CN112950328A - Combined object recommendation method, device, system and storage medium - Google Patents
Combined object recommendation method, device, system and storage medium Download PDFInfo
- Publication number
- CN112950328A CN112950328A CN202110315405.4A CN202110315405A CN112950328A CN 112950328 A CN112950328 A CN 112950328A CN 202110315405 A CN202110315405 A CN 202110315405A CN 112950328 A CN112950328 A CN 112950328A
- Authority
- CN
- China
- Prior art keywords
- user
- recommendation
- recommended
- type
- object type
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 44
- 238000012549 training Methods 0.000 claims abstract description 51
- 230000009471 action Effects 0.000 claims description 14
- 238000012163 sequencing technique Methods 0.000 abstract description 3
- 230000006399 behavior Effects 0.000 description 60
- 238000012216 screening Methods 0.000 description 26
- 239000013598 vector Substances 0.000 description 25
- 239000000047 product Substances 0.000 description 16
- 238000004364 calculation method Methods 0.000 description 8
- 230000000694 effects Effects 0.000 description 8
- 239000000463 material Substances 0.000 description 5
- 238000010586 diagram Methods 0.000 description 4
- 238000012545 processing Methods 0.000 description 4
- 238000005057 refrigeration Methods 0.000 description 4
- 230000001172 regenerating effect Effects 0.000 description 4
- 238000010801 machine learning Methods 0.000 description 3
- 230000008569 process Effects 0.000 description 3
- 230000000386 athletic effect Effects 0.000 description 2
- 239000007795 chemical reaction product Substances 0.000 description 2
- 238000007477 logistic regression Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000005192 partition Methods 0.000 description 2
- 230000032683 aging Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000013145 classification model Methods 0.000 description 1
- 238000001816 cooling Methods 0.000 description 1
- 238000003066 decision tree Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 239000010985 leather Substances 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 238000000638 solvent extraction Methods 0.000 description 1
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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/953—Querying, e.g. by the use of web search engines
- G06F16/9535—Search customisation based on user profiles and personalisation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Business, Economics & Management (AREA)
- Databases & Information Systems (AREA)
- General Engineering & Computer Science (AREA)
- Bioinformatics & Computational Biology (AREA)
- Accounting & Taxation (AREA)
- Evolutionary Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Computation (AREA)
- Finance (AREA)
- Development Economics (AREA)
- Economics (AREA)
- Marketing (AREA)
- Strategic Management (AREA)
- General Business, Economics & Management (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The present disclosure relates to a combined object recommendation method, apparatus, system, and storage medium. In at least one embodiment of the present disclosure, when a combined object is recommended, a corresponding recommendation model is established for each object type by distinguishing different object types, and then, for a recommended user, a recommendation ranking list of objects in each object type is obtained by using the recommendation model corresponding to each object type. Therefore, different objects of the same object type have the same attribute characteristics, and the condition that the characteristic value is null does not exist, so that the training characteristics of the recommendation model are changed from sparse to dense, and the training convergence of the recommendation model is accelerated. In addition, redundant features are removed, noise interference is reduced, and the recommendation sequencing result of the recommendation model is more accurate. In addition, the objects are selected from the recommendation ranking lists corresponding to the object types, and different recommendation ranking lists are fused into a combined object recommendation result, so that the effectiveness and diversity of object recommendation are improved.
Description
Technical Field
The embodiment of the disclosure relates to the technical field of machine learning, in particular to a combined object recommendation method, device, system and storage medium.
Background
With the aging of the machine learning algorithm in industrial application, more and more companies begin to build product recommendation systems belonging to the company, so that the requirements of different users on product personalized recommendation are met, the users can obtain favorite products more conveniently, the products can obtain higher click rate in the exposure display stage, the user activity of the client is improved, and the product purchase rate can be further improved.
Applying machine learning in a product recommendation system requires collecting attribute features of the recommended product as part of the training features of the model. When different types of products are recommended (namely, combined product recommendation), because different types of products have different characteristics, the attribute characteristics of a plurality of types of products are directly spliced to be used as training characteristics.
But when a product does not have the characteristics of the rest of the products, the characteristic value is null. The fact that the value of the multidimensional characteristic dimension is empty can lead to the fact that the characteristic vector of the product becomes sparse, so that the calculated amount and the storage amount of the model can be increased, some unnecessary noise can be introduced, and the effect of the model is reduced. Therefore, the traditional single model recommendation method has poor effect in recommending the combined product.
Disclosure of Invention
In order to solve at least one problem of the prior art, at least one embodiment of the present disclosure provides a combined object recommendation method, apparatus, system, and storage medium.
In a first aspect, an embodiment of the present disclosure provides a combined object recommendation method, where a combined object includes multiple objects, and the multiple objects belong to at least two object types, the method includes:
establishing a corresponding recommendation model for each object type;
and aiming at a recommended user, respectively obtaining a recommended ordered list of objects in each object type by using a recommendation model corresponding to each object type, and selecting the objects from the recommended ordered list corresponding to each object type to obtain a combined object recommendation result.
In some embodiments, prior to establishing the corresponding recommendation model for each object type, the method further comprises:
acquiring one or more attribute characteristics of each object;
all objects with the same quantity of attribute features and the same definition of the attribute features are divided into the same object type.
In some embodiments, establishing a corresponding recommendation model for each object type comprises:
acquiring a user behavior data set, wherein each piece of user behavior data is data describing a specific behavior of a user on an object;
dividing the user behavior data set into data sets corresponding to different object types according to the related different object types;
constructing a model training sample of each object type based on data sets of different object types;
and training a preset model based on the model training sample of each object type to obtain a recommended model corresponding to each object type.
In some embodiments, each piece of user behavior data comprises: user ID, object ID, behavior occurrence time and exposure object set;
the model training samples include: user ID, object ID, action occurrence time, tag.
In some embodiments, for a recommended user, obtaining a recommended ordered list of objects in each object type by using a recommendation model corresponding to each object type includes:
outputting a matching score of the object ID and the recommended user ID in each object type by using a recommendation model corresponding to each object type;
and sorting the matching scores of the object IDs in the object types and the recommended user IDs from high to low to obtain a recommended sorted list of the objects in the object types.
In some embodiments, selecting an object from the recommendation ranking list corresponding to each object type to obtain a combined object recommendation result includes:
acquiring a behavior data set of a recommended user;
determining the user type of the recommended user based on the behavior data set of the recommended user;
and selecting the objects from the recommendation ranking list corresponding to each object type to obtain a combined object recommendation result based on the user type of the recommended user.
In some embodiments, determining the user type of the recommended user based on the behavior data set of the recommended user includes:
determining the object type of each object with behavior in the behavior data set of the recommended user, and counting the number of the object types;
and determining the user type of the recommended user based on the number of the object types.
In some embodiments, selecting an object from the recommendation ranking list corresponding to each object type to obtain a combined object recommendation result based on the user type of the recommended user includes:
acquiring one or more attribute feature data of each user in each user type;
determining one or more statistical characteristic data of each user based on the behavior data of each user in each user type;
and selecting the object from the recommendation ranking list corresponding to each object type to obtain a combined object recommendation result based on the one or more attribute feature data of each user, the one or more statistical feature data of each user and the user type of the recommended user.
In some embodiments, selecting an object from the recommendation ranking list corresponding to each object type to obtain a combined object recommendation result based on the one or more attribute feature data of each user, the one or more statistical feature data of each user, and the user type of the recommended user includes:
determining at least two feature mean vectors based on the one or more attribute feature data of each user and the one or more statistical feature data of each user, wherein the vector quantity of the at least two feature mean vectors is the same as the quantity of the at least two object types;
determining the distance between the recommended user and each feature mean vector;
determining the preference probability of the recommended user to each object type based on the distance between the recommended user and each feature mean vector;
and selecting the objects for multiple times from the recommendation ranking list corresponding to each object type based on the preference probability of the recommended user to each object type to obtain a combined object recommendation result.
In some embodiments, selecting the object from the recommended sorted list corresponding to each object type for a plurality of times based on the preference probability of the recommended user for each object type includes:
for any of the picks:
and generating one or more random probabilities, screening a preference probability from the preference probabilities of all object types based on the one or more random probabilities, and taking an object from a recommended ordered list of the object types corresponding to the screened preference probability.
In some embodiments, generating one or more random probabilities and, based on the one or more random probabilities, screening the preference probabilities for all object types comprises:
generating a random probability;
screening preference probability larger than the random probability to obtain a screening set;
judging whether the number in the screening set is greater than 1;
if the probability is larger than 1, generating the random probability again, and updating the screening set;
and repeating the steps of judging, regenerating and updating until the number in the screening set is equal to 1, and stopping generating the random probability.
In a second aspect, an embodiment of the present disclosure further provides a combined object recommendation apparatus, where the combined object includes a plurality of objects, and the plurality of objects belong to at least two object types, the apparatus includes:
the establishing unit is used for establishing a corresponding recommendation model for each object type;
and the recommending unit is used for respectively obtaining a recommended ordered list of the objects in each object type by using the recommending model corresponding to each object type for a recommended user, and selecting the objects from the recommended ordered list corresponding to each object type to obtain a combined object recommending result.
In some embodiments, the apparatus further comprises:
an acquisition unit configured to acquire one or more attribute features of each object;
the dividing unit is used for dividing all the objects with the same attribute feature quantity and attribute feature definition into the same object type;
and the establishing unit is used for establishing a corresponding recommendation model for each object type after the dividing unit divides the object types.
In some embodiments, the establishing unit is to:
acquiring a user behavior data set, wherein each piece of user behavior data is data describing a specific behavior of a user on an object;
dividing the user behavior data set into data sets corresponding to different object types according to the related different object types;
constructing a model training sample of each object type based on data sets of different object types;
and training a preset model based on the model training sample of each object type to obtain a recommended model corresponding to each object type.
In some embodiments, each piece of user behavior data comprises: user ID, object ID, behavior occurrence time and exposure object set;
the model training samples include: user ID, object ID, action occurrence time, tag.
In some embodiments, the obtaining, by the recommending unit, a recommended ordered list of the objects in each object type by using the recommendation model corresponding to each object type for one recommended user includes:
the recommending unit outputs a matching score between the object ID in each object type and the recommended user ID by using a recommending model corresponding to each object type;
and the recommending unit sorts the matching scores of the object IDs in the object types and the recommended user IDs from high to low to obtain a recommended sorted list of the objects in the object types.
In some embodiments, the obtaining of the combined object recommendation result by the recommending unit selecting the object from the recommendation ranking list corresponding to each object type includes:
the method comprises the steps that a recommending unit obtains a behavior data set of a recommended user;
the recommending unit determines the user type of the recommended user based on the behavior data set of the recommended user;
and the recommending unit selects the objects from the recommendation ranking list corresponding to each object type to obtain a combined object recommendation result based on the user type of the recommended user.
In some embodiments, the recommending unit determining the user type of the recommended user based on the behavior data set of the recommended user includes:
the recommendation unit determines the object type of each object with behavior in the behavior data set of the recommended user and counts the number of the object types;
the recommending unit determines the user type of the recommended user based on the number of the object types.
In some embodiments, the selecting, by the recommending unit, an object from the recommendation ranking list corresponding to each object type to obtain a combined object recommendation result based on the user type of the recommended user includes:
the recommendation unit acquires one or more attribute feature data of each user in each user type;
the recommendation unit determines one or more statistical characteristic data of each user based on the behavior data of each user in each user type;
and the recommending unit selects the object from the recommendation ranking list corresponding to each object type to obtain a combined object recommendation result based on the one or more attribute feature data of each user, the one or more statistical feature data of each user and the user type of the recommended user.
In some embodiments, the selecting, by the recommending unit, an object from the recommendation ranking list corresponding to each object type to obtain a combined object recommendation result based on the one or more attribute feature data of each user, the one or more statistical feature data of each user, and the user type of the recommended user includes:
the recommendation unit determines at least two feature mean vectors based on one or more attribute feature data of each user and one or more statistical feature data of each user, wherein the vector quantity of the at least two feature mean vectors is the same as the quantity of the at least two object types;
the recommending unit determines the distance between the recommended user and each feature mean vector;
the recommending unit determines the preference probability of the recommended user to each object type based on the distance between the recommended user and each feature mean vector;
and the recommending unit selects the objects for multiple times from the recommendation ranking list corresponding to each object type based on the preference probability of the recommended user to each object type to obtain a combined object recommendation result.
In some embodiments, the selecting, by the recommending unit, the object from the recommended sorted list corresponding to each object type for a plurality of times based on the preference probability of the recommended user for each object type includes:
for any of the picks:
the recommendation unit generates one or more random probabilities, screens out a preference probability from the preference probabilities of all object types based on the one or more random probabilities, and takes one object from a recommendation ranking list of the object type corresponding to the screened preference probability.
In some embodiments, the recommending unit generates one or more random probabilities, and the screening of the preference probabilities for all object types based on the one or more random probabilities comprises:
generating a random probability;
screening preference probability larger than the random probability to obtain a screening set;
judging whether the number in the screening set is greater than 1;
if the probability is larger than 1, generating the random probability again, and updating the screening set;
and repeating the steps of judging, regenerating and updating until the number in the screening set is equal to 1, and stopping generating the random probability.
In a third aspect, the disclosed embodiments also provide a system including at least one computing device and at least one storage device storing instructions, where the instructions, when executed by the at least one computing device, cause the at least one computing device to perform the steps of the combined object recommendation method according to any one of the embodiments of the first aspect.
In a fourth aspect, the present disclosure also provides a non-transitory computer-readable storage medium for storing a program or instructions, which when executed by at least one computing device, causes the at least one computing device to perform the steps of the combined object recommendation method according to any one of the embodiments of the first aspect.
It can be seen that, in at least one embodiment of the present disclosure, when a combined object is recommended, a corresponding recommendation model can be established for each object type by distinguishing different object types, and then, for a recommended user, a recommendation ranking list of objects in each object type is obtained by using the recommendation model corresponding to each object type. Therefore, different objects of the same object type have the same attribute characteristics, and the condition that the characteristic value is empty does not exist, so that the training characteristics of the recommendation model are changed from sparse to dense, the training convergence of the recommendation model is accelerated, and the calculation amount and the storage amount of the recommendation model are reduced.
In addition, the recommendation model corresponds to different objects of the same object type, when the recommendation model is trained, redundant features of the objects which do not belong to the object type are removed, noise interference is reduced, the recommendation effect is further improved, and the recommendation sequencing result of the recommendation model is more accurate.
In addition, the objects are selected from the recommendation ranking list corresponding to each object type, and different recommendation ranking lists are fused into the combined object recommendation result, so that the multi-model results can be used simultaneously, the multi-type objects can be effectively ranked, the multi-type objects are combined and recommended, and the effectiveness and diversity of object recommendation are improved.
Drawings
To more clearly illustrate the technical solutions of the embodiments of the present disclosure, the drawings used in the description of the embodiments or the prior art will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present disclosure, and other drawings can be obtained by those skilled in the art according to the drawings.
FIG. 1 is an exemplary flowchart of a method for recommending a combined object according to an embodiment of the disclosure;
FIG. 2 is an exemplary block diagram of a combined object recommendation device provided by an embodiment of the present disclosure;
fig. 3 is an exemplary block diagram of a system including at least one computing device and at least one storage device storing instructions provided by embodiments of the present disclosure.
Detailed Description
In order that the above objects, features and advantages of the present disclosure can be more clearly understood, the present disclosure will be described in further detail with reference to the accompanying drawings and examples. It is to be understood that the embodiments described are only a few embodiments of the present disclosure, and not all embodiments. The specific embodiments described herein are merely illustrative of the disclosure and are not intended to be limiting. All other embodiments derived by one of ordinary skill in the art from the described embodiments of the disclosure are intended to be within the scope of the disclosure.
It is noted that, in this document, relational terms such as "first" and "second," and the like, may be used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions.
A scene is recommended for a combined object, which generally comprises multiple types of objects, and the objects (items) comprise physical commodities and virtual commodities. Physical goods such as tangible goods like clothing, appliances, wearing, etc., virtual goods such as intangible goods like video, music, content, services (take-out), etc. Among them, clothing and electric appliances are different types of objects. The same type of object may also include sub-types, for example, a garment includes two sub-types of clothing and pants.
In the conventional single model recommendation method, when a combined object is recommended, for example, the combined object is a sports shoe and a refrigerator, attribute features of the sports shoe comprise sizes, attribute features of the refrigerator comprise volumes, and attribute features of a plurality of types of objects are usually directly spliced to serve as training features when the single model is trained, for example, the sizes of the sports shoe and the volumes of the refrigerator are spliced, because the sports shoe does not have a volume, the volume of the sports shoe is empty, and similarly, the sizes of the refrigerator are empty, so that the multidimensional feature dimension is empty, so that the feature vector is sparse, the calculation amount and the storage amount of the model are increased, and unnecessary noise is introduced, for example, the feature that the sports shoe has a volume is noise per se, and the effect of the model is reduced.
Therefore, for a combined object recommendation scenario, embodiments of the present disclosure provide a combined object recommendation method, apparatus, system, and storage medium. In at least one embodiment of the present disclosure, when a combined object is recommended, a corresponding recommendation model may be established for each object type by distinguishing different object types, and then, for a recommended user, a recommendation ranking list of objects in each object type is obtained by using the recommendation model corresponding to each object type. Therefore, different objects of the same object type have the same attribute characteristics, and the condition that the characteristic value is empty does not exist, so that the training characteristics of the recommendation model are changed from sparse to dense, the training convergence of the recommendation model is accelerated, and the calculation amount and the storage amount of the recommendation model are reduced.
In addition, the recommendation model corresponds to different objects of the same object type, when the recommendation model is trained, redundant features of the objects which do not belong to the object type are removed, noise interference is reduced, the recommendation effect is further improved, and the recommendation sequencing result of the recommendation model is more accurate.
In addition, the objects are selected from the recommendation ranking list corresponding to each object type, and different recommendation ranking lists are fused into the combined object recommendation result, so that the multi-model results can be used simultaneously, the multi-type objects can be effectively ranked, the multi-type objects are combined and recommended, and the effectiveness and diversity of object recommendation are improved.
Fig. 1 is an exemplary flowchart of a combined object recommendation method provided in an embodiment of the present disclosure, where a combined object includes a plurality of objects, and the plurality of objects belong to at least two object types. For example, the combined object is a sports shoe belonging to the object type of "footwear" and a refrigerator belonging to the object type of "refrigerator". As shown in fig. 1, the combined object recommendation method includes steps 101 and 102:
in step 101, a corresponding recommendation model is established for each object type.
For example, a corresponding footwear recommendation model is established for an object type of "footwear", and when the footwear recommendation model is trained, because different objects (sports shoes, leather shoes, high-heeled shoes, and the like) of the object type of "footwear" have the same attribute features (size, brand, material, and the like), the situation that feature values are empty does not exist, so that the training features of the footwear recommendation model are changed from sparse to dense, which is beneficial to accelerating the training convergence of the footwear recommendation model and reducing the calculation amount and storage amount of the footwear recommendation model.
In addition, the shoe recommendation model corresponds to different objects of the object type of 'shoes', when the shoe recommendation model is trained, redundant features of the objects which do not belong to the 'shoes' are removed, noise interference is reduced, the recommendation effect is further improved, and the recommendation sorting result of the shoe recommendation model is more accurate.
For another example, a corresponding refrigerator recommendation model is established for an object type of "refrigerator", and when the refrigerator recommendation model is trained, different objects (vertical refrigerator, horizontal refrigerator, desktop refrigerator, etc.) of the object type of "refrigerator" have the same attribute features (capacity, refrigeration mode, appearance, use, etc.), so that the situation that the feature values are empty does not exist, the training features of the refrigerator recommendation model are changed from sparse to dense, the convergence of the refrigerator recommendation model training is accelerated, and the calculation amount and the storage amount of the refrigerator recommendation model are reduced.
In addition, the refrigerator recommendation model corresponds to different objects of the type of the refrigerator, when the refrigerator recommendation model is trained, the redundant characteristics of the objects which do not belong to the refrigerator are removed, noise interference is reduced, the recommendation effect is further improved, and the recommendation sorting result of the refrigerator recommendation model is more accurate.
In step 102, for a recommended user, a recommendation model corresponding to each object type is used to obtain a recommendation ranking list of objects in each object type, and an object is selected from the recommendation ranking list corresponding to each object type to obtain a combined object recommendation result.
For example, a recommended ranked list of objects in an object type of "footwear" is obtained using a footwear recommendation model; a recommended ranked list of objects in an object type "refrigerator" is obtained using a refrigerator recommendation model. After the recommended ranking list of the 'shoes' and the recommended ranking list of the 'refrigerator' are obtained, objects can be selected from the two recommended ranking lists, the two recommended ranking lists are fused into a combined object recommendation result, the two recommended ranking lists can be used at the same time, the objects of the object type of the 'shoes' and the objects of the object type of the 'refrigerator' can be effectively ranked, the objects of the object type of the 'shoes' and the object type of the 'refrigerator' can be combined and recommended, and the effectiveness and diversity of object recommendation are improved.
In some embodiments, the present disclosure provides a partition method for object types, which partitions object types according to attribute features of the objects. Specifically, before establishing a corresponding recommendation model for each object type in step 101, one or more attribute features of each object are obtained; and all objects with the same quantity of attribute features and the same definition of the attribute features are divided into the same object type.
For example, athletic footwear has three attribute features: size, brand, material; while high-heeled shoes also have three attribute features: size, brand, material. The sports shoes and the high-heeled shoes have the same number of attribute features (three attribute features) and the same definition of the attribute features (size, brand and material). In this way, athletic shoes and high-heeled shoes are classified into the same object type, i.e., "footwear".
For another example, a vertical refrigerator has four attribute features: capacity, refrigeration mode, shape, use; while the horizontal refrigerator has four attribute features: capacity, refrigeration mode, shape, and use. The vertical refrigerator and the horizontal refrigerator have the same number of attribute features (four attribute features) and the same definition of the attribute features (capacity, refrigeration mode, shape and application). In this way, the vertical refrigerator and the horizontal refrigerator are classified into the same object type, i.e., "refrigerator".
In the present embodiment, any attribute feature of different object types is different, and for example, any one of three attribute features (size, brand, material) of an object type such as "footwear" is different from any one of four attribute features (capacity, cooling method, shape, use) of an object type such as "refrigerator".
In addition, if the number of the attribute features of the two objects is different, or if the number of the attribute features of the two objects is the same but at least one attribute feature with different definition exists, the two objects belong to different object types.
In some embodiments, the establishing a corresponding recommendation model for each object type in step 101 specifically includes steps 1011 to 1014:
1011. acquiring a user behavior data set, wherein each piece of user behavior data is data describing a specific behavior of a user on an object. The specific behavior is, for example, a selection operation such as clicking.
In this embodiment, the user behavior data may be collected in real time through a front-end product page of the client, for example, each piece of user behavior data includes: user ID, object ID, action occurrence time, exposure object set. The exposure object can be understood as an object displayed on a front-end product page of the client, and the exposure object set comprises an object clicked by a user and an object not clicked by the user but exposed. The acquired user behavior data set is shown in table 1, and in table 1, the click time is the behavior occurrence time.
TABLE 1 user behavior data set
| User ID | Object ID | Time of click | Set of exposed objects |
| U1 | P1 | T1 | P1,P2,P3,P4,P5 |
| U2 | P2 | T2 | P1,P2,P3,P4,P5 |
| U2 | P3 | T3 | P1,P2,P3,P4,P5 |
| U3 | P3 | T4 | P1,P2,P3,P4,P5 |
| U3 | P4 | T5 | P1,P2,P3,P4,P5 |
| U4 | P5 | T5 | P1,P2,P3,P4,P5 |
1012. The user behavior data set is divided into data sets corresponding to different object types according to the different object types involved.
In this embodiment, considering that the obtained user behavior data set includes objects belonging to different object types, in order to establish a corresponding recommendation model for each object type, the user behavior data set needs to be sorted to obtain data sets corresponding to different object types.
For example, in table 1, if P1, P2, P3 belong to the same object type a, and P4 and P5 belong to the same object type B, the user behavior data set shown in table 1 can be divided into data sets corresponding to object type a and data sets corresponding to object type B. The data set of object type a is shown in table 2, and the data set of object type B is shown in table 3.
TABLE 2 object type A data set
| User ID | Object ID | Time of click | Set of exposed objects |
| U1 | P1 | T1 | P1,P2,P3,P4,P5 |
| U2 | P2 | T2 | P1,P2,P3,P4,P5 |
| U2 | P3 | T3 | P1,P2,P3,P4,P5 |
| U3 | P3 | T4 | P1,P2,P3,P4,P5 |
TABLE 3 data set of object type B
| User ID | Object ID | Time of click | Set of exposed objects |
| U3 | P4 | T5 | P1,P2,P3,P4,P5 |
| U4 | P5 | T5 | P1,P2,P3,P4,P5 |
1013. Model training samples for each object type are constructed based on the data sets for the different object types.
In this embodiment, if each piece of data in the data set of the object type includes: user ID, object ID, action occurrence time and exposure object set, and the constructed model training sample comprises: user ID, object ID, action occurrence time, tag. The action occurrence time is, for example, click time; the label takes the value 1 or 0.
If the label value is 1, the constructed model training sample is a positive sample and represents an exposed object clicked by a user; and if the value of the label is 0, the constructed model training sample is a negative sample and represents an exposed object which is not clicked by the user.
For example, based on the data set for object type A shown in Table 2, a model training sample for object type A can be constructed, as shown in Table 4. Based on the data set for object type B shown in Table 3, model training samples for object type B can be constructed, as shown in Table 5.
TABLE 4 model training samples for object type A
| User ID | Object ID | Time of click | Label (R) |
| U1 | P1 | T1 | 1 |
| U1 | P2 | T1 | 0 |
| U1 | P3 | T1 | 0 |
| U2 | P1 | T2 | 0 |
| U2 | P2 | T2 | 1 |
| U2 | P3 | T2 | 0 |
| U3 | P1 | T4 | 1 |
| U3 | P2 | T4 | 0 |
| U3 | P3 | T4 | 0 |
TABLE 5 model training samples for object type B
| User ID | Object ID | Time of click | Label (R) |
| U3 | P4 | T5 | 1 |
| U3 | P5 | T5 | 0 |
| U4 | P4 | T5 | 0 |
| U4 | P5 | T5 | 1 |
It should be noted that the model training samples of the object type a shown in table 4 are only partial samples, not all samples, and the model training samples of the object type B shown in table 5 are only partial samples, not all samples.
In some embodiments, the model training samples include in addition to: user ID, object ID, action occurrence time and label, and further comprises: attribute feature information of the user and attribute feature information of the object. The attribute features of the user include, but are not limited to: gender, age, occupation, marital status, annual income and city (or home address), etc.
When a model training sample is constructed, the attribute characteristic information of a user and the attribute characteristic information of an object can be added into the model training sample by splicing the user information table and the object information table. The user information table comprises attribute characteristic information of a user, and the object information table comprises attribute characteristic information of an object.
For example, the model training samples of object type a shown in table 4 are shown in table 6 after the user information table and the object information table are spliced. The model training samples for object type B shown in table 5 are shown in table 7 after the user information table and the object information table are spliced.
TABLE 6 model training samples for object type A
| User ID | Attribute features of users | Object ID | Attribute features of objects | Time of click | Label (R) |
| U1 | F11To F61 | P1 | f11To f31 | T1 | 1 |
| U1 | F11To F61 | P2 | f12To f32 | T1 | 0 |
| U1 | F11To F61 | P3 | f13To f33 | T1 | 0 |
| U2 | F12To F62 | P1 | f11To f31 | T2 | 0 |
| U2 | F12To F62 | P2 | f12To f32 | T2 | 1 |
| U2 | F12To F62 | P3 | f13To f33 | T2 | 0 |
| U3 | F13To F63 | P1 | f11To f31 | T4 | 1 |
| U3 | F13To F63 | P2 | f12To f32 | T4 | 0 |
| U3 | F13To F63 | P3 | f13To f33 | T4 | 0 |
TABLE 7 model training samples for object type B
| User ID | Attribute features of users | Product ID | Attribute features of objects | Time of click | Label (R) |
| U3 | F13To F63 | P4 | f44To f74 | T5 | 1 |
| U3 | F13To F63 | P5 | f45To f75 | T5 | 0 |
| U4 | F14To F64 | P4 | f44To f74 | T5 | 0 |
| U4 | F14To F64 | P5 | f45To f75 | T5 | 1 |
In tables 6 and 7, F1 to F6 represent 6 attribute features of the user, for example: gender, age, occupation, marital status, annual income and city. The subscripts of F1 through F6 are used to distinguish different users. f1 to f3 represent attribute features of the object type a, f4 to f7 represent attribute features of the object type B, and f1 to f7 are different from each other, and subscripts of f1 to f7 are used to distinguish different objects.
1014. And training a preset model based on the model training sample of each object type to obtain a recommended model corresponding to each object type.
In this embodiment, the preset model may be a Tree model such as a GBDT (Gradient Boosting Decision Tree) model, a Regression model such as an LR (Logistic Regression) model, or a classification model.
When the preset model is trained, inputting model training samples of the object type into the preset model, and adjusting parameters of the preset model until a training target is reached, wherein the training target comprises two types: for positive samples, the output of the preset model converges to 1, and for negative samples, the output of the preset model converges to 0.
After a preset model is trained to obtain a recommendation model corresponding to each object type, for a recommendation model corresponding to a certain object type, for a recommended user, the matching score (score) of the object corresponding to the object type and the recommended user can be estimated by using the recommendation model, and the matching score value is larger than 0 and smaller than 1. Specifically, a recommended user ID, attribute characteristics of the recommended user, an object ID and attribute characteristics of the object are input into a recommendation model, and a matching score (score) of the recommended user ID and the object ID is estimated by the recommendation model.
In some embodiments, in step 102, for a recommended user, the obtaining of the recommended ordered list of the objects in each object type by using the recommendation model corresponding to each object type includes the following steps 1021 and 1022:
1021. and outputting a matching score between the object ID in each object type and the recommended user ID by using a recommendation model corresponding to each object type.
1022. And sorting the matching scores of the object IDs in the object types and the recommended user IDs from high to low to obtain a recommended sorted list of the objects in the object types.
After the recommended ordered list of the objects in each object type is obtained, the objects can be selected from the recommended ordered list corresponding to each object type to obtain a combined object recommendation result.
In some embodiments, the selecting an object from the recommendation ranking list corresponding to each object type in step 102 to obtain a combined object recommendation result specifically includes the following steps 1023 to 1025:
1023. and acquiring a behavior data set of the recommended user. The behavior data set of the recommended user can comprise historical behavior data of the recommended user and can also comprise real-time behavior data of the recommended user.
1024. And determining the user type of the recommended user based on the behavior data set of the recommended user.
In the embodiment, the object type of each object with behavior in the behavior data set of the recommended user is determined, and the number of the object types is counted; and determining the user type of the recommended user based on the number of the object types.
For example, an object type a and an object type B are divided according to attribute characteristics that the object has, and accordingly, 4 user types can be divided as shown in table 8.
TABLE 8 user type partitioning
| Type of user | User behavior |
| Y1 | Acting only on objects of A |
| Y2 | Acting only on B's objects |
| Y3 | Simultaneous behavior on objects of A and B |
| Y4 | No action occurred on objects of both A and B |
It should be noted that, in the present embodiment, two object types are taken as an example, and 4 user types can be divided, and if the object types are more than two, more user types can be divided based on the number of object types where actions occur.
1025. Based on the user type of the recommended user, selecting an object from the recommendation ranking list corresponding to each object type to obtain a combined object recommendation result, which may specifically include the following steps S1 to S3:
and S1, acquiring one or more attribute feature data of each user in each user type. In the present embodiment, the attribute feature data of the user is, for example, F1 to F6 shown in table 6 and table 7.
And S2, determining one or more statistical characteristic data of each user based on the behavior data of each user in the user types.
In this embodiment, one or more statistical characteristic data for each user may be determined based on a historical set of user behavior data. The statistical feature data are, for example, sliding window statistical features of the number of object clicks within 3 days, 7 days, 14 days, 30 days, and 60 days of history, and are denoted as H1 to H5. Wherein, the sliding window can be understood as that 7 days are increased by 4 days on the basis of 3 days, 14 days are increased by 7 days on the basis of 7 days, and the like. In some embodiments, the statistical characteristic data may also be counted in other manners, and the embodiment does not limit a specific statistical manner.
And S3, selecting the object from the recommendation ranking list corresponding to each object type to obtain a combined object recommendation result based on the one or more attribute feature data of each user, the one or more statistical feature data of each user and the user type of the recommended user.
In some embodiments, step S3 may include steps S31 to S34 as follows:
and S31, determining at least two feature mean vectors based on the one or more attribute feature data of each user and the one or more statistical feature data of each user, wherein the number of the at least two feature mean vectors is the same as that of the at least two object types.
Taking two object types (a and B) as an example, the user type is divided into 4 as shown in table 8: Y1-Y4, calculating 5 sliding window characteristics of all users in Y1 and Y2 aiming at the object of A, and marking as Ha 1-Ha 5; and 5 sliding window features of all users in Y1 and Y2 for the object of B are calculated and recorded as Hb 1-Hb 5.
After Ha1 to Ha5 and Hb1 to Hb5 were obtained, Ha1 to Ha5 and Hb1 to Hb5 were normalized (for example, Min-Max normalization) to obtain numerical values representing Ha1 to Ha5 and Hb1 to Hb 5. Thus, the average values of Y1 and Y2 on F1-F6, Ha 1-Ha 5 and Hb 1-Hb 5 can be calculated to obtain two characteristic mean vectors V1 and V2.
For example, the user u1 in Y1 has only three features F1, Ha1 and Hb1, which take values of 0,1.2 and 0.7 respectively; the user u2 in Y1 only has three characteristics F1, Ha1 and Hb1, and the values are 1, 0.9 and 0.5 respectively; then, V1 is [0.5,1.05,0.6], as shown in Table 9.
Table 9 calculation of the feature mean vector V1
| F1 | Ha1 | Hb1 | |
| u1 | 0 | 1.2 | 0.7 |
| u2 | 1 | 0.9 | 0.5 |
| Mean value | 0.5 | 1.05 | 0.6 |
And S32, determining the distance between the recommended user and each feature mean vector.
For example, for any user u in Y3, the euclidean distance calculation method is used to calculate the distances from u to V1 and V2, denoted as d1 and d2, and the calculation examples are as follows:
suppose a user u1 in Y3, which has only three features F1, Ha1, Hb1, corresponds to a feature vector of [0,1.2,0.7]The characteristic mean vector V1 is [0.5,1.05,0.6]]Then the distance d1 from user u1 to V1 is: v (0-0.5)2+(1.2-1.05)2+(0.7-0.6)2＝0.5315。
And S33, determining the preference probability of the recommended user to each object type based on the distance between the recommended user and each feature mean vector.
For example, the probability of the recommended user's preference for object type a is: d1/(d1+ d 2); the preference probability of the recommended user to the object type B is as follows: d2/(d1+ d 2).
And S34, selecting the objects for multiple times from the recommendation ranking list corresponding to each object type based on the preference probability of the recommended user to each object type, and obtaining a combined object recommendation result.
For any of the picks:
and generating one or more random probabilities, screening a preference probability from the preference probabilities of all object types based on the one or more random probabilities, and taking an object from a recommended ordered list of the object types corresponding to the screened preference probability.
Specifically, for any selection, a random probability is generated first; secondly, screening out preference probabilities larger than the random probability to obtain a screening set (the screening set is a set of preference probabilities); further judging whether the preference probability quantity in the screening set is greater than 1; if the probability is larger than 1, generating the random probability again, and updating the screening set; and repeating the steps of judging, regenerating and updating until the number in the screening set is equal to 1, and stopping generating the random probability, so that the only preference probability in the screening set is the screened preference probability.
In this embodiment, the top-ranked object is taken from the recommended sorted list of the object types corresponding to the screened preference probabilities. Thus, the combined object recommendation result can be obtained by selecting the object for multiple times.
It should be noted that, taking the user types Y1 to Y3 as examples, the combined object recommendation result with the user type Y3 is obtained through the above steps S31 to S34, that is, when the user type of the recommended user is Y3, the combined object recommendation result is obtained through the above steps S31 to S34.
If the user type of the recommended user is Y1 (or Y2), taking the recommendation ranking list corresponding to A (or B) as a final recommendation result; if the user type of the recommended user is Y4, the object sales amount is calculated by using the object trading table, and the final recommendation result is obtained according to the object sales amount from high to low.
It is noted that, for simplicity of description, the foregoing method embodiments are described as a series of acts or combination of acts, but those skilled in the art will appreciate that the disclosed embodiments are not limited by the order of acts described, as some steps may occur in other orders or concurrently with other steps in accordance with the disclosed embodiments. In addition, those skilled in the art can appreciate that the embodiments described in the specification all belong to alternative embodiments.
Fig. 2 is an exemplary block diagram of a combined object recommendation apparatus according to an embodiment of the disclosure, in which the combined object includes a plurality of objects, and the plurality of objects belong to at least two object types. As shown in fig. 2, the combined object recommendation apparatus includes: a creating unit 21 and a recommending unit 22.
The establishing unit 21 is used for establishing a corresponding recommendation model for each object type;
and the recommending unit 22 is configured to, for a recommended user, respectively obtain a recommended ordered list of objects in each object type by using the recommendation model corresponding to each object type, and select an object from the recommended ordered list corresponding to each object type to obtain a combined object recommendation result.
In some embodiments, the apparatus further comprises:
an obtaining unit 23, configured to obtain one or more attribute features of each object;
a dividing unit 24, configured to divide all the objects with the same number of attribute features and the same definition of attribute features into the same object type;
the establishing unit 21 is configured to establish a corresponding recommendation model for each object type after the dividing unit 24 divides the object types.
In some embodiments, the establishing unit 21 is configured to:
acquiring a user behavior data set, wherein each piece of user behavior data is data describing a specific behavior of a user on an object;
dividing the user behavior data set into data sets corresponding to different object types according to the related different object types;
constructing a model training sample of each object type based on data sets of different object types;
and training a preset model based on the model training sample of each object type to obtain a recommended model corresponding to each object type.
In some embodiments, each piece of user behavior data comprises: user ID, object ID, behavior occurrence time and exposure object set;
the model training samples include: user ID, object ID, action occurrence time, tag.
In some embodiments, the obtaining, by the recommending unit 22, a recommended ordered list of the objects in each object type by using the recommendation model corresponding to each object type for one recommended user includes:
the recommending unit 22 outputs a matching score between the object ID in each object type and the recommended user ID by using a recommending model corresponding to each object type;
the recommending unit 22 sorts the matching scores of the object IDs in the object types and the recommended user IDs from high to low to obtain a recommended sorted list of the objects in the object types.
In some embodiments, the obtaining of the combined object recommendation result by the recommending unit 22 selecting the object from the recommendation ranking list corresponding to each object type includes:
the recommending unit 22 acquires a behavior data set of a recommended user;
the recommending unit 22 determines the user type of the recommended user based on the behavior data set of the recommended user;
the recommending unit 22 selects an object from the recommendation ranking list corresponding to each object type to obtain a combined object recommendation result based on the user type of the recommended user.
In some embodiments, the recommending unit 22 determines the user type of the recommended user based on the behavior data set of the recommended user includes:
the recommending unit 22 determines the object type of each object with behavior in the behavior data set of the recommended user, and counts the number of the object types;
the recommending unit 22 determines the user type of the recommended user based on the number of object types.
In some embodiments, the obtaining, by the recommending unit 22, a combined object recommendation result by selecting an object from the recommendation ranking list corresponding to each object type based on the user type of the recommended user includes:
the recommending unit 22 acquires one or more attribute feature data of each user in each user type;
the recommending unit 22 determines one or more statistical characteristic data of each user based on the behavior data of each user in each user type;
the recommending unit 22 selects an object from the recommendation ranking list corresponding to each object type to obtain a combined object recommendation result based on the one or more attribute feature data of each user, the one or more statistical feature data of each user and the user type of the recommended user.
In some embodiments, the selecting, by the recommending unit 22, an object from the recommendation ranking list corresponding to each object type to obtain a combined object recommendation result based on the one or more attribute feature data of each user, the one or more statistical feature data of each user, and the user type of the recommended user includes:
the recommending unit 22 determines at least two feature mean vectors based on the one or more attribute feature data of each user and the one or more statistical feature data of each user, wherein the vector quantity of the at least two feature mean vectors is the same as the quantity of the at least two object types;
the recommending unit 22 determines the distance between the recommended user and each feature mean vector;
the recommending unit 22 determines the preference probability of the recommended user for each object type based on the distance between the recommended user and each feature mean vector;
the recommending unit 22 selects the objects from the recommendation ranking list corresponding to each object type for multiple times based on the preference probability of the recommended user for each object type, and obtains a combined object recommendation result.
In some embodiments, the selecting, by the recommending unit 22, the object from the recommended ordered list corresponding to each object type for a plurality of times based on the preference probability of the recommended user for each object type includes:
for any of the picks:
the recommending unit 22 generates one or more random probabilities, screens out a preference probability from the preference probabilities of all object types based on the one or more random probabilities, and takes an object from the recommended sorted list of object types corresponding to the screened preference probability.
In some embodiments, the recommending unit 22 generates one or more random probabilities, and based on the one or more random probabilities, sifting out one preference probability from the preference probabilities for all object types includes:
generating a random probability;
screening preference probability larger than the random probability to obtain a screening set;
judging whether the number in the screening set is greater than 1;
if the probability is larger than 1, generating the random probability again, and updating the screening set;
and repeating the steps of judging, regenerating and updating until the number in the screening set is equal to 1, and stopping generating the random probability.
The specific details of the above embodiments of the combined object recommendation device may refer to the embodiments of the combined object recommendation method, and are not repeated here to avoid repetition.
In some embodiments, the division of each unit in the combined object recommendation device is only one logical function division, and there may be another division manner when the division is actually implemented, for example, at least two units in the combined object recommendation device may be implemented as one unit; each unit in the combination object recommendation apparatus may be divided into a plurality of sub-units. It will be understood that the various units or sub-units may be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the implementation. Skilled artisans may implement the described functionality in varying ways for each particular application.
Fig. 3 is an exemplary block diagram of a system including at least one computing device and at least one storage device storing instructions provided by embodiments of the present disclosure. In some embodiments, the system may be used for big data processing, and the at least one computing device and the at least one storage device may be deployed in a distributed manner, making the system a distributed data processing cluster.
As shown in fig. 3, the system includes: at least one computing device 301, at least one storage device 302 storing instructions. It will be appreciated that the storage 302 in this embodiment may be either volatile memory or nonvolatile memory, or may include both volatile and nonvolatile memory.
In some embodiments, storage 302 stores elements, executable units, or data structures, or a subset thereof, or an expanded set thereof as follows: an operating system and an application program.
The operating system includes various system programs, such as a framework layer, a core library layer, a driver layer, and the like, and is used for implementing various basic tasks and processing hardware-based tasks. The application programs, including various application programs such as a Media Player (Media Player), a Browser (Browser), etc., are used to implement various application tasks. The program for implementing the combined object recommendation method provided by the embodiment of the present disclosure may be included in an application program.
In the embodiment of the present disclosure, at least one computing device 301 is configured to execute the steps of the combined object recommendation method provided by the embodiment of the present disclosure by calling a program or an instruction stored in at least one storage device 302, which may be, in particular, a program or an instruction stored in an application program.
The combined object recommendation method provided by the embodiment of the present disclosure may be applied to the computing device 301, or implemented by the computing device 301. The computing device 301 may be an integrated circuit chip having signal processing capabilities. In implementation, the steps of the above method may be performed by integrated logic circuits of hardware or instructions in the form of software in the computing device 301. The computing device 301 may be a general-purpose Processor, a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), an off-the-shelf Programmable Gate Array (FPGA) or other Programmable logic device, a discrete Gate or transistor logic device, or a discrete hardware component. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like.
The steps of the combined object recommendation method provided by the embodiment of the present disclosure may be directly implemented by a hardware decoding processor, or implemented by a combination of hardware and software units in the hardware decoding processor. The software elements may be located in ram, flash, rom, prom, or eprom, registers, among other storage media that are well known in the art. The storage medium is located in a storage device 302, and the computing device 301 reads information in the storage device 302 and completes the steps of the method in combination with hardware thereof.
Embodiments of the present disclosure also provide a non-transitory computer-readable storage medium storing a program or instructions, where the program or instructions, when executed by at least one computing device, cause the at least one computing device to perform the steps of the embodiments of the combined object recommendation method, and in order to avoid repeated descriptions, the steps are not repeated herein. The computing device may be the computing device 301 shown in fig. 3.
It should be noted that, in this document, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising … …" does not exclude the presence of other identical elements in a process, method, article, or apparatus that comprises the element.
Those skilled in the art will appreciate that while some embodiments described herein include some features included in other embodiments, rather than others, combinations of features of different embodiments are meant to be within the scope of the disclosure and form different embodiments.
Those skilled in the art will appreciate that the description of each embodiment has a respective emphasis, and reference may be made to the related description of other embodiments for those parts of an embodiment that are not described in detail.
Although the embodiments of the present disclosure have been described in conjunction with the accompanying drawings, those skilled in the art may make various modifications and variations without departing from the spirit and scope of the present disclosure, and such modifications and variations fall within the scope defined by the appended claims.
Claims (10)
1. A combined object recommendation method, wherein the combined object comprises a plurality of objects, and the plurality of objects belong to at least two object types, the method comprising:
establishing a corresponding recommendation model for each object type;
and aiming at a recommended user, respectively obtaining a recommended ordered list of objects in each object type by using a recommendation model corresponding to each object type, and selecting the objects from the recommended ordered list corresponding to each object type to obtain a combined object recommendation result.
2. The method of claim 1, wherein prior to said establishing a corresponding recommendation model for each object type, the method further comprises:
acquiring one or more attribute characteristics of each object;
all objects with the same quantity of attribute features and the same definition of the attribute features are divided into the same object type.
3. The method of claim 1, wherein the establishing a corresponding recommendation model for each object type comprises:
acquiring a user behavior data set, wherein each piece of user behavior data is data describing a specific behavior of a user on an object;
dividing the user behavior data set into data sets corresponding to different object types according to the related different object types;
constructing model training samples of each object type based on the data sets of different object types;
and training a preset model based on the model training sample of each object type to obtain a recommended model corresponding to each object type.
4. The method of claim 3, wherein the each piece of user behavior data comprises: user ID, object ID, behavior occurrence time and exposure object set;
the model training samples include: user ID, object ID, action occurrence time, tag.
5. The method of claim 1, wherein the obtaining, for one recommended user, the recommended ordered list of the objects in each object type using the recommendation model corresponding to each object type comprises:
outputting a matching score of the object ID and the recommended user ID in each object type by using a recommendation model corresponding to each object type;
and sorting the matching scores of the object IDs in the object types and the recommended user IDs from high to low to obtain a recommended sorted list of the objects in the object types.
6. The method of claim 1, wherein the selecting the object from the recommendation ranking list corresponding to each object type to obtain the combined object recommendation comprises:
acquiring a behavior data set of the recommended user;
determining a user type of the recommended user based on the behavior data set of the recommended user;
and selecting the object from the recommendation ranking list corresponding to each object type to obtain a combined object recommendation result based on the user type of the recommended user.
7. The method of claim 6, wherein the determining the user type of the recommended user based on the set of behavior data of the recommended user comprises:
determining the object type of each object with behavior in the behavior data set of the recommended user, and counting the number of the object types;
determining a user type of the recommended user based on the number of object types.
8. A combined object recommendation apparatus, wherein the combined object includes a plurality of objects, and the plurality of objects belong to at least two object types, the apparatus comprising:
the establishing unit is used for establishing a corresponding recommendation model for each object type;
and the recommending unit is used for respectively obtaining a recommended ordered list of the objects in each object type by using the recommending model corresponding to each object type for a recommended user, and selecting the objects from the recommended ordered list corresponding to each object type to obtain a combined object recommending result.
9. A system comprising at least one computing device and at least one storage device storing instructions that, when executed by the at least one computing device, cause the at least one computing device to perform the steps of the combined object recommendation method of any of claims 1-7.
10. A non-transitory computer-readable storage medium, wherein the non-transitory computer-readable storage medium stores a program or instructions which, when executed by at least one computing device, causes the at least one computing device to perform the steps of the combined object recommendation method of any one of claims 1-7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110315405.4A CN112950328A (en) | 2021-03-24 | 2021-03-24 | Combined object recommendation method, device, system and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110315405.4A CN112950328A (en) | 2021-03-24 | 2021-03-24 | Combined object recommendation method, device, system and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112950328A true CN112950328A (en) | 2021-06-11 |
Family
ID=76228039
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110315405.4A Pending CN112950328A (en) | 2021-03-24 | 2021-03-24 | Combined object recommendation method, device, system and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112950328A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114036398A (en) * | 2021-11-30 | 2022-02-11 | 北京百度网讯科技有限公司 | Content recommendation and ranking model training method, device, equipment and storage medium |
| CN114707074A (en) * | 2022-06-06 | 2022-07-05 | 深圳尚米网络技术有限公司 | Content recommendation method, device and system |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20160110794A1 (en) * | 2014-10-20 | 2016-04-21 | Yahoo! Inc. | E-commerce recommendation system and method |
| CN109902753A (en) * | 2019-03-06 | 2019-06-18 | 深圳市珍爱捷云信息技术有限公司 | User's recommended models training method, device, computer equipment and storage medium |
| CN110348947A (en) * | 2019-06-13 | 2019-10-18 | 阿里巴巴集团控股有限公司 | Object recommendation method and device |
| CN111027714A (en) * | 2019-12-11 | 2020-04-17 | 腾讯科技（深圳）有限公司 | Artificial intelligence-based object recommendation model training method, recommendation method and device |
| CN112115384A (en) * | 2020-08-14 | 2020-12-22 | 南京航空航天大学 | Time sequence recommendation algorithm based on generation sequencing |
| WO2021047593A1 (en) * | 2019-09-11 | 2021-03-18 | 华为技术有限公司 | Method for training recommendation model, and method and apparatus for predicting selection probability |
-
2021
- 2021-03-24 CN CN202110315405.4A patent/CN112950328A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20160110794A1 (en) * | 2014-10-20 | 2016-04-21 | Yahoo! Inc. | E-commerce recommendation system and method |
| CN109902753A (en) * | 2019-03-06 | 2019-06-18 | 深圳市珍爱捷云信息技术有限公司 | User's recommended models training method, device, computer equipment and storage medium |
| CN110348947A (en) * | 2019-06-13 | 2019-10-18 | 阿里巴巴集团控股有限公司 | Object recommendation method and device |
| WO2021047593A1 (en) * | 2019-09-11 | 2021-03-18 | 华为技术有限公司 | Method for training recommendation model, and method and apparatus for predicting selection probability |
| CN111027714A (en) * | 2019-12-11 | 2020-04-17 | 腾讯科技（深圳）有限公司 | Artificial intelligence-based object recommendation model training method, recommendation method and device |
| CN112115384A (en) * | 2020-08-14 | 2020-12-22 | 南京航空航天大学 | Time sequence recommendation algorithm based on generation sequencing |
Non-Patent Citations (1)
| Title |
|---|
| 郭绍翠;童向荣;杨旭;: "社交网络下基于列表级排序学习的推荐算法", 计算机工程与设计, no. 06, pages 270 - 275 * |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114036398A (en) * | 2021-11-30 | 2022-02-11 | 北京百度网讯科技有限公司 | Content recommendation and ranking model training method, device, equipment and storage medium |
| CN114036398B (en) * | 2021-11-30 | 2024-02-13 | 北京百度网讯科技有限公司 | Content recommendation and ranking model training method, device, equipment and storage medium |
| CN114707074A (en) * | 2022-06-06 | 2022-07-05 | 深圳尚米网络技术有限公司 | Content recommendation method, device and system |
| CN114707074B (en) * | 2022-06-06 | 2022-11-04 | 深圳尚米网络技术有限公司 | Content recommendation method, device and system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Jaiswal et al. | Application of random forest algorithm on feature subset selection and classification and regression | |
| CN107492008B (en) | Information recommendation method and device, server and computer storage medium | |
| WO2018041168A1 (en) | Information pushing method, storage medium and server | |
| CN112950328A (en) | Combined object recommendation method, device, system and storage medium | |
| US10140342B2 (en) | Similarity calculation system, method of calculating similarity, and program | |
| US20130018729A1 (en) | System and method for advertisement placement | |
| US10255300B1 (en) | Automatically extracting profile feature attribute data from event data | |
| Moniz et al. | Evaluation of ensemble methods in imbalanced regression tasks | |
| CN111428138A (en) | Course recommendation method, system, equipment and storage medium | |
| CN106327266B (en) | Data mining method and device | |
| US20070299835A1 (en) | Search engine for software components and a search program for software components | |
| CN106874503B (en) | Method and device for acquiring recommended data | |
| CN108122126B (en) | Method for extending association rule, device using same and computer readable medium | |
| CN112036981B (en) | Method, device, equipment and medium for providing target comparison commodities | |
| CN104778237A (en) | Individual recommending method and system based on key users | |
| CN112085058A (en) | Object combination recall method and device, electronic equipment and storage medium | |
| Chaturvedi et al. | Recommender system for news articles using supervised learning | |
| CN111340071A (en) | System and method for personalized product recommendation using hierarchical bayes | |
| CA3033201A1 (en) | Large scale social graph segmentation | |
| CN112883265A (en) | Information recommendation method and device, server and computer readable storage medium | |
| US10614505B2 (en) | Clustering system, method, and program, and recommendation system | |
| CN113225580B (en) | Live broadcast data processing method and device, electronic equipment and medium | |
| CN113111182A (en) | Information recommendation method and device and computer-readable storage medium | |
| CN116630054A (en) | Financial product recommendation method, equipment and medium | |
| CN111259267B (en) | Distributed hybrid collaborative intelligent recommendation method based on sparsity perception |
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 |