CN106651542B - Article recommendation method and device - Google Patents
Article recommendation method and device Download PDFInfo
- Publication number
- CN106651542B CN106651542B CN201611261973.6A CN201611261973A CN106651542B CN 106651542 B CN106651542 B CN 106651542B CN 201611261973 A CN201611261973 A CN 201611261973A CN 106651542 B CN106651542 B CN 106651542B
- Authority
- CN
- China
- Prior art keywords
- item
- recommended
- similarity
- user
- behavior
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/06—Buying, selling or leasing transactions
- G06Q30/0601—Electronic shopping [e-shopping]
- G06Q30/0631—Item recommendations
Abstract
The application discloses a method and a device for recommending articles, which are used for calculating the article description similarity between each article in the articles to be recommended and a target article; calculating the behavior similarity of the behavior of each user in the user group on each item in the items to be recommended and the behavior of the user on the target item; calculating the quality index of each item in the items to be recommended; calculating the similarity between each item in the items to be recommended and the target item according to the item description similarity, the behavior similarity and the quality index; and selecting recommended articles from the articles to be recommended according to the similarity obtained by calculation. Therefore, when recommending articles for a user, multiple factors are comprehensively considered, and the similarity between the two articles is considered from multiple aspects, so that the calculated similarity between each article in the articles to be recommended and the target article is more accurate.
Description
Technical Field
The present application relates to the field of information technology processing, and in particular, to a method and an apparatus for recommending an article.
Background
The 21 st century is a highly information-oriented era, the internet has become an indispensable part of the life of people, and application software with various functions enriches the life of people.
Currently, internet users can watch videos, shop, listen to music, read, etc. online. However, it is inconvenient for the user to find the favorite information among the network information of various types and huge amount.
In the prior art, items browsed by a user are generally taken as a reference when the items are recommended to the user. Specifically, if the reference article is a, for article B, the text description information of the two articles is obtained, and then the similarity of the text description information between the two articles is calculated. And if the similarity is higher, recommending the B to the user. However, the recommendation method is based on the similarity of the text contents only, the recommendation mode is single, and the calculated similarity between the two articles is inaccurate, so that a new article recommendation scheme is needed.
Disclosure of Invention
The embodiment of the application provides an article recommendation method and device, and aims to solve the problems that in the prior art, when articles are recommended to a user, the recommendation mode is single and the calculated similarity between two articles is inaccurate because the articles are recommended only based on the similarity of text contents.
In one aspect, an embodiment of the present application provides an article recommendation method, including:
calculating the item description similarity between each item in the items to be recommended and the target item;
calculating the behavior similarity of the behavior of each user in the user group on each item in the items to be recommended and the behavior of the user on the target item;
calculating the quality index of each item in the items to be recommended;
calculating the similarity between each item in the items to be recommended and the target item according to the item description similarity, the behavior similarity and the quality index;
and selecting recommended articles from the articles to be recommended according to the similarity obtained by calculation.
In another aspect, an embodiment of the present application provides an article recommendation apparatus, including:
the item description similarity calculation module is used for calculating the item description similarity between each item in the items to be recommended and the target item;
the behavior similarity calculation module is used for calculating the behavior similarity of the behavior of each user in the user group on each article in the articles to be recommended and the behavior of the user on the target article;
the quality index calculation module is used for calculating the quality index of each item in the items to be recommended;
the similarity calculation module is used for calculating the similarity between each item in the items to be recommended and the target item according to the item description similarity, the behavior similarity and the quality index;
and the selection module is used for selecting recommended articles from the articles to be recommended according to the similarity obtained by calculation.
The beneficial effect of this application is as follows: in the technical scheme provided by the embodiment of the application, the item description similarity between each item in the items to be recommended and the target item is calculated; calculating the behavior similarity of the behavior of each user in the user group on each item in the items to be recommended and the behavior of the user on the target item; calculating the quality index of each item in the items to be recommended; calculating the similarity between each item in the items to be recommended and the target item according to the item description similarity, the behavior similarity and the quality index; and selecting recommended articles from the articles to be recommended according to the similarity obtained by calculation. Therefore, when recommending articles for a user, multiple factors such as article description similarity between the articles, behavior similarity of the user to the articles, quality index of the articles and the like are comprehensively considered, and the similarity between the two articles is considered from multiple aspects, so that the calculated similarity between each article in the articles to be recommended and the target article is more accurate.
Drawings
In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed to be used in the description of the embodiments are briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments of the present application, and it is obvious for those skilled in the art to obtain other drawings based on these drawings without creative efforts.
Fig. 1 is a schematic flowchart illustrating a method for recommending an item according to an embodiment of the present application;
fig. 2 is a schematic flowchart illustrating a method for recommending an item according to a second embodiment of the present application;
fig. 3 is a structural diagram of an apparatus for recommending articles according to a third embodiment of the present application;
fig. 4 is a schematic hardware structure diagram of an electronic device according to a fifth method for recommending articles provided in an embodiment of the present application.
Detailed Description
In order to make the objects, technical solutions and advantages of the present application clearer, the present application will be described in further detail with reference to the accompanying drawings, and it is obvious that the described embodiments are only a part of the embodiments of the present application, and not all embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present application.
The first embodiment is as follows:
as shown in fig. 1, a flow chart of a method for recommending an item provided in an embodiment of the present application is schematically illustrated, and the method includes the following steps:
step 101: and calculating the item description similarity between each item in the items to be recommended and the target item.
It should be noted that the article described in the embodiment of the present application refers to a display object on a network, and the display object may be a product or a service. Such as manufactured goods, developed APPs (applications), etc.
Step 102: and calculating the behavior similarity of the behavior of each user in the user group on each item in the items to be recommended and the behavior of the user on the target item.
Wherein, in one embodiment, the behavior includes a purchasing behavior, a browsing behavior, a collecting behavior, a marking behavior, a downloading behavior, and the like. The marking behavior is, for example, approval, giving a score, or the like.
Step 103: and calculating the quality index of each item in the items to be recommended.
Step 104: and calculating the similarity between each item in the items to be recommended and the target item according to the item description similarity, the behavior similarity and the quality index.
Step 105: and selecting recommended articles from the articles to be recommended according to the similarity obtained by calculation.
For ease of understanding, the following further description of an item recommendation method provided in the embodiments of the present application may include the following:
in one embodiment, step 102 can be specifically executed in one of the following two ways:
the first method is as follows:
step A1: acquiring behaviors of each user in the user group on each item to be recommended and a target item; and assigning a behavior coefficient for each behavior.
In specific implementation, the behavior coefficients corresponding to the behaviors can be set and configured according to actual requirements. Of course, the level of the behavior may be set first, for example, corresponding to the shopping website, the order of the behavior levels from high to low may be: purchasing, collecting and browsing; for a website providing an APP service, the order of the behavior levels from high to low can be installation, collection and browsing. Wherein the level of marking behavior may be lower than the collecting behavior but higher than the browsing behavior. It should be noted that, in the specific implementation, the behavior level may be set according to the specific behavior and the need, and the embodiment of the present application does not limit this.
Specifically, when the behavior coefficient is set, the higher the level, the higher the coefficient corresponding to the behavior, and specifically, when the calculation is performed, the behavior coefficient corresponding to the behavior at the highest level may be used. For example, the behavior coefficient corresponding to the purchasing behavior and the downloading behavior is the same and is marked as 3; the behavior coefficients corresponding to the collection behavior and the marking behavior are the same and are marked as 2; marking a behavior coefficient corresponding to the browsing behavior as 1; no behavior is noted as 0.
Wherein the behavior for calculating the behavior similarity is the highest level behavior of the user on the item. For example, if a user downloads an item, the user must browse the item, so the behavior of the user includes browsing behavior and downloading behavior, and the behavior coefficient of the downloading behavior is used in the calculation.
Step A2: for each user in the user group and each item to be recommended in the items to be recommended, calculating the product of the behavior coefficient of the user to the item to be recommended and the behavior coefficient of the target item, and taking the calculated product as the similarity of the behavior of the user to the item to be recommended and the behavior of the user to the target item, which is called as the individual behavior similarity.
For example, the behavior coefficient of the user a to the recommended item a is 3, and the behavior coefficient of the target item is 2. Then the similarity of the behavior of the item to be recommended a and the target item for the user a is 6. By analogy, the individual behavior similarity of each user can be calculated.
Step A3: for each item in the items to be recommended, calculating an average value of the individual behavior similarity of each user in the user group to the item and the target item, and taking the average value as the similarity of the behavior of the user group to the item and the behavior of the user group to the target item, which is called behavior similarity.
Assuming that, for an item a in the items to be recommended, the similarity of the behaviors of each user on the item a and the target item is as shown in table 1, and then the similarity of the behaviors of the user group on the item a and the target item is 1.5.
TABLE 1
Preferably, the user group may be classified, for example, by gender or by age. When the behavior similarity between the behavior of the user group on each article in the article set and the behavior of the target article is calculated, the average value of the individual behavior similarity between the behavior of each article to be recommended by each type of user and the behavior of the target article can be specifically calculated and used as the behavior similarity between the behavior of each article in the article set and the behavior of the target article by the type of user.
The second method comprises the following steps: matrix factorization may be employed to perform data mining to determine behavioral similarity, specifically:
step B1: and forming an article set by the articles to be recommended and the target articles, and acquiring the behaviors of each user in the user group on each article in the article set.
Step B2: aiming at each user in the user group, determining the behavior coefficient of each article in the article set by the user according to the preset corresponding relation between the behavior and the behavior coefficient, determining a behavior matrix representing the behavior coefficient of the article set by the user by taking the behavior coefficient as a matrix element,
assuming that the behavior matrix is determined by using the behavior coefficient of each article by the user as a row vector, the finally determined behavior matrix is shown in table 2. Of course, in specific implementation, the behavior coefficient of the user for each article may also be used as a column vector to determine the behavior matrix.
TABLE 2
| Article 1 | Article 2 | …… | Article N | |
| User 1 | T11 | T12 | …… | T1N |
| User 2 | T21 | T22 | …… | T2N |
| …… | …… | …… | …… | …… |
| User n | Tn1 | Tn3 | …… | TnN |
Step B3: performing matrix decomposition on the user behavior matrix to obtain a decomposition matrix meeting preset conditions; the preset conditions include: the decomposition matrix has n vectors, each vector has d elements, wherein n represents the total number of the articles in the article set, and d represents a preset value.
Wherein the value of d may be an integer between 40 and 200. In specific implementation, the recommendation effect can be determined according to an empirical value and also according to the recommendation effect in the subsequent article recommendation process.
In the specific implementation, the matrix decomposition may be performed on the user behavior matrix by using a method such as an alternating least square method, a trigonometric decomposition method, a QR value decomposition method, or a singular value decomposition method. All methods capable of performing matrix decomposition in the prior art are applicable to the embodiment, and the embodiment of the present application does not limit this.
Step B4: and regarding each item in the items to be recommended, taking the vector product of the corresponding vector of the item in the decomposition matrix and the corresponding vector of the target item in the decomposition matrix as the behavior similarity of the item and the target item.
For example, the user behavior matrix is recorded as Am*nThe ith row element represents the behavior coefficient of the ith user on all the articles in the article set, and the jth column represents the behavior of each user in the user group on the article j. To Am*nPerforming approximate decomposition to obtain Um*d,Vn*dAnd meet the requirements
The similarity S of the behaviors of the item i and the item jij＝ViVjWherein V isiRepresents Vn*dThe ith vector of (2)jRepresents Vn*dThe jth vector of (2). If the item j is one of the items to be recommended, the item i is the target item. Then S calculatedijThat is, the similarity between the behavior of each user on the item j in the user group and the behavior of the user on the target item.
In this way, based on the principle of matrix decomposition for data mining, the behavior similarity of the behavior of each user in the user group on each item in the to-be-recommended items and the behavior of the user on the target item can be calculated, so that the calculated result can more represent the behavior similarity of the popular user on the two items.
In one embodiment, when recommending an item for a user to be recommended (such as a first user), in order to make the recommended item better meet the personalized requirements of the first user and improve the accuracy of recommending the item for the first user and the user experience of the first user, in the embodiment of the present application, the following steps may also be performed:
step C1: and receiving a browsing request of the first user for the target item.
Step C2: and acquiring the personal information of the first user according to the browsing request.
The personal information of the first user comprises registration information of the first user and/or behavior of the first user on the article. The registration information includes, for example, the user's sex, age, taste, and the like. The preference is, for example, a preference input by the user at the time of registration, for example, a preference of a shopping site is clothes such as a skirt and a sweater, a preference of a video registered on a video site is a suspense-type video, a thriller-type video, and a game-type preference is a gun-battle-type game).
Of course, the article recommendation method in the embodiment of the present application is not only applicable to the above listed application scenarios, but also applicable to all scenarios in which articles can be recommended.
Step C3: and calculating the preference value of the first user to each item in the selected recommended items according to the personal information of the first user.
Wherein, when the personal information is the registration information. The user group may be classified according to the registration information, for example, by gender and age. And then, according to the user preferences included in the registration information, counting the preference values of various users for various articles. Assuming that the highest value of the preference values is configured to be 10, if 100 users in a class of users are in total, and 80 users like a skirt, the preference value of the skirt is 8; there are 60 people who prefer knits, then the knits preference value is 6, and so on. And obtaining the preference value of the class of users to various articles. Then, for the first user, a user category to which the first user belongs may be first determined according to the registration information of the first user. For each recommended item selected in step 105, an item category to which the recommended item belongs is determined. And then, taking the preference value corresponding to the item type of the recommended item in the user types to which the first user belongs as the preference value of the first user for the recommended item.
Of course, in specific implementation, the method for classifying the user group, the method for classifying the article, and the method for obtaining the user preference may be implemented in other manners, which are not described herein again.
Step C4: and sorting the selected recommended articles according to the preference values.
The sorting of the selected recommended items according to the preference values is, for example, to calculate, for each selected recommended item, a sum of a behavior similarity between the recommended item and the target item and a preference value of the first user for the recommended item. And then sorted according to the calculated sum. Of course, the preference value of the first user for the recommended item may also be used as a weight value, the product of the behavior similarity between the recommended item and the target item and the weight value thereof is calculated, and then the ranking is performed according to the product.
Of course, it should be noted that, in specific implementation, a scheme for sorting the selected to-be-recommended articles by using the preference value may be set according to actual requirements, and the comparison is not limited in the present application.
Step C5: and recommending to the first user according to the sorted recommended articles.
Wherein, in one embodiment, when the personal information is the behavior of the first user on the items in the item set, step B3 may be specifically performed as:
for each of the recommended items, determining a preference value for the item for the first user according to the following formula:
Puj＝∑k∈(S，u)fukSij
whereinJ is an item in the recommended item, PujThe preference value of the first user to the item, u is the first user, i is the target item, S is the set of items of the item set subjected to the highest action level by the first user, k is the kth item in the item set, fukTo represent the coefficient of behavior of the first user on the kth item, SijIs the behavior similarity of the object j and the object i.
For example, the item set has N items in total, the items of the first user that have performed the highest level of behavior are respectively marked as item 1, item 2, and item 3, the behavior coefficients of the first user for the items are respectively a1, a2, and A3, and the similarity of behavior of each user in the user group for item j and target item i is a 4. The final calculated preference value for item j for the first user is (a1+ a2+ A3) × a 4. Therefore, the preference value of the user to the article is obtained according to the actual preference of the user, so that the determination of the preference value is more accurate.
In one embodiment, in order to make the recommended item better meet the requirement of the user and improve the user experience, in the embodiment of the present application, step 101 (calculating the item description similarity between each item in the items to be recommended and the target item) may be specifically performed as:
step D1: and acquiring the text description of each item in the items to be recommended.
Step D2: and performing word segmentation on the text description.
Step D3: and calculating a word vector of each word obtained after word segmentation.
For example, TF-IDF (a weighting technique for information retrieval and information exploration) may be used to calculate a word vector for each word obtained after word segmentation. Of course, all methods capable of calculating word vectors in the prior art are applicable to the present embodiment and all within the protection scope of the present application, and the present application does not limit this.
Step D4: and calculating the item description similarity between each item in the items to be recommended and a target item according to the word vector of each word.
For example, the item description similarity between two items in the item set can be expressed by a formula
In one embodiment, due to the fact that the quantity of the articles is very large, the quality of the articles, such as cost performance, is different, and if some articles with low quality are recommended to the user, the user is disturbed, and the recommending effect is poor. Therefore, in order to improve the recommendation effect, in the embodiment of the present application, step 103 (calculating the quality index of each of the items to be recommended) may be specifically performed as:
calculating a quality index for each of the items to be recommended according to at least one of: item popularity, item rating, number of item reviews, etc.
Information used for calculating the quality index of the article in the popularity, the grading level of the article, the number of article comments and the like can be normalized; and then taking the sum of the normalized information as the quality index of the article. Therefore, some high-quality articles can be recommended to the user, the time of the user is saved, and the user experience is improved.
In one embodiment, in order to further make the recommended item better meet the requirement of the user and improve the user experience, step 104 (calculating the similarity between each item of the items to be recommended and the target item according to the item description similarity, the behavior similarity, and the quality index) may specifically be performed as:
step E1: determining the weight values corresponding to the item description similarity, the behavior similarity and the quality index.
Step E2: and calculating the similarity between each item in the items to be recommended and the target item in a weighted summation mode according to the determined weight values and the item description similarity, the behavior similarity and the quality index.
Similarly, sorting the selected recommended items according to the preference value may be performed as:
step F1: determining the weight values corresponding to the item description similarity, the behavior similarity, the quality index and the preference value respectively.
Step F2: and calculating the sorting value of the selected recommended articles in a weighted summation mode according to the determined weight value, the article description similarity, the behavior similarity, the quality index and the preference value, and then sorting according to the sorting value.
Then, recommending to the first user according to the sorted recommended item may specifically be performed as: and selecting the recommended articles L before sorting to recommend to the first user, wherein L is a positive integer.
In one embodiment, in order to obtain a better weight value, step F1 may be specifically executed as:
step G1: determining initial weight values corresponding to the item description similarity, the behavior similarity, the quality index and the preference value respectively.
Step G2: taking the initial weight value corresponding to the article description similarity in the obtained initial weight values as a parameter to be adjusted in a comparison test to obtain a first group of weight value test objects; and
taking the initial weight value corresponding to the behavior similarity in the obtained initial weight values as a parameter to be adjusted in a comparison test to obtain a second group of weight value test objects; and the number of the first and second groups,
taking the initial weight value corresponding to the quality index in the obtained initial weight values as a parameter to be adjusted in a contrast test to obtain a third group of weight value test objects; and the number of the first and second groups,
and taking the initial weight value corresponding to the preference value in the obtained initial weight values as a parameter to be adjusted in a comparison test to obtain a fourth group of weight value test objects.
Step G3: and performing a comparison test on each group of weight value test objects, and determining the weight values corresponding to the item description similarity, the behavior similarity, the quality index and the preference value according to the comparison test result.
Wherein, in one embodiment, step G3 may be specifically implemented as:
for each group of weighted value test objects, the following operations are performed:
step H1: taking a preset time length as a statistical period from a specified date; and taking the initial weight value of the group of weight value test objects as an experimental sample in a first statistical period, and counting a feedback value used for expressing a recommendation effect in the first statistical period, wherein the feedback value is any one of the following information: click rate, total click amount, purchase rate, total purchase amount, download rate, and total download amount.
Step H2: adjusting the parameters to be adjusted of the group of weighted value test objects according to a first preset rule; wherein the first predetermined rule is increasing or decreasing.
Step H3: and taking the adjusted group of weighted value test objects as an experimental sample of the next statistical period and counting the feedback value of the experimental sample in the next statistical period.
Step H4: for each statistical period except the first statistical period, if the feedback value of the statistical period is increased relative to the feedback value of the last statistical period, returning to execute the step of adjusting the parameter to be adjusted according to a first preset rule; if the feedback value of the statistical period is decreased relative to the feedback value of the previous statistical period, adjusting the parameters to be adjusted of the group of weighted value test objects according to a second preset rule, and returning to execute the steps of taking the adjusted group of weighted value test objects as the experimental sample of the next statistical period and counting the feedback value of the experimental sample in the next statistical period; if the first preset rule is increasing, the second preset rule is decreasing; if the first predetermined rule is decreasing, the second predetermined rule is increasing.
Step H5: and if the feedback values in a preset number of continuous counting periods meet a preset condition, selecting the adjusted group of weighted value test objects corresponding to the highest feedback values from the continuous counting periods as a group of preferred weighted values.
The preset condition refers to that the maximum difference value of the feedback values in a preset number of continuous statistical cycles is smaller than a preset difference value. That is, it is required that the feedback value is substantially constant for a preset number of consecutive statistical periods.
For example, the initial weight value may be randomly set by gaussian distribution N (0,1), and certainly, other methods for setting the initial weight value may also be used for the comparison test in specific implementation, which is not described herein again. If the initial weight values corresponding to the item description similarity, the behavior similarity, the quality index and the preference value are (2,3,3), the initial weight value is marked as W, as shown in the following table:
TABLE 3
| Test set \ W | W1 | W2 | W3 | W3 |
| First group | 2+(0.5) | 3 | 3 | 1 |
| Second group | 2 | 3+(0.5) | 3 | 1 |
| Third group | 2 | 3 | 3+(0.5) | 1 |
| Fourth group | 2 | 3 | 3 | 1+(0.5) |
Assuming that n days are a statistical period, determining feedback values of each day in the statistical period, and taking the average value of the feedback values of the n days as the feedback value of the statistical period. Then, comparing the feedback value of the previous statistical period, if the feedback value of the current statistical period rises, increasing the corresponding parameter value to be adjusted, for example, increasing by 0.5, and if the feedback value of the current statistical period falls, decreasing the corresponding parameter value to be adjusted, for example, decreasing by 0.5; and may be reset to 1 when reduced to 0.
In summary, the following steps: in the embodiment of the application, the article description similarity between each article in the articles to be recommended and the target article is calculated; calculating the behavior similarity of the behavior of each user in the user group on each item in the items to be recommended and the behavior of the user on the target item; calculating the quality index of each item in the items to be recommended; calculating the similarity between each item in the items to be recommended and the target item according to the item description similarity, the behavior similarity and the quality index; and selecting recommended articles from the articles to be recommended according to the similarity obtained by calculation. Therefore, when recommending articles for a user, multiple factors such as article description similarity among the articles, behavior similarity of the user to the articles, quality index of the articles and the like are comprehensively considered, so that the calculated similarity between each article in the articles to be recommended and the target article is more accurate, the actual requirements of the user can be better met, and the user experience is improved.
Example two
To further facilitate an understanding of the method of item recommendation provided herein, the examples of the present application further illustrate the method. As shown in fig. 2, the method comprises the following steps:
step 201: and calculating the item description similarity between each item in the items to be recommended and the target item.
Step 202: and calculating the behavior similarity of the behavior of each user in the user group on each item in the items to be recommended and the behavior of the user on the target item.
Step 203: and calculating the quality index of each item in the items to be recommended.
Step 204: determining the weight values corresponding to the item description similarity, the behavior similarity and the quality index.
Step 205: and calculating the similarity between each item in the items to be recommended and the target item in a weighted summation mode according to the determined weight values and the item description similarity, the behavior similarity and the quality index.
Step 206: and selecting recommended articles from the articles to be recommended according to the similarity obtained by calculation.
Step 207: and receiving a browsing request of the first user for the target item.
Step 208: and acquiring the personal information of the first user according to the browsing request.
Step 209: and calculating the preference value of the first user to each item in the selected recommended items according to the personal information of the first user.
Step 210: and acquiring a weight value corresponding to the preference value, and calculating the product of the preference value of the first user on each selected recommended item and the weight value of the preference value.
For example, if the weight value of the preference value is P, then for each recommended item, the product of the preference value of the first user for the recommended item and P is calculated.
Step 211: and calculating the product of each item in the selected recommended items and the sum of the behavior similarity of the item and the target item by the first user, and sequencing the selected recommended items according to the sum.
If Q recommended articles are selected, Q sum values are obtained, and sorting is carried out according to the sequence of the sum values from large to small.
Step 212: and sending the recommended articles of the L-th order to the first user.
In the technical scheme provided by the embodiment of the application, the item description similarity between each item in the items to be recommended and the target item is calculated; calculating the behavior similarity of the behavior of each user in the user group on each item in the items to be recommended and the behavior of the user on the target item; calculating the quality index of each item in the items to be recommended; calculating the similarity between each item in the items to be recommended and the target item according to the item description similarity, the behavior similarity and the quality index; and selecting recommended articles from the articles to be recommended according to the similarity obtained by calculation. Therefore, when recommending articles for a user, multiple factors such as article description similarity between the articles, behavior similarity of the user to the articles, quality index of the articles and the like are comprehensively considered, and the similarity between the two articles is considered from multiple aspects, so that the calculated similarity between each article in the articles to be recommended and the target article is more accurate.
Example three:
based on the same inventive concept, the embodiment of the application also provides an article recommendation device, and the article recommendation principle of the device is similar to that of the article recommendation method. For details, reference may be made to the contents of the above-mentioned methods, which are not described herein in detail.
As shown in fig. 3, which is a schematic structural diagram of the apparatus, the apparatus includes:
the item description similarity calculation module 301: the method is used for calculating the item description similarity between each item in the items to be recommended and the target item.
The behavior similarity calculation module 302: and the method is used for calculating the behavior similarity of the behavior of each user in the user group on each item in the items to be recommended and the behavior of the user on the target item.
Quality index calculation module 303: for calculating a quality index for each of the items to be recommended.
The similarity calculation module 304: the recommendation system is used for calculating the similarity between each item in the items to be recommended and the target item according to the item description similarity, the behavior similarity and the quality index.
The selection module 305: and the recommendation device is used for selecting recommended articles from the articles to be recommended according to the similarity obtained by calculation.
Wherein, in one embodiment, the apparatus further comprises:
the receiving module is used for receiving a browsing request of a first user for the target object;
the personal information acquisition module is used for acquiring the personal information of the first user according to the browsing request;
a preference value calculation module for calculating a preference value of the first user for each item of the selected recommended items according to the personal information of the first user;
the sorting module is used for sorting the selected recommended articles according to the preference values;
and the recommending module is used for recommending to the first user according to the sorted recommended articles.
Wherein, in one embodiment, the behavior comprises a purchasing behavior, a browsing behavior, a collecting behavior, a marking behavior, a downloading behavior.
In one embodiment, the item description similarity calculation module specifically includes:
the text description acquisition unit is used for acquiring the text description of each article in the articles to be recommended;
the word segmentation unit is used for segmenting the word description;
the word vector calculating unit is used for calculating the word vector of each word obtained after word segmentation;
and the article description similarity calculation unit is used for calculating the article description similarity between each article in the articles to be recommended and the target article according to the word vector of each word.
In an embodiment, the quality index calculating module is specifically configured to:
calculating a quality index for each of the items to be recommended according to at least one of: item popularity, item rating, number of item reviews.
In summary, in the apparatus for recommending an article provided in the embodiment of the present application, the article description similarity calculation module calculates the article description similarity between each article in the articles to be recommended and the target article; a behavior similarity calculation module calculates the behavior similarity of the behavior of each user in the user group on each item in the items to be recommended and the behavior of the user on the target item; the quality index calculation module calculates the quality index of each item in the items to be recommended; the similarity calculation module calculates the similarity between each item in the items to be recommended and the target item according to the item description similarity, the behavior similarity and the quality index; and the selection module selects recommended articles from the articles to be recommended according to the similarity obtained by calculation. Therefore, when recommending articles for a user, multiple factors such as article description similarity between the articles, behavior similarity of the user to the articles, quality index of the articles and the like are comprehensively considered, and the similarity between the two articles is considered from multiple aspects, so that the calculated similarity between each article in the articles to be recommended and the target article is more accurate.
Example four
An embodiment of the present application provides a non-volatile computer storage medium, where the computer storage medium stores computer-executable instructions, and the computer-executable instructions may execute the method for recommending an item in any method embodiment.
EXAMPLE five
Fig. 4 is a schematic hardware structure diagram of an electronic device for performing a method for recommending an item according to a fifth embodiment of the present application, where as shown in fig. 4, the electronic device includes:
one or more processors 410 and a memory 420, with one processor 410 being an example in fig. 4. The electronic device performing the method of item recommendation may further include: an input device 430 and an output device 440.
The processor 410, the memory 420, the input device 430, and the output device 440 may be connected by a bus or other means, such as the bus connection in fig. 4.
The memory 420 is a non-volatile computer-readable storage medium, and can be used for storing non-volatile software programs, non-volatile computer-executable programs, and modules, such as program instructions/modules corresponding to the method for recommending an item in the embodiment of the present application (for example, the item description similarity calculation module 301, the behavior similarity calculation module 302, the quality index calculation module 303, the similarity calculation module 304, and the selection module 305 shown in fig. 3). The processor 410 executes various functional applications of the server and data processing by executing nonvolatile software programs, instructions and modules stored in the memory 420, so as to implement the method for recommending articles according to the above method embodiment.
The memory 420 may include a storage program area and a storage data area, wherein the storage program area may store an operating system, an application program required for at least one function; the storage data area may store data created according to use of the device recommended by the item, and the like. Further, the memory 420 may include high speed random access memory, and may also include non-volatile memory, such as at least one magnetic disk storage device, flash memory device, or other non-volatile solid state storage device. In some embodiments, memory 420 may optionally include memory located remotely from processor 410, which may be connected to the item recommendation device via a network. Examples of such networks include, but are not limited to, the internet, intranets, local area networks, mobile communication networks, and combinations thereof.
The input device 430 may receive input numeric or character information and generate key signal inputs related to user settings and function control of the device for item recommendation. The output device 440 may include a display device such as a display screen.
The one or more modules are stored in the memory 420 and, when executed by the one or more processors 410, perform a method of item recommendation in any of the method embodiments described above.
The product can execute the method provided by the embodiment of the application, and has the corresponding functional modules and beneficial effects of the execution method. For technical details that are not described in detail in this embodiment, reference may be made to the methods provided in the embodiments of the present application.
As will be appreciated by one skilled in the art, embodiments of the present application may be provided as a method, apparatus (device), or computer program product. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.
The present application is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (devices) and computer program products according to embodiments of the application. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
While the preferred embodiments of the present application have been described, additional variations and modifications in those embodiments may occur to those skilled in the art once they learn of the basic inventive concepts. Therefore, it is intended that the appended claims be interpreted as including preferred embodiments and all alterations and modifications as fall within the scope of the application.
It will be apparent to those skilled in the art that various changes and modifications may be made in the present application without departing from the spirit and scope of the application. Thus, if such modifications and variations of the present application fall within the scope of the claims of the present application and their equivalents, the present application is intended to include such modifications and variations as well.
Claims (4)
1. A method for item recommendation, the method comprising:
calculating the item description similarity between each item in the items to be recommended and the target item;
calculating the behavior similarity of the behavior of each user in the user group on each item in the items to be recommended and the behavior of the user on the target item; wherein the behaviors comprise purchasing behavior, browsing behavior, collecting behavior, marking behavior and downloading behavior;
calculating the quality index of each item in the items to be recommended according to at least one item of the popularity, the rating grade and the number of the item comments;
calculating the similarity between each item in the items to be recommended and the target item according to the item description similarity, the behavior similarity and the quality index;
selecting recommended articles from the articles to be recommended according to the similarity obtained by calculation;
receiving a browsing request of a first user for the target item;
acquiring personal information of the first user according to the browsing request;
calculating a preference value of the first user for each item in the selected recommended items according to personal information of the first user;
sorting the selected recommended articles according to the preference values;
and recommending to the first user according to the sorted recommended articles.
2. The method according to claim 1, wherein the calculating of the item description similarity between each of the items to be recommended and the target item comprises:
acquiring the text description of each item in the items to be recommended;
performing word segmentation on the text description;
calculating a word vector of each word obtained after word segmentation;
and calculating the item description similarity between each item in the items to be recommended and a target item according to the word vector of each word.
3. An apparatus for item recommendation, the apparatus comprising:
the item description similarity calculation module is used for calculating the item description similarity between each item in the items to be recommended and the target item;
the behavior similarity calculation module is used for calculating the behavior similarity of the behavior of each user in the user group on each article in the articles to be recommended and the behavior of the user on the target article;
the quality index calculation module is used for calculating the quality index of each item in the items to be recommended;
the similarity calculation module is used for calculating the similarity between each item in the items to be recommended and the target item according to the item description similarity, the behavior similarity and the quality index;
the selection module is used for selecting recommended articles from the articles to be recommended according to the similarity obtained by calculation;
the receiving module is used for receiving a browsing request of a first user for the target object;
the personal information acquisition module is used for acquiring the personal information of the first user according to the browsing request;
a preference value calculation module for calculating a preference value of the first user for each item of the selected recommended items according to the personal information of the first user;
the sorting module is used for sorting the selected recommended articles according to the preference values;
and the recommending module is used for recommending to the first user according to the sorted recommended articles.
4. The apparatus according to claim 3, wherein the item description similarity calculation module specifically includes:
the text description acquisition unit is used for acquiring the text description of each article in the articles to be recommended;
the word segmentation unit is used for segmenting the word description;
the word vector calculating unit is used for calculating the word vector of each word obtained after word segmentation;
and the article description similarity calculation unit is used for calculating the article description similarity between each article in the articles to be recommended and the target article according to the word vector of each word.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611261973.6A CN106651542B (en) | 2016-12-31 | 2016-12-31 | Article recommendation method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611261973.6A CN106651542B (en) | 2016-12-31 | 2016-12-31 | Article recommendation method and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106651542A CN106651542A (en) | 2017-05-10 |
| CN106651542B true CN106651542B (en) | 2021-06-25 |
Family
ID=58838876
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201611261973.6A Active CN106651542B (en) | 2016-12-31 | 2016-12-31 | Article recommendation method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106651542B (en) |
Families Citing this family (21)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107169842A (en) * | 2017-05-31 | 2017-09-15 | 合肥亿迈杰软件有限公司 | A kind of electronic commerce data screening system based on commodity data |
| CN107230136A (en) * | 2017-05-31 | 2017-10-03 | 合肥亿迈杰软件有限公司 | A kind of shopping sequence method for pushing based on big data |
| CN107391692B (en) * | 2017-07-26 | 2023-04-07 | 腾讯科技（北京）有限公司 | Recommendation effect evaluation method and device |
| CN107679943A (en) * | 2017-09-27 | 2018-02-09 | 广州市万表科技股份有限公司 | Intelligent recommendation method and system |
| US11157985B2 (en) * | 2017-11-29 | 2021-10-26 | Ditto Technologies, Inc. | Recommendation system, method and computer program product based on a user's physical features |
| CN108171569B (en) * | 2017-12-11 | 2021-12-21 | 武汉纺织大学 | Feedback type intelligent clothing recommendation method and system |
| CN108109052A (en) * | 2017-12-29 | 2018-06-01 | 广州优视网络科技有限公司 | Article method for pushing, device and server |
| CN110110206B (en) * | 2018-01-17 | 2023-10-31 | 斑马智行网络(香港)有限公司 | Method, device, computing equipment and storage medium for mining and recommending relationships among articles |
| CN108600792B (en) * | 2018-04-02 | 2020-08-04 | 武汉斗鱼网络科技有限公司 | Similarity measurement method, device, equipment and storage medium |
| CN108765051B (en) * | 2018-04-16 | 2022-03-22 | 达而观信息科技（上海）有限公司 | Item recommendation method and device |
| CN108763318B (en) * | 2018-04-27 | 2022-04-19 | 达而观信息科技（上海）有限公司 | Item recommendation method and device |
| CN110473043A (en) * | 2018-05-11 | 2019-11-19 | 北京京东尚科信息技术有限公司 | A kind of item recommendation method and device based on user behavior |
| CN109033233B (en) * | 2018-06-29 | 2021-05-28 | 武汉斗鱼网络科技有限公司 | Live broadcast room recommendation method, storage medium, electronic device and system |
| CN109101563A (en) * | 2018-07-13 | 2018-12-28 | 东软集团股份有限公司 | A kind of object recommendation method, apparatus, medium and equipment |
| CN110738538A (en) * | 2018-07-18 | 2020-01-31 | 北京京东尚科信息技术有限公司 | Method and device for identifying similar articles |
| CN109102127B (en) * | 2018-08-31 | 2021-10-26 | 杭州贝购科技有限公司 | Commodity recommendation method and device |
| CN109754316B (en) * | 2018-12-29 | 2022-07-29 | 深圳Tcl新技术有限公司 | Product recommendation method, product recommendation system and storage medium |
| CN111666481B (en) * | 2019-03-05 | 2024-04-09 | 北京沃东天骏信息技术有限公司 | Data mining method, device, computer readable medium and electronic equipment |
| CN112907312A (en) * | 2019-12-04 | 2021-06-04 | 深圳云天励飞技术有限公司 | Article recommendation method and system for offline user and storage medium |
| CN112182397A (en) * | 2020-10-12 | 2021-01-05 | 北京自如信息科技有限公司 | Article recommendation method and device, electronic equipment and readable storage medium |
| CN113781147A (en) * | 2020-11-17 | 2021-12-10 | 北京沃东天骏信息技术有限公司 | Item recommendation method and device |
Family Cites Families (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103679489A (en) * | 2012-09-10 | 2014-03-26 | 腾讯科技（深圳）有限公司 | Item-based collaborative recommendation method and system |
| CN103514239B (en) * | 2012-11-26 | 2016-12-21 | Tcl美国研究所 | A kind of integrated user behavior and the recommendation method and system of item contents |
| CN103514304B (en) * | 2013-10-29 | 2017-01-18 | 海南大学 | Project recommendation method and device |
| US20150149295A1 (en) * | 2013-11-25 | 2015-05-28 | Kobo Incorporated | Method and system for generating recommendations based on media usage and purchase behavior |
| CN104765751B (en) * | 2014-01-07 | 2019-05-24 | 腾讯科技（深圳）有限公司 | Using recommended method and device |
| CN105608180A (en) * | 2015-12-22 | 2016-05-25 | 北京奇虎科技有限公司 | Application recommendation method and system |
-
2016
- 2016-12-31 CN CN201611261973.6A patent/CN106651542B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN106651542A (en) | 2017-05-10 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106651542B (en) | Article recommendation method and device | |
| CN107451894B (en) | Data processing method, device and computer readable storage medium | |
| CN108960719B (en) | Method and device for selecting products and computer readable storage medium | |
| CN106484777B (en) | Multimedia data processing method and device | |
| CN109543109B (en) | Recommendation algorithm integrating time window technology and scoring prediction model | |
| CN110008397B (en) | Recommendation model training method and device | |
| WO2018121700A1 (en) | Method and device for recommending application information based on installed application, terminal device, and storage medium | |
| WO2015034850A2 (en) | Feature selection for recommender systems | |
| CN107704485A (en) | A kind of position recommends method and computing device | |
| CN108021708B (en) | Content recommendation method and device and computer readable storage medium | |
| CN103761254A (en) | Method for matching and recommending service themes in various fields | |
| CN107423430B (en) | Data processing method, device and computer readable storage medium | |
| CN109903103B (en) | Method and device for recommending articles | |
| CN107423335B (en) | Negative sample selection method for single-class collaborative filtering problem | |
| CN108446297B (en) | Recommendation method and device and electronic equipment | |
| CN111798280B (en) | Multimedia information recommendation method, device and equipment and storage medium | |
| CN112529663A (en) | Commodity recommendation method and device, terminal equipment and storage medium | |
| CN110598120A (en) | Behavior data based financing recommendation method, device and equipment | |
| CN110647683A (en) | Information recommendation method and device | |
| CN111861605A (en) | Business object recommendation method | |
| CN112528164A (en) | User collaborative filtering recall method and device | |
| CN111680213B (en) | Information recommendation method, data processing method and device | |
| CN115982463A (en) | Resource recommendation method, device, equipment and storage medium | |
| US20200098031A1 (en) | Product recommending apparatus and non-transitory computer readable medium | |
| CN109063120B (en) | Collaborative filtering recommendation method and device based on clustering |
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 |