CN114387037A - Retail commodity sales prediction method based on deep learning - Google Patents
Retail commodity sales prediction method based on deep learning Download PDFInfo
- Publication number
- CN114387037A CN114387037A CN202210054835.XA CN202210054835A CN114387037A CN 114387037 A CN114387037 A CN 114387037A CN 202210054835 A CN202210054835 A CN 202210054835A CN 114387037 A CN114387037 A CN 114387037A
- Authority
- CN
- China
- Prior art keywords
- sales
- region
- development
- model
- development state
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0201—Market modelling; Market analysis; Collecting market data
- G06Q30/0202—Market predictions or forecasting for commercial activities
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/06—Buying, selling or leasing transactions
Landscapes
- Business, Economics & Management (AREA)
- Accounting & Taxation (AREA)
- Finance (AREA)
- Strategic Management (AREA)
- Development Economics (AREA)
- Engineering & Computer Science (AREA)
- General Business, Economics & Management (AREA)
- Physics & Mathematics (AREA)
- Marketing (AREA)
- General Physics & Mathematics (AREA)
- Economics (AREA)
- Theoretical Computer Science (AREA)
- Entrepreneurship & Innovation (AREA)
- Data Mining & Analysis (AREA)
- Game Theory and Decision Science (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The embodiment of the specification provides a retail commodity sales prediction method based on deep learning, which comprises the following steps: acquiring a development state vector of a region to be predicted in at least one time period; acquiring a region feature sequence of a region to be predicted based on the development state vector of at least one time period; and processing the region characteristic sequence of the region to be predicted through a sales prediction model, and predicting the commodity sales of the region to be predicted.
Description
Technical Field
The specification relates to the field of sales prediction, in particular to a retail commodity sales prediction method based on deep learning.
Background
With the development of society, factors affecting the sales volume of commodities become increasingly complex, the amount of data related to the sales of commodities and the sales volume becomes enormous, and the manner of predicting the sales volume of commodities based on only a mathematical model gradually fails to satisfy increasingly complex prediction conditions.
Accordingly, it is desirable to provide a retail goods sales prediction method based on deep learning to improve the accuracy and efficiency of prediction of future sales.
Disclosure of Invention
One embodiment of the present specification provides a retail goods sales forecasting method. The retail commodity sales prediction method comprises the following steps: acquiring a development state vector of a region to be predicted in at least one time period; acquiring a region feature sequence of the region to be predicted based on the development state vector of the at least one time segment; and processing the region characteristic sequence of the region to be predicted through a sales prediction model, and predicting the commodity sales of the region to be predicted.
One embodiment of the present description provides a retail goods sales forecasting system. The retail merchandise sales prediction system includes: the device comprises an acquisition module, a prediction module and a prediction module, wherein the acquisition module is used for acquiring a development state vector of an area to be predicted in at least one time period; the processing module is used for acquiring a region feature sequence of the region to be predicted based on the development state vector of the at least one time segment; and the prediction module is used for processing the region characteristic sequence of the region to be predicted through a sales prediction model and predicting the commodity sales of the region to be predicted.
One embodiment of the present specification provides a retail sales forecasting apparatus. The retail commodity sales prediction apparatus comprises a processor for executing the retail commodity sales prediction method of any one of the above.
One of the embodiments of the present specification provides a computer-readable storage medium. The storage medium stores computer instructions, and after the computer reads the computer instructions in the storage medium, the computer executes any one of the retail commodity sales prediction methods.
Drawings
The present description will be further explained by way of exemplary embodiments, which will be described in detail by way of the accompanying drawings. These embodiments are not intended to be limiting, and in these embodiments like numerals are used to indicate like structures, wherein:
FIG. 1 is a schematic diagram of an application scenario of a retail merchandise sales forecasting system in accordance with some embodiments of the present description;
FIG. 2 is an exemplary flow chart of a retail merchandise sales forecasting method according to some embodiments herein;
FIG. 3 is a schematic illustration of a sales prediction model and a method of training the same according to some embodiments of the present description;
FIG. 4 is a schematic diagram of a method of obtaining a first developmental state model according to some embodiments of the present description;
FIG. 5 is an exemplary block diagram of a retail merchandise sales prediction system according to some embodiments of the present description.
Detailed Description
In order to more clearly illustrate the technical solutions of the embodiments of the present disclosure, the drawings used in the description of the embodiments will be briefly described below. It is obvious that the drawings in the following description are only examples or embodiments of the present description, and that for a person skilled in the art, the present description can also be applied to other similar scenarios on the basis of these drawings without inventive effort. Unless otherwise apparent from the context, or otherwise indicated, like reference numbers in the figures refer to the same structure or operation.
It should be understood that "system", "apparatus", "unit" and/or "module" as used herein is a method for distinguishing different components, elements, parts, portions or assemblies at different levels. However, other words may be substituted by other expressions if they accomplish the same purpose.
As used in this specification and the appended claims, the terms "a," "an," "the," and/or "the" are not intended to be inclusive in the singular, but rather are intended to be inclusive in the plural, unless the context clearly dictates otherwise. In general, the terms "comprises" and "comprising" merely indicate that steps and elements are included which are explicitly identified, that the steps and elements do not form an exclusive list, and that a method or apparatus may include other steps or elements.
Flow charts are used in this description to illustrate operations performed by a system according to embodiments of the present description. It should be understood that the preceding or following operations are not necessarily performed in the exact order in which they are performed. Rather, the various steps may be processed in reverse order or simultaneously. Meanwhile, other operations may be added to the processes, or a certain step or several steps of operations may be removed from the processes.
In the field of retail goods, since the sales volume of goods is influenced by various factors, demands for the sales volume in different areas and different time periods are changed, but in order to supply goods required by consumers in time when demands occur, the sales of goods need to be predicted.
In some embodiments, the supply is typically based on historical sales for an area. For example, the supply is based on the previous sales. When the historical sales condition reflects good sales of the commodities, the quantity of the commodities is consciously increased when the commodities are supplied next time; conversely, when the commodity is lost, the quantity of the commodity is reduced at the next supply.
However, the accuracy of determining future offerings based on historical sales is not high, and often the goods are out of stock or lost due to misforecasting. Also, the sales of goods are affected by various factors, and thus, the difficulty of sales prediction is greatly increased.
In view of this, some embodiments of the present disclosure provide a retail goods sales prediction method and system based on deep learning, so as to improve the accuracy and practicability of sales prediction.
FIG. 1 is a schematic diagram of an application scenario of a retail merchandise sales forecasting system according to some embodiments of the present description. As shown in fig. 1, an application scenario of the retail goods sales forecasting system based on deep learning may include a server 110, a processor 120, a storage device 130, a user terminal 140, a network 150, a to-be-tested history 160-1 of a to-be-tested area 160 and/or a reference history 170-1 of a reference area 170, and the like.
The retail goods sales prediction system may be used in a sales service platform. In some embodiments, the system comprises a sales service platform for retail goods. Such as e-commerce platforms, vending machines, etc. The retail commodity sales predicting system can make a prediction of the retail commodity sales amount by the retail commodity sales predicting system disclosed in the present specification.
The server 110 may communicate with the processor 120, the storage device 130, and the user terminal 140 through the network 150 to provide various functions of retail goods sales volume prediction, and the storage device 130 may store all information of the retail goods sales process, for example, characteristics of sales regions, marketing strategy characteristics, goods characteristics, and the like. In some embodiments, the user terminal 140 may send the information of the commodity and the information of the area to be predicted to the server 110, and receive feedback information about the sales amount of the commodity in the area to be predicted from the server 110. The server 110 may obtain information related to the sales of retail goods, process it, and transmit sales volume prompting information to the user terminal 140. The above information transfer relationship between the devices is merely an example, and the present application is not limited thereto.
The server 110 may be used to manage historical sales data and process data and/or information from at least one component of the present system or an external data source (e.g., a cloud data center). In some embodiments, the server 110 may be a single server or a group of servers. The set of servers can be centralized or distributed (e.g., the servers 110 can be a distributed system), can be dedicated, or can be serviced by other devices or systems at the same time. In some embodiments, the server 110 may be regional or remote. In some embodiments, the server 110 may be implemented on a cloud platform, or provided in a virtual manner.
The network 150 may connect the various components of the system and/or connect the system with external resource components. The network 150 allows communication between the various components and with other components outside the system. For example, processor 120 may obtain, via network 150, a history 160-1 of the area under test 160 and a reference history 170-1 of the reference area 170. In some embodiments, the processor 120 may also transmit the predicted sales of the goods to the user terminal 140 via the network 150.
The historical conditions to be tested 160-1 may include historical sales related conditions for the area to be tested. In some embodiments, the historical conditions to be measured can be used to predict the sales volume of goods in the area to be measured. For specific contents of predicting the sales volume of the goods in the area to be tested, refer to fig. 2 and the related description thereof, which are not repeated herein. In some embodiments, the historical conditions to be measured may include commodity sales, per-capita GDP, education, industry distribution, per-capita age and/or per-capita tax, market characteristics, and the like over one or more time periods. In some embodiments, the past condition of a certain period of time may be intercepted as the historical condition to be measured. For example, sales of goods in the previous year, the previous three years, etc.
The area to be measured 160 may refer to an area for which commodity sales amount prediction is required. In some embodiments, the region under test may be any region. Including but not limited to a cell, county, administration, district, and/or province, etc. In some embodiments, the area to be measured may be determined based on characteristics of the article. For example, one or more cells may be used as regions for predicting good selling goods. For example, a product with a small demand amount may be predicted in a region of a district or province.
The reference history 170-1 may include a reference area historically associated with sales. The reference history may be similar to the history to be measured. For example, the per-person GDP of the reference area, the educational level of the reference area, the industry distribution of the reference area, the per-person age and/or per-person tax of the reference area, and the market characteristics of the reference area, etc.
The reference area 170 may refer to an area for referring to the sales amount of the goods. For example, an area for reference sales volume for the area to be measured. In some embodiments, the historical condition of commodity sales for the reference area may be similar to the historical condition of commodity sales for the area under test.
FIG. 2 is an exemplary flow chart of a retail merchandise sales forecasting method according to some embodiments of the present description. As shown in fig. 2, the process 200 may include one or more of the following steps. In some embodiments, one or more steps in flow 200 may be performed by server 130 in fig. 1.
For specific contents of the region to be predicted, refer to fig. 1 and the related description thereof, which are not repeated herein.
The development status vector may be a sequence of vectors that refer to a social development status over a period of time. In some embodiments, the progression state vector sequence may be a vector sequence associated with the underlying progression feature. The base development features may include one or more pieces of information that affect the state of social development, including but not limited to, average person GDP, education, industry distribution, and/or average person income, among others. In some embodiments, the sequence of development status vectors may be used to characterize the social development status, thereby predicting sales of retail goods based on the social development status. In some embodiments, the sequence of development state vectors may characterize the extent to which individual social underlying development features have an impact on the sales of the retail commodity over a period of time.
In some embodiments, a sequence of development state vectors may be obtained based on a time period of basic development characteristics. For example, a past month sequence of development state vectors is obtained based on a plurality of basic development features of the past month. In some embodiments, a sequence of development state vectors may be obtained based on the underlying development features for a plurality of time periods. For example, according to the basic development characteristics of each month in the past 3 months, the development state vectors of the past 3 months are respectively obtained, and the development state vectors of the 3 months are weighted and summed to determine the final development state vector of the past 3 months. In some embodiments, multiple progression state vector sequences may also be obtained. For example, a plurality of development state vectors are acquired from a plurality of time segments. In some embodiments, the time period may also be other lengths of time including, but not limited to, days, months, and/or years, etc. In some embodiments, the time period may be selected in various possible ways, including but not limited to continuous selection or decimation.
In some embodiments, the sequence of development state vectors may be obtained by a first development state model. For example, the base development features are input into a first development state model, which the model processes to output a development state vector, wherein the machine learning model includes, but is not limited to, an LSTM model, a BERT model, and/or an RNN model, etc.
In some embodiments, the base development characteristics of the region to be predicted in at least one time segment may be processed by the first development state model, and a development state vector sequence of at least one time segment may be determined. In some embodiments, the elements in the development state vector determined by the first development state model correspond to basic development features, for example, the averaged human GDP, education degree, industry distribution, and income of 2017 in the a region are input into the first development state model to obtain a development state vector (a1, a2, A3, a4), where a1, a2, A3, and a4 correspond to the averaged human GDP, education degree, industry distribution, and income, respectively, and the values of a1, a2, A3, and a4 may characterize the influence degree of each basic development feature in the a region on commodity sales.
For specific contents of the first development state model, refer to fig. 3 and its related description, which are not repeated herein.
The regional characteristic sequence may include characteristics relating to sales over a period of time for the region. In some embodiments, the elements in the sequence of region features may include one or more of a sales volume for at least one time period, and related features of the sales volume for the at least one time period, and a development status vector for the at least one time period, and the like. For example, the characteristic sequence of the A region 2017-2019 region may be
Wherein different columns represent different characteristics of the area, for example, A, B, C can be a basic development characteristic, D, E can represent a sales characteristic, F can represent a related characteristic of the sales, etc., and different rows represent different times within a segment at any time, for example, the above area characteristic sequence shows the characteristics of the area A in 2017 and 2019.
The sales volume in a time period may refer to the sales of the goods over a certain period of time. For example, the sales volume of article a in 2017. In some embodiments, the sales of the commodities may be counted in various feasible ways, including but not limited to sales of certain types of commodities, sales of certain commodities, joint sales of certain several commodities, and the like. In some embodiments, the sales of the goods may be obtained in various feasible ways, including but not limited to obtaining the sales of the dealers by region, and the like.
The related characteristics of the sales amount may include a variation of the sales amount during the period, a size of the sales amount, a difference in the sales amount of different kinds of commodities, a relation of the sales amount to a selling price of the commodity, a relation of the sales amount to a season, and/or a relation of the sales amount to a holiday, etc. In some embodiments, the characteristics related to sales may be obtained in various possible ways.
In some embodiments, the sequence of region features may be obtained based on the quantity of sales, the quantity of sales related features, and the sequence of development state vectors over at least one time period. In some embodiments, the processing module 520 may perform concatenation on at least one of a development state vector sequence, a sales feature sequence, and a sales related feature sequence in the same time period of the same region, and the region feature sequence may be obtained through concatenation of the feature sequences. For example, for region A, the progression state vector sequence is
The characteristic sequence of the sales volume is
The characteristic sequence related to the sales volume is
The region characteristic sequence of the A region is the result of the concatenation of the above sequences, i.e.
The above description is only an example and not a limitation, and the sales characteristic sequence and the sales related characteristic sequence may include at least one characteristic, which may be determined according to actual situations.
In some embodiments, before the processing module 520 splices the feature sequences, the development state vector sequence, the sales feature sequence, and the sales-related feature sequence in the same time period in the same region may be normalized to make different elements of each feature sequence in the same scale.
In some embodiments, the sequence of region characteristics may further include market characteristics of the region to be predicted for at least one time period, the market characteristics including at least one of marketing strategies, commodity characteristics, and market other participant characteristics. For example, the market characteristic sequence of the A region may be
Wherein M, N, O can characterize marketing separatelyPolicies, merchandise characteristics, market other participant characteristics.
In some embodiments, the processing module 520 may concatenate the market feature sequence with the sales volume, the sales volume related features, and the development status vector for at least one time period to obtain a region feature sequence. For example, the market feature sequence of the a region is spliced with the development state vector sequence, the sales feature sequence, and the sales-related feature sequence to obtain a region feature sequence including the market feature vector
The market characteristics may refer to market conditions within the area. For example, the market characteristics may include marketing strategies, merchandise characteristics, and/or other market participant characteristics, etc. within the area to be predicted. In some embodiments, the market characteristics may be obtained based on various possible ways. For example, the obtaining module 510 may obtain market characteristics from the user terminal 140 based on user input, may read market characteristics from the storage device 130, and may also obtain market characteristics via the network 150.
Marketing strategies may include strategies for selling goods. For example, sales promotion, price reduction, new product, advance reservation, premium item recommendation, sales promotion strength, price, and other means related to the marketing of the items. In some embodiments, marketing strategies for different areas may differ.
The characteristics of the article may include the characteristics of the article itself as well as the sales characteristics of the article. The characteristics of the commodity may include the type of the commodity, the technology of the commodity, the use of the commodity, and/or the selling object of the commodity. The sales characteristics of the goods may include conversion rates, sales volumes, number of orders, sales volume growth rates, order growth rates, and the like.
Other participants may include competitors, consumers, and/or market regulators, etc. The characteristics of the competitors can comprise marketing strategies of the competitors, characteristics of competitive commodities and the like; consumer characteristics may include consumer purchasing preferences, etc.; market regulatory may include levels of market regulatory, etc.
In some embodiments, marketing strategies, merchandise features, and other market participant features for an area may be obtained in various possible ways, including but not limited to utilizing big data acquisition, and the like.
And 230, processing the region characteristic sequence of the region to be predicted through a sales prediction model, and predicting the commodity sales of the region to be predicted.
For the details of the sales prediction model, refer to fig. 3 and the related description thereof, which are not repeated herein.
In some embodiments, the sales of the goods for one or more time periods may be predicted. For example, sales of goods on day 1, day 2, and/or day 3 in the future. In some embodiments, the prediction module 530 may predict based on a single good, for example, predicting the sales of a certain brand a towel. In some embodiments, the prediction module 530 may predict a certain type of good, for example, predict sales of all branded tissues in the market.
Some embodiments in the present specification improve accuracy and efficiency of predicted commodity sales by predicting commodity sales using a sales prediction model. Some embodiments of the present description may better obtain social development characteristics of an area to be predicted by converting data related to a social development state into a development state vector. Some embodiments of the present description generate a region feature sequence of a region to be predicted by combining features related to sales volume with a development state vector, and can consider objective and subjective conditions of the region to be predicted, so as to better predict sales volume of a commodity and improve practicality and accuracy of prediction.
FIG. 3 is a schematic diagram of a sales prediction model and a method of training the same according to some embodiments of the present description.
In some embodiments, the sales prediction model may be an LSTM model. For example, the model may input a regional characteristic sequence for a plurality of time periods into the LSTM model, process the input regional characteristic sequence, and output a predicted sales volume of the product associated with each time period.
As shown in fig. 3, the input of the sales prediction model 320 may be a regional feature sequence 310 comprising regional features for a plurality of time periods; the output may be a predicted sales 330 of the goods for some period of time in the future. In some embodiments, the sales prediction model may be obtained from one or more components in the retail merchandise sales system or an external source via network 150. For example, the sales prediction model may be trained in advance by the processor 120 and stored in the storage device 130. In some embodiments, the sales prediction model may be generated by a machine learning algorithm. The machine learning algorithm may include an artificial neural network algorithm, a deep learning algorithm, a decision tree algorithm, an association rule algorithm, a support vector machine algorithm, and/or a clustering algorithm, among others.
In some embodiments, the regional signature sequence may be input into an LSTM model, processed through LSTM, and output a predicted sales of the commodity.
As shown in FIG. 3, in some embodiments, the initial sales prediction model 340 may be trained based on sales training data 350, resulting in a sales prediction model 320. Wherein the sales training data 350 can be derived based on the first historical data 360. The first historical data may include data historically related to social development status of the area to be predicted and sales of goods to be predicted. The first history data 360 may be obtained based on various feasible manners.
In some embodiments, the processing module 520 may process the first historical data to obtain historical basic development characteristics, process the historical basic development characteristics based on the first development state model to generate a corresponding historical development state vector sequence, obtain a historical region feature sequence in the first historical data by processing the historical development state vector sequence, and use the historical region feature sequence as sales training data. For specific contents of obtaining the region feature sequence of one or more time periods, refer to fig. 2 and the related description thereof, which are not described herein again. In some embodiments, the sales volume of the retail goods related to the time period of the sales volume training data in the first historical data may be used as a label for the sales volume training data. The sales training data can be a region feature sequence of a region to be predicted; the label of the sales training data may be the sales volume of the retail item to be forecasted.
In some embodiments, sales training data may be input into an initial sales prediction model, the model outputting a predicted sales of the goods, and determining a loss function associated with the sales based on a difference between the predicted sales of the goods and the tag; and minimizing a loss function by a gradient descent method and the like, and determining a final initial sales prediction model as a sales prediction model.
In some embodiments of the description, the sales volume is predicted by combining the social development state and the market characteristics, and the model which is most matched with the current regional development state characteristics is determined according to the development state, so that the sales volume of the region to be predicted can be predicted more accurately; by using the LSTM model to process the region feature sequence, the output commodity sales volume can be related to the sequence of the region feature sequence input model, and the accuracy of the predicted commodity sales volume is further improved.
In some embodiments, there may be multiple sales prediction models, which may be determined by the processing module 520 based on the clustering of the sequence of development state vectors. As mentioned above, the sales prediction model can be obtained by training based on the first historical region feature sequence, and the historical region feature sequence is related to the historical development state vector sequence. The development state vector sequences can be classified, and corresponding sales prediction models are trained based on the development state vector sequences corresponding to each class.
In some embodiments, the processing module 520 may cluster the development state vector sequence through a clustering algorithm to obtain a plurality of cluster centers, each cluster center representing a region development category. And for each region development category, determining a corresponding historical region feature sequence based on the corresponding historical development state vector sequence, and training a sales prediction model based on the corresponding historical region feature. It will be appreciated that there may be one sales prediction model for each cluster center (i.e., each region development type). For example, the processing module 520 clusters at least one development state vector sequence, such as the development state vector sequence 1, the development state vector sequences 2, … …, and the development state vector sequence N, by using a clustering algorithm, to obtain three cluster centers in total, i.e., the development state vector sequence 2, the development state vector sequence 3, and the development state vector sequence 5, where each cluster center represents three different region development categories. For each of the three different regional development categories, the clustering center may be used as its corresponding development state vector sequence, e.g., directly using the development state vector sequence 2, the development state vector sequence 3, and the development state vector sequence 5 as the development state vector sequences of the three different regional development categories, respectively; or the development state vector sequence a, the development state vector sequence b and the development state vector sequence c can be obtained by respectively averaging all the development state vector sequences under each category, and the development state vector sequence a, the development state vector sequence b and the development state vector sequence c are respectively used as the development state vector sequences of three different regional development categories. For each of the three different regional development categories, the historical development state vector sequence included in the regional development category is processed to obtain a corresponding historical regional characteristic sequence, and sales prediction models are respectively trained for the three regional development types based on the obtained historical regional characteristic sequences to finally obtain three different sales prediction models. And processing the historical development state vector sequence to obtain a corresponding historical region characteristic sequence, respectively training a sales prediction model for X, Y, Z regional development types based on the obtained historical region characteristic sequence, and finally obtaining three different sales prediction models.
In some embodiments, when predicting sales of a certain area, a sales prediction model used for prediction may be determined based on similarity between the development state vector sequence of the area and the historical development state vector sequences corresponding to a plurality of cluster centers. For example, a sales prediction model corresponding to the cluster center with the highest similarity is obtained.
The type of clustering algorithm may include a variety, for example, the clustering algorithm may include K-Means clustering, density-based clustering method (DBSCAN), and the like.
In some embodiments, the development status vectors may be clustered in other ways, and a sales prediction model may be determined based on the clustering results. In some embodiments, multiple sales prediction models may be combined into one overall sales prediction model.
FIG. 4 is a schematic diagram of a method of obtaining a first developmental state model according to some embodiments of the present description. In some embodiments, the obtaining of the first evolving state model may be performed by the processing module 520.
In some embodiments, the first development state model 410-1 may be a machine learning model for obtaining a sequence of development state vectors based on underlying development features of a region. In some embodiments, the input of the first evolving state model may be the basic evolving characteristics of the first sample a region at any time period, and the output may be the first evolving state vector sequence of the a region at the end time point of the time period. In some embodiments, the output of the first developmental state model may also be referred to as a sequence of developmental state vectors. In some embodiments, the first evolving state model may be an LSTM model.
In some embodiments, the initial first development state model 412-1, the initial second development state model 412-2, and the initial similarity model 421 may be jointly trained based on training data to obtain the first development state model 410-1, the second development state model 410-2, and the similarity model 420; wherein the second state of development model and the first state of development model share parameters.
In some embodiments, the second development state model may obtain a second development state vector sequence of the B region at the end time point of the time period based on the basic development characteristics of the B region of the second sample at any time period. In some embodiments, the second state of development model and the first state of development model have the same model structure, both sharing parameters.
In some embodiments, the second development state model and the first development state model may correspond to time periods with different time and the same duration in the same area, for example, the first development state model corresponds to the development state of the a area 2015-2017, and the second development state model corresponds to the development state of the a area 2018-2020.
In some embodiments, the second development state model and the first development state model may correspond to the same time period of different regions, for example, the first development state model corresponds to the development state of the a region 2015-2017, and the second development state model corresponds to the development state of the B region 2015-2017.
In some embodiments, the time series of the second evolution state model and the first evolution state model are the same length, and the time periods may be different. For example, the time series corresponding to the region A in the first sample is 2015-2017, the time series corresponding to the region B in the second sample is 2018-2020, and both the time series lengths are not 3 years, but correspond to different years.
In some embodiments, the similarity model may determine a correlation of sales trends between the two reference areas based on the first development status vector and the second development status vector. In some embodiments, the similarity model may be a neural network model.
In some embodiments, the output of the first development state model and the output of the second development state model may be input to a similarity model, and the first development state model, the second development state model, and the similarity model may be obtained by joint training. For example, inputting a first sample basic development characteristic 411-1 corresponding to a first sample reference region into an initial first development state model 412-1 to obtain a sample first development state vector sequence 402-1, inputting a second sample basic development characteristic 411-2 corresponding to a second sample reference region into an initial second development state model 412-2 to obtain a sample second development state vector sequence 402-2, inputting the sample first development state vector sequence 402-1 and the sample second development state vector sequence 402-2 into an initial similarity model 421 to determine a sample development similarity 422, constructing a loss function based on the sample development similarity 422 determined by the label and the initial similarity model, and updating parameters of the initial first development state model 412-1, the initial second development state model 412-2 and the initial similarity model 421 at the same time, and obtaining the trained first development state model, second development state model and similarity model until a preset condition is met (for example, the loss function is smaller than a threshold value or converges, and the like).
The first development state model is obtained through the training mode, the problem that labels are difficult to obtain when the first development state model is trained independently is solved under some conditions, and the first development state model can well obtain the development state vectors of the determined reference region and the region to be predicted.
In some embodiments, the label for the joint training of the first developmental state model, the second developmental state model, and the similarity model may be a similarity between the historical first developmental state vector and the historical second developmental state vector.
In some embodiments, the processor 120 may calculate a sequence of change rates of the sales data of the at least one sample reference region in the time period of the duration to be predicted, for example, if the sales volume of a certain commodity in 2017 is increased by 5% compared to the last year, 5% is the change rate of the sales data of the commodity in 2017, and if the sales volume of the commodity in 2018 is decreased by 0.2% compared to 2017, the change rate of the sales data of the commodity in 2018 is-0.2%, and if the change rate of the sales data of the commodity in 2019 is 2%, the sequence of change rates of the sales data of the commodity in 2017 and 2019 is { 5%, -0.2%, 2% }. And determining the correlation degree based on the vector distance of the change rate sequence of the at least one sample reference area as a training label. Wherein the rate of change may be a first derivative of the sales data dispersion over a time period of the duration to be predicted for the at least one sample reference region.
In some embodiments, the training samples may include multiple sets of training data, each set of training data including: a first sample basic development characteristic and a second sample basic development characteristic. Wherein the training data may be from historical development data of the sample reference region.
Generally, the amount of development information in a social area over a period of time is huge, and it is difficult to directly determine the correlation of development characteristics with changes in sales trends. In some embodiments of the present description, a development state vector sequence is determined by using a first development state model and a second development state model, and then a model representing a social development state can be effectively determined based on regional historical development information in a correlation determination manner, so that features can be more effectively extracted.
FIG. 5 is an exemplary block diagram of a retail merchandise sales prediction system according to some embodiments of the present description. As shown in fig. 5, the system 500 includes one or more of the following modules.
An obtaining module 510, configured to obtain a development state vector of the area to be predicted in at least one time period. For specific content of obtaining the development state vector of the region to be predicted, refer to fig. 2 and the related description thereof, which are not described herein again.
The processing module 520 is configured to obtain a region feature sequence of the region to be predicted based on the development state vector of at least one time segment. For more details of obtaining the region feature sequence based on the development state vector, refer to fig. 2 and the related description thereof, which are not repeated herein.
The predicting module 530 is configured to process the region feature sequence of the region to be predicted through a sales predicting model, and predict the commodity sales of the region to be predicted. For more details about predicting the sales volume of the commodity, refer to fig. 2 and the related description thereof, which are not described herein again.
Having thus described the basic concept, it will be apparent to those skilled in the art that the foregoing detailed disclosure is to be regarded as illustrative only and not as limiting the present specification. Various modifications, improvements and adaptations to the present description may occur to those skilled in the art, although not explicitly described herein. Such modifications, improvements and adaptations are proposed in the present specification and thus fall within the spirit and scope of the exemplary embodiments of the present specification.
Also, the description uses specific words to describe embodiments of the description. Reference throughout this specification to "one embodiment," "an embodiment," and/or "some embodiments" means that a particular feature, structure, or characteristic described in connection with at least one embodiment of the specification is included. Therefore, it is emphasized and should be appreciated that two or more references to "an embodiment" or "one embodiment" or "an alternative embodiment" in various places throughout this specification are not necessarily all referring to the same embodiment. Furthermore, some features, structures, or characteristics of one or more embodiments of the specification may be combined as appropriate.
Additionally, the order in which the elements and sequences of the process are recited in the specification, the use of alphanumeric characters, or other designations, is not intended to limit the order in which the processes and methods of the specification occur, unless otherwise specified in the claims. While various presently contemplated embodiments of the invention have been discussed in the foregoing disclosure by way of example, it is to be understood that such detail is solely for that purpose and that the appended claims are not limited to the disclosed embodiments, but, on the contrary, are intended to cover all modifications and equivalent arrangements that are within the spirit and scope of the embodiments herein. For example, although the system components described above may be implemented by hardware devices, they may also be implemented by software-only solutions, such as installing the described system on an existing server or mobile device.
Similarly, it should be noted that in the preceding description of embodiments of the present specification, various features are sometimes grouped together in a single embodiment, figure, or description thereof for the purpose of streamlining the disclosure aiding in the understanding of one or more of the embodiments. This method of disclosure, however, is not intended to imply that more features than are expressly recited in a claim. Indeed, the embodiments may be characterized as having less than all of the features of a single embodiment disclosed above.
Numerals describing the number of components, attributes, etc. are used in some embodiments, it being understood that such numerals used in the description of the embodiments are modified in some instances by the use of the modifier "about", "approximately" or "substantially". Unless otherwise indicated, "about", "approximately" or "substantially" indicates that the number allows a variation of ± 20%. Accordingly, in some embodiments, the numerical parameters used in the specification and claims are approximations that may vary depending upon the desired properties of the individual embodiments. In some embodiments, the numerical parameter should take into account the specified significant digits and employ a general digit preserving approach. Notwithstanding that the numerical ranges and parameters setting forth the broad scope of the range are approximations, in the specific examples, such numerical values are set forth as precisely as possible within the scope of the application.
For each patent, patent application publication, and other material, such as articles, books, specifications, publications, documents, etc., cited in this specification, the entire contents of each are hereby incorporated by reference into this specification. Except where the application history document does not conform to or conflict with the contents of the present specification, it is to be understood that the application history document, as used herein in the present specification or appended claims, is intended to define the broadest scope of the present specification (whether presently or later in the specification) rather than the broadest scope of the present specification. It is to be understood that the descriptions, definitions and/or uses of terms in the accompanying materials of this specification shall control if they are inconsistent or contrary to the descriptions and/or uses of terms in this specification.
Finally, it should be understood that the embodiments described herein are merely illustrative of the principles of the embodiments of the present disclosure. Other variations are also possible within the scope of the present description. Thus, by way of example, and not limitation, alternative configurations of the embodiments of the specification can be considered consistent with the teachings of the specification. Accordingly, the embodiments of the present description are not limited to only those embodiments explicitly described and depicted herein.
Claims (10)
1. A retail goods sales forecasting method based on deep learning, comprising:
acquiring a development state vector sequence of a region to be predicted in at least one time period;
acquiring a region feature sequence of the region to be predicted based on the development state vector sequence of the at least one time segment;
and processing the region characteristic sequence of the region to be predicted through a sales prediction model, and predicting the commodity sales of the region to be predicted.
2. The retail merchandise sales forecasting method of claim 1, the obtaining the sequence of development state vectors for at least one time segment of the area to be forecasted comprising:
and processing the basic development characteristics of the region to be predicted in the at least one time period through a first development state model, and determining a development state vector sequence of the at least one time period.
3. The retail merchandise sales forecasting method of claim 2, obtaining the first developmental state model, comprising:
performing combined training on an initial first development state model, an initial second development state model and an initial similarity model based on training data to obtain the first development state model, the second development state model and the similarity model; wherein the second state of development model and the first state of development model share parameters.
4. The retail merchandise sales forecasting method of claim 1, the elements in the regional characteristic sequence comprising sales volumes for at least one time period and related characteristics of the sales volumes for the at least one time period and a development state vector sequence for the at least one time period;
the region feature sequence further comprises market features of the region to be predicted in the at least one time period, wherein the market features comprise at least one of marketing strategies, commodity features and market other participant features.
5. A retail merchandise sales prediction system based on deep learning, comprising:
the device comprises an acquisition module, a prediction module and a prediction module, wherein the acquisition module is used for acquiring a development state vector sequence of a region to be predicted in at least one time period;
the processing module is used for acquiring a region feature sequence of the region to be predicted based on the development state vector sequence of the at least one time period;
and the prediction module is used for processing the region characteristic sequence of the region to be predicted through a sales prediction model and predicting the commodity sales of the region to be predicted.
6. The retail merchandise sales prediction system of claim 5, the acquisition module further to:
and processing the basic development characteristics of the region to be predicted in the at least one time period through a first development state model, and determining a development state vector sequence of the at least one time period.
7. The retail merchandise sales prediction system of claim 6, the acquisition module further to acquire the first developmental state model, comprising:
performing combined training on an initial first development state model, an initial second development state model and an initial similarity model based on training data to obtain the first development state model, the second development state model and the similarity model; wherein the second state of development model and the first state of development model share parameters.
8. The retail merchandise sales forecasting system of claim 5, elements of the regional characteristic sequence comprising an amount of sales over at least one time period, and associated characteristics of the amount of sales over the at least one time period, and a development status vector for the at least one time period;
the region feature sequence further comprises market features of the region to be predicted in the at least one time period, wherein the market features comprise at least one of marketing strategies, commodity features and market other participant features.
9. A retail commodity sales prediction apparatus based on deep learning, comprising a processor for executing the retail commodity sales prediction method of any one of claims 1 to 4.
10. A computer readable storage medium storing computer instructions which, when read by a computer, cause the computer to perform the retail merchandise sales prediction method of any one of claims 1 to 4.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210054835.XA CN114387037A (en) | 2022-01-18 | 2022-01-18 | Retail commodity sales prediction method based on deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210054835.XA CN114387037A (en) | 2022-01-18 | 2022-01-18 | Retail commodity sales prediction method based on deep learning |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114387037A true CN114387037A (en) | 2022-04-22 |
Family
ID=81202841
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210054835.XA Pending CN114387037A (en) | 2022-01-18 | 2022-01-18 | Retail commodity sales prediction method based on deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114387037A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115994788A (en) * | 2023-03-20 | 2023-04-21 | 北京永辉科技有限公司 | Data processing analysis method and device |
| CN116882902A (en) * | 2023-09-06 | 2023-10-13 | 酒仙网络科技股份有限公司 | Storage management optimization method, system and storage medium based on purchase and sale information of wine |
-
2022
- 2022-01-18 CN CN202210054835.XA patent/CN114387037A/en active Pending
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115994788A (en) * | 2023-03-20 | 2023-04-21 | 北京永辉科技有限公司 | Data processing analysis method and device |
| CN116882902A (en) * | 2023-09-06 | 2023-10-13 | 酒仙网络科技股份有限公司 | Storage management optimization method, system and storage medium based on purchase and sale information of wine |
| CN116882902B (en) * | 2023-09-06 | 2023-11-07 | 酒仙网络科技股份有限公司 | Storage management optimization method, system and storage medium based on purchase and sale information of wine |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Kharfan et al. | A data-driven forecasting approach for newly launched seasonal products by leveraging machine-learning approaches | |
| CN107563705A (en) | Household electrical appliances product safety stock and the system and method ordered goods again are analyzed using big data | |
| CN111178624B (en) | New product demand prediction method | |
| US20210103858A1 (en) | Method and system for model auto-selection using an ensemble of machine learning models | |
| CN114387037A (en) | Retail commodity sales prediction method based on deep learning | |
| CN110555578B (en) | Sales prediction method and device | |
| CN111861759B (en) | Matching method and system for product and customer group | |
| Cui et al. | A stochastic multi-item replenishment and delivery problem with lead-time reduction initiatives and the solving methodologies | |
| CN111667307B (en) | Method and device for predicting financial product sales volume | |
| AU2020260401A1 (en) | Prospect recommendation | |
| CN112529628A (en) | Client label generation method and device, computer equipment and storage medium | |
| ABIDAR et al. | Predicting Customer Segment Changes to Enhance Customer Retention: A Case Study for Online Retail using Machine Learning | |
| CN116304929A (en) | Financial manipulation recognition method and device based on A-stock market | |
| CN115545307A (en) | Goods allocation method, device, equipment and storage medium | |
| Kiran et al. | Analysis on the prediction of sales using various machine learning testing algorithms | |
| CN115936184A (en) | Load prediction matching method suitable for multi-user types | |
| CN114881694A (en) | Automatic replenishment method, system, electronic device, storage medium, and program product | |
| Vaitkus et al. | Electrical spare parts demand forecasting | |
| Kurz et al. | The top-dog index: A new measurement for the demand consistency of the size distribution in pre-pack orders for a fashion discounter with many small branches | |
| Kabanova et al. | ABC-XYZ inventory analysis accounting for change points | |
| US20200211036A1 (en) | Method and system for machine learning based item matching by considering user mindset | |
| Hervert-Escobar et al. | Optimal pricing model based on reduction dimension: A case of study for convenience stores | |
| Frankeová et al. | Fuzzy-expert system for customer behavior prediction | |
| CN112862602B (en) | User request determining method, storage medium and electronic device | |
| Ebrahimi et al. | A predictive analytics approach to improve the dealers-manufacturer relationship in the after-sales service network; case study in the automotive industry |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |