CN107943932B

CN107943932B - Item recommendation method, storage device and terminal

Info

Publication number: CN107943932B
Application number: CN201711176705.9A
Authority: CN
Inventors: 陶胜
Original assignee: Guangzhou Huya Information Technology Co Ltd
Current assignee: Guangzhou Huya Information Technology Co Ltd
Priority date: 2017-11-22
Filing date: 2017-11-22
Publication date: 2020-06-09
Anticipated expiration: 2037-11-22
Also published as: CN107943932A

Abstract

The invention provides a class recommendation method, a storage device and a terminal, wherein the method comprises the following steps: obtaining a category watching record of a target user in a preset time period, wherein the category watching record comprises a plurality of watched categories and corresponding watching duration; obtaining the preference degree of the target user to each category according to the watching duration corresponding to the plurality of categories; generating a first power set element according to a source article class set consisting of a plurality of article classes; generating a first identification list according to the preference degree of the categories contained in the first power set element; determining a target candidate set from a preset data set according to the first power set element and the first identification list; and selecting categories from the target candidate set and recommending the categories to the target user. The embodiment of the invention greatly improves the calculation efficiency of the category recommendation.

Description

Item recommendation method, storage device and terminal

Technical Field

The invention relates to the technical field of computers, in particular to a class recommendation method, a storage device and a terminal.

Background

Generally, the category refers to a category or a category of an object, and for example, in the field of game technology, the category is a name of a game, and the category is a category such as hero union, royal glory, and the like. The category information is complex and various, and users cannot obtain the information really needed by themselves when facing a large amount of information, so that the recommendation of categories to the users is particularly important. When the item classes are recommended to the users, the recommended item classes for all the users are the same without different treatment, and are found in an intelligent direction at present, and the item classes are recommended according to the own interests and hobbies of the users, namely a commonly-spoken personalized recommendation system. One important step is to push the categories which the user does not see to the user, and the categories are realized by collaborative filtering.

Collaborative filtering, a model for recommending items to users in data mining, includes collaborative filtering based on item similarity and collaborative filtering based on user similarity. Collaborative filtering based on article similarity based on the perspective of articles, more recommended to users are similar within the interest range of the users. From the perspective of users, collaborative filtering based on user similarity is performed, and in addition to satisfying the former function, it is also able to expand the interests of the users themselves, provide the users with multiple categories that may be interested in, and improve the probability of the users being retained by the application.

In the collaborative filtering based on the user similarity, a plurality of users most similar to the user are calculated for each user, and then categories which are not viewed by the user are selected from the categories which are viewed by the users and recommended to the user. The number of applied users is generally large, and some active users applied for one month have tens of millions, and selecting a plurality of users closest to the users involves comparing every two users in a user group, and actually making a Cartesian product, so that when the number of users is large, the amount of calculation of the Cartesian product is huge so that the calculation is almost impossible, and the category recommendation cannot be performed.

Disclosure of Invention

The invention provides a category recommendation method, a storage device and a terminal aiming at the defects of the existing mode, and aims to solve the problems that the calculation efficiency of category recommendation is low or even the calculation cannot be carried out in the prior art so as to improve the calculation efficiency of category recommendation.

According to a first aspect, an embodiment of the present invention provides a method for recommending categories, including the steps of:

obtaining a category watching record of a target user in a preset time period, wherein the category watching record comprises a plurality of watched categories and corresponding watching duration;

obtaining the preference degree of the target user to each category according to the watching duration corresponding to the plurality of categories;

generating a first power set element according to a source article class set consisting of a plurality of article classes;

generating a first identification list according to the preference degree of the categories contained in the first power set element;

determining a target candidate set from a preset data set according to the first power set element and the first identification list, wherein the data set is used for describing a candidate set corresponding to each power set element and the identification list thereof, and the candidate set comprises a category to be recommended;

and selecting categories from the target candidate set and recommending the categories to the target user.

In one embodiment, the determining the target candidate set from the preset data set includes:

acquiring a group corresponding to the target user, wherein the group is used for representing an idempotent element and an identification list thereof;

and determining the candidate set corresponding to the grouping as a target candidate set.

In an embodiment, before the obtaining of the group corresponding to the target user, the method further includes:

acquiring a category watching record of each user in a preset time period;

obtaining the preference degree of each user to each category according to the watching duration corresponding to a plurality of categories;

obtaining power set elements of each user according to a source item set consisting of a plurality of items;

generating an identification list of each power set element according to the preference degree of the categories contained in each power set element of each user;

deleting the categories contained in the power set elements from the source category set of each user to obtain a complementary set of each power set element;

grouping according to the power set elements and the identification list, and determining the grouping in each grouping neighborhood;

and obtaining a candidate set corresponding to each group according to the user and the complementary set of each group and the users and the complementary sets of the groups in the respective adjacent domains.

In one embodiment, the determining the packets in the respective packet neighborhoods comprises:

selecting a group from a set consisting of the groups as a current group;

acquiring a group of which the power set element in the set is the same as the power set element of the current group;

summing absolute values of differences between each identifier in the obtained identifier list of the group and the identifier at the corresponding position in the identifier list of the current group, and if the sum is less than or equal to a preset threshold value, determining that the group is located in the neighborhood of the current group;

and returning to the step of selecting one group from the set consisting of all the groups as the current group until the groups in the neighborhood of all the groups are determined.

In one embodiment, the candidate set corresponding to each group further includes a number of users corresponding to the category to be recommended, where the category to be recommended includes the category in the group complement and the category in the group complement in the neighborhood thereof, and the number of users is the sum of the number of people in the group who view the category to be recommended and the number of people in the group in the neighborhood thereof who view the category to be recommended.

In one embodiment, the selecting and recommending the categories from the target candidate set to the target user includes:

deleting the categories contained in the source category set of the target user from the target candidate set to obtain a pending candidate set;

obtaining a recommendation sequence of the categories in the undetermined candidate set according to the total preference degree of the first power set element corresponding to the undetermined candidate set and the number of users corresponding to the categories in the undetermined candidate set, wherein the total preference degree is the sum of the preference degrees of the categories in the power set element;

and selecting the categories from the undetermined candidate set according to the recommendation sequence of the categories in the undetermined candidate set and recommending the categories to the target user.

In one embodiment, the obtaining the recommended order of the categories in the pending candidate set includes:

ordering the candidate sets to be determined according to the sequence of the total preference degrees of the first power set elements from top to bottom;

sorting the categories in the candidate set to be determined according to the sequence of the number of users from top to bottom;

and obtaining a recommendation sequence of the categories in the pending candidate set according to the positions of the sorted categories, wherein the recommendation sequence is used for indicating that the categories closer to the front position have higher priority.

In one embodiment, the generating a first identification list according to the preference degree of the categories contained in the first power set element comprises: and generating a first identifier list of the first power set element according to the range to which the preference degree of the categories contained in the first power set element belongs and the identifiers corresponding to the preset ranges.

In one embodiment, the item viewing record further comprises a viewing date corresponding to a viewing duration;

the obtaining of the preference degree of the target user to each category according to the watching duration corresponding to the plurality of categories comprises:

summing the products of each watching duration and the respective set data weight in the same category to obtain the weighted watching duration of each category, wherein the data weight is increased along with the increase of the watching date corresponding to the watching duration;

and obtaining the preference degree of the target user to each category according to the ratio of the weighted watching duration of each category to the sum of the weighted watching durations of all the categories.

Embodiments of the present invention also provide, according to the second aspect, a storage device having stored thereon a computer program which, when executed by a processor, performs the steps of any of the methods described above.

Embodiments of the present invention also provide, according to a third aspect, a terminal, which includes a memory, a processor, and a computer program stored on the memory and executable on the processor, and when the processor executes the computer program, the processor implements the steps of any one of the above-mentioned methods.

According to the class recommendation method, the storage device and the terminal, because the number of classes watched by the user is generally in a controllable range, the first power set element is obtained by adopting a mode of performing power set multiplication on a source class set consisting of a plurality of classes, and then the target candidate set is found according to the first power set element and the first identification list thereof and the class recommendation is performed, so that the situation that Cartesian multiplication is directly performed on pairwise comparison of the users in the user group is effectively avoided, the calculation efficiency of class recommendation is greatly improved, and better personalized recommendation is realized.

Furthermore, the candidate set of each group comprises the data in the group and the data in the group in the neighborhood of the group, and when the data set formed by the candidate set of each group is searched for a target candidate set by using a first power set element obtained in a power set propagation mode and a first identification list thereof, the purest collaborative filtering algorithm based on the user similarity can be approximately restored, the time consumption is about 1 hour, the calculation efficiency of the class recommendation is further improved, and the accuracy reaches 72%.

Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention.

Drawings

The foregoing and/or additional aspects and advantages of the present invention will become apparent and readily appreciated from the following description of the embodiments, taken in conjunction with the accompanying drawings of which:

FIG. 1 is a flowchart illustrating a method for recommending classes according to an embodiment of the present invention;

fig. 2 is a schematic structural diagram of a terminal according to an embodiment of the present invention.

Detailed Description

Reference will now be made in detail to embodiments of the present invention, examples of which are illustrated in the accompanying drawings, wherein like or similar reference numerals refer to the same or similar elements or elements having the same or similar function throughout. The embodiments described below with reference to the drawings are illustrative only and should not be construed as limiting the invention.

As used herein, the singular forms "a", "an", "the" and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. It will be further understood that the terms "comprises" and/or "comprising," when used in this specification, specify the presence of stated features, integers, steps, operations, elements, and/or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and/or groups thereof. It will be understood that when an element is referred to as being "connected" or "coupled" to another element, it can be directly connected or coupled to the other element or intervening elements may also be present. Further, "connected" or "coupled" as used herein may include wirelessly connected or wirelessly coupled. As used herein, the term "and/or" includes all or any element and all combinations of one or more of the associated listed items.

It will be understood by those skilled in the art that, unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs. It will be further understood that terms, such as those defined in commonly used dictionaries, should be interpreted as having a meaning that is consistent with their meaning in the context of the prior art and will not be interpreted in an idealized or overly formal sense unless expressly so defined herein.

As used herein, a "terminal" includes both devices that are wireless signal receivers, devices that have only wireless signal receivers without transmit capability, and devices that have receive and transmit hardware, devices that have receive and transmit hardware capable of performing two-way communication over a two-way communication link, as will be understood by those skilled in the art. Such a device may include: a cellular or other communication device having a single line display or a multi-line display or a cellular or other communication device without a multi-line display; PCS (Personal Communications Service), which may combine voice, data processing, facsimile and/or data communication capabilities; a PDA (Personal Digital Assistant), which may include a radio frequency receiver, a pager, internet/intranet access, a web browser, a notepad, a calendar and/or a GPS (Global Positioning System) receiver; a conventional laptop and/or palmtop computer or other device having and/or including a radio frequency receiver. As used herein, a "terminal" may be portable, transportable, installed in a vehicle (aeronautical, maritime, and/or land-based), or situated and/or configured to operate locally and/or in a distributed fashion at any other location(s) on earth and/or in space. The "terminal" used herein may also be a communication terminal, a web-enabled terminal, a music/video playing terminal, such as a PDA, an MID (Mobile Internet Device) and/or a Mobile phone with music/video playing function, and may also be a smart tv, a set-top box, etc.

It is necessary to first make the following instructive description of the application environment of the present invention and its technical concept.

The item recommendation method, the storage device and the terminal provided by the invention can be applied to Hive. Hive is a data warehouse tool based on Hadoop, can map Structured data files into a database table, provides a simple SQL (Structured Query Language) Query function, and can convert SQL statements into tasks to run on Hadoop. Among them, Hadoop is a distributed system infrastructure developed by Apache foundation, and SQL is a special purpose programming language, a database query and programming language, used to access data and query, update, and manage database systems. It should be understood that the invention is not limited to use in hives, and that the invention can be used in any tool as conditions are appropriate.

The technical conception of the invention is as follows: the power set type multiplication and the Cartesian product in the field are used for approximately restoring the purest coordination filtering based on the user, so that the product screening can be accurately and quickly carried out, and better product personalized recommendation is realized.

The following describes in detail specific embodiments of the item recommendation method, the storage device, and the terminal according to the present invention with reference to the accompanying drawings.

As shown in fig. 1, in one embodiment, a method for recommending categories includes the steps of:

s110, obtaining a category watching record of a target user in a preset time period, wherein the category watching record comprises a plurality of watched categories and corresponding watching duration.

The target user is the user for whom the item class needs to be recommended. The preset time period can be set according to actual needs, and considering that when data is calculated, the latest complete data is only yesterday, because today is not finished yet, the preset time period is generally set to a time period before today, for example, the preset time period is the latest 21 days, the data of the latest 21 days is obtained to be used as an input data source, and subsequent calculation of category recommendation is performed.

The category is a category or category of the object viewed. Taking the field of live game broadcasting as an example, a user watches a main broadcast, the content of the main broadcast is a category, the category refers to the alias of the game, such as hero union, royal glory and the like, and generally, many main broadcasts are in live broadcast under each category. It should be understood that the categories may be otherwise defined and the invention is not limited thereto. The category viewing record generally includes a viewed category and a corresponding viewing duration, and optionally, the category viewing record may further include a viewing date and other fields corresponding to the viewing duration, which is not limited in the present invention.

In order to better understand the above category viewing records, a specific category viewing record is taken as an example for explanation. As shown in table 1, the item viewing record includes the user, the viewing duration, and the viewed item, and further includes an optional viewing date field and a main broadcast field, specifically, the item on the main broadcast [ ayyud 01] is glory for the royal, and the duration for the user [ user01] to view the glory for the main broadcast on the main broadcast royal at 4, 8 days 2017 is 1 hour.

TABLE 1 field description of item viewing records

S120, obtaining the preference degree of the target user to each category according to the watching duration corresponding to the plurality of categories.

The preference degree is used for describing the degree that a user likes a certain category. If only one category is obtained in step S110, in this step, only the preference degree of the target user for the category needs to be calculated according to the viewing duration of the category, and if multiple categories are obtained in step S110, in this step, for each category, the preference degree of the target user for each category needs to be determined according to the corresponding viewing duration.

S130, generating a first power set element according to a source class set composed of a plurality of classes.

The power set is a set, in which the input and output are set, and is a set family formed by all subsets (including full set and empty set) in the original set. When X is a set, | X | ═ k (the number of elements in the set X is k, the same applies below), and when the power set is generated, each element in X has two choices, and appears or does not appear, the number of sets in the power set of X is the k power of 2. For example, X { }, { a }, { B }, { C }, { a, B }, { a, C }, { B, C }, { a, B, C } }, where: is an empty set, with 8 sets in total.

For a target user, a set of the viewed categories is regarded as a source category set, an idempotent set is produced based on the source category set, and elements in the idempotent set (still a set) are regarded as idempotent elements. Since the power set element includes the empty set therein, the empty set in the generated power set element may be optionally deleted. If only one item is included in the set of source items, only one first power set element is generated. If the source item set comprises at least two items, a plurality of first power set elements are generated.

S140, generating a first identification list according to the preference degree of the categories contained in the first power set element.

The first idempotent element has one or more categories, and the preference degree corresponding to each category has been determined in step S120, so that the identifier list corresponding to the idempotent element can be generated according to the categories included in the idempotent element and the preference degrees of the categories. Each identifier in the identifier list identifies the preference degree of the corresponding position item in the one-to-one power set element, for example, if one power set element contains three items, the preference degrees of the three items are (a, B, C), and then the identifier list is generated to be (a1, B1, C1). In addition, the identification list may be a sequence number list, or may be a list in other forms, and the sequence number list may further be an equal part sequence number list or a non-equal part sequence number list, which is not limited in the present invention.

S150, determining a target candidate set from a preset data set according to the first power set element and the first identification list, wherein the data set is used for describing a candidate set corresponding to each power set element and the identification list thereof, and the candidate set comprises categories to be recommended.

The preset data set stores each power set element and the candidate set corresponding to the identifier list thereof, as shown in table 2, the record is recorded in the data set, the power set element, the identifier list and the candidate set in the record are in a one-to-one correspondence relationship, and the candidate set includes the category to be recommended. The same power set elements and identification lists as the first power set elements and the first identification list are found from the data set, and then the found candidate set is determined as a target candidate set.

TABLE 2A record in the data set

And S160, selecting categories from the target candidate set and recommending the categories to the target user.

The target candidate set may include some or all categories that have been viewed by the target user, so that the categories in the source category set of the target user need to be deleted, and then a preset number of categories are selected according to the set recommendation rule and recommended to the target user.

The following describes each step in detail with reference to examples.

In step S120, there are many ways to determine the preference degree of the target user for each category, and the following description is made in conjunction with two specific embodiments.

Considering that the longer the viewing time of a certain category is, the more preferred the user is for that category, in one embodiment, the preference level of each category may be determined according to the following formula:

preference degree (the user's category) viewing duration/sum (all viewing durations of the user)

The weighting is needed considering that data closer to today is more reflective of the user's current interests, e.g. yesterday's data is more reflective of the user's current interests than the previous 21 st day's data. Therefore, in an embodiment, the obtaining, according to the viewing durations corresponding to a plurality of categories, the preference degrees of the target user for the respective categories includes:

s1201, products of all the watching durations and the set data weights in the same category are summed to obtain the weighted watching durations of all the categories, wherein the data weights are increased along with the increase of the watching dates corresponding to the watching durations.

The data weight may be calculated in a variety of ways, for example, in one embodiment, the data weight may be calculated by the following formula:

the weight of the data on the previous ith day is (31-i +1)/31i is 1

The ith previous day is the number of days from the current date. According to the above formula, the weight of the data of the previous 21 st day is 11/31, the weight of the data of the previous 20 th day is 12/31, the numerator of the weight of every other day is increased by 1, and so on, and the weight of the previous 1 st day (i.e. yesterday) is 31/31.

It should be understood that the present invention is not limited to the above-mentioned manner of calculating the data weight, and the user may also calculate the data weight in other manners as long as the trend of the data weight is increased as the viewing date corresponding to the viewing duration becomes larger.

Next, based on the item viewing record obtained in step S110, two dimensions of date and anchor need to be removed, and only the following are left: the user, the category, the weighted watching time length and the preference degree, the primary key is the user and the category, and the calculation formula of the weighted watching time length is as follows:

weighted viewing duration sum (in the same category as the user, the data weight of the previous ith day sum the viewing duration of the previous ith day, i 1.., n)

S1202, obtaining the preference degree of the target user to each category according to the ratio of the weighted watching duration of each category to the sum of the weighted watching durations of all the categories.

For each category, the user's preference for it is determined using the following equation:

preference degree (the user's category) weighted viewing duration/sum (weighted viewing duration of all categories of the user)

For any one user, the sum of his preference levels equals 100%, which is then processed using these 3 fields: user, category, preference level, weighted viewing duration exist only as a middle field. In addition, for the convenience of subsequent calculation, optionally, the unit positioning% of the preference degree is rounded to be a positive integer. As shown in table 3, a specific example of the field information obtained after the above steps are performed on table 1 is shown, and at this time, only three fields are included: user, category, and level of preference.

TABLE 3 prepared data records

Field(s)	Type (B)	Examples of the invention
			User' s	String	user01
Articles and the like	String	Hero alliance
			Degree of preference	int	20 in units of%

After the data is prepared, there is preference degree data of the user under the category he or she watches in step S130 and step S140. Performing power set type reproduction on a source product set consisting of the products watched by the target user to generate each power set element, and then generating a corresponding identification list according to the preference degree of the products contained in each power set element.

There are many ways to generate the identification list, for example, in one embodiment, the generating a first identification list according to the preference degree of the categories included in the first power set element includes: and generating a first identifier list of the first power set element according to the range to which the preference degree of the categories contained in the first power set element belongs and the identifiers corresponding to the preset ranges.

The value range of the preference degree is 0-100, wherein the value range is mathematical noun, and the value range changed by variable change in the classical definition of the function is called the value range of the function. The value range is divided into various ranges (equal or unequal), and the unique corresponding identification of each range is set, so that the identification corresponding to each class can be determined according to the range of each class, and then the identification list of the idempotent element can be obtained according to the class contained by the idempotent element and the identification of the class.

To illustrate by a specific example, the value range of the preference degree is 0-100, 15 equal parts are carried out, and the serial number of each equal part is a positive integer from 1 to 15 and is marked as the serial number of the equal part. Then, the number of the parts of preference degree 50 is 7, and the number of the parts of preference degree 70 is 10. If a power set element includes two categories, the first category has a preference of 50 and the second category has a preference of 70, then the list of equal part numbers for the power set element is (7, 10).

In addition, in consideration of recommendation of subsequent varieties and generation of a data set, it is further required to obtain a total preference degree and a complementary set of the power set element, where the total preference degree is a sum of preference degrees of each variety in the power set element, and the complementary set is a set obtained by subtracting the power set element from the source variety set, for example, a source variety set is { variety a, variety B, and variety C }, and one power set element is { variety a }, then the complementary set of the power set element is { variety B, variety C }.

The power set has data of how many rows it has to propagate, the fields of each row of data are shown in table 4, the primary key of which is user, power set element:

TABLE 4 intermediate data records after power set derivation

The core idea of deriving based on the power set is that if two users are very similar, the two users are necessarily very similar on a common set consisting of a plurality of categories, and the category set is necessarily an element of the power set of the source category sets of the two users, so that power set type multiplication is performed, and then the power set elements are grouped, so that the two users are bound to establish a relationship. Although the number of the power set is exponentially increased, on one hand, the number of the categories watched by most users does not exceed 15 and is within a controllable range, and on the other hand, if the data of the categories watched by the users are many, the category recommendation can be performed by taking a plurality of power set elements with the highest [ total preference degree of the power set elements ].

In step S150, in one embodiment, the determining the target candidate set from the preset data set includes: acquiring a group corresponding to the target user, wherein the group is used for representing an idempotent element and an identification list thereof; and determining the candidate set corresponding to the grouping as a target candidate set. And obtaining a first power set element and a first identification list of the target user, wherein each group is represented by the corresponding power set element and the identification list thereof, so that the group with the power set element same as the first power set element and the identification list same as the first identification list is searched, and the candidate set corresponding to the group is the target candidate set.

Therefore, in an embodiment, before obtaining the group corresponding to the target user, the method further includes:

and S040, obtaining the item watching records of each user in a preset time period.

The individual users may include the target user or may be other users than the target user. The preset time period can be set according to actual needs, and considering that when data is calculated, the latest complete data is only yesterday, because today is not finished yet, the preset time period is generally set to a time period before today, for example, the preset time period is the latest 21 days, the data of the latest 21 days is obtained to be used as an input data source, and subsequent calculation of category recommendation is performed.

The category is a category or category of the object viewed. Taking the field of live game broadcasting as an example, a user watches a main broadcast, the content of the main broadcast is a category, the category refers to the alias of the game, such as hero union, royal glory and the like, and generally, many main broadcasts are in live broadcast under each category. It should be understood that the categories may be otherwise defined and the invention is not limited thereto. The category viewing record generally includes a category to be viewed and a corresponding viewing duration, and optionally, the category viewing record may further include a viewing date and other fields corresponding to the viewing duration, which may be specifically referred to in table 1, which is not limited by the present invention.

And S050, obtaining the preference degree of each user to each category according to the watching duration corresponding to the plurality of categories.

There are many ways to determine the preference of each user for each category, and two specific embodiments are described below.

Considering that the longer the viewing time of a certain category is, the more preferred the user is to the category, in one embodiment, the preference degree of each user for each category may be determined according to the following formula:

The weighting is needed considering that data closer to today is more reflective of the user's current interests, e.g. yesterday's data is more reflective of the user's current interests than the previous 21 st day's data. Therefore, in an embodiment, the obtaining, according to the viewing durations corresponding to the multiple categories, preference degrees of the respective users for the respective categories includes:

s0501, summing products of each watching time length and each set data weight in the same category to obtain weighted watching time length of each category of each user, wherein the data weight is increased along with the increase of watching date corresponding to the watching time length.

the weight of the data on the previous ith day is (31-i +1)/31i is 1

Next, based on the item class viewing record acquired in step S040, two dimensions of date and anchor need to be removed, leaving only: the user, the category, the weighted watching time length and the preference degree, the primary key is the user and the category, and the calculation formula of the weighted watching time length is as follows:

And S0502, obtaining preference degrees of the users to the categories according to the ratio of the weighted watching duration of the categories to the sum of the weighted watching durations of all the categories.

For any one user, the sum of his preference levels equals 100%, which is then processed using these 3 fields: user, category, preference level, weighted viewing duration exist only as a middle field. In addition, for the convenience of subsequent calculation, optionally, the unit positioning% of the preference degree is rounded to be a positive integer.

S060, according to a source class set composed of a plurality of classes, obtaining power set elements of each user.

And respectively performing power set type reproduction on a source item set consisting of the items watched by each user to generate power set elements of each user.

And S070, generating an identification list of each power set element according to the preference degree of the categories of the power set elements of each user.

For example, in an embodiment, the generating the identification list of each power set element according to the preference degree of the categories included in each power set element of each user includes: and generating an identification list of each power set element according to the range to which the preference degree of the categories contained in each power set element of each user belongs and the preset identification corresponding to each range.

The value range of the preference degree is 0-100, the value range is divided into various ranges (equal or unequal), and the unique corresponding identifier of each range is set, so that the identifier corresponding to each class can be determined according to the range of each class, and the identifier list of the power set element can be obtained according to the classes contained in the power set element and the identifiers of the classes.

S080, deleting the categories contained in the power set elements from the source category set of each user to obtain the complement of each power set element.

For a certain power set element of a certain user, the complement set corresponding to the power set element can be obtained by subtracting the item class contained in the power set element from the source item class set of the user. By analogy, the complement of each power set element of each user can be obtained.

The power set has how many elements to reproduce the data of how many rows, and the fields of each row of data are shown in table 4, and the primary key is the user, power set element.

S090, grouping is carried out according to the power set elements and the identification list, and grouping in each grouping neighborhood is determined.

This and subsequent steps are essentially cartesian products within the neighborhood. Grouping is carried out by taking the power set element and the identification list (namely, taking the 2 fields as main keys to carry out group by operation), and the data with the same power set element and the same identification list are divided into the same group.

Neighborhood: mathematical terms, the distance to a known point is no greater than the set of all points for which a positive number is known. There are many ways to determine the packets within each packet neighborhood, for example, in one embodiment, the determining the packets within each packet neighborhood comprises:

s0901, selecting one group from the set consisting of the respective groups as a current group.

One packet is selected from the respective packets as a current packet currentLine.

S0902, the grouping of the power set elements in the set is obtained, wherein the power set elements are the same as the power set elements of the current grouping.

And if the power set element of one group sLine is the same as the power set element of the current group currentLine, acquiring the data of the group sLine.

S0903, summing absolute values of differences between each identifier in the obtained identifier list of the group and the identifier at the corresponding position in the identifier list of the current group, and determining that the group is located in the neighborhood of the current group if the sum is less than or equal to a preset threshold value.

Power set element, identification List, the data in these 2 fields (List type) are in one-to-one correspondence, their lengths are the same, let the length of the field be len. When judging whether the sLine falls in the currentLine neighborhood, a threshold needs to be specified, optionally, the type is a positive integer, and the conditions are as follows:

sum (abs (splane. logo list) get (i) -currentline. logo list get (i)), i ═ 1, 2.., len) < ═ threshold

Get (i) is the ith data of the obtained tag list, abs is the calculated absolute value, and sum is the sum of these len absolute values.

By performing the above judgment on each sPLINE, the packets in all neighborhoods of currentLine can be obtained.

S0904, the step of selecting one group from the set consisting of all groups as the current group is executed until the groups in all the group neighborhoods are determined.

And reselecting a packet as the current packet, and sequentially executing the steps to obtain the packets in the neighborhood of all the packets.

S100, obtaining a candidate set corresponding to each group according to the users and the complementary sets of each group and the users and the complementary sets of the groups in the adjacent regions.

Because each group is divided according to the power set element and the identification list, the candidate set corresponding to each group is the candidate set corresponding to each power set element and the identification list thereof, and the data set can be obtained through the steps.

In a specific implementation, for the grouped users and complementary sets, each (deduplicated) class in the complementary sets and the number of people who have viewed the class (among the grouped users) can be calculated and stored by using a field [ candidate set ], the type of which is Map < String, Integer ], to generate the data shown in table 5, the primary key is a power set element and an identification list, and then, a field [ extended candidate set ] is calculated, and the [ extended candidate set ] is also a candidate set of each group.

TABLE 5 intermediate data records after power set derivation

The first three fields in the table may be calculated before the field [ extended candidate set ] is calculated. The calculation rule for the field [ extended candidate set ] is:

step 1, recording the current record of the table (the set of each group) as currentLine, recording the record with the same [ power set element ] as currentLine in the table as samePowerLine.1, wherein N is the number of records;

step 2, firstly, loading currentline candidate set into currentline expanded candidate set;

step 3, traversing each record of the samePowerLine series, recording as sLine, and if the sLine falls into the neighborhood of currentLine, loading the sLine candidate set into a currentLine extended candidate set (if the records with the same key, adding the value); after traversal, the computational logic for field [ extended candidate set ] is complete.

It should be understood that the present invention is not limited to the above-described embodiments, as long as an extended candidate set of the present packet is obtained from the candidate set of the present packet and the candidate set of the packets in the neighborhood.

When searching according to the power set element and the identification list of the target user, optionally, Join operation is performed on the data (such as table 4) and the data set (such as table 5) of the target user, the associated field is the power set element and the identification list, and the result is shown in table 6. For each line record in the data set, denoted as line231, if the target user's power set element, identification list, and the corresponding field of line231 are the same, then the extended candidate set of line231 is determined to be the target candidate set.

TABLE 6 intermediate data records after power set derivation

In step S160, as shown in table 6, the extended candidate set may include some or all of the categories that the target user has already viewed, so that the categories in the target user' S source category set need to be deleted therein, and the result is stored with the field [ pending candidate set ]. In addition, rules for sorting the categories need to be set. Thus, in one embodiment, the selecting and recommending an item from a target candidate set to the target user comprises:

s1601, deleting the categories contained in the source category set of the target user from the target candidate set to obtain a pending candidate set.

Because there may be a plurality of target user power set elements, there may be a plurality of target candidate sets obtained, and for each target candidate set, the categories included in the target user source category set are deleted, so that the pending candidate set corresponding to each target candidate set can be obtained.

S1602, obtaining a recommendation sequence of the categories in the undetermined candidate set according to the total preference degree of the first power set element corresponding to the undetermined candidate set and the number of users corresponding to the categories in the undetermined candidate set, wherein the total preference degree is the sum of the preference degrees of the categories in the power set element.

The recommendation sequence of the categories is determined by two factors, one is the total preference degree of the first power set element, the higher the field is, the more comprehensively the corresponding power set element can reflect the interest and hobbies of the target user, and the higher the probability that the corresponding [ undetermined candidate set ] is liked and watched by the target user is; the other is the height of the number of users (i.e., value) corresponding to the category, and the value represents how many users viewed the key (category), so the category with the high value is selected preferentially.

There are various implementations of determining the recommended order of the categories according to the above two factors, and the following description will be made with reference to two embodiments.

s1602a, sorting the candidate sets to be determined according to the order of the total preference degree of the first power set elements from top to bottom.

As shown in table 6, the corresponding multiple pending candidate sets are ordered from top to bottom in the order of the field [ total preference degree of power set element ] from high to low.

S1602b, the categories in the candidate set are sorted in the order of the number of users from top to bottom.

As shown in table 6, for each pending candidate set, the internal multiple categories also need to be sorted from top to bottom according to the order of value.

S1602c, obtaining a recommendation sequence of the to-be-determined candidate concentrated categories according to the positions of the sorted categories, wherein the recommendation sequence is used for indicating that the categories closer to the front position have higher priority.

The above is done in two orders:

1 st order: ordering field [ total preference degree of power set element ], because the higher the field is, the more comprehensively the corresponding power set element can reflect the interest and hobbies of the user, and the higher the probability that the corresponding [ pending candidate set ] is liked and watched by the user is;

2 nd ordering: sort field [ pending candidate set ], in which field value represents how many users viewed a key (item class), so the item class with the value high is preferably chosen.

Then after sorting, the position of a certain category is more advanced, and the priority is higher to recommend to the target user.

For a specific example, it is assumed that after the above sorting, the positions of the categories are as follows:

< type 1, type 2, type 3>, < type 4, type 5, type 6>

Wherein, item class 1, item class 2, item class 3> represents a pending candidate set, item class 4, item class 5, item class 6> represents another pending candidate set, and then the item class recommendation sequence is from top to bottom: item 1, item 2, item 3, item 4, item 5, item 6.

In another embodiment, the recommended order of the categories may also be determined according to the following formula:

total preference degree of first power set element and number of users corresponding to item class

Assuming that the total preference degree of the first power set element corresponding to a category is 20, and the number of users corresponding to the category is 2, the recommendation order (recommendation degree) of the category is 40, so as to calculate the recommendation order of each category in the pending candidate set, and then it can be determined which categories are preferentially recommended according to the size of the recommendation order. Generally, the higher the recommendation order value, the higher the priority of the corresponding item recommendation.

S1603, selecting the categories from the to-be-determined candidate set according to the recommendation sequence of the categories in the to-be-determined candidate set and recommending the categories to the target user.

For the recommendation sequence determined by the sorting mode, all data of a target user are obtained, sorting is carried out according to the field [ the total preference degree of the power set element ] from high to low, the field [ the candidate set to be determined ] is sequentially obtained, and the grades are selected from front to back in the field until the grades are selected, so that the recommendation of the grades is completed.

And for the recommendation sequence determined by the product mode, acquiring all data of the target user, and sequentially selecting the item classes from top to bottom according to the recommendation sequence until the item classes are selected enough, thereby completing the recommendation of the item classes.

If under certain circumstances, the user cannot select the popular item which is not watched by the target user.

The invention also proposes a storage device on which a computer program is stored which, when being executed by a processor, carries out the steps of any one of the methods described above. The storage device includes, but is not limited to, any type of disk including floppy disks, hard disks, optical disks, CD-ROMs, and magneto-optical disks, ROMs (Read-Only memories), RAMs (Random access memories), EPROMs (EraSable Programmable Read-Only memories), EEPROMs (Electrically EraSable Programmable Read-Only memories), flash memories, magnetic cards, or optical cards. That is, a storage device includes any medium that stores or transmits information in a form readable by a device (e.g., a computer). Which may be a read-only memory, magnetic or optical disk, or the like.

The invention also proposes a terminal comprising a memory, a processor and a computer program stored on the memory and executable on the processor, the processor implementing the steps of any of the methods described above when executing the program.

Fig. 2 is a block diagram of a part of the terminal according to the present invention, and for convenience of description, only the part related to the embodiment of the present invention is shown. The terminal can be a terminal device including a mobile phone, a tablet computer, a notebook computer, a desktop computer and the like which can watch videos, listen to FM or music and watch news or novels. The following takes a desktop computer as an example:

referring to fig. 2, the desktop computer includes a processor, a memory, an input unit, a display unit, and the like. Those skilled in the art will appreciate that the desktop configuration shown in FIG. 2 is not intended to be limiting of all desktop computers, and may include more or less components than those shown, or some components in combination. The memory may be used to store a computer program and various functional modules, and the processor may execute various functional applications and data processing of the desktop computer by operating the computer program stored in the memory. The memory may mainly include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application program required by at least one function (such as a category recommendation function), and the like; the storage data area may store data created from use of the desktop computer (such as a genre viewing record, etc.), and the like. Further, the memory 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 volatile solid state storage device.

The input unit can be used for receiving the item viewing record input by the user and generating signal input related to the user setting and function control of the desktop computer. Specifically, the input unit may include a touch panel and other input devices. The touch panel can collect touch operations of a user on or near the touch panel (for example, operations of the user on or near the touch panel by using any suitable object or accessory such as a finger, a stylus and the like) and drive the corresponding connecting device according to a preset program; other input devices may include, but are not limited to, one or more of a physical keyboard, function keys (e.g., play control keys, switch keys, etc.), a trackball, a mouse, a joystick, and the like. The display unit may be used to display information input by a user or information provided to the user and various menus of the desktop computer. The display unit may take the form of a liquid crystal display, an organic light emitting diode, or the like. The processor is a control center of the desktop computer, connects various parts of the whole computer by using various interfaces and lines, and executes various functions and processes data by operating or executing software programs and/or modules stored in the memory and calling data stored in the memory.

In order to illustrate the technical effects of the present invention, a specific example is given below.

The date is 5 months and 16 days as the critical point

The user group trains based on the data of 21 days, wherein the total viewing time length on a certain category is 5min in 2017-04-26-05-16 (21 days in total), and a new category is viewed on 17 days in 5 months.

Definition of new class, new class watched by user on day 5, month 17, is that the watching duration of day 5, month 17 > is 5min and (in 2017-04-26 to 05-16: the class has not been watched or the total watching duration of the class in the time period <5min)

Aiming at data, a power set multiplication + field is used for making a Cartesian product, the purest coordination filtering based on users is approximately restored, a recommendation list of 15 new categories is generated for each user of a user group, and then a test numerator, a test denominator and a model effect are calculated:

the test denominator is count (user, new class viewed by user on day 5, 17)

The test molecule is count (user, new class viewed by user on day 5, 17 and inside the recommendation list)

Model effect test numerator/test denominator 442845/615236%

It should be understood that, although the steps in the flowcharts of the figures are shown in order as indicated by the arrows, the steps are not necessarily performed in order as indicated by the arrows. The steps are not performed in the exact order shown and may be performed in other orders unless explicitly stated herein. Moreover, at least a portion of the steps in the flow chart of the figure may include multiple sub-steps or multiple stages, which are not necessarily performed at the same time, but may be performed at different times, which are not necessarily performed in sequence, but may be performed alternately or alternately with other steps or at least a portion of the sub-steps or stages of other steps.

The foregoing is only a partial embodiment of the present invention, and it should be noted that, for those skilled in the art, various modifications and decorations can be made without departing from the principle of the present invention, and these modifications and decorations should also be regarded as the protection scope of the present invention.

Claims

1. A method for recommending categories, comprising the steps of:

determining a target candidate set from a preset data set according to the first power set element and the first identification list, wherein the data set is used for describing a candidate set corresponding to each power set element and the identification list thereof, and the candidate set comprises a category to be recommended; the candidate set is obtained according to the item class which is contained in the first power set element and deleted from the source item class set of the target user, and the complement set of the first power set element is obtained according to the target user, the complement set and grouped users and complement sets in the neighborhood of the target user;

2. The item recommendation method according to claim 1, wherein said determining a target candidate set from a preset data set comprises:

3. The item recommendation method according to claim 2, wherein before the obtaining of the group corresponding to the target user, the method further comprises:

acquiring a category watching record of each user in a preset time period;

4. The item recommendation method of claim 3, wherein said determining the groupings within each grouping neighborhood comprises:

selecting a group from a set consisting of the groups as a current group;

5. The item recommendation method of claim 3, wherein the candidate set corresponding to each group further comprises a number of users corresponding to the item to be recommended, wherein the item to be recommended comprises the item in the supplementary group and the item in the supplementary group in the neighborhood thereof, and the number of users is the sum of the number of people in the group who view the item to be recommended and the number of people in the neighborhood thereof who view the recommended item.

6. The item recommendation method of claim 5, wherein said selecting and recommending an item from a target candidate set to said target user comprises:

7. The item recommendation method according to claim 6, wherein said obtaining a recommendation order of the items in the pending candidate set comprises:

8. The item recommendation method according to any one of claims 1 to 7,

generating a first identification list according to the preference degree of the categories contained in the first power set element, wherein the first identification list comprises: generating a first identifier list of the first power set element according to the range to which the preference degree of the categories contained in the first power set element belongs and the identifiers corresponding to the preset ranges;

and/or the presence of a gas in the gas,

the item watching record also comprises a watching date corresponding to the watching duration;

9. A storage device having a computer program stored thereon, which program, when being executed by a processor, is adapted to carry out the steps of the method of any one of claims 1 to 8.

10. A terminal comprising a memory, a processor and a computer program stored on the memory and executable on the processor, characterized in that the steps of the method according to any of claims 1 to 8 are implemented when the processor executes the program.