CN115392947A

CN115392947A - Demand prediction method and device

Info

Publication number: CN115392947A
Application number: CN202210868881.3A
Authority: CN
Inventors: 沈泽飞; 周仁浩; 吴菲菲
Original assignee: Alibaba China Co Ltd
Current assignee: Hangzhou Alibaba Overseas Internet Industry Co ltd
Priority date: 2022-07-22
Filing date: 2022-07-22
Publication date: 2022-11-25

Abstract

The embodiment of the specification provides a demand forecasting method and a demand forecasting device, wherein the demand forecasting method comprises the following steps: the method comprises the steps of obtaining a plurality of attribute information of an object and behavior information of a user aiming at the object, traversing a pre-constructed decision tree according to the attribute information, determining target attribute information matched with the decision tree, determining a fine classification of the object according to the target attribute information, and predicting object demand corresponding to the fine classification in a target time period according to the fine classification and the behavior type. The method comprises the steps of analyzing a plurality of attribute information of an object by utilizing a decision tree algorithm, accurately positioning the fine classification of the object meeting the user requirements, accurately predicting the demand of the object of the fine classification according to the behavior type of the behavior information, obtaining effective reference data, and providing a clear and definite planning scheme for a direct supplier of the object.

Description

Demand prediction method and device

Technical Field

The embodiment of the specification relates to the technical field of electronic commerce planning, in particular to a demand forecasting method.

Background

With the development of internet technology, object providers and commodity suppliers can more conveniently provide rich services and commodities for users, dealers and franchisees through web pages, applications, applets and the like.

Due to different requirements of the user, the distributor and the joining party for services and goods, a B2B mode (Business-to-Business electronic commerce mode) and a B2C mode (Business-to-customer electronic commerce mode) are correspondingly generated, and the operation modes, targets, management methods and the like of the two modes are different. For example, a single user purchases a smaller number of premium, inexpensive goods and services, while dealers and franchisees focus on low cost, warehoused, profitable goods.

Accordingly, how to accurately position the service and commodity classification according to different operation modes, targets, management methods and the like of different modes to further understand the demands of users, and further accurately determine the demand quantity of the service and commodity corresponding to the classification, and the demand quantity is used as effective reference data to provide a clear and definite planning scheme for a direct supplier of an object, which is a problem to be solved. Therefore, a demand forecasting method for accurately responding to the user demand is needed.

Disclosure of Invention

In view of this, the embodiments of the present disclosure provide a demand prediction method. One or more embodiments of the present specification also relate to a demand forecasting method, a demand forecasting apparatus, a computing device, a computer-readable storage medium, and a computer program, so as to solve the technical drawbacks of the prior art.

According to a first aspect of embodiments herein, there is provided a demand prediction method including:

acquiring a plurality of attribute information of an object and behavior information of a user aiming at the object, wherein the behavior information comprises a behavior type;

traversing a pre-constructed decision tree according to a plurality of attribute information, and determining target attribute information matched with the decision tree;

determining the fine classification of the object according to the target attribute information;

and predicting the object demand corresponding to the fine classification in the target time period according to the fine classification and the behavior type.

According to a second aspect of embodiments herein, there is provided a commodity demand prediction method including:

acquiring a plurality of attribute information of a commodity and behavior information of a user for the commodity, wherein the behavior information comprises a behavior type;

traversing a pre-constructed decision tree according to the attribute information, and determining target attribute information matched with the decision tree;

determining the detailed categories of the commodities according to the target attribute information;

and predicting the commodity demand corresponding to the fine classification in the target time period according to the fine classification and the behavior type.

According to a third aspect of the embodiments herein, there is provided a demand prediction apparatus including:

the device comprises a first acquisition module, a second acquisition module and a third acquisition module, wherein the first acquisition module is configured to acquire a plurality of attribute information of an object and behavior information of a user for the object, and the behavior information comprises a behavior type;

the first target attribute information determining module is configured to traverse a pre-constructed decision tree according to the plurality of attribute information and determine target attribute information matched with the decision tree;

a first fine category determination module configured to determine a fine category of the object according to the target attribute information;

and the first prediction module is configured to predict the object demand corresponding to the fine classification in the target time period according to the fine classification and the behavior type.

According to a fourth aspect of embodiments herein, there is provided a product demand predicting device including:

the second acquisition module is configured to acquire a plurality of attribute information of the commodity and behavior information of the user for the commodity, wherein the behavior information comprises a behavior type;

the second target attribute information determining module is configured to traverse a pre-constructed decision tree according to the plurality of attribute information and determine target attribute information matched with the decision tree;

the second fine category determining module is configured to determine the fine category of the commodity according to the target attribute information;

and the second prediction module is configured to predict the commodity demand corresponding to the fine classification in the target time period according to the fine classification and the behavior type.

According to a fifth aspect of embodiments herein, there is provided a computing device comprising:

a memory and a processor;

the memory is used for storing computer-executable instructions, and the processor is used for executing the computer-executable instructions, and the computer-executable instructions are executed by the processor to realize the steps of the demand forecasting method or the commodity demand forecasting method.

According to a sixth aspect of embodiments herein, there is provided a computer-readable storage medium storing computer-executable instructions that, when executed by a processor, implement the steps of the demand prediction method or the commodity demand prediction method described above.

According to a seventh aspect of the embodiments of the present specification, there is provided a computer program, wherein when the computer program is executed in a computer, the computer is caused to execute the steps of the demand amount prediction method or the commodity demand amount prediction method described above.

In the embodiment of the present specification, multiple attribute information of an object and behavior information of a user for the object are obtained, a pre-constructed decision tree is traversed according to the multiple attribute information, target attribute information matched with the decision tree is determined, a fine classification of the object is determined according to the target attribute information, and an object demand amount corresponding to the fine classification in a target time period is predicted according to the fine classification and the behavior type. The method comprises the steps of analyzing a plurality of attribute information of an object by utilizing a decision tree algorithm, accurately positioning the fine classification of the object meeting the user requirements, accurately predicting the demand of the object of the fine classification according to the behavior type of the behavior information, obtaining effective reference data, and providing a clear and definite planning scheme for a direct supplier of the object.

Drawings

FIG. 1 is a system flow diagram of a demand forecasting method provided by one embodiment of the present disclosure;

fig. 2A is a schematic diagram of a click behavior information generation display of a demand forecasting method according to an embodiment of the present specification;

fig. 2B is a schematic diagram of a purchase behavior information generation display of a demand forecasting method according to an embodiment of the present disclosure;

FIG. 2C is a schematic diagram illustrating a collection behavior information generation display of a demand forecasting method according to an embodiment of the present disclosure;

FIG. 3A is a flow chart of a demand forecasting method provided by one embodiment of the present disclosure;

FIG. 3B is a schematic diagram of a decision tree in a demand forecasting method according to an embodiment of the present disclosure;

FIG. 3C is a schematic diagram illustrating a decision tree construction in a demand forecasting method according to an embodiment of the present disclosure;

FIG. 4 is a flow chart of a method for forecasting demand for a commodity according to an embodiment of the present disclosure;

FIG. 5 is a flowchart of a process of a method for forecasting demand for goods on an e-commerce platform according to an embodiment of the present disclosure;

FIG. 6 is a process flow diagram of a method for red sea or blue sea prediction according to an embodiment of the present disclosure;

fig. 7 is a schematic structural diagram of a demand prediction apparatus according to an embodiment of the present disclosure;

fig. 8 is a schematic structural diagram of a commodity demand predicting device according to an embodiment of the present disclosure;

fig. 9 is a block diagram of a computing device according to an embodiment of the present disclosure.

Detailed Description

In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present specification. This description may be embodied in many different forms and should not be construed as limited to the embodiments set forth herein, as those skilled in the art will be able to make and use the present disclosure without departing from the spirit and scope of the present disclosure.

The terminology used in the description of the one or more embodiments is for the purpose of describing the particular embodiments only and is not intended to be limiting of the description of the one or more embodiments. As used in one or more embodiments of the present specification and the appended claims, the singular forms "a," "an," and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. It should also be understood that the term "and/or" as used in one or more embodiments of the present specification refers to and encompasses any and all possible combinations of one or more of the associated listed items.

It will be understood that, although the terms first, second, etc. may be used herein in one or more embodiments to describe various information, these information should not be limited by these terms. These terms are only used to distinguish one type of information from another. For example, a first can also be referred to as a second and, similarly, a second can also be referred to as a first without departing from the scope of one or more embodiments of the present description. The word "if," as used herein, may be interpreted as "at … …" or "at … …" or "in response to a determination," depending on the context.

First, the noun terms to which one or more embodiments of the present specification relate are explained.

Object category: the classification of services and goods, including the major categories of system architecture, i.e., categories of services and goods, by which e-commerce platforms often design, plan web pages, service catalogs and goods catalogs of applications or applets, e.g., "furniture-sofa-leather sofa". The method also comprises finely divided subclasses, and the subclasses correspond to the requirements of the user and are obtained after understanding the requirement driving and purchasing behaviors of the user. E-commerce platforms often provide users with more accurate searches, shopping guides, etc. through fine categories. Such as "satay shoes", "log furniture", etc.

And (4) category planning: the dealers and the franchisees set the object types as planning units on the basis of knowing and grasping the user requirements, and develop corresponding planning schemes by analyzing data such as the demand, the supply and the search heat of the object types. In an e-commerce scenario, a planning scheme for object category categories mainly includes a series of dimensions such as stocking, warehousing, drainage, marketing activities and the like.

Red sea, blue sea: the red sea and the blue sea are used for describing the supply and demand saturation condition of one object class, the blue sea market refers to the condition that the supply is not enough, and the red sea market refers to the condition that the supply is larger than the demand. The blue sea market generally has a better development prospect, dealers and franchisers can increase the goods input of corresponding object categories, the red sea market generally has a poorer development prospect, and the dealers and the franchisers need to make corresponding marketing strategies to reduce the storage amount in order to reduce the cost.

Method for Entity word Recognition (NER), named Entity Recognition: the method is characterized in that words (entities) with specific meanings in the text are recognized, wherein the words mainly comprise names of people, place names, mechanism names, proper nouns, attribute names, attribute values and the like, and the words needing to be recognized are marked in the text sequence.

At present, the object classification of service and commodity is taken as the basis of the e-commerce platform operation. For the B2C model, due to the rapid development of information supply platforms such as various short videos and social media, a user can obtain search hot words about a target object on the short videos and the social media, and then perform related searches on a web page, an application program and an applet of an e-commerce platform to obtain corresponding services and commodities. For example, a user searches a ranking list of hot words through a certain social media recently to obtain a search hot word for a certain clothes style, namely, national wind Chinese uniform, and the user inputs the search hot word, namely, the national wind Chinese uniform, on an e-commerce platform to search to obtain corresponding clothes.

However, for such a search hot word, the e-commerce platform mainly identifies an entity word used for describing a target object in the search hot word based on an entity word identification method, and then correspondingly pre-labels tag information of a plurality of objects stored on the e-commerce platform, determines a degree of association between the tag information and the entity word, and further obtains a plurality of corresponding search results to be displayed to a user, so that the user can select the target object from the corresponding search results. By analyzing the search hot words input by the user and the target object selected finally, the object type corresponding to the target object can be defined by searching the hot words, and further analysis is carried out on the demand of the object type. Continuing with the above example, the search hot word "Guofeng Han clothing" is recognized by the physical word recognition method to obtain the physical words "Guofeng" and "Han clothing", and the label information "Tang system", "Song system", "Chinese" pannier "," pannier "," left selected "and" right "etc. associated with the physical words" Han clothing "are obtained to obtain the label information" Tang system "and" Song system "associated with the physical words" Guofeng ", and the label information" Chinese "associated with the physical words" Guofeng "is obtained to obtain a plurality of corresponding search results to be presented to the user for selection.

In this way, the object class is defined and the demand analysis is performed based on the analysis at the text level. However, since the search hotword is often a subjective fuzzy definition, the object category determined by using the definition cannot allow the object provider to accurately understand the requirements of merchants such as the dealer and the franchisee in a specialized and structured mode such as B2B, and accordingly, an accurate object is provided to allow the dealer and the franchisee to meet the requirements of the user, so that an effective analysis cannot be given to the demand of the object category, an effective reference data is obtained, and a clear planning scheme is provided for the direct supplier of the object. In the B2B mode, the object category needs to be defined by more specialized and structured information such as style, type, size, and material, so that the object provider can more accurately understand what target object the user needs to correspond to, and for the above example, the object category corresponding to "national wind chinese clothing" should be: the style is as follows: tang system, song Zhi, ming system, version: a variety of different forms of the skirt including , right and water sleeves, size: code balancing, material: cotton and silk.

In view of the above problems, the present specification provides a demand forecasting method, and the present specification also relates to a commodity demand forecasting method, a demand forecasting apparatus, a commodity demand forecasting apparatus, a computing device, and a computer-readable storage medium, which are described in detail one by one in the following embodiments.

Referring to fig. 1, fig. 1 is a system flow chart illustrating a demand forecasting method according to an embodiment of the present disclosure.

As shown in fig. 1, the analysis back end obtains attribute information of an object and behavior information of a user for the object from an object database, traverses a pre-constructed decision tree according to the attribute information of the object to obtain matched target attribute information, defines and obtains an object category based on the target attribute information, analyzes the object category according to the behavior information of the user for the object to obtain a demand of the object category, obtains a supply of the object category through the object database, and performs predictive analysis on a supply-demand relationship of the object category by using the demand of the object category and the supply of the object category to obtain an analysis result of the supply-demand relationship.

Referring to fig. 2A to 2C, fig. 2A is a schematic diagram illustrating a click behavior information generation display of a demand forecasting method according to an embodiment of the present disclosure.

As shown in fig. 2A, a user inputs a search word in a search box of the e-commerce platform, clicks a "search" button to obtain a plurality of commodities (commodities 1 to 8), each commodity has a commodity picture and corresponding commodity attribute information (attribute information 1 to 4), and the user clicks a corresponding target commodity to generate click behavior information, which includes attribute information and a behavior type of the target commodity: clicking and recording the clicking behavior information into an object database.

Fig. 2B is a schematic diagram illustrating a purchase behavior information generation display of a demand forecasting method according to an embodiment of the present disclosure.

As shown in fig. 2B, the user enters a product detail page by clicking the product 3 picture in fig. 2A, where the product detail page includes the product 3 picture, price, attribute information (attribute information 1-4), number selection, purchase control and collection control, and the user generates purchase behavior information by clicking the purchase control, including attribute information and behavior type of the target product: and purchasing and recording the purchasing behavior information into the object database.

Fig. 2C is a schematic diagram illustrating a collection behavior information generation display of a demand forecasting method according to an embodiment of the present disclosure.

As shown in fig. 2C, the user enters a product detail page by clicking the product 3 picture in fig. 2A, where the product detail page includes the product 3 picture, price, attribute information (attribute information 1-4), number selection, purchase control and collection control, and the user generates collection behavior information by clicking the collection control, including attribute information and behavior type of the target product: and collecting and recording the collection behavior information into an object database.

Referring to fig. 3A-3B, fig. 3A shows a flowchart of a demand forecasting method according to an embodiment of the present disclosure, which specifically includes the following steps:

step 302: acquiring a plurality of attribute information of an object and behavior information of a user aiming at the object, wherein the behavior information comprises a behavior type;

the embodiment of the specification is applied to B2B mode analysis of an e-commerce platform, wherein direct providers of users, services and objects and indirect providers of the services and the objects exist in a B2B mode. The user is a consumer of the service and the object, the direct provider of the service and the object is a merchant for the consumer to purchase the corresponding service and the commodity through the e-commerce platform, and the indirect provider of the service and the object is a supplier for providing the service and the commodity for the direct provider. For example, a consumer of the e-commerce platform is a user, a resident merchant on the e-commerce platform is a direct provider of goods, and a provider of goods is an indirect provider, and for example, a consumer of the service platform is a user, a service broker on the service platform is a direct provider, a provider of services is an indirect provider, and for renting and purchasing services, a tenant or a buyer on the house e-commerce platform is a user, a broker brand is a direct provider, and a landlord or a homeowner is an indirect provider.

The e-commerce platform provides a B2C mode for a user, and simultaneously correspondingly provides a B2B mode for a direct provider of services and objects, and can provide three or more platforms for a shopping platform, a house renting, an automobile renting, a book renting, making friends and the like in one transaction.

The objects are services and goods provided to users (consumers).

The attribute information is attribute type information and attribute value information of a service or a product, and for example, for a certain product, the product includes attribute type information such as a size, a place of origin, and a color, and the corresponding XL size, XX country, red color, and the like are attribute value information.

The behavior information is information recorded by the back end of the e-commerce platform after the user operates the object on the e-commerce platform, and comprises identity information, time information, identification information of the object, behavior type information and the like of the user. And after the E-commerce platform records the behavior information, the behavior information is stored in a corresponding object database. For example, a user purchases a certain commodity on an e-commerce platform, and the e-commerce platform records identity information, purchase time information, identification information of the commodity, and a behavior type of the user: and obtaining behavior information according to the information such as purchase, and storing the behavior information into a commodity database of the E-commerce platform.

The behavior type is a type preset by the e-commerce platform and used for identifying the operation of the user on the object, for example, the user performs collection operation on the object, and the corresponding behavior type is as follows: and (6) storing.

Acquiring a plurality of attribute information of an object and behavior information of a user aiming at the object, wherein the specific mode is as follows: according to the requirements of the user, acquiring a plurality of attribute information of the object and behavior information of the user aiming at the object from a pre-constructed object database.

Correspondingly, the object is determined by sampling a plurality of candidate objects, and a plurality of attribute information of the object and behavior information of the user for the object are acquired from a pre-constructed object database according to the object identification of the object. It should be noted that, acquiring behavior information of a user for an object specifically includes: and recording the behavior information of the user aiming at the object in the target time period from a pre-constructed object database.

Exemplarily, the user's requirement is "punk wind", the candidate object is 100 sports shoes, 100 sports shoes are sampled, 10 sports shoes are determined, according to the identification information of the 10 sports shoes, the attribute information (style 1: color (red), vamp material (lacquer surface), lacing mode (shoelace); style 2: color (black), vamp material (lacquer surface), lacing mode (shoelace) … …) of the 10 sports shoes and the behavior information of the user on the 10 sports shoes in one month are obtained from the commodity database of the e-commerce platform, wherein the behavior information includes behavior types such as clicking, purchasing and collecting.

By acquiring a plurality of attribute information of the object, a traversal index is provided for determining the attribute information of the object in the subsequent decision tree, the accuracy of the subsequently determined attribute information of the object is ensured, the behavior information of the user aiming at the object and including the behavior type is acquired, and a data basis is laid for predicting the demand quantity of the object in the detailed classification in the subsequent process.

Step 304: traversing a pre-constructed decision tree according to a plurality of attribute information, and determining target attribute information matched with the decision tree;

the decision tree is a classification model which reversely deduces attribute information of the object according to the result so as to accurately classify the object. Reference attribute information of the object is recorded in the decision tree. For example, a user who rents an e-commerce platform from a car needs to purchase a car for home use, and the importance of the car is as follows: four cylinders, automatic gear, 1.6L displacement, home-made brand, and the corresponding decision tree is a hierarchical structure of (household car) -cylinder number (four cylinders) -gearbox type (automatic gear) -displacement (1.6L) -brand (home-made). The specific decision tree traversal method and the construction method are described in detail later.

The target attribute information is information matched with both the attribute type information and the attribute value information of the plurality of attribute information in the decision tree, and comprises target attribute type information and target attribute value information.

According to the attribute information, traversing a pre-constructed decision tree, and determining target attribute information matched with the decision tree, wherein the specific mode is as follows: and comparing the plurality of attribute information with reference attribute information recorded in a pre-constructed decision tree one by one to determine matched reference attribute information as target attribute information.

Exemplarily, the attribute information of 10 sports shoes (style 1: color (red), vamp material (painted surface), and lacing method (shoelace); style 2: color (black), vamp material (painted surface), and lacing method (shoelace) … …) is the reference attribute information recorded in the decision tree: the shoe upper is characterized in that the shoe upper is made of a material (painted surface) and is in a lacing mode (shoelace), after attribute information and reference attribute information of 10 styles of sports shoes are respectively paired one by one, and matched target attribute information is determined to be the shoe upper material (painted surface) and the lacing mode (shoelace).

And traversing the pre-constructed decision tree according to the plurality of attribute information, and determining the target attribute information matched with the decision tree. And through traversing the decision tree, target attribute information with structuredness and specialization is obtained, and information support is provided for subsequently determining the fine classification of the object, so that the fine classification of the object has the characteristics of structuredness and specialization.

Step 306: determining the fine classification of the object according to the target attribute information;

the fine classification of the object is a fine classification of the object corresponding to the user's needs. The fine categories of the objects are obtained by splicing according to the target attribute information, for example, the requirement of the user is 'father shoes', the attribute information of the object, which is 'father shoes', is analyzed through the steps 302-304, and the obtained target attribute information of the father shoes is as follows: sole (thick), vamp material (cortex), frenulum mode (shoelace), vamp style (multiaspect concatenation), obtain corresponding careful classification and do: the father shoes, the soles (thickness), the vamp materials (leather), the lacing modes (shoelaces) and the vamp styles (multi-face splicing).

According to the target attribute information, determining the detailed classification of the object, wherein the specific mode is as follows: and splicing to obtain the fine classification of the object according to the target attribute category information and the target attribute value information of the target attribute information.

Exemplarily, the target attribute information of the 10 sports shoes for "punk" required by the user is the material of the upper (paint surface), the lacing method (shoelace), and according to the target attribute category information (material of the upper, lacing method) and the target attribute value information (paint surface, shoelace), the fine category corresponding to "punk" is obtained by splicing: puncheon, vamp material (painted surface), and lacing (shoelace).

According to the target attribute information, the fine classification of the object is determined, the fine classification of the object is defined by more specialized and structured information, the classification of the object is more accurately positioned, and meanwhile, the analysis object is determined for the object demand corresponding to the fine classification in the subsequent prediction target time period.

Step 308: and predicting the object demand corresponding to the fine classification in the target time period according to the fine classification and the behavior type.

The object demand is the demand of the user on the e-commerce platform for each object under the fine classification in the target time period. For example, the detail classification is the father shoes, the soles (thick), the vamp materials (leather), the lacing modes (shoelaces) and the vamp styles (multi-surface splicing), and the demand of users on the e-commerce platform for the 'father shoes' of each style in the detail classification in the next three months is met.

Predicting the object demand corresponding to the fine classification in the target time period according to the fine classification and the behavior type, wherein the specific mode is as follows: and analyzing the historical behavior data trend of the object under the fine classification according to the behavior type, predicting behavior data corresponding to each behavior type in the target time period, and obtaining the object demand corresponding to the fine classification in the target time period according to the predicting behavior data.

The historical behavior data is information recorded by the back end of the e-commerce platform after the user operates the object on the e-commerce platform in the historical time period, and the historical behavior data comprises identity information, time information, identification information of the object, behavior type information and the like of the user.

Because the demand amounts corresponding to different behavior types are different, for example, a user may only browse a commodity and not make a purchase by clicking, that is, there is no demand for the commodity, and the user has a greater possibility to make a purchase by collecting, but also does not make a purchase, that is, there is a greater possible demand for the commodity, and the user makes a purchase by purchasing, that is, determining to make a purchase, and thus there is a demand for the commodity. Therefore, it is necessary to perform prediction for different behavior types to ensure accuracy of the predicted object demand.

Illustratively, according to the type of behavior (click, buy, favorite), the sub-categories on the e-commerce platform are: trend analysis is carried out on historical behavior data corresponding to each sports shoe in the modes of punning, vamp material (paint surface) and tying (shoelace), and prediction behaviors corresponding to behavior types under the subdivision categories are obtained: the click trend is increased by 0.7% daily, the purchase trend is increased by 1% daily, and the collection trend is 0.4%, and the object demand corresponding to the category in the next month is obtained as follows: 32000 is two.

In the embodiment of the present specification, multiple attribute information of an object and behavior information of a user for the object are obtained, a pre-constructed decision tree is traversed according to the multiple attribute information, target attribute information matched with the decision tree is determined, a fine classification of the object is determined according to the target attribute information, and an object demand amount corresponding to the fine classification in a target time period is predicted according to the fine classification and the behavior type. The method comprises the steps of analyzing a plurality of attribute information of an object by utilizing a decision tree algorithm, accurately positioning the fine classification of the object meeting the user requirements, accurately predicting the demand of the object of the fine classification according to the behavior type of the behavior information to obtain effective reference data, formulating a planning scheme for a direct supplier of the object, and providing clear, definite and accurate reference data.

Optionally, step 302 includes the following specific steps:

respectively acquiring attribute information and behavior information generated by each platform from a plurality of platforms;

and normalizing the attribute information and the behavior information generated by each platform to obtain a plurality of attribute information of the normalized object and the behavior information of the user aiming at the object.

The plurality of platforms include a current e-commerce platform to which the demand prediction method is applied and other platforms other than the e-commerce platform.

The method comprises the following steps of respectively acquiring attribute information and behavior information generated by each platform from a plurality of platforms, wherein the specific mode is as follows: and acquiring the attribute information and the behavior information generated by the current platform from the object database of the current platform, and acquiring the attribute information and the behavior information generated by other platforms from other platforms except the current platform by using an information acquisition tool.

Normalizing the attribute information and the behavior information generated by each platform to obtain a plurality of attribute information of normalized objects and behavior information of users aiming at the objects, wherein the specific mode is as follows: and according to the information mapping relation between the other platform and the current platform, mapping the attribute information and the behavior information generated by the other platform to the attribute information and the behavior information of the current platform to obtain a plurality of attribute information of the normalized object and the behavior information of the user aiming at the object.

The information mapping relation comprises at least one mapping relation among the format of the information, the measurement unit of the information and the attribute type information in the attribute information. The method for establishing the information mapping relation comprises the following steps: and determining the object with the mapping relation by using an image recognition model according to the image information of the object on the other platform and the image information of the object on the current platform, and establishing the information mapping relation according to the object identifier of the object with the mapping relation. Further comprising: and establishing an information mapping relation according to the unified identification information of the objects on other platforms and the object on the current platform. For example, the information mapping relationship is established by using uniform identification information such as the commodity item number of the commodity on the other platform and the current platform, the barcode on the commodity package, and the IMEI (International Mobile Equipment Identity) of the electronic Equipment such as the Mobile phone. The format of the corresponding information, for example, the information format on other platforms is two-dimensional code, and the information format on the current platform is character-through coding. The measurement unit of the corresponding information, for example, the attribute value information in the attribute information is an english measurement unit on another platform, and an european measurement unit is higher on the current platform.

Illustratively, attribute information (weight (500 g), size (10 cm × 5 cm)) and behavior information (click) generated by the current platform are acquired from the object database of the current platform, and attribute information (mass (1.1 pound) size (3.9in × 1.9in)) and behavior information (click) generated by other platforms are acquired from platforms other than the current platform by using the information acquisition tool. And according to the information mapping relation (weight-mass, centimeter-inch, click-click) between the other platforms and the current platform, mapping the attribute information and the behavior information generated by the other platforms to the attribute information and the behavior information of the current platform to obtain normalized attribute information (weight (500 g), size (10cm x 5cm)) of the object and behavior information (click) of the user aiming at the object.

The method comprises the steps of respectively obtaining attribute information and behavior information generated by each platform from the plurality of platforms, and carrying out normalization processing on the attribute information and the behavior information generated by each platform to obtain a plurality of attribute information of normalized objects and behavior information of users aiming at the objects. By acquiring attribute information and behavior information from a plurality of platforms, the information quantity for determining the fine classification and predicting the object demand is enriched, so that the determined fine classification and the predicted object demand are more accurate, the obtained fine classification and the predicted object demand can be applied to analysis of different platforms, and the applicability is improved; and the attribute information and the behavior information generated by each platform are subjected to normalization processing, so that the accuracy of the determined fine classification and the predicted object demand is further ensured.

Optionally, before step 306, the following specific steps are further included:

acquiring a search hot word;

and performing attribute identification on the search hot words to obtain target attribute information.

The search hot words are keywords which correspond to the requirements of the user at present and meet preset conditions on each information supply platform. The information providing platform may be an e-commerce platform in the above embodiments, and the information providing platform may also be a platform for providing information, such as a social platform, a news platform, a video platform, and the like. The preset condition may be a keyword exceeding a certain search amount threshold on the information providing platform, or a keyword of the top N on the search popularity list. For example, for a certain social platform, in the top 5 search ranking in the week, there is a search hot word "a-style decoration style" corresponding to the user's needs.

The method for acquiring the search hot words comprises the following specific steps: and acquiring keywords meeting preset conditions from other information supply platforms by using a keyword acquisition tool to serve as search hot words.

Optionally, before performing attribute identification on the search hotword, the method further includes:

and carrying out normalization processing of multiple languages on the search hot words to obtain the search hot words of the target language. Specifically, the search hot word is an "a-style decoration style", the search hot words "a's decoration style" and "file Decorativo De a" in other languages are acquired from the information provision platform, and a similar keyword "a-style decoration style" is obtained through a pre-trained similar keyword conversion model and is used as the search hot word in the target language.

Carrying out attribute identification on the search hot words to obtain target attribute information, wherein the specific mode is as follows: and identifying the entity words used for describing the target objects in the search hot words based on an entity word identification method, determining the association degree between the label information and the entity words corresponding to the label information of a plurality of objects stored on a pre-labeled E-commerce platform, and determining the entity words with the association degree larger than a preset threshold value as target attribute information.

The method for identifying the entity word can be a method for determining the attribute category information and/or attribute value information of the representation in the search hot word by using a knowledge graph, or a method for identifying the entity word by using a pre-trained neural network model to obtain the attribute category information and/or attribute value information of the search hot word and further determining the attribute information as the target attribute information. The entity word recognition model may be a BERT (Bidirectional Encoder Representation from transforms) model, a Long Short-Term Memory (LSTM) model, and other basic models and derivative models.

Exemplarily, a keyword acquisition tool is used for acquiring a search hot word "a formula" corresponding to a user requirement from the top 5 in the search hot list of the week from a social platform, and based on a method for identifying the entity word, the entity word used for describing a target object in the search hot word is identified to obtain the entity words "a formula", "decoration", "style", and corresponding to label information "european style", "roman column", "fireplace", "vault", "decoration" of a plurality of decoration shops stored on a pre-labeled e-commerce platform, and the association degree between the label information and the entity word is determined, and the entity words "roman column", "fireplace" and "vault" with the association degree greater than a preset threshold value are determined as target attribute information.

And acquiring the search hot words, and performing attribute identification on the search hot words to obtain target attribute information. Through analysis based on a text layer, the search hot words are analyzed to obtain target attribute information, the target attribute information is further enriched, the follow-up determined fine categories are more accurate, and further the demand prediction of the objects of the fine categories is more accurate.

Optionally, after step 306, the following specific steps are further included:

acquiring a search trend corresponding to the search hot word;

and predicting the object demand corresponding to the fine classification in the target time period according to the search trend.

The search trend is the variation trend of the search heat degree of the search heat words in the historical time periods on each information supply platform. The search trend can be expressed in the form of "historical unit time period-search volume", for example, the search trend of "a-style decoration style" in a historical week on a certain news platform is: the first day is 32 ten thousand, the second day is 34 ten thousand, the third day is 39 ten thousand, the fourth day is 63 ten thousand, the fifth day is 78 ten thousand, the sixth day is 116 ten thousand and the seventh day is 140 ten thousand. The search trend can also be expressed by the increasing and decreasing rate of the search volume in the historical time period, for example, the search trend of the "a-style decoration style" in one historical week on a certain news platform is as follows: the search volume increased 337% over the past week. The search trend can also be expressed by "historical time period-search ranking", for example, the search trend of "a-style finishing style" in a historical week on a certain news platform is: first day-43 th, second day-38 th, third day-31 th, fourth day-17 th, fifth day-12 th, sixth day-3 rd, and seventh day-1 st. The search trend may also be represented by a search popularity bar ranking ramp rate over a historical period of time. For example, the search trend of "a-style Fitment style" over a historical week on a certain news platform is: the search ranking of the popularity charts in the past week is promoted by 42.

Acquiring a search trend corresponding to the search hot word, wherein the specific mode is as follows: and acquiring a search trend corresponding to the search hot word from the information supply platform by using a trend acquisition tool.

According to the search trend, predicting the object demand corresponding to the fine classification in the target time period, wherein the specific mode is as follows: and predicting the predicted search amount of the fine categories in the target time period according to the search trend, and determining the object demand amount corresponding to the fine categories in the target time period according to the predicted search amount and the object search conversion rate of the fine categories.

The search conversion rate is the conversion rate between the search and the demand of the object under the fine classification based on the historical data statistics. For example, historical data shows a 7% conversion between precious metal jewelry and demand.

Exemplarily, the search hot word is the A-type decoration style, attribute recognition is carried out on the A-type decoration style to obtain target attribute information, and then a fine classification of the A-type decoration style is correspondingly obtained. The search trend of the 'A-type decoration style' in a historical week on a certain news platform is obtained as follows: the method comprises the steps of firstly, obtaining a search trend, namely, a first day of 32 ten thousand, a second day of 34 ten thousand, a third day of 39 ten thousand, a fourth day of 63 ten thousand, a fifth day of 78 ten thousand, a sixth day of 116 ten thousand and a seventh day of 140 ten thousand, predicting the predicted search amount of the 'A-type decoration style' in the next week to be 510 ten thousand according to the search trend, and determining the object demand amount corresponding to the fine classification in the target time period to be 25 ten thousand according to the predicted search amount and the fine classification target object search conversion rate to be 5%.

And acquiring a search trend corresponding to the search hot words, and predicting the object demand corresponding to the fine classification in the target time period according to the search trend, so that the efficiency of predicting the object demand is improved.

Optionally, the behavior information further comprises historical behavior data; the fine categories correspond to a plurality of behavior types;

correspondingly, step 308 includes the following specific steps:

trend analysis is carried out on historical behavior data generated by each behavior type corresponding to the subdivision classification, and prediction behavior data corresponding to each behavior type in a target time period are obtained;

and weighting the predicted behavior data corresponding to each behavior type in the target time period according to the behavior weight corresponding to each behavior type to obtain the object demand corresponding to the subdivision classification in the target time period.

Trend analysis is carried out on historical behavior data generated by various behavior types corresponding to the subdivision categories to obtain predicted behavior data corresponding to various behavior types in a target time period, and the specific mode is as follows: and counting historical behavior data generated by each behavior type corresponding to the subdivision classification objects in the historical time period, and obtaining predicted behavior data corresponding to each behavior type in the target time period by utilizing a pre-configured trend analysis method.

The specific trend analysis method can be a linear regression analysis method and can also be a mathematical model modeling analysis. The specific statistical and analysis tool is a data processing tool.

Illustratively, within a month, the fine categories are counted: historical behavior data generated by behavior types (clicking, purchasing and collecting) corresponding to the father shoes in the modes of father shoes, soles (thickness), vamp materials (leather), vamp materials (shoelaces) and vamp styles (multi-surface splicing) are obtained by a linear regression analysis method, a linear regression function corresponding to clicking is F1 (x), a corresponding linear regression function is purchased to be F2 (x), a corresponding linear regression function is collected to be F3 (x), the object quantity of historical behavior data of each row as the type is taken as an independent variable x, predicted behavior data Y1, Y2 and Y3 corresponding to each row as the type in the next month are obtained, and the predicted behavior data Y1, Y2 and Y3 corresponding to each row as the type in the next month are weighted according to behavior weights omega 1, omega 2 and omega 3 corresponding to each row as the type in the next month to obtain the object demand quantity Z corresponding to the subdivision class in the target time period.

And performing trend analysis on historical behavior data generated by each behavior type corresponding to the subdivision category to obtain predicted behavior data corresponding to each behavior type in the target time period, and weighting the predicted behavior data corresponding to each behavior type in the target time period according to the behavior weight corresponding to each behavior type to obtain the object demand amount corresponding to the subdivision category in the target time period. And performing trend analysis on the historical behavior data by corresponding to each behavior type to obtain more detailed predicted behavior data, and weighting the more detailed predicted behavior data according to the weight of each behavior type to ensure the accuracy of the object demand corresponding to the subdivision category in the target time period.

Optionally, after step 308, the following specific steps are further included:

inquiring an object database according to the fine categories to obtain historical object supply quantities corresponding to the fine categories;

predicting the object supply amount corresponding to the subdivision classification in the target time period according to the historical object supply amount;

and predicting and obtaining the object supply and demand relationship corresponding to the fine classification in the target time period according to the object supply quantity and the object demand quantity.

The object database is a database for storing attribute information of the object and behavior information of the user on the object by taking the object as a storage unit for the e-commerce platform.

The historical object supply amount is a supply amount determined according to historical data of a direct supplier of the object on the object. The number of objects obtained from the indirect provider of objects for the direct provider of objects.

The object supply and demand relationship corresponding to the fine classification is a supply and demand relationship aiming at each object under the fine classification, and comprises two conditions of supply more than demand and supply less than demand, the fine classification is determined to be blue sea corresponding to the condition of supply more than demand, and the fine classification is determined to be red sea corresponding to the condition of supply less than demand.

Inquiring an object database according to the fine categories to obtain historical object supply quantities corresponding to the fine categories, wherein the specific mode is as follows: and inquiring the object database according to the object identification of each object under the fine classification to obtain the historical object supply amount corresponding to the fine classification.

Predicting the object supply amount corresponding to the fine classification in the target time period according to the historical object supply amount, wherein the specific mode is as follows: and counting historical object supply amount, and predicting the object supply amount corresponding to the subdivision classification in the target time period by using a pre-configured prediction method. The prediction method is a linear regression analysis method. The specific statistical and analysis tool is a data processing tool.

According to the object supply quantity and the object demand quantity, predicting and obtaining the object supply and demand relation corresponding to the fine classification in the target time period, wherein the specific mode is as follows: and (3) constructing a mathematical model according to the fine categories, taking the object supply quantity and the object demand quantity as input, taking the target time period as a constraint condition, and predicting to obtain the object supply and demand relationship corresponding to the fine categories in the target time period. The specific mathematical model construction method can be constructed by using a mathematical model analysis software tool.

Illustratively, the fine categories are: according to the commodity numbers of the mature shoes, the soles (the thicknesses), the vamp materials (the leather), the lacing modes (the shoelaces) and the vamp styles (the multiface splicing), historical object supply quantities obtained from a mature shoe manufacturer by direct suppliers corresponding to the fine categories are obtained by inquiring an object database according to the commodity numbers of the mature shoes under the fine categories, after the historical object supply quantities are counted, the object supply quantities corresponding to the fine categories in the next month are predicted to be V by using a pre-configured linear regression analysis method, a mathematical model M is constructed by using a mathematical model analysis software tool according to the fine categories, the object supply quantities V and the object demand quantities Z are used as input, the next month is used as a constraint condition, the object supply relation corresponding to the fine categories, namely the mature shoes, the soles (the thicknesses), the vamp materials (the leather), the shoelaces modes (the shoelaces) and the vamp styles (the multiface splicing) in the next month is predicted to be smaller than the mathematic supply relation in the blue sea.

And querying the object database according to the fine categories to obtain historical object supply quantities corresponding to the fine categories, predicting object supply quantities corresponding to the fine categories in the target time period according to the historical object supply quantities, and predicting to obtain object supply and demand relations corresponding to the fine categories in the target time period according to the object supply quantities and the object demand quantities. According to the historical object supply quantity corresponding to the fine categories, the object supply quantity corresponding to the fine categories in the target time period is obtained through prediction, reference data are provided for analyzing the object supply and demand relationship of the fine categories, the object supply and demand relationship corresponding to the fine categories in the target time period is obtained through prediction according to the object supply quantity and the object demand quantity, and the accurate object supply and demand relationship is obtained, so that a direct supplier of the object can formulate a more targeted planning scheme according to the supply and demand relationship of the objects of the fine categories, the cost is saved, and the efficiency is improved.

Optionally, the decision tree includes a plurality of decision nodes, each decision node represents a decision condition for the attribute information, and edges between the decision nodes represent decision results for the attribute information;

correspondingly, step 304 includes the following specific steps:

matching each attribute information with each decision node and each edge in the decision tree according to the plurality of attribute information to determine a target decision node and a target edge;

and reading the target decision condition of the target decision node and the target decision result of the target edge to obtain target attribute information.

The decision nodes are nodes recorded with attribute type information of object attribute information, and the edges connected among the decision nodes are recorded with attribute value information of the object attribute information.

The target decision condition is reference attribute category information in the reference attribute information of the object recorded on the target decision node.

And the target decision result is the reference attribute value information in the reference attribute information of the object recorded on the target edge.

Fig. 3B is a schematic diagram illustrating a decision tree in a demand forecasting method according to an embodiment of the present disclosure.

As shown in fig. 3B, the decision tree constructed for shoes has 6 levels, the decision node of the first level is a shoe, the decision node of the second level is a sole, the decision node of the third level is a material, the decision node of the fourth level is a color, the decision node of the fifth level is a brand, and the decision node of the sixth level is a purchase or click. The decision results of the edges between the second level and the third level are thick bases, the decision results of the edges between the third level and the fourth level are cortex and cloth texture, the decision results of the edges between the fourth level and the fifth level are black and blue, and the decision results of the edges between the fifth level and the sixth level are XX cards and YY cards. And traversing each decision node and each edge through the behavior data of the object by the user, so that the behavior process of the user can be restored, and the attribute information of the object is determined.

Matching each attribute information with each decision node and each decision edge in the decision tree according to a plurality of attribute information to determine a target decision node and a target edge, wherein the specific mode is as follows: and matching the attribute type information and the attribute value information of the attribute information with each decision node and edge in the decision tree one by one, and determining the consistent decision node and edge as a target decision node and a target edge.

Reading a target decision condition of a target decision node and a target decision result of a target edge to obtain target attribute information, wherein the specific mode is as follows: reading the reference attribute type information in the reference attribute information of the object recorded on the target decision node, reading the reference attribute value information in the reference attribute information of the object recorded on the target edge to obtain the target attribute type information and the target attribute value information, and splicing the target attribute type information and the target attribute value information according to the hierarchical relationship of the decision tree and the corresponding relationship between the target attribute type information and the target attribute value information to obtain the target attribute information.

Exemplarily, attribute information (style 1: color (red), vamp material (painted surface), lacing mode (shoelace)), and style 2: color (black), vamp material (painted surface), lacing mode (shoelace) … …, are matched with each decision node and edge in a sports shoe decision tree, a target decision node and a target edge are determined, reference attribute type information (vamp material, lacing mode) in reference attribute information of an object recorded on the target decision node is read, reference attribute value information (painted surface, shoelace) in reference attribute information of the object recorded on the target edge is read, target attribute type information (vamp material, lacing mode) and target attribute value information (painted surface, shoelace) are obtained, and the target attribute type information and the target attribute value information are spliced according to the hierarchical relationship of the decision tree and the corresponding relationship therebetween, so that target attribute information material (painted surface), lacing mode (shoelace) is obtained.

And matching each attribute information with each decision node and each decision edge in the decision tree according to the plurality of attribute information, determining a target decision node and a target edge, reading a target decision condition of the target decision node and a target decision result of the target edge, and obtaining target attribute information. By matching the decision nodes and the edges, the behavior process of the user on the object is accurately restored, the target decision conditions of the target decision nodes and the target decision results of the target edges are read, the target attribute information of the object is accurately analyzed, and reference data are provided for subsequently determining the detailed classification.

Optionally, before step 304, the following specific steps are further included:

acquiring a sample set, wherein the sample set comprises a plurality of attribute information of a sample object and a sample behavior type;

obtaining attribute categories of first attribute information, and calculating information entropy of the first attribute information, wherein the first attribute information is any one of a plurality of attribute information;

determining the hierarchical relation of each decision node according to the information entropy of each attribute information;

and constructing a decision tree according to the hierarchical relation of each decision node.

The sample set is a set formed by historical behavior data of a plurality of users on sample objects.

The information entropy is the uncertainty describing each decision node, i.e. the number of edges corresponding to each decision node. The specific information calculation formula is shown in formula 1:

h = -plogp equation 1

Wherein H represents the information entropy of the attribute information, and p represents the proportion of the number of attribute categories of the attribute information to the total number.

Determining the hierarchical relationship of each decision node according to the information entropy of each attribute information, wherein the specific mode is as follows: and arranging the decision nodes corresponding to the attribute information from top to bottom according to the information entropy of the attribute information.

Optionally, the depth of the decision tree is set according to a project rule corresponding to each sample object in the sample set. For example, the sample object is a certain brand of computer, and the brand of computer is known to be all notebook computers and all use the same operating system, i.e., the corresponding decision nodes do not need to be set, the depth of the decision tree is reduced, and overfitting in the subsequent process of traversing the decision tree is avoided.

Constructing a decision tree according to the hierarchical relation of each decision node, wherein the specific mode is as follows: and constructing an initial decision tree according to the hierarchical relation of each decision node, and deleting edges corresponding to sample objects in the historical behavior data which do not meet a quantity threshold value by analyzing the number of the sample objects in the historical behavior data corresponding to each decision node to obtain the decision tree.

Illustratively, a sample set I is obtained, first attribute information of a plurality of sample objects is determined from the sample set: colors (red, blue, white and black), materials (cortex, cloth texture and silk texture), production areas (southeast Asia and middle east), years (2018,2019,2020,2021,2022), information entropies of the first attribute information are calculated to be 0.5,0.73,1,0.27 according to attribute categories of the first attribute information, and decision nodes corresponding to the attribute information are arranged from top to bottom according to the information entropies of the attribute information: the method comprises the steps of establishing an initial decision tree according to the hierarchical relation of decision nodes, the year, the place of production, the material and the color, and deleting the edge years (2018) corresponding to the sample objects in the historical behavior data which do not meet the quantity threshold value by analyzing the number of the sample objects in the historical behavior data corresponding to the decision nodes to obtain the decision tree.

Fig. 3C is a schematic diagram illustrating a decision tree construction in a demand forecasting method according to an embodiment of the present disclosure.

As shown in fig. 3C, a sample set is obtained, the depth of the decision tree is set to 5 according to the item rule corresponding to each sample object in the sample set, the attribute information of a plurality of sample objects is determined from the sample set, the information entropy of each attribute information is calculated, the decision nodes corresponding to each attribute information are arranged from top to bottom to construct an initial decision tree, and the decision tree is obtained by analyzing the number of sample objects in the historical behavior data corresponding to each decision node and deleting the corresponding decision nodes in the historical behavior data which do not satisfy the number threshold. The decision tree comprises a plurality of levels of decision nodes: a first level decision node, a second level decision node, a third level decision node 1, a third level decision node 2, a fourth level decision node 1, a fourth level decision node 2, a fifth level decision node 1, and a fifth level decision node 2.

The method comprises the steps of obtaining a sample set, wherein the sample set comprises a plurality of attribute information of a sample object and a sample behavior type, obtaining attribute categories of first attribute information, calculating information entropy of the first attribute information, determining the hierarchical relation of decision nodes according to the information entropy of the attribute information, and constructing a decision tree according to the hierarchical relation of the decision nodes. The decision tree is constructed through the sample set, a foundation is laid for subsequent traversal to obtain the target attribute nodes, the hierarchical relation of the decision tree is determined according to the information entropy to construct the decision tree, the structural performance of the obtained target attribute information is guaranteed, and a foundation is laid for subsequent determination of accurate detailed classification and prediction demand.

Referring to fig. 4, fig. 4 is a flowchart illustrating a method for predicting demand of a commodity according to an embodiment of the present disclosure, which specifically includes the following steps:

step 402: acquiring a plurality of attribute information of a commodity and behavior information of a user for the commodity, wherein the behavior information comprises a behavior type;

step 404: traversing a pre-constructed decision tree according to a plurality of attribute information, and determining target attribute information matched with the decision tree;

step 406: determining the detailed classification of the commodity according to the target attribute information;

step 408: and predicting the commodity demand corresponding to the fine classification in the target time period according to the fine classification and the behavior type.

In the embodiment of the present specification, the objects in the embodiment of fig. 3 are limited to be commodities, and the corresponding demand prediction methods are consistent, so details are not repeated here.

The demand forecasting method will be further described below with reference to fig. 5, by taking the application of the demand forecasting method provided in the present specification to the commodities of the e-commerce platform as an example. Fig. 5 is a flowchart illustrating a processing procedure of a method for predicting a demand of a commodity applied to an e-commerce platform according to an embodiment of the present disclosure, and specifically includes the following steps.

Step 502: according to user requirements, acquiring a plurality of attribute information of corresponding commodities and behavior information of a user for the corresponding commodities from a plurality of e-commerce platforms;

and normalizing the attribute information and the behavior information of the E-commerce platforms to obtain a plurality of attribute information of corresponding commodities and behavior information of the user for the corresponding commodities.

Step 504: matching each attribute information with each decision node and edge in the decision tree according to a plurality of attribute information of the corresponding commodity, and determining a target decision node and a target edge;

step 506: target attribute category information recorded by a target decision node and target attribute value information recorded by a target edge to obtain first target attribute information;

step 508: acquiring a search hot word, and performing attribute identification on the search hot word to obtain second target attribute information;

step 510: splicing the user requirements, the first target attribute information and the second target attribute information to obtain the fine categories of the corresponding commodities;

step 512: predicting commodity demand corresponding to the fine classification in the target time period according to the fine classification and the behavior type in the behavior information;

step 514: inquiring a commodity database according to the fine categories to obtain historical commodity supply quantities corresponding to the fine categories;

step 516: predicting commodity supply quantities corresponding to the subdivision categories in the target time period according to the historical commodity supply quantities;

step 518: establishing a mathematical model corresponding to the fine classification by using a mathematical model analysis software tool according to the fine classification;

the method comprises the following steps: taking the object supply quantity and the object demand quantity as input, taking the target time period as a constraint condition, analyzing the mathematical model by using a mathematical model analysis software tool, and predicting to obtain commodity supply and demand relations corresponding to the fine categories in the target time period;

step 520: and determining the fine classification as red sea or blue sea according to the object supply and demand relationship.

In the embodiment of the specification, a plurality of attribute information of a commodity is analyzed by using a decision tree algorithm, first target attribute information is determined, a text analysis attribute identification method is combined to obtain second target attribute information, and the information quantity of the target attribute information is enriched, so that the fine classification of the commodity meeting the requirements of a user can be more finely and accurately positioned, the commodity demand quantity of the fine classification in a target time period is predicted, on the basis, the commodity supply quantity of the fine classification in the target time period is predicted according to the historical commodity supply quantity, a mathematical model analysis software tool is used for accurately predicting the commodity supply and demand relation corresponding to the fine classification in the target time period, the fact that the fine classification is red sea or blue sea is determined, the fine classification is accurately predicted for a direct supplier of the commodity, the accurate prediction is carried out according to the behavior type of behavior information, effective reference data is obtained, a clear and accurate reference basis is provided for a targeted planning scheme of the direct supplier of the object.

Corresponding to the above-mentioned demand forecasting method embodiment, fig. 6 is a schematic processing flow diagram of a forecasting method for red sea or blue sea according to an embodiment of the present disclosure.

As shown in fig. 6, behavior data of a user of a first e-commerce platform for a commodity and behavior data of a user of a second e-commerce platform for the commodity are acquired according to user requirements, where the behavior data of the user for the commodity includes a plurality of attribute information of the commodity and behavior information of the user for the corresponding commodity. And acquiring the search hot words of the information supply platform, wherein the search hot words comprise the search hot words of the E-commerce platform and the search hot words of other information supply platforms. Normalizing the plurality of attribute information of the commodities of the second e-commerce platform, wherein the normalization processing mode comprises the following steps: data cleaning, format conversion, attribute mapping and commodity mapping. And traversing the decision tree according to the plurality of attribute information of the first e-commerce platform and the normalized attribute information of the commodity of the second e-commerce platform to obtain first target attribute information. And after normalization processing is carried out on the search hot words of the information supply platform, attribute identification is carried out to obtain second target attribute information. And splicing the user requirements, the first target attribute information and the second target attribute information to obtain fine categories of the commodities, and constructing mathematical models of the fine categories. And predicting the commodity demand of the detailed categories according to the detailed categories of the commodities and the behavior information of the user aiming at the target commodity. And carrying out trend analysis on the historical commodity supply amount corresponding to the fine classification, and predicting the commodity supply amount of the fine classification. And taking the commodity supply amount corresponding to the fine category and the commodity demand amount corresponding to the fine category as input, taking the target time period as a constraint condition, outputting a commodity supply and demand relationship corresponding to the fine category, and determining whether the fine category is red sea or blue sea according to the commodity supply and demand relationship.

Corresponding to the above embodiment of the demand forecasting method, the present specification further provides an embodiment of a demand forecasting device, and fig. 7 shows a schematic structural diagram of a demand forecasting device provided in an embodiment of the present specification. As shown in fig. 7, the apparatus includes:

a first obtaining module 702 configured to obtain a plurality of attribute information of an object and behavior information of a user for the object, wherein the behavior information includes a behavior type;

a first target attribute information determination module 704 configured to traverse a pre-constructed decision tree according to a plurality of attribute information, and determine target attribute information matched with the decision tree;

a first fine category determining module 706 configured to determine a fine category of the object according to the target attribute information;

and the first prediction module 708 is configured to predict the object demand corresponding to the fine classification in the target time period according to the fine classification and the behavior type.

Optionally, the first obtaining module 702 is further configured to:

the method comprises the steps of respectively obtaining attribute information and behavior information generated by each platform from the plurality of platforms, and carrying out normalization processing on the attribute information and the behavior information generated by each platform to obtain a plurality of attribute information of normalized objects and behavior information of users aiming at the objects.

Optionally, the apparatus further comprises:

and the attribute identification module is configured to acquire the search hot words, and perform attribute identification on the search hot words to obtain target attribute information.

Optionally, the apparatus further comprises:

an opportunistic market analysis module configured to

And acquiring a search trend corresponding to the search hot words, and predicting the object demand corresponding to the fine classification in the target time period according to the search trend.

correspondingly, the first prediction module 708 is further configured to:

and performing trend analysis on historical behavior data generated by each behavior type corresponding to the subdivision category to obtain predicted behavior data corresponding to each behavior type in the target time period, and weighting the predicted behavior data corresponding to each behavior type in the target time period according to the behavior weight corresponding to each behavior type to obtain the object demand corresponding to the subdivision category in the target time period.

Optionally, the apparatus further comprises:

and the third prediction module is configured to query the object database according to the fine categories, obtain historical object supply quantities corresponding to the fine categories, predict object supply quantities corresponding to the fine categories in the target time period according to the historical object supply quantities, and predict and obtain object supply and demand relations corresponding to the fine categories in the target time period according to the object supply quantities and the object demand quantities.

correspondingly, the first target attribute information determining module 704 is further configured to:

Optionally, the apparatus further comprises:

the decision tree building module is configured to obtain a sample set, wherein the sample set comprises a plurality of attribute information of a sample object and a sample behavior type, attribute categories of first attribute information are obtained, an information entropy of the first attribute information is calculated, the first attribute information is any one of the attribute information, a hierarchical relation of each decision node is determined according to the information entropy of each attribute information, and a decision tree is built according to the hierarchical relation of each decision node.

The above is an illustrative arrangement of a demand prediction apparatus of the present embodiment. It should be noted that the technical solution of the demand predicting device and the technical solution of the demand predicting method belong to the same concept, and for details of the technical solution of the demand predicting device, which are not described in detail, reference may be made to the description of the technical solution of the demand predicting method.

Corresponding to the above embodiment of the method for predicting the demand for goods, the present specification further provides an embodiment of a device for predicting the demand for goods, and fig. 8 shows a schematic structural diagram of a device for predicting the demand for goods provided in an embodiment of the present specification. As shown in fig. 8, the apparatus includes:

a second obtaining module 802, configured to obtain a plurality of attribute information of the product and behavior information of the user for the product, where the behavior information includes a behavior type;

a second target attribute information determining module 804 configured to traverse a pre-constructed decision tree according to the plurality of attribute information, and determine target attribute information matched with the decision tree;

a second fine category determination module 806 configured to determine a fine category of the commodity according to the target attribute information;

and the second prediction module 808 is configured to predict the commodity demand corresponding to the fine classification in the target time period according to the fine classification and the behavior type.

In the embodiment of the description, a plurality of attribute information of a commodity and behavior information of a user for the commodity are acquired, a pre-constructed decision tree is traversed according to the plurality of attribute information, target attribute information matched with the decision tree is determined, a fine classification of the commodity is determined according to the target attribute information, and commodity demand corresponding to the fine classification in a target time period is predicted according to the fine classification and the behavior type. The method comprises the steps of analyzing a plurality of attribute information of the commodities and behavior information of a user aiming at the commodities by utilizing a decision tree algorithm, accurately positioning the fine classification of the commodities meeting the requirements of the user, accurately predicting the commodity demand quantity of the fine classification according to the behavior type of the behavior information, obtaining effective reference data, and providing a clear and definite planning scheme for a direct supplier of the object.

The above is a schematic configuration of a commodity demand predicting apparatus according to the present embodiment. It should be noted that the technical solution of the product demand predicting apparatus and the technical solution of the product demand predicting method described above belong to the same concept, and details of the technical solution of the product demand predicting apparatus, which are not described in detail, can be referred to the description of the technical solution of the product demand predicting method described above.

Fig. 9 shows a block diagram of a computing device according to an embodiment of the present specification. Components of the computing device 900 include, but are not limited to, a memory 910 and a processor 920. The processor 920 is coupled to the memory 910 via a bus 930, and a database 950 is used to store data.

Computing device 900 also includes an access device 940, examples of which include a Public Switched Telephone Network (PSTN), a Local Area Network (LAN), a Wide Area Network (WAN), a Personal Area Network (PAN), or a combination of communication networks such as the internet, access device 940 enabling computing device 900 to communicate … … via one or more networks 960. Access device 940 may include one or more of any type of Network Interface (e.g., a Network Interface Controller (NIC)) whether wired or Wireless, such as an IEEE802.11 Wireless Local Area Network (WLAN) Wireless Interface, a worldwide Interoperability for Microwave Access (Wi-MAX) Interface, an ethernet Interface, a Universal Serial Bus (USB) Interface, a cellular Network Interface, a bluetooth Interface, a Near Field Communication (NFC) Interface, and so forth.

In one embodiment of the present description, the above-described components of computing device 900, as well as other components not shown in FIG. 9, may also be connected to each other, such as by a bus. It should be understood that the block diagram of the computing device architecture shown in FIG. 9 is for purposes of example only and is not limiting as to the scope of the description. Those skilled in the art may add or replace other components as desired.

Computing device 900 may be any type of stationary or mobile computing device, including a mobile computer or mobile computing device (e.g., tablet computer, personal digital assistant, laptop computer, notebook computer, netbook, etc.), mobile phone (e.g., smartphone), wearable computing device (e.g., smartwatch, smart glasses, etc.), or other type of mobile device, or a stationary computing device such as a desktop computer or PC. Computing device 900 may also be a mobile or stationary server.

The processor 920 is configured to execute computer-executable instructions, which when executed by the processor implement the steps of the demand forecasting method or the commodity demand forecasting method.

The above is an illustrative scheme of a computing device of the present embodiment. It should be noted that the technical solution of the computing device is the same as the technical solution of the demand forecasting method and the commodity demand forecasting method, and details of the technical solution of the computing device, which are not described in detail, can be referred to the description of the technical solution of the demand forecasting method and the commodity demand forecasting method.

An embodiment of the present specification also provides a computer-readable storage medium storing computer-executable instructions which, when executed by a processor, implement the steps of the demand forecasting method or the commodity demand forecasting method.

The above is an illustrative scheme of a computer-readable storage medium of the present embodiment. It should be noted that the technical solution of the storage medium is the same concept as the technical solution of the demand prediction method and the commodity demand prediction method, and details of the technical solution of the storage medium, which are not described in detail, can be referred to the description of the technical solution of the demand prediction method and the commodity demand prediction method.

An embodiment of the present specification further provides a computer program, wherein when the computer program is executed in a computer, the computer is caused to execute the steps of the demand amount prediction method or the commodity demand amount prediction method.

The above is an illustrative scheme of a computer program of the present embodiment. It should be noted that the technical solution of the computer program is the same concept as the technical solution of the demand prediction method and the commodity demand prediction method, and details of the technical solution of the computer program, which are not described in detail, can be referred to the description of the technical solution of the demand prediction method and the commodity demand prediction method.

The foregoing description has been directed to specific embodiments of this disclosure. Other embodiments are within the scope of the following claims. In some cases, the actions or steps recited in the claims can be performed in a different order than in the embodiments and still achieve desirable results. In addition, the processes depicted in the accompanying figures do not necessarily require the particular order shown, or sequential order, to achieve desirable results. In some embodiments, multitasking and parallel processing may also be possible or may be advantageous.

The computer instructions comprise computer program code, which may be in the form of source code, object code, an executable file or some intermediate form, etc. The computer readable medium may include: any entity or device capable of carrying computer program code, recording medium, U.S. disk, removable hard disk, magnetic disk, optical disk, computer Memory, read-Only Memory (ROM), random Access Memory (RAM), electrical carrier wave signals, telecommunications signals, software distribution media, and the like.

It should be noted that, for the sake of simplicity, the foregoing method embodiments are described as a series of acts, but those skilled in the art should understand that the present embodiment is not limited by the described acts, because some steps may be performed in other sequences or simultaneously according to the present embodiment. Further, those skilled in the art should also appreciate that the embodiments described in this specification are preferred embodiments and that acts and modules referred to are not necessarily required for an embodiment of the specification.

In the above embodiments, the descriptions of the respective embodiments have respective emphasis, and for parts that are not described in detail in a certain embodiment, reference may be made to related descriptions of other embodiments.

The preferred embodiments of the present specification disclosed above are intended only to aid in the description of the specification. Alternative embodiments are not exhaustive or limit the invention to the precise embodiments disclosed. Obviously, many modifications and variations are possible in light of the above teaching. The embodiments were chosen and described in order to best explain the principles of the embodiments and the practical application, to thereby enable others skilled in the art to best understand and utilize the embodiments. The specification is limited only by the claims and their full scope and equivalents.

Claims

1. A demand prediction method comprising:

2. The method of claim 1, wherein the obtaining of the plurality of attribute information of the object and the behavior information of the user for the object comprises:

and normalizing the attribute information and the behavior information generated by each platform to obtain a plurality of attribute information of normalized objects and the behavior information of the users aiming at the objects.

3. The method of claim 1, further comprising, prior to said determining a fine classification of the object based on the target attribute information:

acquiring a search hot word;

4. The method of claim 3, further comprising, after said determining a fine classification of the object according to the target attribute information:

acquiring a search trend corresponding to the search hot word;

5. The method of claim 1, the behavior information further comprising historical behavior data; the fine categories correspond to a plurality of behavior types;

predicting the object demand corresponding to the fine classification in the target time period according to the fine classification and the behavior type, wherein the predicting comprises the following steps:

and weighting the predicted behavior data corresponding to each behavior type in the target time period according to the behavior weight corresponding to each behavior type to obtain the object demand corresponding to the subdivision category in the target time period.

6. The method according to any one of claims 1-5, further comprising, after said predicting an object demand amount corresponding to said fine category within a target time period according to said fine category and said behavior type:

querying an object database according to the fine categories to obtain historical object supply quantities corresponding to the fine categories;

predicting the object supply quantity corresponding to the fine category in the target time period according to the historical object supply quantity;

and predicting to obtain the object supply and demand relationship corresponding to the fine classification in the target time period according to the object supply quantity and the object demand quantity.

7. The method of claim 1, wherein the decision tree comprises a plurality of decision nodes, each decision node representing a decision condition on attribute information, edges between decision nodes representing decision results on attribute information;

the step of traversing a pre-constructed decision tree according to the attribute information to determine target attribute information matched with the decision tree comprises the following steps:

matching each attribute information with each decision node and each edge in the decision tree according to the attribute information to determine a target decision node and a target edge;

8. The method of claim 1, further comprising, before said traversing a pre-constructed decision tree based on the plurality of attribute information to determine target attribute information matching the decision tree:

obtaining an attribute category of first attribute information, and calculating an information entropy of the first attribute information, wherein the first attribute information is any one of the plurality of attribute information;

and constructing a decision tree according to the hierarchical relation of the decision nodes.

9. A commodity demand prediction method includes:

determining the detailed classification of the commodity according to the target attribute information;

and predicting the commodity demand corresponding to the fine category in a target time period according to the fine category and the behavior type.

10. A demand prediction apparatus comprising:

the first target attribute information determining module is configured to traverse a pre-constructed decision tree according to the plurality of attribute information, and determine target attribute information matched with the decision tree;

11. A commodity demand predicting device includes:

the system comprises a second acquisition module, a first storage module and a second storage module, wherein the second acquisition module is configured to acquire a plurality of attribute information of a commodity and behavior information of a user for the commodity, and the behavior information comprises a behavior type;

the second target attribute information determining module is configured to traverse a pre-constructed decision tree according to the plurality of attribute information, and determine target attribute information matched with the decision tree;

a second fine category determination module configured to determine a fine category of the commodity according to the target attribute information;

and the second prediction module is configured to predict the commodity demand amount corresponding to the fine category in the target time period according to the fine category and the behavior type.

12. A computing device, comprising:

a memory and a processor;

the memory is configured to store computer-executable instructions and the processor is configured to execute the computer-executable instructions, which when executed by the processor, implement the steps of the demand forecasting method of any one of claims 1 to 8 or the commodity demand forecasting method of claim 9.

13. A computer-readable storage medium storing computer-executable instructions which, when executed by a processor, implement the steps of the demand prediction method of any one of claims 1 to 8 or the commodity demand prediction method of claim 9.