CN118012920A - User portrait label quick matching method based on big data - Google Patents
User portrait label quick matching method based on big data Download PDFInfo
- Publication number
- CN118012920A CN118012920A CN202410417349.9A CN202410417349A CN118012920A CN 118012920 A CN118012920 A CN 118012920A CN 202410417349 A CN202410417349 A CN 202410417349A CN 118012920 A CN118012920 A CN 118012920A
- Authority
- CN
- China
- Prior art keywords
- recommended
- content
- score
- target
- value
- 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.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 82
- 238000010276 construction Methods 0.000 claims description 17
- 238000004364 calculation method Methods 0.000 claims description 6
- 238000012216 screening Methods 0.000 claims description 5
- 230000001174 ascending effect Effects 0.000 claims description 3
- 238000003058 natural language processing Methods 0.000 claims description 3
- 238000012913 prioritisation Methods 0.000 claims 1
- 230000000694 effects Effects 0.000 abstract description 7
- 230000006399 behavior Effects 0.000 description 27
- 238000004590 computer program Methods 0.000 description 4
- 238000010586 diagram Methods 0.000 description 4
- 235000013305 food Nutrition 0.000 description 3
- 230000007547 defect Effects 0.000 description 2
- 235000021168 barbecue Nutrition 0.000 description 1
- 238000010411 cooking Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 239000000446 fuel Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 235000014102 seafood Nutrition 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Landscapes
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention relates to the technical field of information recommendation, and discloses a user portrait tag quick matching method based on big data; constructing a user portrait based on historical recommended content, calculating interest scores, generating to-be-recommended content, calculating to-be-recommended scores, generating score difference values based on the interest scores and the to-be-recommended scores, marking target scores, and matching the target scores to the to-be-recommended content from the to-be-recommended content; compared with the prior art, the method has the advantages that an accurate user portrait can be constructed based on the comprehensive labels, the interest score of the user portrait is calculated, meanwhile, the score to be recommended of the content to be recommended in the database is compared with the interest score, the target recommended content meeting the interest degree of the user portrait is quickly and accurately matched, and the target recommended content is orderly arranged under the limitation of the arrangement priority, so that the target recommended content can be orderly, orderly and accurately recommended to the user, and the quick and accurate matching effect of the user portrait labels is realized.
Description
Technical Field
The invention relates to the technical field of information recommendation, in particular to a user portrait tag quick matching method based on big data.
Background
User portraits are descriptions and summaries of a particular user population, typically based on a series of attributes and features to portray the user's behavior, interests, preferences, and other relevant information, and by matching user portraits tags, the user population can be more precisely located and identified, optimizing product design, marketing campaigns, and service offerings, and making accurate recommendations for content of interest to the user.
The Chinese patent application with the application publication number of CN117539881A discloses an artificial intelligent e-commerce pushing system based on block chains and big data, wherein a user portrait construction module is used for acquiring user attributes, constructing label portraits for users and generating user portraits; the store label building module is used for acquiring store attributes, building label images of stores and generating store image images; the distributed storage module is used for decentralizing and storing the user image and the shop image through a blockchain technology; the intention collecting module is used for collecting the purchase intention of the user; the off-line pushing module is used for collecting real-time positioning information of a user and store positioning information, matching stores according to purchase intention of the user and user images, generating commodity recommendation information, and judging pushing time of the commodity recommendation information according to the distance between the matched store positioning and the real-time positioning of the user; the online pushing module is used for collecting user state information, judging the idle condition of the user, matching commodity recommendation information according to the purchase intention of the user and the user image, and judging the pushing time of the commodity recommendation information according to the idle condition of the user; the repetition rate of pushing information can be reduced, and the occurrence of message fatigue of a user is prevented;
the prior art has the following defects:
The existing user portrait tag matching can recommend interesting contents to a user by comparing the similarity of recommended contents and user portraits, and the interesting contents are recommended to the user, when the quantity of the interesting contents obtained after the user portrait tag matching is large, a phenomenon that the recommendation sequence is disordered easily occurs when a large quantity of interesting contents are recommended to the user in a concentrated mode, so that the problem of disorder in the recommendation process is caused, orderly recommendation can not be performed according to the interestingness of the user on the contents, and the recommendation accuracy of the interesting contents is reduced.
In view of the above, the present invention proposes a method for fast matching user portrait tags based on big data to solve the above-mentioned problems.
Disclosure of Invention
In order to overcome the defects in the prior art, the invention provides the following technical scheme for achieving the purposes: the quick matching method of the user portrait tag based on big data is applied to a data server and comprises the following steps:
S1: screening historical recommended content of a user from a database, and acquiring a comprehensive label based on the historical recommended content;
S2: constructing a user portrait based on the comprehensive label based on the construction criterion, and calculating an interest score of the user portrait;
s3: marking a label to be recommended in a database, generating content to be recommended based on the label to be recommended, and calculating a score to be recommended of the content to be recommended;
s4: generating a scoring difference value based on the interest score and the score to be recommended, comparing the scoring difference value with a preset scoring difference threshold value, and marking a target score based on a comparison result;
S5: matching the target recommended content from the content to be recommended based on the target score, and ranking the target recommended content based on the ranking priority.
Further, the integrated tags include type tags, content tags, and behavior tags.
Further, the construction criteria are: a sub-region contains at most a first semantic meaning or a second semantic meaning;
the user portrait construction method comprises the following steps:
Establishment of Dividing the image unit into three independent subareas, namely a first subarea, a second subarea and a third subarea;
Identifying semantics of type tags and content tags by natural language processing techniques, respectively obtaining First semantic sum/>A second semantic meaning;
Filling a first semantic meaning and a second semantic meaning into the first sub-region and the second sub-region, respectively, based on the construction criteria;
Filling behavior tags corresponding to the first semantics and the second semantics into a third subarea to obtain The filled picture units;
Will be The filled picture elements are arranged and are shown in/>And establishing a ring-shaped and closed outline at the periphery of each image unit to obtain the user image.
Further, the interest score calculating method includes:
recognition by natural semantic recognition techniques The behavior labels in the third subarea of the portrait units are used for obtaining the watching time length, the comment word number, the forwarding times and the clicking times;
respectively counting the number of watching time length, comment word number, forwarding times and clicking times to obtain A plurality of viewing time length values,Numerical value of individual comment word,/>Sum of forwarding order values/>A number of click times;
Will be Individual viewing duration values,/>Numerical value of individual comment word,/>Sum of forwarding order values/>After the click times are given with different weight factors, generating/>Scoring the individual units;
The expression of the unit score is:
；
In the method, in the process of the invention, For/>Individual Unit score,/>For/>Value of individual viewing duration,/>For/>Numerical value of individual comment word,/>For/>Number of forwarding times,/>For/>A number of clicks,/>、/>、/>、/>Is a weight factor;
Will be The unit scores are arranged in descending order from large to small, and the unit scores located at the first name are marked as interest scores.
Further, the labels to be recommended comprise a label to be recommended type, a label to be recommended content and a label to be recommended behavior;
The method for generating the content to be recommended comprises the following steps:
Querying a database The format attribute and the content attribute of each content file are obtained respectively/>Personal type to be recommended tag/>Content tags to be recommended;
Respectively marking and marking Personal type to be recommended tag/>Obtaining/>, wherein the watching time length, the comment word number, the forwarding times and the clicking times correspond to the content labels to be recommendedThe behavior labels to be recommended;
Will be Individual type tags to be recommended,/>Content tags to be recommended/>The behavior labels to be recommended are correspondingly combined one by one to obtain/>And the content to be recommended.
Further, the calculation method of the score to be recommended comprises the following steps:
Respectively count The number of watching time length, comment word number, forwarding times and clicking times in the content to be recommended is obtainedTime length value to be watched,/>Numerical value of each word to be reviewed,/>Sum of the number of times to be forwarded/>A number of times to be clicked;
Will be Time length value to be watched,/>Numerical value of each word to be reviewed,/>Sum of the number of times to be forwarded/>After the number of times to be clicked is given different weight factors, generating/>Scoring to be recommended;
The expression of the score to be recommended is:
；
In the method, in the process of the invention, For/>Score to be recommended,/>For/>The value of the time length to be watched,/>Is the firstNumerical value of each word to be reviewed,/>For/>Numerical value of number of times to be forwarded,/>For/>And a number of times to click.
Further, the method for generating the scoring difference value comprises the following steps:
Will be The scores to be recommended are compared with the interest scores one by one to obtain/>A score difference;
the expression of the scoring difference is:
；
In the method, in the process of the invention, For/>Score difference,/>Scoring the interests;
the marking method of the target score comprises the following steps:
Will be Score difference/>Respectively with a preset scoring difference threshold/>Comparing;
When (when) Greater than or equal to/>When (1)The individual score differences are marked as target scores;
When (when) Less than/>When (1)The individual score differences are not marked as target scores.
Further, the matching method of the target recommended content comprises the following steps:
when the target score is unique, marking the content to be recommended corresponding to the target score as target recommended content;
When the target score is not the same, the method will />, Corresponding to the individual target scoresThe individual content to be recommended is marked/>The individual targets recommend content.
Further, the arrangement priority is: the priority of the watching duration value is first, the priority of the comment word value is second, the priority of the clicking time value is third, and the priority of the forwarding time value is fourth.
Further, the method for arranging the target recommended content comprises the following steps:
Sequentially marking A watching time length value, a comment word value, a clicking time value and a forwarding time value in the target recommended content;
When (when) When the sizes of the individual viewing duration values are not consistent, the method comprises the step of/>The values of the watching time periods are arranged in descending order from big to small;
Arranging in descending order The values of the watching duration are numbered in ascending order in sequence, and the/>, is numbered according to the number pairsThe target recommended contents are sequentially arranged;
When (when) When the sizes of the individual viewing duration values are consistent, the method comprises the following steps of/>The comment word values are arranged in descending order from big to small;
Arranging in descending order The numerical values of the comment words are numbered in sequence, and the number pairs/>, according to the number pairsThe target recommended contents are sequentially arranged;
When (when) Individual viewing duration value sum/>When the values of the comment words are consistent in size, the method comprises the step of/>The click times are arranged in descending order from big to small;
Arranging in descending order The click times are numbered in sequence and the number pairs/>, according to the number pairsThe target recommended contents are sequentially arranged;
When (when) Individual viewing duration values,/>Numerical sum/>, of individual comment wordsWhen the values of the clicking times are consistent, the pair/>The target recommended content is randomly arranged.
The user portrait tag quick matching method based on big data has the technical effects and advantages that:
The method comprises the steps of screening historical recommended contents of a user from a database, acquiring comprehensive labels based on the historical recommended contents, constructing a user portrait based on the comprehensive labels based on a construction criterion, calculating interest scores of the user portrait, marking the labels to be recommended in the database, generating the contents to be recommended based on the labels to be recommended, calculating the scores to be recommended of the contents to be recommended, generating a scoring difference value based on the interest scores and the scores to be recommended, comparing the scoring difference value with a preset scoring difference threshold value, marking a target score based on a comparison result, matching the target recommended content from the contents to be recommended based on the target score, and arranging the target recommended content based on arrangement priority; compared with the prior art, the method has the advantages that an accurate user portrait can be constructed based on the comprehensive labels, the interest score of the user portrait is calculated, meanwhile, the score to be recommended of the content to be recommended in the database is compared with the interest score, the target recommended content meeting the interest degree of the user portrait is quickly and accurately matched, and the target recommended content is orderly arranged under the limitation of the arrangement priority, so that the target recommended content can be orderly, orderly and accurately recommended to the user, the quick and accurate matching effect of the user portrait label is realized, and the content of interest can be accurately recommended to the user.
Drawings
FIG. 1 is a flow chart of a quick matching method of user portrait tags based on big data provided in embodiment 1 of the present invention;
FIG. 2 is a schematic diagram of a quick matching system for user portrait tags based on big data according to embodiment 2 of the present invention;
fig. 3 is a schematic diagram of a structural schematic diagram of an electronic device according to embodiment 3 of the present invention;
fig. 4 is a schematic structural diagram of a computer readable storage medium according to embodiment 4 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 be within the scope of the invention.
Example 1: referring to fig. 1, the method for fast matching user portrait tag based on big data in this embodiment is applied to a data server, and includes:
S1: screening historical recommended content of a user from a database, and acquiring a comprehensive label based on the historical recommended content;
the historical recommended content refers to a set of all contents recommended to and watched by a user in a past historical time period, and when the historical recommended content is screened, the content needs to be ensured to be recommended and watched by the user, namely, some unviewed content can be removed, so that each historical recommended content can contain a required comprehensive label;
the comprehensive label refers to comprehensive information which is generated when a user views historical recommended content and can represent various aspects such as the content interest level, the content category, the operation information and the like of the user, and the comprehensive label can be obtained by obtaining the comprehensive label, so that the label can be accurately represented, and the subsequent matching operation of the label is convenient;
The comprehensive labels comprise type labels, content labels and behavior labels;
The type labels refer to expression forms in which information in the historical recommended content is watched and understood by a user, and when the information in the same recommended content is watched and understood by the user through different expression forms, different effects are generated on the user, so that corresponding labels are generated; the type label is obtained by referring to the attribute format of the file corresponding to the history recommended content;
The type label includes, but is not limited to, video and graphics context, and is exemplified by, when the type label is video, the attribute format of the file corresponding to the history recommended content is MP4, AVI, FLV, MOV, etc., and when the type label is graphics context, the attribute format of the file corresponding to the history recommended content is BMP, JPG, PNG, TIF, GIF, etc.;
The content labels refer to meanings expressed by information in the historical recommended content, the content label of each historical recommended content may or may not be unique, and meanings expressed by different content labels are different;
The content tag includes, but is not limited to, food, travel and automobile, and by way of example, when the content tag is food, file information in the history recommended content is hot pot, cooking, barbecue, seafood, etc., when the content tag is travel, file information in the history recommended content is seaside scenic spot, mountain scenic spot, ancient city scenic spot, etc., and when the content tag is automobile, file information in the history recommended content is fuel car, electric car, new car, second car, etc.;
The behavior labels refer to operation behavior data performed by a user when the user views and knows the historical recommended content, and the behavior labels corresponding to each historical recommended content are inconsistent, so that the interestingness of the user for each historical recommended content is reflected, the behavior labels comprise, but are not limited to, viewing time, comment word number and forwarding times, the behavior data such as the viewing time, comment word number and forwarding times are integrally displayed in one historical recommended content, so that the behavior labels are integral, and the behavior labels of the historical recommended content are, for example, 30 seconds in viewing time, 15 words in comment word number and 3 forwarding times;
S2: constructing a user portrait based on the comprehensive label based on the construction criterion, and calculating an interest score of the user portrait;
When the comprehensive label is obtained, the user portrait is required to be constructed based on the comprehensive label, so that the user portrait can accurately, truly and comprehensively reflect the content of interest of the user, thereby being used as a comparison basis for the matching of the subsequent labels, and when the user portrait is constructed, a specific construction criterion is required to be relied on so as to ensure that the comprehensive label can construct the user portrait meeting the requirements;
the construction criteria are: a sub-region contains at most a first semantic meaning or a second semantic meaning;
the user portrait construction method comprises the following steps:
Establishment of Dividing the image unit into three independent subareas, namely a first subarea, a second subarea and a third subarea; the blank portrait units refer to portrait units with zero current used capacity, and one portrait unit corresponds to a subset of a certain type of labels, content labels and behavior labels of a user, so that one portrait unit can contain one type of labels, one content label and one behavior label, and therefore, the user portrait can be divided into a plurality of independent subsets, the effect of integrating the portrait into zero is achieved, and the user portrait can be quickly and accurately constructed;
Identifying semantics of type tags and content tags by natural language processing techniques, respectively obtaining First semantic sum/>A second semantic meaning; the first semantics refer to content information referred to in the type tag, including but not limited to video and graphics; the second semantic refers to content information referred to in the content tags, including but not limited to food, travel, and automobiles;
Filling a first semantic meaning and a second semantic meaning into the first sub-region and the second sub-region, respectively, based on the construction criteria;
Filling behavior tags corresponding to the first semantics and the second semantics into a third subarea to obtain The filled picture units;
Will be The filled picture elements are arranged and are shown in/>The periphery of each image unit is provided with an annular and closed outline, so that a user image is obtained; all the image units can be packaged in a concentrated manner through the annular and closed outline, so that all the image units can be integrated in a closed area, a complete user image is assembled, the phenomenon that the image units are lost or omitted is avoided, and the authenticity and the comprehensiveness of the user image are improved;
the interest score is a numerical representation of the user's interest level in the portrait unit with the highest interest level in the user image, and when the interest score is obtained, the content interest level indicated by the portrait unit corresponding to the interest score is indicated to be the highest by the user;
The interest score calculating method comprises the following steps:
recognition by natural semantic recognition techniques The behavior labels in the third subarea of the portrait units are used for obtaining the watching time length, the comment word number, the forwarding times and the clicking times;
respectively counting the number of watching time length, comment word number, forwarding times and clicking times to obtain A plurality of viewing time length values,Numerical value of individual comment word,/>Sum of forwarding order values/>A number of click times; the number of the watching time length, the comment word number, the forwarding times and the clicking times is counted, so that the interest degree of the user for the specific content can be intuitively and accurately represented, and the interest degree is used as a data base of interest scoring;
Will be Individual viewing duration values,/>Numerical value of individual comment word,/>Sum of forwarding order values/>After the click times are given with different weight factors, generating/>Scoring the individual units; the unit score refers to a numerical value representation capable of representing the interestingness of the content indicated by each portrait unit in the user portrait, and the interestingness of each portrait unit can be represented;
The expression of the unit score is:
；
In the method, in the process of the invention, For/>Individual Unit score,/>For/>Value of individual viewing duration,/>For/>Numerical value of individual comment word,/>For/>Number of forwarding times,/>For/>A number of clicks,/>、/>、/>、/>Is a weight factor;
Wherein, Exemplary,/>Is 0.32,/>Is 0.25/>The total number of the components was 0.26,0.17; It should be noted that, the size of the weight factor is a specific value obtained by quantizing each data, so that the subsequent comparison is convenient, and the size of the weight factor depends on the number of the watching duration value, the comment word value, the forwarding order value and the clicking order value and the corresponding weight factor is preliminarily set by a person skilled in the art for each group of the watching duration value, the comment word value, the forwarding order value and the clicking order value;
Will be The unit scores are arranged in descending order from big to small, and the unit scores positioned at the first name are marked as interest scores;
s3: marking a label to be recommended in a database, generating content to be recommended based on the label to be recommended, and calculating a score to be recommended of the content to be recommended;
the label to be recommended refers to the label attribute of the content file to be recommended in the database, and is used for carrying out exclusive representation on each file to be recommended;
the labels to be recommended comprise a label to be recommended type, a label to be recommended content and a label to be recommended behavior;
The content to be recommended is content which is generated by the label of the type to be recommended, the label of the content to be recommended and the label of the behavior to be recommended and is ready for recommendation to the user in the database;
The method for generating the content to be recommended comprises the following steps:
Querying a database The format attribute and the content attribute of each content file are obtained respectively/>Personal type to be recommended tag/>Content tags to be recommended;
Respectively marking and marking Personal type to be recommended tag/>Obtaining/>, wherein the watching time length, the comment word number, the forwarding times and the clicking times correspond to the content labels to be recommendedThe behavior labels to be recommended;
Will be Individual type tags to be recommended,/>Content tags to be recommended/>The behavior labels to be recommended are correspondingly combined one by one to obtain/>Content to be recommended;
the score to be recommended refers to the numerical representation of the interestingness in the content to be recommended by the user, and each content to be recommended corresponds to a specific score to be recommended, so that the interestingness of the content to be recommended is numerically represented;
The calculation method of the score to be recommended comprises the following steps:
Respectively count The number of watching time length, comment word number, forwarding times and clicking times in the content to be recommended is obtainedTime length value to be watched,/>Numerical value of each word to be reviewed,/>Sum of the number of times to be forwarded/>A number of times to be clicked;
Will be Time length value to be watched,/>Numerical value of each word to be reviewed,/>Sum of the number of times to be forwarded/>After the number of times to be clicked is given different weight factors, generating/>Scoring to be recommended;
The expression of the score to be recommended is:
；
In the method, in the process of the invention, For/>Score to be recommended,/>For/>The value of the time length to be watched,/>Is the firstNumerical value of each word to be reviewed,/>For/>Numerical value of number of times to be forwarded,/>For/>And a number of times to click.
S4: generating a scoring difference value based on the interest score and the score to be recommended, comparing the scoring difference value with a preset scoring difference threshold value, and marking a target score based on a comparison result;
The scoring difference value is a numerical representation of the difference between the interest score and the score to be recommended, and when the scoring difference value is larger, the difference between the interest score and the score to be recommended is larger, the user's interest level of the content to be recommended is not matched with the interest level of the user image, and the user's interest level of the content to be recommended is lower;
the scoring difference generating method comprises the following steps:
Will be The scores to be recommended are compared with the interest scores one by one to obtain/>A score difference;
the expression of the scoring difference is:
；
In the method, in the process of the invention, For/>Score difference,/>Scoring the interests;
The target score refers to an interest score corresponding to the score to be recommended within a preset range, so that the interest degree of the user on the target recommended content corresponding to the target score meets the expectation, and the user can be matched with the recommended content which accords with the user portrait from the content to be recommended, and then the user is recommended;
the marking method of the target score comprises the following steps:
Will be Score difference/>Respectively with a preset scoring difference threshold/>Comparing; the preset scoring difference threshold is a numerical basis capable of distinguishing the magnitude of the scoring difference, so that the scoring difference is divided into a range meeting the expected range of the user interest and a range not meeting the expected range of the user interest; the preset scoring difference threshold value is obtained by acquiring a great number of target scores which meet the expected range of the user interest degree and do not meet the expected range of the user interest degree, and then obtaining the average value of the target scores;
When (when) Greater than or equal to/>Description of the first embodimentThe difference of the scores is greater than or equal to a preset score difference threshold, at the moment, the/>The interest degree of the content to be recommended corresponding to the score difference value reaches the user expected range, and the first/>The individual score differences are marked as target scores;
When (when) Less than/>Description of the first embodimentThe score difference is smaller than a preset score difference threshold, at the moment/>If the interest degree of the content to be recommended corresponding to the score difference value does not reach the user expected range, the first/>The individual score differences are not marked as target scores.
S5: matching the target recommended content from the content to be recommended based on the target score, and ranking the target recommended content based on the ranking priority;
the target recommended content is to-be-recommended content corresponding to the target score, so that the target recommended content can meet interest expectations of users, and the quick matching effect of the user image labels is achieved;
The matching method of the target recommended content comprises the following steps:
When the target score is unique, the number of the contents to be recommended corresponding to the target score is unique, and the contents to be recommended corresponding to the target score are marked as target recommended contents;
when the target scores are not the same, the number of the contents to be recommended corresponding to the target scores is not the same, at least two contents exist, and the target scores need to be matched The individual target scores are respectively corresponding to/>Marking the behavior labels to be recommended, and matching the behavior labels to be recommended with marked/>, from the content to be recommended/>, Corresponding to the individual target scoresThe behavior labels to be recommended;
Will be />, Corresponding to the individual target scoresThe individual content to be recommended is marked/>Recommending content by each target;
When the target recommended content is matched, the target recommended content is required to be recommended to a user according to a certain sequence, so that the user can orderly and quickly receive the target recommended content and watch and operate the target recommended content, and in order to ensure that a plurality of target recommended contents are orderly and tidy in recommendation, the target recommended content is required to be provided with a priority;
The arrangement priority is an arrangement rule for distinguishing the level of the watching time length value, the comment word value, the forwarding order value and the clicking order value in the content to be recommended corresponding to the target score, so that different content to be recommended can be orderly arranged;
The method comprises the steps that the user reflects the interestingness of the content to be recommended through the values of the watching duration value, the comment word value, the forwarding order value and the clicking order value, so that the behavior label to be recommended is used as a basis of the ranking priority, when the interestingness of the content to be recommended is higher, the watching duration of the user is longer, the ranking priority of the watching duration value is highest, when the watching duration is longer, the number of words of the user for commenting the content to be recommended is more, the ranking priority of the comment word value is second, when the number of the comment word is more, the clicking operation of the content to be recommended is more, the ranking priority of the clicking order value is third, and the ranking priority of the rest forwarding order value is fourth;
In summary, the ranking priorities are: the priority of the watching duration value is first, the priority of the comment word value is second, the priority of the click time value is third, and the priority of the forwarding time value is fourth;
the method for arranging the target recommended content comprises the following steps:
Sequentially marking A watching time length value, a comment word value, a clicking time value and a forwarding time value in the target recommended content;
When (when) When the sizes of the watching duration values are inconsistent, comparing/>, one by oneThe size of the individual viewing duration values will/>The values of the watching time periods are arranged in descending order from big to small;
Arranging in descending order The values of the watching duration are numbered in ascending order in sequence, and the/>, is numbered according to the number pairsThe target recommended contents are sequentially arranged;
When (when) When the sizes of the watching duration values are consistent, comparing/>, one by oneThe size of the individual comment word values will/>The comment word values are arranged in descending order from big to small;
Arranging in descending order The numerical values of the comment words are numbered in sequence, and the number pairs/>, according to the number pairsThe target recommended contents are sequentially arranged;
When (when) Individual viewing duration value sum/>When the values of the comment words are consistent, comparing/>, one by oneThe size of the click times value will/>The click times are arranged in descending order from big to small;
Arranging in descending order The click times are numbered in sequence and the number pairs/>, according to the number pairsThe target recommended contents are sequentially arranged;
When (when) Individual viewing duration values,/>Numerical sum/>, of individual comment wordsWhen the click times are consistent, the method indicates that the watching time length value, the comment word value, the forwarding times and the click times corresponding to the target scores are consistent, and the target scores are the same as the click timesRandomly arranging the recommended content of each target;
In the embodiment, historical recommended content of a user is screened out from a database, comprehensive labels are obtained based on the historical recommended content, a user portrait is constructed based on the comprehensive labels based on a construction criterion, interest scores of the user portrait are calculated, labels to be recommended in the database are marked, the contents to be recommended are generated based on the labels to be recommended, the scores to be recommended of the contents to be recommended are calculated, score difference values are generated based on the interest scores and the scores to be recommended, the score difference values are compared with a preset score difference threshold value, a target score is marked based on a comparison result, the target recommended content is matched from the contents to be recommended based on the target score, and the target recommended content is arranged based on arrangement priority; compared with the prior art, the method has the advantages that an accurate user portrait can be constructed based on the comprehensive labels, the interest score of the user portrait is calculated, meanwhile, the score to be recommended of the content to be recommended in the database is compared with the interest score, the target recommended content meeting the interest degree of the user portrait is quickly and accurately matched, and the target recommended content is orderly arranged under the limitation of the arrangement priority, so that the target recommended content can be orderly, orderly and accurately recommended to the user, the quick and accurate matching effect of the user portrait label is realized, and the content of interest can be accurately recommended to the user.
Example 2: referring to fig. 2, a part of the description of embodiment 1 is not described in detail in this embodiment, and a user portrait tag quick matching system based on big data is provided, which is applied to a data server and is used for implementing a user portrait tag quick matching method based on big data, and includes a comprehensive tag acquisition module, an interest score calculation module, a score calculation module to be recommended, a target score marking module and a target recommended content arrangement module, wherein the modules are connected by a wired or wireless network manner:
the comprehensive label acquisition module is used for screening historical recommended content of a user from the database and acquiring a comprehensive label based on the historical recommended content;
The interest score calculation module is used for constructing the user portrait based on the comprehensive label based on the construction criterion and calculating the interest score of the user portrait;
the to-be-recommended score calculating module is used for marking to-be-recommended labels in the database, generating to-be-recommended content based on the to-be-recommended labels, and calculating to-be-recommended scores of the to-be-recommended content;
the target score marking module is used for generating a score difference value based on the interest score and the score to be recommended, comparing the score difference value with a preset score difference threshold value and marking the target score based on a comparison result;
And the target recommended content arrangement module is used for matching the target recommended content from the content to be recommended based on the target score and arranging the target recommended content based on the arrangement priority.
Example 3: referring to fig. 3, the disclosure provides an electronic device, including a processor and a memory;
wherein the memory stores a computer program for the processor to call;
The processor executes the user portrait tag quick matching method based on big data by calling the computer program stored in the memory.
Since the electronic device described in this embodiment is an electronic device used to implement the method for quickly matching a user portrait tag based on big data in embodiment 1 of the present application, based on the method for quickly matching a user portrait tag based on big data described in this embodiment of the present application, those skilled in the art can understand the specific implementation of the electronic device and various modifications thereof, so how to implement the method in this embodiment of the present application for this electronic device will not be described in detail herein. As long as the person skilled in the art implements the electronic device adopted by the user portrait tag quick matching method based on big data in the embodiment of the application, the electronic device belongs to the scope of the application to be protected.
Example 4: referring to fig. 4, the present embodiment disclosure provides a computer readable storage medium having stored thereon a computer program that is erasable;
When the computer program is run, the method for realizing the quick matching of the user portrait labels based on big data is executed.
The foregoing is merely illustrative of the present invention, and the present invention is not limited thereto, and any person skilled in the art will readily recognize that variations or substitutions are within the scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.
Claims (10)
1. The quick matching method of the user portrait tag based on big data is applied to a data server and is characterized by comprising the following steps:
S1: screening historical recommended content of a user from a database, and acquiring a comprehensive label based on the historical recommended content;
S2: constructing a user portrait based on the comprehensive label based on the construction criterion, and calculating an interest score of the user portrait;
s3: marking a label to be recommended in a database, generating content to be recommended based on the label to be recommended, and calculating a score to be recommended of the content to be recommended;
s4: generating a scoring difference value based on the interest score and the score to be recommended, comparing the scoring difference value with a preset scoring difference threshold value, and marking a target score based on a comparison result;
S5: matching the target recommended content from the content to be recommended based on the target score, and ranking the target recommended content based on the ranking priority.
2. The big data based user portrayal tab quick match method of claim 1, wherein the integrated tab includes a type tab, a content tab and a behavior tab.
3. The big data based user portrayal label fast matching method according to claim 2, characterized in that the construction criteria are: a sub-region contains at most a first semantic meaning or a second semantic meaning;
the user portrait construction method comprises the following steps:
Establishment of Dividing the image unit into three independent subareas, namely a first subarea, a second subarea and a third subarea;
Identifying semantics of type tags and content tags by natural language processing techniques, respectively obtaining First semantic sum/>A second semantic meaning;
Filling a first semantic meaning and a second semantic meaning into the first sub-region and the second sub-region, respectively, based on the construction criteria;
Filling behavior tags corresponding to the first semantics and the second semantics into a third subarea to obtain The filled picture units;
Will be The filled picture elements are arranged and are shown in/>And establishing a ring-shaped and closed outline at the periphery of each image unit to obtain the user image.
4. The big data based user portrayal label fast matching method according to claim 3, wherein the interest score calculating method comprises:
recognition by natural semantic recognition techniques The behavior labels in the third subarea of the portrait units are used for obtaining the watching time length, the comment word number, the forwarding times and the clicking times;
respectively counting the number of watching time length, comment word number, forwarding times and clicking times to obtain Individual viewing duration values,/>Numerical value of individual comment word,/>Sum of forwarding order values/>A number of click times;
Will be Individual viewing duration values,/>Numerical value of individual comment word,/>Sum of forwarding order values/>After the click times are given with different weight factors, generating/>Scoring the individual units;
The expression of the unit score is:
；
In the method, in the process of the invention, For/>Individual Unit score,/>For/>Value of individual viewing duration,/>For/>The number of the comment words,For/>Number of forwarding times,/>For/>A number of clicks,/>、/>、/>、/>Is a weight factor;
Will be The unit scores are arranged in descending order from large to small, and the unit scores located at the first name are marked as interest scores.
5. The big data-based user portrait tag quick matching method according to claim 4, wherein the tags to be recommended include a type tag to be recommended, a content tag to be recommended, and a behavior tag to be recommended;
The method for generating the content to be recommended comprises the following steps:
Querying a database The format attribute and the content attribute of each content file are obtained respectively/>Personal type to be recommended tag/>Content tags to be recommended;
Respectively marking and marking Personal type to be recommended tag/>Obtaining/>, wherein the watching time length, the comment word number, the forwarding times and the clicking times correspond to the content labels to be recommendedThe behavior labels to be recommended;
Will be Individual type tags to be recommended,/>Content tags to be recommended/>The behavior labels to be recommended are correspondingly combined one by one to obtain/>And the content to be recommended.
6. The quick matching method for user portrait tags based on big data according to claim 5, wherein said calculation method for scores to be recommended includes:
Respectively count Obtaining the number of watching time length, comment word number, forwarding times and clicking times in the content to be recommendedTime length value to be watched,/>Numerical value of each word to be reviewed,/>Sum of the number of times to be forwarded/>A number of times to be clicked;
Will be Time length value to be watched,/>Numerical value of each word to be reviewed,/>Sum of the number of times to be forwarded/>After the number of times to be clicked is given different weight factors, generating/>Scoring to be recommended;
The expression of the score to be recommended is:
；
In the method, in the process of the invention, For/>Score to be recommended,/>For/>The value of the time length to be watched,/>For/>Numerical value of each word to be reviewed,/>For/>Numerical value of number of times to be forwarded,/>For/>And a number of times to click.
7. The quick matching method for user portrait tags based on big data according to claim 6, wherein the generating method for scoring differences includes:
Will be The scores to be recommended are compared with the interest scores one by one to obtain/>A score difference;
the expression of the scoring difference is:
；
In the method, in the process of the invention, For/>Score difference,/>Scoring the interests;
the marking method of the target score comprises the following steps:
Will be Score difference/>Respectively with a preset scoring difference threshold/>Comparing;
When (when) Greater than or equal to/>When (1)The individual score differences are marked as target scores;
When (when) Less than/>When (1)The individual score differences are not marked as target scores.
8. The big data based user portrait tag quick matching method according to claim 7, wherein the matching method of the target recommended content includes:
when the target score is unique, marking the content to be recommended corresponding to the target score as target recommended content;
When the target score is not unique />, Corresponding to the individual target scoresThe individual content to be recommended is marked/>The individual targets recommend content.
9. The big data based user portrayal label fast matching method according to claim 8, wherein the prioritization is: the priority of the watching duration value is first, the priority of the comment word value is second, the priority of the clicking time value is third, and the priority of the forwarding time value is fourth.
10. The big data based user portrayal label fast matching method according to claim 9, wherein the method of ranking the target recommended content comprises:
Sequentially marking A watching time length value, a comment word value, a clicking time value and a forwarding time value in the target recommended content;
When (when) When the sizes of the individual viewing duration values are not consistent, the method comprises the step of/>The values of the watching time periods are arranged in descending order from big to small;
Arranging in descending order The values of the watching duration are numbered in ascending order in sequence, and the/>, is numbered according to the number pairsThe target recommended contents are sequentially arranged;
When (when) When the sizes of the individual viewing duration values are consistent, the method comprises the following steps of/>The comment word values are arranged in descending order from big to small;
Arranging in descending order The numerical values of the comment words are numbered in sequence, and the number pairs/>, according to the number pairsThe target recommended contents are sequentially arranged;
When (when) Individual viewing duration value sum/>When the values of the comment words are consistent in size, the method comprises the step of/>The click times are arranged in descending order from big to small;
Arranging in descending order The click times are numbered in sequence and the number pairs/>, according to the number pairsThe target recommended contents are sequentially arranged;
When (when) Individual viewing duration values,/>Numerical sum/>, of individual comment wordsWhen the values of the clicking times are consistent, the pair/>The target recommended content is randomly arranged.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410417349.9A CN118012920A (en) | 2024-04-09 | 2024-04-09 | User portrait label quick matching method based on big data |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410417349.9A CN118012920A (en) | 2024-04-09 | 2024-04-09 | User portrait label quick matching method based on big data |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN118012920A true CN118012920A (en) | 2024-05-10 |
Family
ID=90956710
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202410417349.9A Pending CN118012920A (en) | 2024-04-09 | 2024-04-09 | User portrait label quick matching method based on big data |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN118012920A (en) |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105824923A (en) * | 2016-03-17 | 2016-08-03 | 海信集团有限公司 | Movie and video resource recommendation method and device |
| CN106651533A (en) * | 2016-12-29 | 2017-05-10 | 合肥华凌股份有限公司 | User behavior-based personalized product recommendation method and apparatus |
| WO2018032790A1 (en) * | 2016-08-16 | 2018-02-22 | 武汉斗鱼网络科技有限公司 | Weighted k-nearest-neighbor scoring-based live broadcast room recommendation method and system |
| CN109543111A (en) * | 2018-11-28 | 2019-03-29 | 广州虎牙信息科技有限公司 | Recommendation information screening technique, device, storage medium and server |
| CN109933720A (en) * | 2019-01-29 | 2019-06-25 | 汕头大学 | A kind of dynamic recommendation method based on user interest Adaptive evolution |
| CN110334202A (en) * | 2019-03-28 | 2019-10-15 | 平安科技（深圳）有限公司 | User interest label construction method and relevant device based on news application software |
| CN110674410A (en) * | 2019-10-08 | 2020-01-10 | 北京物灵科技有限公司 | User portrait construction and content recommendation method, device and equipment |
-
2024
- 2024-04-09 CN CN202410417349.9A patent/CN118012920A/en active Pending
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105824923A (en) * | 2016-03-17 | 2016-08-03 | 海信集团有限公司 | Movie and video resource recommendation method and device |
| WO2018032790A1 (en) * | 2016-08-16 | 2018-02-22 | 武汉斗鱼网络科技有限公司 | Weighted k-nearest-neighbor scoring-based live broadcast room recommendation method and system |
| CN106651533A (en) * | 2016-12-29 | 2017-05-10 | 合肥华凌股份有限公司 | User behavior-based personalized product recommendation method and apparatus |
| CN109543111A (en) * | 2018-11-28 | 2019-03-29 | 广州虎牙信息科技有限公司 | Recommendation information screening technique, device, storage medium and server |
| CN109933720A (en) * | 2019-01-29 | 2019-06-25 | 汕头大学 | A kind of dynamic recommendation method based on user interest Adaptive evolution |
| CN110334202A (en) * | 2019-03-28 | 2019-10-15 | 平安科技（深圳）有限公司 | User interest label construction method and relevant device based on news application software |
| CN110674410A (en) * | 2019-10-08 | 2020-01-10 | 北京物灵科技有限公司 | User portrait construction and content recommendation method, device and equipment |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US10223454B2 (en) | Image directed search | |
| US9959467B2 (en) | Image processing client | |
| US9830522B2 (en) | Image processing including object selection | |
| US9058611B2 (en) | System and method for advertising using image search and classification | |
| US9575995B2 (en) | Image processing methods | |
| US10825048B2 (en) | Image processing methods | |
| US20220207274A1 (en) | Client Based Image Analysis | |
| US20150261806A1 (en) | Computerized internet search system and method | |
| US20160342624A1 (en) | Image Tagging System | |
| CA2885880C (en) | Image processing including object selection | |
| CA2885858A1 (en) | Image tagging system | |
| CA2885879C (en) | Image processing methods | |
| US9639867B2 (en) | Image processing system including image priority | |
| CA2885874A1 (en) | Image processing system including image priority | |
| CN114297499A (en) | Information display method and device, computer equipment and storage medium | |
| US9569465B2 (en) | Image processing | |
| US20170315676A1 (en) | Dynamic content insertion | |
| US8281245B1 (en) | System and method of preparing presentations | |
| CN118012920A (en) | User portrait label quick matching method based on big data | |
| KR102429104B1 (en) | Product catalog automatic classification system based on artificial intelligence | |
| US20150220569A1 (en) | Priority Based Image Processing Methods | |
| CN113822566A (en) | Business assessment processing method and device, computer equipment and storage medium | |
| CN113254775A (en) | Credit card product recommendation method based on client browsing behavior sequence | |
| CA2885835A1 (en) | Image processing server | |
| US11282104B2 (en) | Provision device, provision method and non-transitory computer readable storage medium |
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 |