CN110544118B

CN110544118B - Sales prediction method, sales prediction device, medium and computing equipment

Info

Publication number: CN110544118B
Application number: CN201910787851.8A
Authority: CN
Inventors: 王珺; 郭训力; 吕韬; 王文豹
Original assignee: Alibaba China Co Ltd
Current assignee: Alibaba China Co Ltd
Priority date: 2019-08-23
Filing date: 2019-08-23
Publication date: 2023-07-14
Anticipated expiration: 2039-08-23
Also published as: CN110544118A

Abstract

The embodiment of the invention provides a sales prediction method. The method comprises the following steps: clustering at least a part of the products to obtain at least one category; determining a first category to which the first product belongs in the at least one category; acquiring a sales trend function of the long-term product based on historical sales change trends of the long-term product in the first category in N stages; calculating M category time lifting factors of the first category, wherein the M category time lifting factors are in one-to-one correspondence with M phases in the N phases; and calculating the predicted sales of the first product in the stage to be predicted based on the historical sales of the first product in the S stages closest to the current stage and the class time lifting factors corresponding to the stage to be predicted. The method can solve the problem of inaccurate demand prediction caused by insufficient historical sales data of new products. The embodiment of the invention also provides a sales predicting device, a medium and a computing device.

Description

Sales prediction method, sales prediction device, medium and computing equipment

Technical Field

The embodiment of the invention relates to the technical field of Internet, in particular to a sales prediction method, a sales prediction device, a sales prediction medium and a calculation device.

Background

This section is intended to provide a background or context to the embodiments of the invention that are recited in the claims. The description herein is not admitted to be prior art by inclusion in this section.

With the development of the internet+, online purchasing activities are increasingly frequent. A large amount of supply and sales data can be accumulated in the operation of the online procurement platform. By mining information in these supply and sales data, the operation of the online procurement platform can be optimized, for example, inventory cost and operating efficiency optimization can be achieved.

However, for products newly marketed and sold by an online purchasing platform (hereinafter referred to as new products), the historical supply and sales data is often insufficient, and how to furthest mine valuable information based on the insufficient data so as to scientifically and reasonably predict sales of the new products and more reasonably arrange inventory conditions of the new products is a problem to be solved.

Disclosure of Invention

In this context, embodiments of the present invention desire to provide a sales prediction method and apparatus that can predict future sales of new products by predicting sales data of long-term products in online purchasing platforms, so that future sales of new products can be predicted more scientifically and reasonably.

In a first aspect of an embodiment of the present invention, a sales prediction method is provided. The method comprises the following steps: clustering at least a part of the products to obtain at least one category; determining a first category to which a first product belongs in the at least one category, wherein the first category includes long-term products, the long-term products being one or more products whose sales time meets a predetermined condition; acquiring a sales trend function of the long-term product based on historical sales change trends of the long-term product in N stages, wherein the N stages comprise N time-sequential basic timing units, and N is an integer greater than or equal to 2; calculating M class time lifting factors of the first class, wherein M is a positive integer less than or equal to N, and the M class time lifting factors are in one-to-one correspondence with M phases in the N phases; each of the M category time promotion factors is characterized by a correspondence between historical sales of the long-term product and trend sales calculated using the sales trend function in a corresponding stage; and calculating the predicted sales of the first product in the stage to be predicted based on the historical sales of the first product in S stages closest to the current stage and the class time lifting factors corresponding to the stage to be predicted, wherein S is a positive integer less than or equal to M, the M stages comprise the S stages, and the stage to be predicted is any one or more stages corresponding to the M stages in a future stage.

In one embodiment of the invention, the calculating the M category time boost factors for the first category includes: counting M stage history sales of the long-term product, which are in one-to-one correspondence with the M stages; calculating M stage trend sales of the long-term product, which are in one-to-one correspondence with the M stages, through the sales trend function; obtaining M category trending sales based on the ratio of the M phase historical sales to the phase historical sales and the phase trend sales corresponding to the same phase in the M phase trend sales; and carrying out normalization processing on the trending sales of the M categories to obtain the M category time lifting factors.

In one embodiment of the present invention, the normalizing the trending sales of the M categories to obtain the M category time lifting factors includes: calculating the average value of the trending sales of the M categories to obtain the average trending sales of the categories; and dividing the M category trending sales volume by the category average trending sales volume respectively to obtain the M category time lifting factors.

In one embodiment of the present invention, the clustering at least a portion of the products to obtain at least one category includes: hierarchical clustering is performed based on static features and dynamic features of the at least one part of products, wherein the static features are used for describing static attributes of the at least one part of products, the dynamic features are used for describing dynamic attributes of the at least one part of products generated in a circulation link based on interactive behaviors of users, and the hierarchical clustering comprises multi-level categories with father-son relations.

In one embodiment of the present invention, the calculating the M category time boost factors for the first category further includes: when the class time lifting factors which do not meet the stability conditions exist in the M class time lifting factors, acquiring a stage corresponding to the class time lifting factors which do not meet the stability conditions, and acquiring a stage in which the lifting factors are undetermined; and determining a class time lifting factor meeting the stability condition corresponding to the lifting factor pending stage in the parent class of the first class as a class time lifting factor corresponding to the lifting factor pending stage of the first class. In one embodiment of the invention, the stability condition is set to a predetermined range of values.

In one embodiment of the present invention, the calculating the predicted sales of the first product in the stage to be predicted based on the historical sales of the first product in the S stages closest to the current stage and the class time lifting factor corresponding to the stage to be predicted includes: calculating the sum of the historical sales of the S stages closest to the current stage; selecting S category time lifting factors corresponding to the S phases one by one from the M category time lifting factors; calculating the sum of the S category time lifting factors to obtain the sum of the lifting factors of the first product in the S stages; dividing the sum of the historical sales of the S stages closest to the current time by the sum of the lifting factors to obtain the reference sales of the first product in each stage of the S stages closest to the current time; determining that the stage to be predicted corresponds to a first stage of M stages, and determining a class time lifting factor corresponding to the first stage as a class time lifting factor corresponding to the stage to be predicted; and determining the predicted sales of the first product in the stage to be predicted according to the reference sales and the class time lifting factors corresponding to the stage to be predicted.

In one embodiment of the invention, the basic timing unit is set to month and the N phases are set to N consecutive months.

In a second aspect of embodiments of the present invention, a sales predicting apparatus is provided. The device comprises a clustering module, a category determining module, a trend function obtaining module, a category time lifting factor calculating module and a predicting module. The clustering module is used for clustering at least one part of products to obtain at least one category. The category determination module is configured to determine a first category to which a first product belongs in the at least one category, wherein the first category includes long-term products, the long-term products being one or more products whose sales time satisfies a predetermined condition. The trend function acquisition module is used for acquiring sales trend functions of the long-term products based on historical sales change trends of the long-term products in N stages, wherein the N stages comprise N basic timing units which are consecutive in time, and N is an integer greater than or equal to 2. The class time lifting factor calculation module is used for calculating M class time lifting factors of the first class, wherein M is a positive integer less than or equal to N, and the M class time lifting factors are in one-to-one correspondence with M phases in the N phases; each of the M category time promotion factors is characterized by a correspondence between historical sales of the long-term product and trend sales calculated using the sales trend function over a corresponding phase. The prediction module is configured to calculate, based on the historical sales of the first product in S stages closest to a current stage and a class time lifting factor corresponding to a stage to be predicted, a predicted sales of the first product in the stage to be predicted, where S is a positive integer less than or equal to M, the M stages include the S stages, and the stage to be predicted is any one or more stages corresponding to the M stages in a future stage.

In one embodiment of the present invention, the category time lifting factor calculation module includes a statistics sub-module, a trend sales calculation sub-module, a category trending sales calculation sub-module, and a category time lifting factor calculation sub-module. The statistics submodule is used for counting M stage history sales of the long-term product, wherein the M stage history sales correspond to the M stages one by one. The trend sales computing operator module is used for computing M-stage trend sales of the long-term product, which are in one-to-one correspondence with the M stages, through the sales trend function. The class trending sales calculation operator module is used for obtaining M class trending sales based on the ratio of the M stage historical sales to the stage historical sales and the stage trend sales corresponding to the same stage in the M stage trend sales. And the class time lifting factor calculation sub-module is used for carrying out normalization processing on the trending sales of the M classes to obtain the M class time lifting factors.

In one embodiment of the present invention, the class time lifting factor calculating submodule is specifically configured to calculate an average value of the M class trending sales to obtain class average trending sales, and divide the M class trending sales by the class average trending sales to obtain the M class time lifting factors.

In one embodiment of the present invention, the clustering module is specifically configured to perform hierarchical clustering based on static features and dynamic features of the at least a portion of products, where the static features are used to describe static attributes of the at least a portion of products, the dynamic features are used to describe dynamic attributes of the at least a portion of products generated in a circulation link based on interactive behaviors of users, and the hierarchical clustering includes multiple levels of categories with parent-child relationships.

In one embodiment of the present invention, the category time lifting factor calculation module further includes a trace back sub-module. The tracing submodule is used for obtaining a stage corresponding to the class time lifting factor which does not meet the stability condition when the class time lifting factor which does not meet the stability condition exists in the M class time lifting factors, obtaining a lifting factor undetermined stage, and determining the class time lifting factor which meets the stability condition and corresponds to the lifting factor undetermined stage in the father class of the first class as the class time lifting factor which corresponds to the lifting factor undetermined stage of the first class. In one embodiment of the invention, the stability condition is set to a predetermined range of values.

In one embodiment of the invention, the prediction module comprises a prediction sales volume sum calculation sub-module, a category time lifting factor selection sub-module, a prediction lifting factor sum calculation module, a reference sales volume determination sub-module, a stage to be predicted category time lifting factor determination module and a prediction sales volume determination sub-module. The prediction sales sum calculation submodule is used for calculating the sum of the historical sales of the S stages closest to the current time. The class time lifting factor selecting submodule is used for selecting S class time lifting factors corresponding to the S phases one by one from the M class time lifting factors. The prediction lifting factor sum calculating module is used for calculating the sum of the S category time lifting factors to obtain the lifting factor sum of the first product in the S stages. The reference sales determination submodule is used for dividing the sum of the historical sales of the S stages closest to the current time by the sum of the lifting factors to obtain the reference sales of the first product in each stage of the S stages closest to the current time. The class time lifting factor determining module is configured to determine that the to-be-predicted stage corresponds to a first stage of the M stages, and determine a class time lifting factor corresponding to the first stage as a class time lifting factor corresponding to the to-be-predicted stage. The predicted sales volume determining submodule is used for determining the predicted sales volume of the first product in the stage to be predicted according to the reference sales volume and the category time lifting factors corresponding to the stage to be predicted.

In a third aspect of embodiments of the present invention, a computing device is provided. The computing device includes one or more storage units, and one or more processing units. The one or more memory units store executable instructions. The one or more processing units execute the executable instructions to implement the method as described above.

In a fourth aspect of embodiments of the present invention, there is provided a computer readable storage medium having stored thereon executable instructions which, when executed by a processor, cause the processor to perform a method as described above.

According to the method, the device, the medium and the computing equipment provided by the embodiment of the invention, the problem of inaccurate sales prediction caused by insufficient new product historical sales data is solved to a certain extent, and the accuracy of new product sales prediction can be improved, so that a more scientific and reasonable sales plan can be formulated, risks caused by high inventory or shortage of goods are avoided, and effective support is provided for operation services of an online purchasing platform.

Drawings

The above, as well as additional purposes, features, and advantages of exemplary embodiments of the present invention will become readily apparent from the following detailed description when read in conjunction with the accompanying drawings. Several embodiments of the present invention are illustrated by way of example, and not by way of limitation, in the figures of the accompanying drawings and in which:

fig. 1 schematically shows an application scenario of a sales prediction method and apparatus according to an embodiment of the present invention;

FIG. 2 schematically illustrates a flow chart of a sales prediction method according to an embodiment of the present invention;

FIG. 3 schematically shows a structural diagram of clustering results of hierarchical clustering in a sales volume prediction method according to an embodiment of the present invention;

FIG. 4 schematically illustrates a flow chart of a method of calculating M category time lifting factors for a first category in a sales prediction method according to an embodiment of the invention;

FIG. 5 schematically illustrates a flow diagram of a method for normalizing M categories trending sales according to an embodiment of the present invention;

FIG. 6 schematically illustrates a flow chart of a method for predicting sales in a sales prediction method according to an embodiment of the invention;

FIG. 7 schematically illustrates a block diagram of a sales predicting apparatus according to an embodiment of the present invention;

FIG. 8 schematically illustrates a block diagram of the category time boost factor calculation module of FIG. 7, in accordance with one embodiment of the present invention;

FIG. 9 schematically illustrates a block diagram of the prediction module of FIG. 7, in accordance with one embodiment of the present invention;

FIG. 10 schematically illustrates a schematic diagram of a program product adapted to implement a sales prediction method according to an embodiment of the present invention;

FIG. 11 schematically illustrates a block diagram of a computing device suitable for implementing a sales prediction method according to an embodiment of the present invention.

In the drawings, the same or corresponding reference numerals indicate the same or corresponding parts.

Detailed Description

The principles and spirit of the present invention will be described below with reference to several exemplary embodiments. It should be understood that these embodiments are presented merely to enable those skilled in the art to better understand and practice the invention and are not intended to limit the scope of the invention in any way. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

Those skilled in the art will appreciate that embodiments of the invention may be implemented as a system, apparatus, device, method, or computer program product. Accordingly, the present disclosure may be embodied in the following forms, namely: complete hardware, complete software (including firmware, resident software, micro-code, etc.), or a combination of hardware and software.

According to an embodiment of the invention, a sales prediction method, a sales prediction device, a sales prediction medium and a computing device are provided.

In this context, it is to be understood that the terms involved:

category time boost factor: the method for characterizing the influence of short-term seasonal factors on sales of products in a class in a stage can be represented by a ratio of historical sales of long-term products in the class in the stage to trend sales calculated according to a sales trend function.

Furthermore, any number of elements in the figures is for illustration and not limitation, and any naming is used for distinction only and not for any limiting sense.

The principles and spirit of the present invention are explained in detail below with reference to several representative embodiments thereof.

Summary of The Invention

The inventors have found that sales of products at each stage (e.g., monthly or quarterly) are affected by both long-term trending factors and short-term seasonal factors. For example, sales of a product increase month by month from long-term trends, but may fluctuate during one or more months.

The trend factor can be represented by a sales trend function obtained by regression of historical sales data of the product. The meaning of regression is to find the trend of sales of products over time. The short-term seasonal factor may be represented, for example, by a relative state to the trending factor (e.g., short-term seasonal factor index), such as by a ratio of historical sales at a stage to trend sales calculated from a regression trend function at that stage.

Thus, if the sales trend function of the product and the short-term seasonal factor index of the stage to be predicted can be obtained when the sales of the product are predicted, the sales of the product in the stage to be predicted can be predicted.

For example, if the sales time of a product (e.g., a long-term product) is long enough, the data amount is large, and the sales tendency of the long-term product is remarkable. When the sales of the long-term product is predicted, the sales of the long-term product can be obtained according to the sales trend function of the long-term product, and then the sales of the long-term product can be predicted by multiplying the sales of the long-term product by the short-term seasonal factor index corresponding to the to-be-predicted stage.

Or, for example, if the sales time of a product (e.g., new product) is too short, the amount of data that can be returned is too small, and the tendency is not obvious, in which case it can be considered that the sales of the product remains substantially at the current sales level for a short period of time with little tendency to rise or fall. In this way, the sales of the new product can be predicted as the current sales level multiplied by the short-term seasonal factor index corresponding to the phase to be predicted, in combination with the influence of the short-term seasonal factor.

However, since the historical sales data of new products is less, the acquisition of short-term seasonal factor index thereof can also become a problem. In this regard, the inventors believe that the short-term seasonal factor index of other long-term products that have some similarity to the short-term sales regular fluctuations of new products may be used instead. Furthermore, based on the statistical principle and big data analysis, a large number of products in the online purchasing platform can be clustered based on sales volume characteristics and the like, and then the short-term seasonal factor index of the new product is replaced by the short-term seasonal factor index of the category to which the new product belongs. Based on such a concept, the short-term seasonal factor index of finding a new product can be converted into a problem of first clustering a large number of products, and then finding the short-term seasonal factor index (i.e., the class time boost factor herein) of the products in the class to which the new product belongs.

Having described the basic principles of the present invention, various non-limiting embodiments of the invention are described in detail below.

Application scene overview

Reference is first made to fig. 1.

Fig. 1 schematically shows an application scenario of a sales prediction method and apparatus according to an embodiment of the present invention.

As shown in fig. 1, the application scenario includes a terminal device 11 and a server (cluster) 12. The terminal device 11 and the server (cluster) 12 are connected via a network. The network may be a local area network, wide area network, mobile internet, etc., and may include various connection types, such as wired, wireless communication links, etc.

The terminal device 11 may be, but not limited to, a portable device (e.g., a mobile phone, a tablet, a notebook, etc.), or a personal computer (PC, personal Computer). A user may interact with a server (cluster) 12 via a network using a terminal device 11 to receive or send messages or the like. Various client applications, such as shopping applications, web browser applications, etc. (by way of example only) may be installed on the terminal device 11. The user may perform an online product purchase activity by performing a ordering operation or the like with a client application installed in the terminal device 11.

The server (cluster) 12 may be a server (cluster) providing various services, such as a background server (merely an example) providing support for a user with shopping-type applications and the like used by the terminal device 11. The background server can analyze and process the received data such as the user request and the like, and feed back the processing result to the terminal equipment. For example, the server (cluster) 12 may process order information (e.g., products in an order, attributes of products, sales of products, etc.) generated by a user making an order with the terminal device 11.

According to an embodiment of the present invention, the server (cluster) 12 may also store, and process various data generated by the user making online purchases with the terminal device 11, and use the data to predict future sales of individual products, etc. by the method according to the embodiments of the present disclosure.

It should be noted that the method for sales prediction provided by the embodiment of the present invention may be performed by the server (cluster) 12. Accordingly, sales prediction apparatus, media, or computing devices provided by embodiments of the present invention may be generally disposed in a server (cluster) 12. Alternatively, the sales prediction method provided by the embodiment of the present invention may be performed by another server or server cluster that is different from the server (cluster) 12 and is capable of communicating with the terminal device 11 and/or the server (cluster) 12. Accordingly, the sales predicting apparatus, medium or computing device provided by the embodiments of the present invention may also be provided in other servers or server clusters that are different from the server (cluster) 12 and are capable of communicating with the terminal device 11 and/or the server (cluster) 12.

It should be understood that the number of terminal devices, networks and servers in fig. 1 is merely illustrative. There may be any number of terminal devices, networks, and servers, as desired for implementation.

Exemplary method

A sales prediction method according to an exemplary embodiment of the present invention will be described below with reference to fig. 2 to 6 in conjunction with the application scenario of fig. 1. It should be noted that the above application scenario is only shown for the convenience of understanding the spirit and principle of the present invention, and the embodiments of the present invention are not limited in any way. Rather, embodiments of the invention may be applied to any scenario where applicable.

Fig. 2 schematically shows a flow chart of a sales prediction method according to an embodiment of the invention.

As shown in fig. 2, the sales prediction method may include operations S210 to S250.

At least a portion of the products are clustered to obtain at least one category in operation S210. For example, the at least a part of the products may be products sold in an online purchasing platform (hereinafter referred to as a platform) in which the server (cluster) 12 provides service support. The clustering method used in clustering at least a portion of the products in operation S210 may be any clustering method, such as K-means clustering, gaussian mixture clustering, hierarchical clustering, or the like.

Hierarchical clustering may be used in operation S210 according to one embodiment of the present invention.

Fig. 3 schematically shows a schematic structural diagram of a clustering result 300 of hierarchical clustering in the sales volume prediction method according to an embodiment of the present invention.

As shown in FIG. 3, the clustering result 300 of hierarchical clustering includes multiple levels of categories with parent-child relationships. Hierarchical clustering can be classified into aggregate (aggregate) hierarchical clustering and split (divve) hierarchical clustering. The split hierarchical clustering uses the concept of "top-down", where all samples are first considered as the same class, and then the classes are divided into smaller classes by iteration until there is only one sample in each sample. The aggregation hierarchical clustering adopts the idea of 'bottom-up', each sample is firstly regarded as a different class, and two or a plurality of closest classes are combined into a new class through repeated iteration until all the last samples belong to the same class.

In one embodiment of the invention, a hierarchical aggregation cluster may be employed. For example, each product in at least a portion of the clusters may be considered as a class, then the similarity statistics between the classes are determined, and two or more classes closest to each other are selected and combined into a new class, the similarity statistics between the new class and other classes are calculated, and two or more classes closest to each other are selected and combined into a new class until all the objects participating in the clusters are combined into a total class (e.g., class 1 of one stage in fig. 3). After hierarchical clustering is finished, a plurality of layers can be selected as required to serve as a clustering result. For example, the clustering result 300 of FIG. 3 includes 3 levels of categories with parent-child relationships, including total categories.

In one embodiment of the present invention, the process of hierarchical clustering at least a portion of the products in the platform may be hierarchical clustering based on static features and dynamic features of at least a portion of the products in the platform. The static characteristics are used for describing static properties of at least one part of products, and the dynamic characteristics are used for describing dynamic properties of at least one part of products generated based on interactive behaviors of users in a circulation link. For example, static and dynamic features for each product in at least a portion of the products in the platform (e.g., all of the products sold) may be collected first. After collecting the static characteristics and the dynamic characteristics of the products, hierarchical clustering is carried out on all the products sold in the platform.

The static characteristics of the product may include, for example, the title of the product, the name of the category to which the product belongs, the brand, country of origin, seasonal labels, the gender of the user for which the product is intended, the price class of the last category in which the price of the product is located, the applicable age, etc.

The dynamic characteristics of the product may include characteristics extracted from data of a user's access to the product, such as a number of times a user accessed the product details page in one or more historical stages in the past, a number of times the user accessed the product review page, a number of times the user shared the product details page to a WeChat friend, a number of times the user viewed the product details module in the product details page, a number of times the user reached the product details page directly from the search results page, a number of times the user reached the product details page directly from the shopping cart page, a number of times the user added the product to the shopping cart, or a number of times the user collected the product, etc. Alternatively, the dynamic characteristics of the product may also include statistical data characteristics resulting from a large number of users accessing the product, such as the number of purchases, the number of orders purchased, the average price paid, or the average discount rate for the product in one or more historical stages in the past.

The clustering result 300 illustrated in fig. 3, including 3 levels, is only one illustration. In other embodiments, if there are more product categories in the online procurement platform, 4 layers, 5 layers, or even more layers may be aggregated. The last layer is the layer with the finest clustering granularity.

In operation S220, a first category to which a first product belongs in the at least one category is determined, wherein the first category includes long-term products, which are one or more products whose sales time satisfies a predetermined condition.

The predetermined condition satisfied by the sales time of the long-term product may be determined according to the operation of the platform or the size of the total amount of data. For example, in one embodiment, a product in the platform that is now on sale and that has been on sale two years ago may be determined to be a long-term product.

In one embodiment, the first product is included in the at least a portion of the products when clustered in operation S210, such that the first category determined in operation S220 may be a category in which the first product is clustered. In another embodiment, the first product may not be included in the at least a portion of the products when the clustering is performed in operation S210. For example, the first product is a new product that has not yet been sold when the products in the platform are clustered. For this case, the characteristics (e.g., static characteristics and dynamic characteristics) of the first product may be extracted when the first category is determined in operation S220, and then the first product may be classified into one category according to the characteristics of the first product.

The first category is a category that includes long-term products. According to the embodiment of the invention, if the classified category of the first product does not contain long-term products, the category containing the long-term products can be found out by searching the parent class of the category in the category result obtained by hierarchical clustering and used as the first category. In the category results obtained by other clustering methods, a category including long-term products can be obtained by combining the adjacent categories as a first category.

Of course, in a clustering result 300, such as illustrated in fig. 3, that is obtained through hierarchical clustering, the first category may not be unique. If the first product is categorized into three categories 312 in the finest granularity of classification, and three categories 312 contain long-term products, three categories 312 may be preferentially selected as the first category; or if the tertiary category 312 does not contain long-term products, but the secondary category 11 contains long-term products, the secondary category 11 may be selected as the first category. For clarity of description, the first category is described below as a three-level category 312.

In operation S230, a sales trend function of the long-term product is obtained based on the historical sales change trend of the long-term product in N stages, where the N stages include N basic timing units that are consecutive in time, where N is an integer greater than or equal to 2. The basic timing unit can be hours, days, weeks, months, quarters, years, etc., and is specifically determined according to the actual analysis requirements.

According to one embodiment of the invention, the basic timing unit is set to month and the N phases are set to N consecutive months. For example, when N takes a value of 12, N stages are 12 consecutive months, which is exactly one year. When N is taken for 24, N stages are continuous for 24 months, namely two years.

One example of a calculation of the sales trend function for a long-term product is illustrated by the product in the platform now being sold and the time of sale being two years ago, taking the last 24 months as an example.

At least a portion of the products (e.g., all products) in the platform that are now on sale and that are at a time of day two years ago are first sought, thereby finding all long-term products in the platform. Next, whether long-term products are included in each category in the clustered results 300 is screened. For a class with long-term products, for example, may be labeled 1, for which the characterization is subsequently calculated. While for a class that does not contain long-term products, for example, it may be marked 0, the characterization will not subsequently calculate for this class.

All long-term products within the platform are then acquired for historical sales measured in days over the past 24 months (e.g., which may correspond to two full years). For example, the total monthly sales for all long-term products within each of the categories labeled 1 described above may be counted month by month, resulting in a category total monthly sales.

For example, assume that there are four products A, B, C and D in the three-level category 312 in the clustered result 300, where A, B, C is a long-term product. The total monthly sales for the three-level category 312 category is the sum of the total monthly sales for each of the A, B, C three products in the three-level category 312, and can be calculated by the following equation:

category month total sales for tertiary category 312 = category average day sales day of the month.

Wherein, the daily average sales of the category is the sum of the daily average sales of the A, B, C products.

In practical application, the influence of random factors such as backorder and sales promotion on sales tendency of products can be eliminated in advance when the daily average sales of the products is calculated. Taking the daily average sales of A as an example, the monthly historical sales of A in a certain month can be obtained by statistics; wherein if the month is out of stock or is promoted greatly, the out-of-stock or the sales of the commodity A in the period of promotion are removed from the historical sales of the month of A, and the reserved month sales of the commodity A are obtained. The remaining monthly sales for A are then divided by the remaining days after the number of days of the month from which the major or backorder was removed, resulting in a daily average sales for product A.

The sales trend function for the long-term product A, B, C as a whole in the three-level category 312 is next calculated. After obtaining the total sales of the three-level class 312 over the last 24 months, the overall sales trend function for the long-term products A, B, C in the three-level class 312 may be obtained by trend line fitting. Trendline fitting is a long-term trend of time-series data represented by a trendline, which reflects the trend (e.g., increase, decrease, or remain unchanged) of a particular set of data over a period of time.

In one embodiment, linear regression may be utilized to calculate the sales trend function for the long-term products A, B, C in the three-level class 312 as a whole. Linear regression is a regression analysis that uses linear regression equations to model the relationship between one or more independent and dependent variables. For example, 24 months in the last two years may be numbered chronologically, numbered 1 through 24, with a linear regression of month numbers using the total sales of category months for three-level category 312 calculated month by month. For example, a sales trend function y=ax+b is obtained, where x is the month number and y is the total sales for the class months of the three-level class 312. It will be appreciated that linear regression is just one example of a fit in order to find the relationship between sales and months. In other embodiments, other fitting methods may also be used, which will not be described in detail.

In operation S240, M class time lifting factors of the first class are calculated, where M is a positive integer less than or equal to N, and the M class time lifting factors are in one-to-one correspondence with M phases of the N phases. Each of the M category time promotion factors is characterized by a correspondence between historical sales of the long-term product and trend sales calculated using the sales trend function over a corresponding phase.

In one embodiment, N is taken 24 and M is taken 12. That is, the sales trend function of the long-term product in the three-stage class 312 may be fitted using the data of the last two years, and the class time lifting factor may be calculated by calculating only the class time lifting factor corresponding to each of 12 months in one year in consideration of the periodicity of time, to obtain the class time lifting factor table referring to table 1 below.

TABLE 1

The class time boost factor for each stage in table 1 is used to characterize the short-term seasonal factor index of the products in the three-level class 312. For clarity of description, the following will be described by way of example using the example of N for 24 and M for 12.

According to an embodiment of the present invention, the calculating M class time lifting factors of the first class (i.e., the third class 312) in operation S240 may further include obtaining a stage corresponding to the class time lifting factor that does not satisfy the stability condition when there is a class time lifting factor that does not satisfy the stability condition in the M class time lifting factors, and obtaining a stage in which the lifting factor is pending; and determining a class time lifting factor meeting the stability condition, which corresponds to the lifting factor pending stage, in the parent class of the first class 312 as a class time lifting factor, which corresponds to the lifting factor pending stage, of the first class 312.

In one embodiment, the stability condition is set to a predetermined range of values. The predetermined range of values may be empirically set (e.g., 0.4-4). When the class time lifting factor exceeds this range of values, the calculated month factor may be considered very unstable, in which case the class time lifting factor satisfying the stability condition may be looked up from the parent class of the first class 312.

Assuming that the calculated class time of the three-level class 312 for a certain month (e.g., 2 months) increases by a factor a2=0.1, a2 < 0.4, and the stability condition is not satisfied. At this time, it can be checked whether the class time boost factor of 2 months of the parent class of the tertiary class 312 (i.e., the secondary class 11) satisfies the stability condition; if so, assigning a value of a class time boost factor of 2 months for the secondary class 11 to A2; if not, tracing to whether the class time lifting factor of 2 months of the parent class 1 of the secondary class 11 meets the stability condition. And if the category result has a plurality of levels, gradually tracing by pushing until a category time lifting factor meeting the stability condition is found.

In operation S250, based on the historical sales of the first product in S stages closest to the current stage and the class time lifting factors corresponding to the stages to be predicted, the predicted sales of the first product in the stages to be predicted are calculated, where S is a positive integer less than or equal to M, the M stages include the S stages, and the stages to be predicted are any one or more stages corresponding to the M stages in the future stages.

For example, when the first product is a new product, the amount of data that can be regressed is too small. In this case, it is considered that the influence of the trend factor on the sales change of the product in a future period of time is small (it is considered that there is no trend of rising or falling of the sales of the product in a short period of time). In this way, during prediction, the current sales level of the first product can be calculated as the reference sales according to the historical sales conditions of the S stages closest to the current time point, and then the sales of the first product in the stage to be predicted can be predicted by multiplying the reference sales by the class time lifting factor corresponding to the stage to be predicted.

Fig. 4 schematically shows a flowchart of a method for calculating M category time lifting factors of a first category in operation S240 in a sales volume prediction method according to an embodiment of the present invention.

As shown in fig. 4, operation S240 may include operations S241 to S244.

In operation S241, M stage history sales of the long-term product, which are in one-to-one correspondence with the M stages, are counted. For example, the historical sales for each month in the past 12 months for the long-term product A, B, C in the tertiary category 312 as a whole (i.e., the category month total sales for the aforementioned tertiary category 312) are counted.

In operation S242, M-stage trend sales of the long-term product, which are in one-to-one correspondence with the M stages, are calculated by the sales trend function. For example, the trending sales for each month over the past 12 months for the long-term product A, B, C in the three-level category 312 as a whole is calculated by the sales trending function y=ax+b.

In operation S243, based on the ratio of the historical sales of M phases to the historical sales of phases and the trend sales of phases corresponding to the same phase in the trend sales of M phases, the de-trend sales of M categories are obtained. For example, the three-level category 312 includes category trending sales per month in the past 12 months=historical sales per month counted in operation S241/trend sales per month counted in operation S242.

In operation S244, the trending sales amounts of the M categories are normalized, so as to obtain the M category time lifting factors. One embodiment of the normalization process may be referred to as illustrated in fig. 5.

Fig. 5 schematically illustrates a flowchart of a method for normalizing M category trending sales in operation S244 according to an embodiment of the present invention.

As shown in fig. 5, operation S244 may include operation S501 and operation S502.

In operation S501, an average value of the class trending sales is calculated, so as to obtain a class average trending sales. In operation S502, the M category detrending sales are divided by the category average detrending sales, respectively, to obtain the M category time lifting factors.

For example, M is 12, class average trending sales = sum of class trending sales of last 12 months of the year/12;

category time improvement factor per month = category trending sales per month last year/category average trending sales per month.

Fig. 6 schematically shows a flowchart of a method of predicting sales in operation S250 in a sales prediction method according to an embodiment of the present invention.

As shown in fig. 6, operation S250 may include operations S251 to S256 according to an embodiment of the present invention.

First, in operation S251, a sum of the historic sales of the S stages closest to the current is calculated. For example, assuming a value of 3 for S, the Total Sales for the last three months of the three-level category 312 is calculated. Assuming that the current month is 8, the last three months are 5, 6, 7.

Then, in operation S252, S category time lifting factors corresponding to the S phases one to one are selected from the M category time lifting factors. For example, category time boost factors A5, A6, A7 of 5,6,7 three months may be selected from table 1.

Next, in operation S253, the sum of the S category time lifting factors is calculated, so as to obtain the sum of the lifting factors of the first product in the S phases. For example, calculate the lifting factor integrated SUM (5, 6, 7) =a5+a6+a7 for three months 5,6, 7.

Then, in operation S254, the reference sales of the first product in each of the S stages closest to the current stage is obtained by dividing the sum of the historic sales of the S stages closest to the current stage by the sum of the lifting factors. For example, reference samples=Total salts/SUM (5, 6, 7).

Then, in operation S255, it is determined that the stage to be predicted corresponds to a first stage of the M stages, and the class time lifting factor corresponding to the first stage is determined as the class time lifting factor corresponding to the stage to be predicted.

And then in operation S256, determining the predicted sales of the first product in the stage to be predicted according to the reference sales and the category time promotion factors corresponding to the stage to be predicted. The phase to be predicted may be one phase of the M phases, or a plurality of phases. For example, the phase to be predicted may be 8 months-one month, or the phase to be predicted may be a third quarter (including 8, 9, 10 months).

If the stage to be predicted is 8 months and one month, it can be determined that the class time lifting factor of 8 months is A8 according to table 1. So that sales = Reference samples A8 for the first product for 8 months are predicted in operation S256.

If the phase to be predicted is the third quarter, including three months of 8, 9 and 10. From table 1, it can be determined that the class time boost factors of three months 8, 9, 10 are A8, A9, a10, respectively. The predicted sales of 8, 9, 10 months each can be calculated by using the reference sales and the category time lifting factors A8, A9, A10, and then the predicted sales of 8, 9, 10 months each are added to obtain the predicted sales of the third quarter.

According to the sales volume predicting method provided by the embodiment of the invention, the accuracy of new sales volume prediction can be greatly improved, and the problem of inaccurate demand prediction caused by insufficient data of new historical sales volume is solved to a certain extent. Therefore, the method can assist the operation to reasonably make a sales plan and avoid risks brought by high inventory.

According to the sales volume prediction method provided by the embodiment of the invention, the whole process is automatically realized, the manual intervention is small, the labor cost required for making a sales plan is greatly reduced, and the risk caused by errors generated by manual experience calculation in the past is avoided.

According to the sales predicting method provided by the embodiment of the invention, sales prediction can be automatically performed on all products, so that the blank of sales prediction of products which are not concerned by operation can be filled, and scientific formulation of sales plans and inventory plans can be assisted.

Exemplary apparatus

Having described the method of the exemplary embodiment of the present invention, next, a sales predicting apparatus of the exemplary embodiment of the present invention will be described with reference to fig. 7 to 9.

Fig. 7 schematically shows a block diagram of a sales predicting apparatus 700 according to an embodiment of the present invention.

As shown in fig. 7, the sales predicting apparatus 700 may include a clustering module 710, a category determining module 720, a trend function obtaining module 730, a category time lifting factor calculating module 740, and a predicting module 750.

The clustering module 710 may, for example, perform operation S210 for clustering at least a portion of the products to obtain at least one category. According to one embodiment of the present invention, the clustering module 710 may be specifically configured to perform hierarchical clustering based on static features and dynamic features of the at least a portion of the products, where the static features are used to describe static attributes of the at least a portion of the products, and the dynamic features are used to describe dynamic attributes of the at least a portion of the products generated during a circulation link based on interactive behaviors of users, and the hierarchical clustering includes multiple levels of categories with parent-child relationships.

The category determination module 720 may, for example, perform operation S220 for determining a first category to which a first product belongs in the at least one category, wherein the first category includes long-term products, which are one or more products whose sales time satisfies a predetermined condition.

The trend function obtaining module 730 may, for example, perform operation S230, configured to obtain a sales trend function of the long-term product based on the historical sales variation trends of the long-term product in N stages, where the N stages include N basic time units that are consecutive in time, where N is an integer greater than or equal to 2. According to an embodiment of the invention, the basic timing unit is set to months and the N phases are set to N consecutive months, for example 24 consecutive months.

The class time lifting factor calculation module 740 may, for example, perform operation S240, configured to calculate M class time lifting factors of the first class, where M is a positive integer less than or equal to N, and the M class time lifting factors are in one-to-one correspondence with M phases of the N phases; each of the M category time promotion factors is characterized by a correspondence between historical sales of the long-term product and trend sales calculated using the sales trend function over a corresponding phase. For example, n=24, m=12, N phases may be 24 months in two years, and M phases may be 12 months in one year.

The prediction module 750 may, for example, perform operation S250, configured to calculate, based on the historical sales of the first product in S stages closest to the current stage and the class time lifting factor corresponding to the stage to be predicted, the predicted sales of the first product in the stage to be predicted, where S is a positive integer less than or equal to M, and the M stages include the S stages, and the stage to be predicted is any one or more stages corresponding to the M stages in a future stage.

Fig. 8 schematically illustrates a block diagram of the category time lifting factor calculation module 740 in fig. 7 according to an embodiment of the invention.

As shown in fig. 8, the category time lifting factor calculation module 740 may include a statistics sub-module 741, a trend sales calculation sub-module 742, a category trending sales calculation sub-module 743, and a category time lifting factor calculation sub-module 744.

The statistics sub-module 741 may perform, for example, operation S241 for counting M-stage history sales of the long-term product that are in one-to-one correspondence with the M stages.

The trend sales computing operator module 742 may, for example, perform operation S242 for computing M-stage trend sales of the long-term product, which are one-to-one corresponding to the M stages, through the sales trend function;

The category trending sales calculation operator module 743 may, for example, perform operation S243, configured to obtain M category trending sales based on a ratio of the M category historical sales to a category historical sales and a category trend sales corresponding to the same category in the M category trend sales; and

the class time lifting factor calculation sub-module 744 may, for example, perform operation S244, configured to normalize the trending sales of the M classes to obtain the M class time lifting factors. According to an embodiment of the present invention, the class time lifting factor calculating sub-module 744 may specifically perform operations S501 and S502, configured to calculate an average value of the class trending sales, obtain class average trending sales, and divide the class average trending sales by the class average trending sales to obtain the class time lifting factors.

According to another embodiment of the present invention, the category time lifting factor calculation module 740 further includes a trace back sub-module 745. The trace back sub-module 745 is configured to obtain a stage corresponding to the class time lifting factor that does not satisfy the stability condition when there is a class time lifting factor that does not satisfy the stability condition in the M class time lifting factors, obtain a stage with a pending lifting factor, and determine, as the class time lifting factor corresponding to the pending lifting factor stage, a class time lifting factor that satisfies the stability condition in a parent class of the first class and corresponds to the pending lifting factor stage. According to one embodiment of the invention, the stability condition is set to a predetermined range of values.

Fig. 9 schematically illustrates a block diagram of the prediction module 750 of fig. 7 according to an embodiment of the present invention.

As shown in fig. 9, the prediction module 750 may include a prediction sales volume calculation sub-module 751, a category time lifting factor selection sub-module 752, a prediction lifting factor sum calculation module 753, a reference sales volume determination sub-module 754, a stage to be predicted category time lifting factor determination module 755, and a prediction sales volume determination sub-module 756.

The prediction sales sum calculation sub-module 751 may, for example, perform operation S251 for calculating a sum of the historical sales for the S phases that are currently closest to the current.

The class time lifting factor selection sub-module 752 may, for example, perform operation S252 for selecting S class time lifting factors corresponding to the S phases one by one from the M class time lifting factors.

The prediction lifting factor sum calculating module 753 may, for example, perform operation S253, configured to calculate the sum of the S category time lifting factors, to obtain a sum of the lifting factors of the first product in the S phases.

The reference sales determination sub-module 754 may, for example, perform operation S254 for obtaining a reference sales for the first product at each of the S stages that are currently closest to the first product by dividing the sum of the historical sales for the S stages that are currently closest to the first product by the sum of the lifting factors.

The to-be-predicted stage class time lifting factor determining module 755 may, for example, perform operation S255, configured to determine that the to-be-predicted stage corresponds to a first stage of the M stages, and determine a class time lifting factor corresponding to the first stage as a class time lifting factor corresponding to the to-be-predicted stage.

The predicted sales determination submodule 756 may, for example, perform operation S256 for determining a predicted sales of the first product at the stage to be predicted based on the reference sales and a category time boost factor corresponding to the stage to be predicted.

Exemplary Medium

Having described the method and apparatus of an exemplary embodiment of the present invention, next, a medium suitable for implementing the sales prediction method of an exemplary embodiment of the present invention will be described with reference to fig. 10.

Embodiments of the present disclosure also provide a computer-readable storage medium having stored thereon executable instructions that, when executed by a processor, cause the processor to perform the sales prediction method described with reference to fig. 2-6.

In some possible implementations, aspects of the invention may also be implemented in the form of a program product comprising program code for causing a computing apparatus to perform operations in a sales prediction method according to various exemplary embodiments of the invention as described in the above-mentioned "exemplary methods" section of the specification, when the program product is run on the computing apparatus, e.g. the computing apparatus may perform operation S210 as shown in fig. 2, clustering at least a portion of the products to obtain at least one category; operation S220, determining a first category to which a first product belongs in the at least one category, wherein the first category includes long-term products, which are one or more products whose sales time satisfies a predetermined condition; operation S230, obtaining a sales trend function of the long-term product based on the historical sales change trends of the long-term product in N stages, where the N stages include N basic time units that are consecutive in time, where N is an integer greater than or equal to 2; operation S240, calculating M class time lifting factors of the first class, where M is a positive integer less than or equal to N, where the M class time lifting factors are in one-to-one correspondence with M phases of the N phases; each of the M category time promotion factors is characterized by a correspondence between historical sales of the long-term product and trend sales calculated using the sales trend function in a corresponding stage; and operation S250, calculating a predicted sales amount of the first product in the stage to be predicted based on the historical sales amount of the first product in S stages closest to the current stage and a class time lifting factor corresponding to the stage to be predicted, where S is a positive integer less than or equal to M, the M stages include the S stages, and the stage to be predicted is any one or more stages corresponding to the M stages in a future stage.

The program product may employ any combination of one or more readable media. The readable medium may be a readable signal medium or a readable storage medium. The readable storage medium can be, for example, but not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or a combination of any of the foregoing. More specific examples (a non-exhaustive list) of the readable storage medium would include the following: an electrical connection having one or more wires, a portable disk, a hard disk, random Access Memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

As shown in FIG. 10, a program product 1000 suitable for implementing a sales prediction method according to an embodiment of the present invention is described, which may employ a portable compact disc read only memory (CD-ROM) and include program code, and may be run on a computing device, such as a personal computer. However, the program product of the present invention is not limited thereto, and in this document, a readable storage medium may be any tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device.

The readable signal medium may include a data signal propagated in baseband or as part of a carrier wave with readable program code embodied therein. Such a propagated data signal may take any of a variety of forms, including, but not limited to, electro-magnetic, optical, or any suitable combination of the foregoing. A readable signal medium may also be any readable medium that is not a readable storage medium and that can communicate, propagate, or transport a program for use by or in connection with an instruction execution system, apparatus, or device.

Program code embodied on a readable medium may be transmitted using any appropriate medium, including but not limited to wireless, wireline, optical fiber cable, RF, etc., or any suitable combination of the foregoing.

Program code for carrying out operations of the present invention may be written in any combination of one or more programming languages, including an object oriented programming language such as Java, C++ or the like and conventional procedural programming languages, such as the "C" programming language or similar programming languages. The program code may execute entirely on the user's computing device, partly on the user's device, as a stand-alone software package, partly on the user's computing device, partly on a remote computing device, or entirely on the remote computing device or server. In the case of remote computing devices, the remote computing device may be connected to the user computing device through any kind of network, including a Local Area Network (LAN) or a Wide Area Network (WAN), or may be connected to an external computing device (e.g., connected through the Internet using an Internet service provider).

Exemplary computing device

Having described the methods, apparatus and media of exemplary embodiments of the present invention, a computing device suitable for implementing the sales prediction method of exemplary embodiments of the present invention is next described with reference to FIG. 11.

Embodiments of the present invention also provide a computing device. The computing device includes one or more storage units, and one or more processing units. The one or more memory units store executable instructions. The one or more processing units execute the executable instructions to implement the sales prediction method described with reference to fig. 2-6

The embodiment of the invention also provides a computing device. Those skilled in the art will appreciate that the various aspects of the invention may be implemented as a system, method, or program product. Accordingly, aspects of the invention may be embodied in the following forms, namely: an entirely hardware embodiment, an entirely software embodiment (including firmware, micro-code, etc.) or an embodiment combining hardware and software aspects may be referred to herein as a "circuit," module "or" system.

In some possible implementations, a computing device according to the invention may include at least one processing unit, and at least one memory unit. Wherein the storage unit stores program code that, when executed by the processing unit, causes the processing unit to perform the operations in the sales prediction method according to various exemplary embodiments of the present invention described in the section "exemplary method" above of the present specification. For example, the processing unit may perform operation S210 as shown in fig. 2, clustering at least a portion of the products to obtain at least one category; operation S220, determining a first category to which a first product belongs in the at least one category, wherein the first category includes long-term products, which are one or more products whose sales time satisfies a predetermined condition; operation S230, obtaining a sales trend function of the long-term product based on the historical sales change trends of the long-term product in N stages, where the N stages include N basic time units that are consecutive in time, where N is an integer greater than or equal to 2; operation S240, calculating M class time lifting factors of the first class, where M is a positive integer less than or equal to N, where the M class time lifting factors are in one-to-one correspondence with M phases of the N phases; each of the M category time promotion factors is characterized by a correspondence between historical sales of the long-term product and trend sales calculated using the sales trend function in a corresponding stage; and operation S250, calculating a predicted sales amount of the first product in the stage to be predicted based on the historical sales amount of the first product in S stages closest to the current stage and a class time lifting factor corresponding to the stage to be predicted, where S is a positive integer less than or equal to M, the M stages include the S stages, and the stage to be predicted is any one or more stages corresponding to the M stages in a future stage.

A computing device 1100 suitable for implementing the sales prediction method according to an embodiment of the present invention is described below with reference to fig. 11. The computing device 1100 shown in fig. 11 is merely an example and should not be taken as limiting the functionality and scope of use of embodiments of the present invention.

As shown in fig. 11, computing device 1100 is in the form of a general purpose computing device. Components of computing device 1100 may include, but are not limited to: the at least one processing unit 1110, the at least one memory unit 1120, a bus 1130 connecting the different system components, including the memory unit 1120 and the processing unit 1110.

Bus 1130 includes a data bus, a control bus, and an address bus.

The storage unit 1120 may include a readable medium in the form of volatile memory, such as Random Access Memory (RAM) 1121 and/or cache memory 1122, and may further include Read Only Memory (ROM) 1123.

Storage unit 1120 may also include a program/utility 1125 having a set (at least one) of program modules 1124, such program modules 1124 including, but not limited to: an operating system, one or more application programs, other program modules, and program data, each or some combination of which may include an implementation of a network environment.

Computing device 1100 can also communicate with one or more external devices 1140 (e.g., keyboard, pointing device, bluetooth device, etc.). Such communication may occur through an input/output (I/0) interface 1150. Moreover, computing device 1100 may also communicate with one or more networks such as a Local Area Network (LAN), a Wide Area Network (WAN), and/or a public network, such as the Internet, through a network adapter 1160. As shown, network adapter 1160 communicates with other modules of computing device 1100 via bus 1130. It should be appreciated that although not shown, other hardware and/or software modules may be used in connection with computing device 1100, including, but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, data backup storage systems, and the like.

It should be noted that although several units/modules or sub-units/modules of the apparatus are mentioned in the above detailed description, this division is merely exemplary and not mandatory. Indeed, the features and functionality of two or more units/modules described above may be embodied in one unit/module in accordance with embodiments of the present invention. Conversely, the features and functions of one unit/module described above may be further divided into ones that are embodied by a plurality of units/modules.

Furthermore, although the operations of the methods of the present invention are depicted in the drawings in a particular order, this is not required to either imply that the operations must be performed in that particular order or that all of the illustrated operations be performed to achieve desirable results. Additionally or alternatively, certain steps may be omitted, multiple steps combined into one step to perform, and/or one step decomposed into multiple steps to perform.

While the spirit and principles of the present invention have been described with reference to several particular embodiments, it is to be understood that the invention is not limited to the disclosed embodiments nor does it imply that features of the various aspects are not useful in combination, nor are they useful in any combination, such as for convenience of description. The invention is intended to cover various modifications and equivalent arrangements included within the spirit and scope of the appended claims.

Claims

1. A sales prediction method, comprising:

clustering at least a part of the products to obtain at least one category;

determining a first category to which a first product belongs in the at least one category, wherein the first category includes long-term products, the long-term products being one or more products whose sales time meets a predetermined condition;

Acquiring a sales trend function of the long-term product based on historical sales change trends of the long-term product in N stages, wherein the N stages comprise N time-sequential basic timing units, and N is an integer greater than or equal to 2;

calculating M class time lifting factors of the first class, wherein M is a positive integer less than or equal to N, and the M class time lifting factors are in one-to-one correspondence with M phases in the N phases; each of the M category time promotion factors is characterized by a correspondence between historical sales of the long-term product and trend sales calculated using the sales trend function in a corresponding stage;

calculating the predicted sales of the first product in the stage to be predicted based on the historical sales of the first product in S stages closest to the current stage and the class time lifting factors corresponding to the stage to be predicted, wherein S is a positive integer less than or equal to M, the M stages comprise the S stages, and the stage to be predicted is any one or more stages corresponding to the M stages in a future stage.

2. The method of claim 1, wherein the calculating M category time boost factors for the first category comprises:

Counting M stage history sales of the long-term product, which are in one-to-one correspondence with the M stages;

calculating M stage trend sales of the long-term product, which are in one-to-one correspondence with the M stages, through the sales trend function;

obtaining M category trending sales based on the ratio of the M phase historical sales to the phase historical sales and the phase trend sales corresponding to the same phase in the M phase trend sales; and

and carrying out normalization processing on the trending sales of the M categories to obtain the M category time lifting factors.

3. The method of claim 2, wherein the normalizing the M category trending sales volumes to obtain the M category time lifting factors comprises:

calculating the average value of the trending sales of the M categories to obtain the average trending sales of the categories;

dividing the M category trending sales volume by the category average trending sales volume to obtain the M category time lifting factors.

4. The method of claim 1, wherein clustering at least a portion of the products to obtain at least one category comprises:

hierarchical clustering is performed based on static features and dynamic features of the at least one part of products, wherein the static features are used for describing static attributes of the at least one part of products, the dynamic features are used for describing dynamic attributes of the at least one part of products generated in a circulation link based on interactive behaviors of users, and the hierarchical clustering comprises multi-level categories with father-son relations.

5. The method of claim 4, wherein the calculating M category time boost factors for the first category further comprises:

when the class time lifting factors which do not meet the stability conditions exist in the M class time lifting factors, acquiring a stage corresponding to the class time lifting factors which do not meet the stability conditions, and acquiring a stage in which the lifting factors are undetermined;

and determining a class time lifting factor meeting the stability condition, which corresponds to the lifting factor pending stage, in the parent class of the first class as a class time lifting factor, which corresponds to the lifting factor pending stage, of the first class.

6. The method of claim 1, wherein the calculating the predicted sales of the first product in the stage to be predicted based on the historical sales of the first product in S stages closest to the current stage and the class time boost factor corresponding to the stage to be predicted comprises:

calculating the sum of the historical sales of the S stages closest to the current stage;

selecting S category time lifting factors corresponding to the S phases one by one from the M category time lifting factors;

calculating the sum of the S category time lifting factors to obtain the sum of the lifting factors of the first product in the S stages;

Dividing the sum of the historical sales of the S stages closest to the current time by the sum of the lifting factors to obtain the reference sales of the first product in each stage of the S stages closest to the current time;

determining that the stage to be predicted corresponds to a first stage of M stages, and determining a class time lifting factor corresponding to the first stage as a class time lifting factor corresponding to the stage to be predicted; and

and determining the predicted sales of the first product in the stage to be predicted according to the reference sales and the class time lifting factors corresponding to the stage to be predicted.

7. The method of claim 1, wherein the base timing unit is set to months and the N phases are set to N consecutive months.

8. A sales predicting apparatus comprising:

the clustering module is used for clustering at least one part of products to obtain at least one category;

a category determination module for determining a first category to which a first product belongs in the at least one category, wherein the first category includes a long-term product, the long-term product being one or more products whose sales time satisfies a predetermined condition;

The trend function acquisition module is used for acquiring sales trend functions of the long-term products based on historical sales change trends of the long-term products in N stages, wherein the N stages comprise N basic timing units which are sequential in time, and N is an integer greater than or equal to 2;

the class time lifting factor calculation module is used for calculating M class time lifting factors of the first class, wherein M is a positive integer less than or equal to N, and the M class time lifting factors are in one-to-one correspondence with M phases in the N phases; each of the M category time promotion factors is characterized by a correspondence between historical sales of the long-term product and trend sales calculated using the sales trend function in a corresponding stage;

the prediction module is used for calculating the predicted sales of the first product in the to-be-predicted stage based on the historical sales of the first product in S stages closest to the current stage and the class time lifting factors corresponding to the to-be-predicted stage, wherein S is a positive integer less than or equal to M, the M stages comprise the S stages, and the to-be-predicted stage is any one or more stages corresponding to the M stages in a future stage.

9. A computing device, comprising:

one or more memory units storing executable instructions;

one or more processing units executing the executable instructions to implement the method of any one of claims 1-7.

10. A computer readable storage medium having stored thereon executable instructions which, when executed by a processor, cause the processor to perform the method according to any of claims 1-7.