (W) i 1 And C i 1 Positive correlation; c (C) i 1 Is A i Corresponding interaction coefficient, C i 1 And A is a i The corresponding interaction parameters are positively correlated;
W 2 for position weight, C i 2 Is A i Corresponding position coefficients; c (C) i 2 And A is a i The matching degree between the corresponding position parameter and the position information contained in the characteristic parameter is positively correlated;
W 3 is a time weight; c (C) i 3 Is A i Corresponding to the time coefficient; c (C) i 3 The following conditions are satisfied:
C i 3 ＝e^[-max(0，T i -1) 2 /k 1 ]the method comprises the steps of carrying out a first treatment on the surface of the Wherein, max () is a maximum function; t (T) i Is A i Corresponding duration of presence, k 1 For adjusting the coefficient;
sequencing each piece of information to be pushed according to the sequence from big to small according to the first correlation value corresponding to each piece of information to be pushed, and generating a first push information sequence;
performing coarse ordering processing on the first n pieces of information to be pushed in the first push information sequence by using a double-tower dssm model to generate a second push information sequence;
using a deep model to carry out precise sorting treatment on the first m pieces of information to be pushed in the second push information sequence, and generating a target push information sequence; wherein z is greater than or equal to n is greater than or equal to m; the number of input parameters of the input model in the coarse ordering process is smaller than the number of input parameters of the input model in the fine ordering process.
3. The method of claim 2, wherein the description parameters corresponding to the information to be pushed include real-time parameters;
after generating the first push information sequence, the method further comprises:
obtaining information to be pushed with the existence time length within a first time length threshold t1 in the first push information sequence, and generating a third push information sequence;
generating a first weight raising coefficient corresponding to each piece of information to be pushed in the third push information sequence according to the real-time parameter corresponding to each piece of information to be pushed in the third push information sequence; d (D) b 1 The following conditions are satisfied:
D b 1 ＝1+k b *0.32; wherein D is b 1 The first weight-raising coefficient corresponding to the b-th information to be pushed in the third pushing information sequence; k (k) b The real-time parameter k corresponding to the b-th information to be pushed in the third pushing information sequence b =0 or k b ＝1；
Generating a second weight raising coefficient corresponding to each piece of information to be pushed in the third push information sequence according to the existence duration corresponding to each piece of information to be pushed in the third push information sequence and the times of user interface display; the second weighting coefficient satisfies the following condition:
D b 2 ＝max(α/(1+β*t b /t1+γ*x b /X) 0.5 1) a step of; wherein α, β, γ are three super parameters, α=1.93, β=1.3, γ=1.4, respectively; d (D) b 2 The second weight-raising coefficient corresponding to the b-th information to be pushed in the third pushing information sequence; t is t b The existence duration corresponding to the b-th information to be pushed in the third pushing information sequence; x is x b The number of times of user interface display corresponding to the b-th information to be pushed in the third pushing information sequence; x is a user interface display threshold;
generating a second correlation value corresponding to each piece of information to be pushed in the third push information sequence according to the first correlation value, the first weight-increasing coefficient and the second weight-increasing coefficient corresponding to each piece of information to be pushed in the third push information sequence; second phaseThe off value satisfies the following condition: d (D) b ＝max(1.2，D b 1 ，D b 2 )*D b xg The method comprises the steps of carrying out a first treatment on the surface of the Wherein D is b The second correlation value corresponding to the b-th information to be pushed in the third pushing information sequence; d (D) b xg The first correlation value corresponding to the b-th information to be pushed in the third pushing information sequence;
and according to the second correlation value corresponding to each piece of information to be pushed in the third push information sequence, sequencing each piece of information to be pushed according to the sequence from big to small, and generating a fourth push information sequence.
4. The method of claim 2, wherein k 1 ＝2*-(scale 2 2 x ln (decay)); wherein scale is the decay scale factor, scale=7; decay is the attenuation coefficient, decay=0.01.
5. The method of claim 2, wherein the target push information sequence is the initial push information sequence;
after generating the target push information sequence, the method further comprises the steps of:
acquiring an initial push information sequence corresponding to a target user; the initial push information sequence comprises a plurality of information to be pushed, and each information to be pushed corresponds to one adjustment information set E 1 、E 2 、…、E c 、…、E v The method comprises the steps of carrying out a first treatment on the surface of the Wherein E is c ＝(E c 1 ，E c 2 )，E c An adjustment information set corresponding to the c-th information to be pushed in the initial push information sequence; v is the total amount of information to be pushed contained in the initial push information sequence; c=1, 2, …, v; e (E) c 1 The sequence number corresponding to the c-th information to be pushed in the initial push information sequence; e (E) c 2 For E c A corresponding set of attributes; the attribute set comprises a tag attribute value, a topic attribute value and a position attribute value corresponding to the information to be pushed;
generating adjustment parameters corresponding to each piece of information to be pushed according to the adjustment information set corresponding to each piece of information to be pushed; the adjustment parameters satisfy the following conditions:
wherein F is c The adjustment parameters corresponding to the c-th information to be pushed in the initial push information sequence; j (J) d 1 For E c 2 The total number of occurrences before the d-th tag attribute value; f (de) is E c 2 The total number of categories of tag attribute values; j (J) f 2 For E c 2 The total number of occurrences before the f-th topic attribute value; f (fe) is E c 2 The total number of categories of the attribute values of the middle topics; j (J) g 3 For E c 2 The total number of occurrences before the g-th position attribute value; f (ge) is E c 2 The total number of categories of the mid-position attribute values; w (w) 1 、w 2 、w 3 W 4 The weights are respectively corresponding to the tag attribute value, the topic attribute value, the position attribute value and the arrangement sequence number;
and sequencing each piece of information to be pushed according to the corresponding adjustment parameters of each piece of information to be pushed from small to large, and generating an optimized push information sequence.
6. The method of claim 5, wherein after generating the adjustment parameter corresponding to each piece of information to be pushed, the method further comprises:
when F c In (a) and (b)
When according to F c The previously generated adjustment parameters generate compensation parameters E c b ；
According to w 4 、E c 1 E and E c b Generating the c-th information to be pushed in the initial push information sequenceCorresponding optimization parameter G c ；G c The following conditions are satisfied:
will G c And taking the information as an adjustment parameter corresponding to the c-th information to be pushed in the initial push information sequence.
7. The method of claim 5, wherein w 4 >max(w 1 ，w 2 ，w 3 )。
8. The method of claim 5, wherein the optimized push information sequence comprises a plurality of push information sub-sequences, each push information sub-sequence comprising the same amount of information to be pushed;
after generating the optimized push information sequence, the method further comprises:
if the B corresponding to the information to be pushed in the current push information subsequence max >Q1 or H max >Q2 or W max >Q3, taking the information to be pushed in the current push information subsequence as information to be replaced; wherein Q1 is a first homogeneity threshold; q2 is a second homogeneity threshold; q3 is a third homogeneity threshold; b (B) max The method comprises the steps that the maximum value of the repeated occurrence times of the tag attribute value of information to be pushed is obtained; h max The maximum value of the repeated occurrence times of the topic attribute value is the topic attribute value; w (W) max A maximum value of the number of repeated occurrences for the position attribute value having;
performing position exchange on the information to be replaced and target replacement information; the target replacement information is information to be pushed in a next push information sub-sequence adjacent to the current push information sub-sequence.
9. A non-transitory computer readable storage medium storing a computer program, wherein the computer program when executed by a processor implements a method of improving accuracy of information pushing according to any of claims 1 to 8.
10. An electronic device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that the processor implements a method of improving the accuracy of information pushing according to any of claims 1 to 8 when executing the computer program.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310444079.6A CN116401460B (en) | 2023-04-23 | 2023-04-23 | Method for improving information push accuracy, storage medium and electronic equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310444079.6A CN116401460B (en) | 2023-04-23 | 2023-04-23 | Method for improving information push accuracy, storage medium and electronic equipment |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116401460A true CN116401460A (en) | 2023-07-07 |
| CN116401460B CN116401460B (en) | 2024-04-05 |
Family
ID=87015988
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310444079.6A Active CN116401460B (en) | 2023-04-23 | 2023-04-23 | Method for improving information push accuracy, storage medium and electronic equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116401460B (en) |
Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20150120742A1 (en) * | 2012-06-21 | 2015-04-30 | Tencent Technology (Shenzhen) Company Limited | Method and system for processing recommended target software |
| US20160004764A1 (en) * | 2014-07-03 | 2016-01-07 | Palantir Technologies Inc. | System and method for news events detection and visualization |
| US20190102652A1 (en) * | 2016-08-31 | 2019-04-04 | Tencent Technology (Shenzhen) Company Limited | Information pushing method, storage medium and server |
| CN110659417A (en) * | 2019-09-12 | 2020-01-07 | 广东浪潮大数据研究有限公司 | Information pushing method and system, electronic equipment and storage medium |
| WO2020093289A1 (en) * | 2018-11-07 | 2020-05-14 | 深圳市欢太科技有限公司 | Resource recommendation method and apparatus, electronic device and storage medium |
| CN112765480A (en) * | 2021-04-12 | 2021-05-07 | 腾讯科技（深圳）有限公司 | Information pushing method and device and computer readable storage medium |
| CN113434762A (en) * | 2021-06-28 | 2021-09-24 | 平安银行股份有限公司 | Association pushing method, device and equipment based on user information and storage medium |
| CN114222252A (en) * | 2022-02-17 | 2022-03-22 | 中航信移动科技有限公司 | Message generation method and device, electronic equipment and storage medium |
| WO2022095382A1 (en) * | 2020-11-03 | 2022-05-12 | 平安科技（深圳）有限公司 | Knowledge graph-based electronic card generating and pushing method and device |
| CN114756742A (en) * | 2022-03-25 | 2022-07-15 | 北京沃东天骏信息技术有限公司 | Information pushing method and device and storage medium |
| CN116070028A (en) * | 2023-02-08 | 2023-05-05 | 百度时代网络技术（北京）有限公司 | Page-based content information pushing method, device, equipment and storage medium |
-
2023
- 2023-04-23 CN CN202310444079.6A patent/CN116401460B/en active Active
Patent Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20150120742A1 (en) * | 2012-06-21 | 2015-04-30 | Tencent Technology (Shenzhen) Company Limited | Method and system for processing recommended target software |
| US20160004764A1 (en) * | 2014-07-03 | 2016-01-07 | Palantir Technologies Inc. | System and method for news events detection and visualization |
| US20190102652A1 (en) * | 2016-08-31 | 2019-04-04 | Tencent Technology (Shenzhen) Company Limited | Information pushing method, storage medium and server |
| WO2020093289A1 (en) * | 2018-11-07 | 2020-05-14 | 深圳市欢太科技有限公司 | Resource recommendation method and apparatus, electronic device and storage medium |
| CN110659417A (en) * | 2019-09-12 | 2020-01-07 | 广东浪潮大数据研究有限公司 | Information pushing method and system, electronic equipment and storage medium |
| WO2022095382A1 (en) * | 2020-11-03 | 2022-05-12 | 平安科技（深圳）有限公司 | Knowledge graph-based electronic card generating and pushing method and device |
| CN112765480A (en) * | 2021-04-12 | 2021-05-07 | 腾讯科技（深圳）有限公司 | Information pushing method and device and computer readable storage medium |
| CN113434762A (en) * | 2021-06-28 | 2021-09-24 | 平安银行股份有限公司 | Association pushing method, device and equipment based on user information and storage medium |
| CN114222252A (en) * | 2022-02-17 | 2022-03-22 | 中航信移动科技有限公司 | Message generation method and device, electronic equipment and storage medium |
| CN114756742A (en) * | 2022-03-25 | 2022-07-15 | 北京沃东天骏信息技术有限公司 | Information pushing method and device and storage medium |
| CN116070028A (en) * | 2023-02-08 | 2023-05-05 | 百度时代网络技术（北京）有限公司 | Page-based content information pushing method, device, equipment and storage medium |
Non-Patent Citations (5)
| Title |
|---|
| 刘毅;钟忺;李琳;: "融合隐性特征的群体推荐方法研究", 计算机科学, no. 03, 15 March 2017 (2017-03-15) * |
| 姚彬修;倪建成;于苹苹;李淋淋;曹博;: "基于多源信息相似度的微博用户推荐算法", 计算机应用, no. 05, 10 May 2017 (2017-05-10) * |
| 崔艳萍;阎知知;王小巍;彭媛;: "互联网信息资源用户获取优化推送仿真研究", 计算机仿真, no. 07, 15 July 2017 (2017-07-15) * |
| 薛亚非;: "基于相似度的多重信息协同过滤算法优化仿真", 计算机仿真, no. 11, 15 November 2019 (2019-11-15) * |
| 郑慧;李冰;陈冬林;刘平峰;: "基于位置簇的移动生活服务个性化推荐技术", 计算机应用, no. 04, 10 April 2015 (2015-04-10) * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116401460B (en) | 2024-04-05 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN116405453A (en) | Information pushing method based on multiple features, storage medium and electronic equipment | |
| US9734261B2 (en) | Context aware query selection | |
| CN110020411B (en) | Image-text content generation method and equipment | |
| EP2438539B1 (en) | Co-selected image classification | |
| CN110830812B (en) | Similar anchor classification model training method, anchor recommendation method and related device | |
| US20230162091A1 (en) | Generating, using a machine learning model, request agnostic interaction scores for electronic communications, and utilization of same | |
| WO2016107354A1 (en) | Method and apparatus for providing user personalised resource message pushing | |
| CN109582904B (en) | Published content modification method, device, server, terminal and storage medium | |
| JP2021182371A (en) | Method, apparatus, electronic device, computer-readable storage medium, and computer program for regulating user emotion | |
| CN113449144B (en) | Video processing method and device and electronic equipment | |
| CN114501064A (en) | Video generation method, device, equipment, medium and product | |
| KR20220083822A (en) | Identification and publication of repeatable queries | |
| CN110889725A (en) | Online advertisement CTR estimation method, device, equipment and storage medium | |
| CN113015010B (en) | Push parameter determination method, device, equipment and computer readable storage medium | |
| CN112926308B (en) | Method, device, equipment, storage medium and program product for matching text | |
| CN113724015A (en) | Method and device for determining target display page, electronic equipment and storage medium | |
| CN112182281A (en) | Audio recommendation method and device and storage medium | |
| CN116401460B (en) | Method for improving information push accuracy, storage medium and electronic equipment | |
| CN112860929A (en) | Picture searching method and device, electronic equipment and storage medium | |
| CN113836388A (en) | Information recommendation method and device, server and storage medium | |
| CN116405452A (en) | Information pushing method, storage medium and electronic equipment | |
| CN113807106B (en) | Translation model training method and device, electronic equipment and storage medium | |
| CN107688956B (en) | Information processing method and server | |
| CN114218431A (en) | Video searching method and device, electronic equipment and storage medium | |
| CN112395490B (en) | Method and device for generating 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 |