Article demand quantity prediction method and device

Abstract

The invention discloses an article demand quantity prediction method and device, and relates to the technical field of data analysis. A specific embodiment of the method comprises the following steps: receiving a prediction request of an article demand quantity, wherein the prediction request indicates an article identifier and a to-be-predicted time period; determining a target demand increasing activity corresponding to the to-be-predicted time period, wherein a time interval between an activity time period corresponding to the target demand increasing activity and the to-be-predicted time period is not greater than a preset threshold value; according to the activity time period and the to-be-predicted time period, determining a time period type of the to-be-predicted time period; determining a target keyword of the article according to the article identifier; according to the target keyword, determining a basic demand quantity corresponding to the article; and according to the time period type and the basic demand quantity, predicting the target demand quantity of the articles in the to-be-predicted time period. According to the embodiment, the accuracy of demand quantity prediction is improved, so that the phenomenon of insufficient article supply caused by low predicted demand quantity and less stock preparation or overstock caused by too high predicted demand quantity and more stock preparation is reduced.

Description

Technical Field

[0001] The present invention relates to the field of data analysis technology, and in particular to a method and device for predicting item demand. Background Art

[0002] In the field of e-commerce, user demand satisfaction rate and product turnover cycle are crucial to e-commerce operations. These two factors depend on the accuracy of product sales forecasts. However, current product sales forecasts generally only refer to recent historical sales, and predict the sales of the next stage based on recent historical sales, so that warehouses can prepare goods according to the predicted sales.

[0003] Since the above method only obtains prediction results based on historical sales, it is easy to lead to low accuracy and poor referenceability of the prediction results. For example, if the next stage is the peak season of user demand, the predicted sales will be lower than the actual demand, while if the referenced historical sales are in the peak season of demand, the predicted sales will be higher than the actual demand. Summary of the invention

[0004] In view of this, the embodiment of the present invention provides a method and device for predicting the demand of an item, by receiving a prediction request for the demand of an item, the prediction request indicates the item identification of the item and the time period to be predicted; determining the target demand increase activity corresponding to the time period to be predicted, the time interval between the activity period corresponding to the target demand increase activity and the time period to be predicted is not greater than a preset threshold; determining the time period type corresponding to the time period to be predicted according to the time relationship between the activity period and the time period to be predicted; the time period type includes one or more of a target time period belonging to the activity period, a first time period earlier than the target time period, and a second time period later than the target time period; determining the target keyword corresponding to the item according to the item identification; determining the basic demand corresponding to the item according to the target keyword; predicting the target demand of the item in the time period to be predicted according to the time period type and the basic demand. Thus, by combining the time period type of the time period to be predicted on the basis of the basic demand, the target demand of the item in the time period to be predicted is predicted, so that the accuracy of the prediction result is improved, thereby reducing the phenomenon of insufficient supply of items caused by low predicted demand and less stock, or inventory backlog caused by excessively high predicted demand and more stock.

[0005] To achieve the above objective, according to one aspect of an embodiment of the present invention, a method for predicting item demand is provided.

[0006] A method for predicting item demand in an embodiment of the present invention includes: receiving a prediction request for item demand, the prediction request indicating an item identification of the item and a time period to be predicted; determining a target demand enhancement activity corresponding to the time period to be predicted, wherein the time interval between the activity time period corresponding to the target demand enhancement activity and the time period to be predicted is not greater than a preset threshold; determining a time period type corresponding to the time period to be predicted based on the time relationship between the activity time period and the time period to be predicted; the time period type includes one or more of a target time period belonging to the activity time period, a first time period earlier than the target time period, and a second time period later than the target time period; determining a target keyword corresponding to the item based on the item identification; determining a basic demand corresponding to the item based on the target keyword; and predicting the target demand for the item in the time period to be predicted based on the time period type and the basic demand.

[0007] Optionally, predicting the target demand for the item in the time period to be predicted based on the time period type and the basic demand includes: when the time period type includes the target time period, determining the target activity factor corresponding to the item based on historical data of the item; determining a first target demand corresponding to the target time period based on the basic demand and the target activity factor; and predicting the target demand based on the first target demand and the basic demand.

[0008] Optionally, determining the target activity factor corresponding to the item based on the historical data of the item includes: determining, based on the historical data, a first actual demand for the item in a historical demand increase activity, first click data in an out-of-stock state, second click data in a non-out-of-stock state, and a demand conversion rate corresponding to the second click data; determining an ideal demand corresponding to the historical demand increase activity based on the first actual demand, the first click data, and the demand conversion rate; determining, based on the historical data, a second actual demand within a preset time period before the historical demand increase activity; and determining the target activity factor based on the ideal demand and the second actual demand.

[0009] Optionally, determining the first target demand corresponding to the target time period based on the basic demand and the target activity factor includes: determining whether the item identifier corresponds to a target price change identifier; the target price change identifier corresponds to the target demand increase activity; in the case where there is a corresponding target price change identifier, determining the target price factor corresponding to the item based on the target price change identifier; determining the first target demand based on the basic demand, the target activity factor and the target price factor.

[0010] Optionally, determining the target price factor corresponding to the item based on the target price change identifier includes: determining the historical price change identifier corresponding to the item in the historical demand increase activity based on the historical data; determining the historical price factor based on the first actual demand in the historical demand increase activity and the second actual demand within a preset time period before the historical demand increase activity; determining the relationship between the historical price change identifier and the historical price factor; determining the target price factor based on the target price change identifier and the relationship.

[0011] Optionally, determining the relationship between the historical price change identifier and the historical price factor; and determining the target price factor based on the target price change identifier and the relationship, includes: using the historical price change identifier and the historical price factor to train a price factor prediction model so that the price factor prediction model learns the relationship between the historical price change identifier and the historical price factor; inputting the target price change identifier into the price factor prediction model to obtain the target price factor.

[0012] Optionally, predicting the target demand based on the first target demand and the basic demand includes: determining whether the time period type includes the first time period and / or the second time period other than the target time period; if including the first time period and / or the second time period, determining the second target demand corresponding to the first time period and / or the second time period based on the basic demand; predicting the target demand based on the first target demand and the second target demand.

[0013] Optionally, determining the basic demand corresponding to the item based on the target keyword includes: obtaining historical data corresponding to the target keyword; determining the historical demand through data smoothing based on the historical data; determining the reference demand corresponding to the time period to be predicted based on sales data in a preset time period adjacent to the time period to be predicted; inputting the reference demand into a basic demand prediction model to obtain the basic demand for the item output by the basic demand prediction model; the basic demand prediction model is trained using the historical demand.

[0014] Optionally, determining the target keyword corresponding to the item based on the item identification includes: using a word segmentation model to segment the item title corresponding to the item identification to obtain initial keywords; screening the initial keywords to obtain candidate keywords; and determining the target keyword based on the candidate keywords.

[0015] Optionally, determining the target keyword according to the candidate keywords includes:

[0016] Determine a time keyword set; combine the time keyword set with the item title as a target item title; calculate the difference between each keyword in the first candidate keyword and other keywords, and the correlation between each keyword and the target item title; determine the target keyword from the first candidate keywords based on the difference and correlation; the target keyword includes the target time keyword related to the commodity in the time keyword set.

[0017] Optionally, determining the target keyword from the first candidate keywords based on the difference and the association includes: calculating the maximum marginal correlation coefficient between each keyword and the target item title based on the difference and the association; and determining whether each keyword is the target keyword based on the maximum marginal correlation coefficient.

[0018] To achieve the above objective, according to another aspect of an embodiment of the present invention, a device for predicting item demand is provided.

[0019] An item demand prediction device according to an embodiment of the present invention includes:

[0020] A receiving module, configured to receive a forecast request for item demand, wherein the forecast request indicates an item identifier of the item and a time period to be forecasted;

[0021] a time period type determination module, used to determine the target demand quantity increase activity corresponding to the time period to be predicted, wherein the time interval between the activity period corresponding to the target demand quantity increase activity and the time period to be predicted is not greater than a preset threshold; and determine the time period type corresponding to the time period to be predicted according to the time relationship between the activity period and the time period to be predicted; the time period type includes one or more of a target time period belonging to the activity period, a first time period earlier than the target time period, and a second time period later than the target time period;

[0022] A keyword mining module, used to determine the target keyword corresponding to the item according to the item identifier;

[0023] An analysis module, used to determine the basic demand corresponding to the item according to the target keyword;

[0024] The prediction module is used to predict the target demand of the item in the time period to be predicted according to the time period type and the basic demand.

[0025] To achieve the above objective, according to another aspect of an embodiment of the present invention, an electronic device for predicting item demand is provided.

[0026] An electronic device for predicting item demand quantity according to an embodiment of the present invention comprises: one or more processors; a storage device for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement an item demand quantity prediction method according to an embodiment of the present invention.

[0027] To achieve the above objective, according to another aspect of an embodiment of the present invention, a computer-readable storage medium is provided.

[0028] A computer-readable storage medium according to an embodiment of the present invention stores a computer program, and when the program is executed by a processor, a method for predicting the demand quantity of an item according to an embodiment of the present invention is implemented.

[0029] An embodiment of the above invention has the following advantages or beneficial effects: by receiving a forecast request for the demand of an item, the forecast request indicates the item ID of the item and the time period to be predicted; determining the target demand increase activity corresponding to the time period to be predicted, and the time interval between the activity period corresponding to the target demand increase activity and the time period to be predicted is not greater than a preset threshold; determining the time period type corresponding to the time period to be predicted according to the time relationship between the activity period and the time period to be predicted; the time period type includes one or more of a target time period belonging to the activity period, a first time period earlier than the target time period, and a second time period later than the target time period; determining the target keyword corresponding to the item according to the item ID; determining the basic demand corresponding to the item according to the target keyword; predicting the target demand of the item in the time period to be predicted according to the time period type and the basic demand. Thus, by combining the time period type of the time period to be predicted on the basis of the basic demand, the target demand of the item in the time period to be predicted is predicted, so that the accuracy of the prediction result is improved, thereby reducing the phenomenon of insufficient supply of items caused by low predicted demand and less stock, or inventory backlog caused by excessively high predicted demand and more stock.

[0030] The further effects of the above-mentioned non-conventional optional manner will be described below in conjunction with the specific implementation manner. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] The accompanying drawings are used to better understand the present invention and do not constitute an improper limitation of the present invention.

[0032] Figure 1 is a schematic diagram of the main steps of a method for predicting item demand according to an embodiment of the present invention;

[0033] Figure 2a is a schematic diagram of the relationship between the two when the time period to be predicted is longer than the activity time period according to an embodiment of the present invention;

[0034] Figure 2b is a schematic diagram of the relationship between the two when the time period to be predicted is not longer than the activity time period according to an embodiment of the present invention;

[0035] Figure 3 is a schematic diagram of the main steps of keyword mining according to an embodiment of the present invention;

[0036] Figure 4 is a schematic diagram of main steps for determining basic demand according to an embodiment of the present invention;

[0037] Figure 5 is a schematic diagram of main steps of determining a target activity factor according to an embodiment of the present invention;

[0038] Figure 6 is a schematic diagram of main steps of determining a target price factor according to an embodiment of the present invention;

[0039] Figure 7 is a schematic diagram of the main steps of a method for predicting item demand according to an embodiment of the present invention;

[0040] Figure 8 is a schematic diagram of main modules of a device for predicting item demand according to an embodiment of the present invention;

[0041] Fig. 9 is an exemplary system architecture diagram to which embodiments of the present invention may be applied;

[0042] Fig.10 It is a schematic diagram of the structure of a computer system of a terminal device or a server suitable for implementing an embodiment of the present invention. DETAILED DESCRIPTION

[0043] The following is a description of exemplary embodiments of the present invention in conjunction with the accompanying drawings, including various details of the embodiments of the present invention to facilitate understanding, which should be considered as merely exemplary. Therefore, it should be recognized by those of ordinary skill in the art that various changes and modifications may be made to the embodiments described herein without departing from the scope and spirit of the present invention. Similarly, for clarity and conciseness, the description of well-known functions and structures is omitted in the following description.

[0044] It should be noted that the collection, collection, updating, analysis, processing, use, transmission, storage and other aspects of user personal information involved in the technical solution of this disclosure are in compliance with the provisions of relevant laws and regulations, are used for legitimate purposes, and do not violate public order and good morals. Necessary measures are taken for user personal information to prevent illegal access to user personal information data and maintain the security of user personal information, network security and national security.

[0045] It should be pointed out that the embodiments of the present invention and the technical features therein may be combined with each other without conflict.

[0046] Figure 1 4 is a schematic diagram of the main steps of the method for predicting item demand according to an embodiment of the present invention.

[0047] like Figure 1 As shown, the method for predicting the demand for goods in the embodiment of the present invention mainly includes the following steps:

[0048] Step S101: receiving a forecast request for item demand, wherein the forecast request indicates an item ID of the item and a time period to be forecasted.

[0049] The embodiment of the present invention takes the retail business of an e-commerce website as an application scenario, and takes the prediction of the sales volume of goods sold on the e-commerce website as an example to elaborate on the method for predicting the demand for goods in the embodiment of the present invention. The main body for executing the method for predicting the demand for goods in the embodiment of the present invention is an item demand prediction system, hereinafter referred to as the system. It should be noted that the meaning of "item" in the embodiment of the present invention is equivalent to "commodity"; "demand" is equivalent to "sales volume".

[0050] The overall idea of ​​the method for predicting the demand for a commodity in an embodiment of the present invention is to predict the basic demand of users for a commodity, that is, the basic sales volume of the commodity, based on historical data, and then determine whether there is a target demand improvement activity, that is, a marketing activity, in the forecast period according to the time period type of the forecast period. If there is a marketing activity, determine the impact of the marketing activity on the sales volume of the commodity in the forecast period, that is, the target activity factor, and determine the impact of the price reduction promotion in the marketing activity on the sales volume of the commodity, that is, the target price factor. Finally, predict the target demand of the commodity in the forecast period based on the basic demand, target activity factor and target price factor of the commodity. This solves the problem in the prior art that the sales volume of the commodity in the future time period is only predicted based on the recent historical sales volume, without considering whether the forecast period is during the marketing activity, or after / before the marketing activity, and without considering the impact of the marketing activity on the sales volume of the commodity in the forecast period.

[0051] The system receives a sales forecast request sent by a user, which indicates the product ID corresponding to the product to be predicted and the time period to be predicted. For example, the system predicts the sales of the product with product ID A within the time period 8.1-8.30. The product corresponding to product ID A is a book cover, book cover or book film wrapped on the outside of the book to protect the book from damage.

[0052] Generally speaking, factors that affect product sales include marketing activities, price reductions, promotions, time or season, etc. Therefore, if you want to get a sales forecast with higher accuracy, you need to take the above factors into account so that the predicted product sales can meet the actual customer demand, and then the warehouse can prepare inventory according to the predicted sales.

[0053] Step S102: determining a target demand quantity increasing activity corresponding to the time period to be predicted, wherein the time interval between the activity time period corresponding to the target demand quantity increasing activity and the time period to be predicted is not greater than a preset threshold.

[0054] In order to obtain accurate sales forecast values, after determining the time period to be predicted, it is also necessary to determine whether there are marketing activities that are close to or overlap with the time period to be predicted, such as the Women's Day sale, the Dragon Boat Festival sale, the back-to-school season sale, etc., based on the marketing plan of the e-commerce website in the future, to determine whether there are marketing activities that can affect the sales of goods in the time period to be predicted. Setting a preset interval means determining whether the marketing activity has an impact on the sales of goods in the prediction period based on the time interval between the prediction period and the time period corresponding to the marketing activity. For example, if the interval between the two time periods is greater than the preset threshold, it is determined that there is no impact, otherwise there is an impact, that is, when predicting the sales of goods in the prediction period, it is necessary to consider the impact of the marketing activity on the predicted sales.

[0055] Step S103: Determine the time period type corresponding to the time period to be predicted based on the time relationship between the activity time period and the time period to be predicted; the time period type includes one or more of a target time period belonging to the activity time period, a first time period earlier than the target time period, and a second time period later than the target time period.

[0056] In order to improve the accuracy of commodity sales forecasting, the characteristics of the period to be predicted can be determined based on the time relationship between the activity period and the period to be predicted, and then the sales of the commodity in the period to be predicted can be more accurately predicted based on the characteristics of the period to be predicted.

[0057] The relationship between the forecast period and the activity period of the marketing activity (i.e., the activity to increase target demand) is shown in Figure 2. Figure 2a shows five kinds of relationships between the two when the period to be predicted is longer than the activity period; Figure 2b Five relationships between the time period to be predicted and the activity period are shown when the time period to be predicted is shorter than the activity period. According to the different time relationships between the time period to be predicted and the activity period in the two cases, it can be concluded that the time period type corresponding to the time period to be predicted includes one or more of a target time period C belonging to the activity period, a first time period Z earlier than the target time period, and a second time period W later than the target time period.

[0058] Step S104: Determine the target keyword corresponding to the item according to the item identifier.

[0059] To predict the sales volume of a product, it is necessary to refer to the historical sales volume of the product. Therefore, whether the historical sales volume data is comprehensive and reliable also has an important impact on the prediction results of the product sales volume. For products with a relatively simple name or name and a relatively clear category, the corresponding historical sales volume can be obtained according to its name and category. For example, "pure milk" can easily obtain its historical sales volume according to the keyword "pure milk" in the dairy category; however, for products with many names or names, such as "book cover", and products with unclear categories, such as "Spring Festival couplets", it is necessary to mine the corresponding keywords and summarize the sales volume data of the product according to the mined keywords. For example, product identification A corresponds to something wrapped on the outside of a book to protect the book from being damaged. Its common name is "book cover", but other keywords can also be obtained through mining, such as "book film", "book cover", "book cover", "book film", etc. Mining target keywords can start from the product title.

[0060] In order to dig out more keywords related to the product, in one embodiment of the present invention, determining the target keyword corresponding to the item based on the item identification includes: using a word segmentation model to segment the item title corresponding to the item identification to obtain initial keywords; screening the initial keywords to obtain candidate keywords; and determining the target keywords based on the candidate keywords.

[0061] According to the most commonly used keywords corresponding to the product identification, some product titles are obtained, such as "Back to school season, Mao Taizi (MAOTAIZI) primary school students' kraft paper self-adhesive book cover, thickened book cover, book film, large, medium and small set of 52 sheets", and the title is input into the word segmentation model, such as the LAC (Lexical Analysis of Chinese, lexical analysis) model, and the following word segmentation results are obtained: "Back to school season", "Mao Taizi", "(", "MAOTAIZI", ")", "primary school students", "kraft paper", "self-adhesive", "book cover", "book cover", "thickened", "book cover", "book film", "large", "medium and small", "set", "52 sheets". Among these keywords, there are some brand names, quantifiers, special symbols and other keywords that do not belong to the product category, such as "52 sheets", "Mao Taizi", etc. These words are filtered out through regular expressions to obtain candidate keywords.

[0062] As mentioned above, the sales of commodities are greatly affected by time. For example, some commodities are used less frequently in daily life, and only in a certain time, the user demand is large. Therefore, the sales of the commodity need to be considered in the special time or special festival. For example, the book cover of the commodity will be affected by the start of school (after the start of school, students will generally give new books with book covers), the rice dumplings will be affected by the Dragon Boat Festival, and the Spring Festival couplets will be affected by the New Year. In this embodiment, such commodities are called time-related commodities. Therefore, when predicting the sales of such commodities, it is necessary to first determine whether the commodity is a time-related commodity. Since e-commerce websites will carry out corresponding marketing activities at such special times or special festivals, such commodities can also be called activity-related commodities. To predict the sales of such commodities, it is necessary to refer to the historical sales of such commodities at the corresponding time or festival in the historical period. Therefore, in addition to obtaining the historical sales according to the keywords representing the commodity use attributes, it is also necessary to consider the time keywords corresponding to the commodity to determine the historical sales of the commodity in the corresponding time.

[0063] In order to determine the historical sales volume of the product at a special time or on a special holiday, it is necessary to filter out target time keywords related to the product from common time keywords. Therefore, in one embodiment of the present invention, determining the target keyword based on the candidate keywords includes: determining a time keyword set; combining the time keyword set with the item title as a target item title; calculating the difference between each keyword in the first candidate keyword and other keywords, and the correlation between each keyword and the target item title; determining the target keyword from the first candidate keywords based on the difference and correlation; the target keyword includes the target time keyword related to the product in the time keyword set.

[0064] The time keyword set includes words related to time or season that often appear in product titles on e-commerce websites, such as "Christmas", "New Year", "warm", "thick", "back to school season", etc. In order to calculate the relevance and difference between product titles, time keyword sets and candidate keywords, they need to be converted into numbers. Therefore, the product title, time keyword set and each candidate keyword are input into a language representation model, such as the SentenceTransformer model based on Bert (Bidirectional Encoder Representation from Transformers), and the title vector of the product title is obtained. Time keyword set vector And the e corresponding to the candidate keywords, the title vector Vector set of keywords with time Weighted to get the target item title vector

[0065] Next, the target keywords to be retained are determined from the candidate keywords through the above vectors. The target keywords include keywords related to the purpose of the product and keywords related to the usage time of the product. Assume that D is the candidate keyword set, S is the currently selected keyword set, D\S is the currently unselected keyword set, λ=0.7 represents the adjustment weight coefficient of the correlation and difference, and the calculation of the marginal correlation coefficient must first calculate the correlation between each candidate keyword vector and the target item title vector, where the correlation between the two vectors is represented by the cosine similarity, that is, Among them, e i is the candidate keyword that is not currently selected; then the difference between the candidate keyword and the selected keyword is calculated, and the difference is characterized by the opposite of the correlation degree. Therefore, the correlation degree between the candidate keyword and the selected keyword is first calculated, expressed as Sim(e i ,e j ), after adding the weight, the correlation between the candidate keyword vector and the target item title vector is Then the difference between the candidate keyword and the selected keyword is The calculation formula of cosine similarity is:

[0066]

[0067] Among them, e i is the candidate keyword that is not currently selected, e j It is the keyword that has been selected among the candidate keywords. i and e j The cosine similarity of two vectors e i and e j The cosine of the angle is used as a measure of e i and e j The difference between the i and e j The closer the cosine value is to 1, the closer the angle between the two vectors is to 0 degrees, and the more similar the two vectors are; the closer the cosine value is to 0, the closer the angle between the two vectors is to 180 degrees, and the more dissimilar the two vectors are.

[0068] In one embodiment of the present invention, determining the target keyword from the first candidate keywords based on the difference and the association includes: calculating the maximum marginal correlation coefficient between each keyword and the target item title based on the difference and the association; and determining whether each keyword is the target keyword based on the maximum marginal correlation coefficient.

[0069] The formula for calculating the marginal correlation coefficient based on the degree of association and the degree of difference is as follows:

[0070]

[0071] Among them, e i is the candidate keyword that is not currently selected, e j is the candidate keyword that has been selected, D is the candidate keyword set, S is the currently selected keyword set, D\S is the currently unselected keyword set, To convert the title vector Vector set of keywords with time The weighted target item title vector is

[0072] According to the above formula, the marginal correlation coefficient corresponding to each candidate keyword is calculated, and the keywords with the highest marginal correlation coefficient ranking are selected, such as "school season", "book cover", "kraft paper", "thickening", "primary school students", and "book cover". The keywords mined from multiple product titles related to the product are gathered together and manually screened again. Time keywords that are obviously irrelevant to the product are removed, such as "thickening", "warmth", etc., and keywords that are irrelevant to sales are removed, such as "primary school students". The final result obtained by screening is "school season", "book cover", "book paper", "book film", "book film", "book cover", and "book cover". The final result obtained by screening is used as the target keyword. According to the target keyword, the historical sales corresponding to the product are determined from the historical data, and the historical sales related to the target time keyword, such as "school season", can be determined from it.

[0073] The following is a further exemplary description of the keyword mining process. Figure 3 As shown in the figure, keyword mining mainly includes the following steps:

[0074] S301: Determine the item title corresponding to the item identifier.

[0075] S302: Segment the item title to obtain initial keywords.

[0076] S303: Screening the initial keywords to obtain candidate keywords.

[0077] S304: Determine a time keyword set.

[0078] S305: vectorize each candidate keyword, time keyword set, and item title.

[0079] S306: Weight the time keyword vector and the item title vector to obtain a target vector.

[0080] S307: Calculate the relevance between each candidate keyword vector and the target vector.

[0081] S308: Calculate the difference between each candidate keyword vector and other candidate keyword vectors.

[0082] S309: Calculate the marginal correlation coefficient corresponding to each candidate keyword according to the relevance and difference.

[0083] S310: Determine a target keyword from candidate keywords according to the edge correlation coefficient.

[0084] The above steps are the main steps involved in keyword mining and do not represent the entire keyword mining process.

[0085] Step S105: Determine the basic demand corresponding to the item according to the target keyword.

[0086] In order to determine the sales volume in non-marketing activity periods based on historical sales data, that is, daily sales volume, in one embodiment of the present invention, determining the basic demand corresponding to the item based on the target keyword includes: obtaining historical data corresponding to the target keyword; determining the historical demand through data smoothing based on the historical data; determining the reference demand corresponding to the time period to be predicted based on the sales data in a preset time period adjacent to the time period to be predicted; inputting the reference demand into a basic demand prediction model to obtain the basic demand of the item output by the basic demand prediction model; the basic demand prediction model is trained using the historical demand.

[0087] Determine the basic demand for the product, that is, the sales volume of the product during daily time after eliminating activity factors. First, obtain the historical sales data corresponding to the product from the historical sales data based on the target keywords, and then determine the historical sales volume related to the target time keywords from the historical sales data of the product based on the target time keywords in the target keywords, that is, the historical sales volume related to the marketing activities, and then eliminate the historical sales volume related to the marketing activities from the historical sales volume of the product, that is, the daily sales volume data in the historical period; then, in the obtained daily sales data, determine the abnormal orders of a single user whose sales volume exceeds the normal value range, eliminate the sales volume corresponding to the abnormal order that exceeds the normal range, and then use the retained sales data to train the basic demand prediction model.

[0088] The basic demand forecasting model includes multiple sub-models, each of which is set with a corresponding weight. During the training process, the total loss function value corresponding to the basic demand forecasting model is calculated according to the weights and loss function values ​​of each sub-model. When the total loss function value is the minimum, the model training operation is stopped and the weights w corresponding to each sub-model are determined. mThe basic demand forecast value of a certain commodity corresponding to each sub-model is The predicted values ​​of multiple sub-models are weighted and summed to obtain the predicted value of the basic demand for the product. The prediction calculation formula for the basic demand is:

[0089]

[0090] Where M represents the number of sub-models, w m represents the weight corresponding to the mth sub-model, represents the basic demand for the product predicted by the mth sub-model, Indicates the basic demand for the product predicted by the basic demand forecasting model.

[0091] For the commodity forecast within the time period to be predicted, determine the sales data within the preset time period adjacent to the time period to be predicted, such as the sales data of the previous 15 days (sales data of the non-active time period), and determine the reference demand corresponding to the time period to be predicted based on the sales data within the preset time period; and use the reference demand as the input of the basic demand forecasting model, and the output value of the basic demand forecasting model based on the input is the basic demand for the commodity.

[0092] Figure 4 The main steps of basic demand forecasting are shown, including:

[0093] S401: Obtain historical demand data corresponding to the item.

[0094] S402: Eliminate the demand data during the marketing campaign from the historical demand data.

[0095] S403: Smoothing the demand corresponding to the abnormal orders in the remaining historical demand data to obtain processed historical basic demand data.

[0096] S404: Use historical basic demand data to train a basic demand forecasting model.

[0097] S405: Determine the recent actual basic demand corresponding to the time period to be predicted.

[0098] S406: Inputting the recent actual basic demand into the trained basic demand prediction model.

[0099] S407: Obtaining the basic demand forecast value of the item output by the basic demand forecast model.

[0100] The basic demand mentioned above refers to the sales volume of goods when there are no daily marketing activities.

[0101] Step S106: predicting the target demand for the item in the time period to be predicted according to the time period type and the basic demand.

[0102] In order to improve the accuracy of sales forecasts, it is necessary not only to refer to daily sales data, but also to determine whether there are any marketing activities in the time period to be predicted. If there are marketing activities, it is necessary to consider the impact of marketing activities on sales. Therefore, in one embodiment of the present invention, the target demand for the item in the time period to be predicted is predicted based on the time period type and the basic demand, including: when the time period type includes the target time period, determining the target activity factor corresponding to the item based on the historical data of the item; determining the first target demand corresponding to the target time period based on the basic demand and the target activity factor; and predicting the target demand based on the first target demand and the basic demand.

[0103] When a marketing activity is included in the time period to be predicted, the impact of the marketing activity on the sales volume of the product, i.e., the target activity factor, is determined based on the sales volume data of the product in previous marketing activities. Then, based on the impact of the marketing activity on the sales volume and the daily sales data, the sales volume corresponding to the active time period in the time period to be predicted is determined, i.e., the first target demand. Then, the inactive time period in the time period to be predicted is determined, i.e., other time periods except the target time period, and the second target demand corresponding to other time periods is determined based on the basic demand. The sum of the first target demand and the second target demand is taken as the target demand corresponding to the time period to be predicted.

[0104] In order to determine the impact of marketing activities on commodity sales based on data from previous marketing activities, in one embodiment of the present invention, determining the target activity factor corresponding to the item based on the historical data of the item includes: determining the first actual demand corresponding to the item in a historical demand increase activity, the first click data in an out-of-stock state, the second click data in a non-out-of-stock state, and the demand conversion rate corresponding to the second click data based on the historical data; determining the ideal demand corresponding to the historical demand increase activity based on the first actual demand, the first click data, and the demand conversion rate; determining the second actual demand within a preset time period before the historical demand increase activity based on the historical data; and determining the target activity factor based on the ideal demand and the second actual demand.

[0105] According to the data of previous marketing activities, the actual sales volume Q (first actual demand volume) in the activities, the page click volume N (first click data) in the out-of-stock state, and the page click volume P (second click data) in the non-out-of-stock state can be determined. The page click conversion rate cvr (demand conversion rate) in the non-out-of-stock state can be calculated based on the page click volume P and its corresponding actual sales volume. According to the actual sales volume Q and the page click conversion rate cvr, the sales volume that can be achieved by the product under the assumption that the product supply is sufficient, that is, there is no out-of-stock situation, that is, the ideal sales volume, so the ideal sales volume during the marketing activity is y = Q + N * cvr. The impact of marketing activities on product sales, that is, the target activity factor, can be determined based on the ratio of the daily average sales during the activity period to the daily average sales within the preset time period before the activity (for example, 28 days). Therefore, the calculation formula of the target activity factor is as follows:

[0106]

[0107] Among them, l is the number of days of marketing activities, y n is the sales volume of a product on the nth day of a marketing campaign, The sales volume of a product on each day in the 28 days before the event (since it is a non-event period, it is expressed as basic sales volume). is the target activity factor.

[0108] Therefore, assuming that the basic sales volume of a product is The corresponding marketing activity factor (target activity factor) is The sales forecast formula for this product in future marketing activities is:

[0109]

[0110] in, The basic sales volume of the product. It is the marketing activity factor (target activity factor).

[0111] Figure 5 The main steps for determining the target activity factor are shown, including:

[0112] S501: Acquire historical data in historical marketing activities corresponding to a product.

[0113] S502: Determine the actual sales volume Q in the activity based on historical data.

[0114] S503: Determine the page click count N in the out-of-stock state.

[0115] S504: Determine the page click volume P of the non-out-of-stock status.

[0116] S505: Determine the actual sales volume corresponding to the page click volume P, and calculate the page click conversion rate cvr.

[0117] S506: Determine the ideal sales volume in the historical activity as Q+N*cvr.

[0118] S507: Determine the pre-event sales volume 28 days before the event.

[0119] S508: Determine the ratio of the ideal sales volume to the sales volume before the activity as the target activity factor.

[0120] According to daily life experience, there is another important point that has a significant impact on the sales volume of goods, namely, price reduction and promotion. Therefore, when making sales forecasts, it is necessary to consider whether a certain product will be discounted and promoted during the forecast period. In one embodiment of the present invention, the first target demand corresponding to the target time period is determined according to the basic demand and the target activity factor, including: determining whether the item identifier corresponds to the target price change identifier; the target price change identifier corresponds to the target demand increase activity; in the case of corresponding to the target price change identifier, determining the target price factor corresponding to the item according to the target price change identifier; determining the first target demand according to the basic demand, the target activity factor and the target price factor.

[0121] Before making sales forecasts, we first need to determine whether the product will be discounted during the forecast period, including price reductions for individual products, discounted packages, free gifts, etc. There are also some promotional methods that can stimulate users' desire to buy, such as marking how much discount the product will get if it is purchased before a certain time point, or when the price will return to the original price, etc. In this embodiment, the presentation of the above promotional methods in the system is collectively referred to as price change identification. If a product is discounted during the forecast period, to predict its sales, we need to determine the impact of the price reduction on the sales of the product, that is, we need to determine the target price factor. To determine the target price factor, we need to find the relationship between the historical promotion logo and the historical price factor from the previous price reduction promotion activities. It is understandable that if a certain product also has a price reduction promotion in the marketing activity, the sales forecast of this product must consider both the target activity factor and the target price factor. Therefore, the target demand forecast formula for a certain product can be expressed as the following formula:

[0122]

[0123] in, As the basic demand, The target activity factor, is the target price factor.

[0124] To determine, in one embodiment of the present invention, the target price factor corresponding to the item is determined according to the target price change identifier, including: determining the historical price change identifier corresponding to the item in the historical demand increase activity according to the historical data; determining the historical price factor according to the first actual demand in the historical demand increase activity and the second actual demand within a preset period before the historical demand increase activity; determining the relationship between the historical price change identifier and the historical price factor; and determining the target price factor according to the target price change identifier and the relationship.

[0125] From all the previous price reduction promotions, determine the historical price change identifier of each price reduction promotion, that is, the promotion method, determine the sales volume during the historical price reduction promotion, that is, the first actual demand, and determine the sales volume before the historical price reduction promotion, that is, the second actual demand, and determine the ratio of the first actual demand to the second actual demand as the historical price factor; based on the historical promotion identifiers and historical price factors in multiple price reduction promotions, determine the relationship between the historical promotion identifiers and the historical price factors, and apply the determined relationship to the sales volume forecast of the goods in the forecast period.

[0126] In one embodiment of the present invention, the determining the relationship between the historical price change identifier and the historical price factor; and determining the target price factor based on the target price change identifier and the relationship, include: using the historical price change identifier and the historical price factor to train a price factor prediction model so that the price factor prediction model learns the relationship between the historical price change identifier and the historical price factor; inputting the target price change identifier into the price factor prediction model to obtain the target price factor.

[0127] The historical promotion identifier, the first actual demand, and the first actual demand in the historical data are input into the price factor prediction model, and the price factor prediction model learns the relationship between the historical promotion identifier and the historical price factor through a large amount of data. Therefore, the target promotion identifier corresponding to the current prediction period is input into the price factor prediction model, and the output value of the price factor prediction model is the target price factor. It can be understood that in the case where the product corresponds to a price reduction promotion, the sales volume of the product in the price reduction promotion period can be determined based on the product of the basic demand and the target price factor.

[0128] Figure 6 The main steps for determining the target price factor are shown, including:

[0129] S601: Determine a historical price change indicator in a historical price reduction promotion based on the historical data of the product.

[0130] S602: Determine the first actual demand quantity in the historical price reduction promotion.

[0131] S603: Determine the second actual demand quantity before the historical price reduction promotion.

[0132] S604: Determine a historical price factor according to the first and second actual demand quantities.

[0133] S605: Using the price factor prediction model to learn the relationship between the historical price change identifier and the historical price factor.

[0134] S606: Determine the target price change indicator corresponding to the time period to be predicted.

[0135] S607: Input the target price change identifier into the price factor prediction model.

[0136] S608: Using the output of the price factor prediction model as the target price factor.

[0137] In order to illustrate how to determine the sales volume of goods in a time period to be predicted when the time period to be predicted includes not only active time periods but also inactive time periods, in one embodiment of the present invention, predicting the target demand based on the first target demand and the basic demand includes: determining whether the time period type includes the first time period and / or the second time period other than the target time period; in the case of including the first time period and / or the second time period, determining the second target demand corresponding to the first time period and / or the second time period according to the basic demand; and predicting the target demand based on the first target demand and the second target demand.

[0138] It has been explained above that in the case of marketing activities and price reduction promotions, the target demand is determined based on the basic demand, the target activity factor, and the target price factor. In the case of no marketing activities or price reduction promotions, the sales volume of the product in the forecast period can be regarded as the same as the daily sales volume in the historical period. Therefore, the basic demand forecast method can be used to determine the sales volume of the product in the non-activity period in the forecast period. Finally, the first target demand corresponding to the target period included in the forecast period is added to the second target demand corresponding to the first period and / or the second period, that is, the target demand corresponding to the product in the forecast period is obtained.

[0139] Figure 7 This is a schematic diagram of the main steps of the method for predicting the demand for goods in an embodiment of the present invention. The forecasting of the demand for goods mainly includes the following steps:

[0140] S701: Receive a demand forecast request for an item within a forecast period.

[0141] S702: Determine the corresponding marketing activities for the time period to be predicted.

[0142] S703: Determine the time period type of the time period to be predicted according to the activity period of the marketing activity.

[0143] S704: Determine the target keyword corresponding to the item.

[0144] S705: Determine the basic demand for the item based on the target keyword.

[0145] S706: Determine the target activity factor according to the time period type.

[0146] S707: Determine the target price factor.

[0147] S708: The product of the basic demand, the target activity factor and the target price factor is determined as the demand for the item in the forecast period, ie, the target demand.

[0148] According to the method for predicting the demand for items according to the embodiment of the present invention, it can be seen that by receiving a prediction request for the demand for items, the prediction request indicates the item identification of the item and the time period to be predicted; determining the target demand increase activity corresponding to the time period to be predicted, and the time interval between the activity period corresponding to the target demand increase activity and the time period to be predicted is not greater than a preset threshold; determining the time period type corresponding to the time period to be predicted according to the time relationship between the activity period and the time period to be predicted; the time period type includes one or more of a target time period belonging to the activity period, a first time period earlier than the target time period, and a second time period later than the target time period; determining the target keyword corresponding to the item according to the item identification; determining the basic demand corresponding to the item according to the target keyword; predicting the target demand for the item in the time period to be predicted according to the time period type and the basic demand. Therefore, by combining the time period type of the time period to be predicted on the basis of the basic demand, the target demand for the item in the time period to be predicted is predicted, so that the accuracy of the prediction result is improved, thereby reducing the phenomenon of insufficient supply of items caused by low predicted demand and less stock, or inventory backlog caused by excessively high predicted demand and more stock.

[0149] Figure 8 4 is a schematic diagram of main modules of an apparatus for predicting item demand according to an embodiment of the present invention.

[0150] like Figure 8 As shown, the device 800 for predicting the demand for goods according to the embodiment of the present invention includes:

[0151] A receiving module 801 is used to receive a forecast request for item demand, wherein the forecast request indicates an item ID of the item and a time period to be forecasted;

[0152] The time period type determination module 802 is used to determine the target demand increase activity corresponding to the time period to be predicted, and the time interval between the activity period corresponding to the target demand increase activity and the time period to be predicted is not greater than a preset threshold; and determine the time period type corresponding to the time period to be predicted according to the time relationship between the activity period and the time period to be predicted; the time period type includes one or more of a target time period belonging to the activity period, a first time period earlier than the target time period, and a second time period later than the target time period;

[0153] A keyword mining module 803 is used to determine a target keyword corresponding to the item according to the item identifier;

[0154] An analysis module 804 is used to determine the basic demand corresponding to the item according to the target keyword;

[0155] The prediction module 805 is used to predict the target demand quantity of the item in the time period to be predicted according to the time period type and the basic demand quantity.

[0156] In one embodiment of the present invention, the prediction module 805 is also used to determine the target activity factor corresponding to the item based on the historical data of the item when the time period type includes the target time period; determine the first target demand corresponding to the target time period based on the basic demand and the target activity factor; and predict the target demand based on the first target demand and the basic demand.

[0157] In one embodiment of the present invention, the prediction module 805 is also used to determine, based on the historical data, a first actual demand corresponding to the item in the historical demand increase activity, first click data in an out-of-stock state, second click data in a non-out-of-stock state, and a demand conversion rate corresponding to the second click data; determine an ideal demand corresponding to the historical demand increase activity based on the first actual demand, the first click data, and the demand conversion rate; determine, based on the historical data, a second actual demand within a preset period before the historical demand increase activity; and determine the target activity factor based on the ideal demand and the second actual demand.

[0158] In one embodiment of the present invention, the prediction module 805 is also used to determine whether the item identifier corresponds to the target price change identifier; the target price change identifier corresponds to the target demand increase activity; in the case of corresponding to the target price change identifier, the target price factor corresponding to the item is determined according to the target price change identifier; the first target demand is determined according to the basic demand, the target activity factor and the target price factor.

[0159] In one embodiment of the present invention, the prediction module 805 is also used to determine the historical price change identifier corresponding to the item in the historical demand increase activity based on the historical data; determine the historical price factor based on the first actual demand in the historical demand increase activity and the second actual demand within a preset time period before the historical demand increase activity; determine the relationship between the historical price change identifier and the historical price factor; and determine the target price factor based on the target price change identifier and the relationship.

[0160] In one embodiment of the present invention, the prediction module 805 is also used to train a price factor prediction model using the historical price change identifier and the historical price factor, so that the price factor prediction model learns the relationship between the historical price change identifier and the historical price factor; and inputs the target price change identifier into the price factor prediction model to obtain the target price factor.

[0161] In one embodiment of the present invention, the prediction module 805 is also used to determine whether the time period type includes the first time period and / or the second time period other than the target time period; if the first time period and / or the second time period are included, determine the second target demand corresponding to the first time period and / or the second time period according to the basic demand; and predict the target demand based on the first target demand and the second target demand.

[0162] In one embodiment of the present invention, the analysis module 804 is also used to obtain historical data corresponding to the target keyword; determine the historical demand through data smoothing based on the historical data; determine the reference demand corresponding to the time period to be predicted based on the sales data in a preset time period adjacent to the time period to be predicted; input the reference demand into a basic demand prediction model to obtain the basic demand of the item output by the basic demand prediction model; the basic demand prediction model is trained using the historical demand.

[0163] In one embodiment of the present invention, the keyword mining module 803 is also used to use a word segmentation model to segment the item title corresponding to the item identifier to obtain initial keywords; screen the initial keywords to obtain candidate keywords; and determine the target keywords based on the candidate keywords.

[0164] In one embodiment of the present invention, the keyword mining module 803 is also used to determine a time keyword set; combine the time keyword set with the item title as a target item title; calculate the difference between each keyword in the first candidate keyword and other keywords, and the correlation between each keyword and the target item title; determine the target keyword from the first candidate keyword based on the difference and correlation; the target keyword includes the target time keyword related to the commodity in the time keyword set.

[0165] In one embodiment of the present invention, the keyword mining module 803 is also used to calculate the maximum marginal correlation coefficient between each keyword and the target item title based on the difference and the correlation; and determine whether each keyword is a target keyword based on the maximum marginal correlation coefficient.

[0166] According to the item demand forecasting device of the embodiment of the present invention, it can be seen that by receiving the forecasting request of the item demand, the forecasting request indicates the item identification of the item and the time period to be predicted; determining the target demand improvement activity corresponding to the time period to be predicted, and the time interval between the activity period corresponding to the target demand improvement activity and the time period to be predicted is not greater than the preset threshold; determining the time period type corresponding to the time period to be predicted according to the time relationship between the activity period and the time period to be predicted; the time period type includes one or more of the target time period belonging to the activity period, the first time period earlier than the target time period, and the second time period later than the target time period; determining the target keyword corresponding to the item according to the item identification; determining the basic demand corresponding to the item according to the target keyword; predicting the target demand of the item in the time period to be predicted according to the time period type and the basic demand. Therefore, by combining the time period type of the time period to be predicted on the basis of the basic demand, the target demand of the item in the time period to be predicted is predicted, so that the accuracy of the prediction result is improved, thereby reducing the phenomenon of insufficient supply of items caused by low predicted demand and less stock, or inventory backlog caused by excessively high predicted demand and more stock.

[0167] Fig. 9 An exemplary system architecture 900 is shown to which the method or device for predicting the demand quantity of an item according to an embodiment of the present invention can be applied.

[0168] like Fig. 9As shown, the system architecture 900 may include historical data storage devices 901, 902, 903, networks 904, 906, a server 905, and prediction terminal devices 907, 908, 909. The networks 904, 906 are used to provide a medium for communication links between the historical data storage devices 901, 902, 903, the server 905, and the prediction terminal devices 907, 908, 909. The networks 904, 906 may include various connection types, such as wired, wireless communication links, or optical fiber cables, etc.

[0169] The historical data storage devices 901, 902, 903 and the prediction terminal devices 907, 908, 909 interact with the server 905 through the networks 904, 906 to receive or send messages, etc. The historical data storage devices 901, 902, 903 are electronic devices used by the e-commerce website to store historical data. After receiving the sales forecast request sent by the prediction terminal devices 907, 908, 909, the electronic device (server 905) used for item demand forecasting obtains the historical sales data of the commodity from the historical data storage devices 901, 902, 903, and sends the predicted target demand to the prediction terminal devices 907, 908, 909. The prediction terminal devices 907, 908, 909 are used to receive the target demand sent by the electronic device (server 905).

[0170] The server 905 may be a server that provides various services, such as a background management server that provides support for determining the corresponding target demand according to the historical sales data after obtaining the historical sales data from the historical data storage devices 901, 902, and 903. The background management server may analyze the obtained historical sales data, and send the target demand to the prediction terminal devices 907, 908, and 909 according to the processing results.

[0171] It should be noted that the item demand quantity prediction method provided in the embodiment of the present invention is generally executed by the server 905 , and accordingly, the item demand quantity prediction device is generally disposed in the server 905 .

[0172] It should be understood that Fig. 9 The number of terminal devices, networks and servers in the embodiment is only for illustration. Any number of terminal devices, networks and servers may be provided according to implementation requirements.

[0173] Reference below Fig.10 , which shows a schematic diagram of the structure of a computer system 1000 of a terminal device suitable for implementing an embodiment of the present invention. Fig.10 The terminal device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present invention.

[0174] like Fig.10As shown, the computer system 1000 includes a central processing unit (CPU) 1001, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 1002 or a program loaded from a storage part 1008 into a random access memory (RAM) 1003. In the RAM 1003, various programs and data required for the operation of the system 1000 are also stored. The CPU 1001, the ROM 1002, and the RAM 1003 are connected to each other via a bus 1004. An input / output (I / O) interface 1005 is also connected to the bus 1004.

[0175] The following components are connected to the I / O interface 1005: an input section 1006 including a keyboard, a mouse, etc.; an output section 1007 including a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker, etc.; a storage section 1008 including a hard disk, etc.; and a communication section 1009 including a network interface card such as a LAN card, a modem, etc. The communication section 1009 performs communication processing via a network such as the Internet. A drive 1010 is also connected to the I / O interface 1005 as needed. A removable medium 1011, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 1010 as needed, so that a computer program read therefrom is installed into the storage section 1008 as needed.

[0176] In particular, according to the embodiments disclosed in the present invention, the process described above with reference to the flowchart can be implemented as a computer software program. For example, the embodiments disclosed in the present invention include a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes a program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from the network through the communication part 1009, and / or installed from the removable medium 1011. When the computer program is executed by the central processing unit (CPU) 1001, the above-mentioned functions defined in the system of the present invention are executed.

[0177] It should be noted that the computer-readable medium shown in the present invention may be a computer-readable signal medium or a computer-readable storage medium or any combination of the above two. The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present invention, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, device or device. In the present invention, a computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, which carries a computer-readable program code. This propagated data signal may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. Computer-readable signal media may also be any computer-readable medium other than computer-readable storage media, which may send, propagate or transmit a program for use by or in conjunction with an instruction execution system, apparatus or device. The program code contained on the computer-readable medium may be transmitted using any appropriate medium, including but not limited to: wireless, wire, optical cable, RF, etc., or any suitable combination of the above.

[0178] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present invention. In this regard, each box in the flow chart or block diagram can represent a module, a program segment, or a part of a code, and the above-mentioned module, program segment, or a part of a code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram or flow chart, and the combination of the boxes in the block diagram or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0179] The modules involved in the embodiments of the present invention may be implemented by software or hardware. The modules described may also be set in a processor, for example, it may be described as: a processor includes a receiving module, a time period type determination module, a keyword mining module, an analysis module, and a prediction module. The names of these modules do not constitute a limitation on the modules themselves in some cases, for example, the prediction module may also be described as "a module for predicting the target demand for the item in the time period to be predicted".

[0180] As another aspect, the present invention also provides a computer-readable medium, which may be included in the device described in the above embodiment; or it may exist independently and not be assembled into the device. The above computer-readable medium carries one or more programs, and when the above one or more programs are executed by a device, the device includes: receiving a prediction request for item demand, the prediction request indicates the item identification of the item and the time period to be predicted; determining the target demand increase activity corresponding to the time period to be predicted, and the time interval between the activity period corresponding to the target demand increase activity and the time period to be predicted is not greater than a preset threshold; determining the time period type corresponding to the time period to be predicted according to the time relationship between the activity period and the time period to be predicted; the time period type includes one or more of a target time period belonging to the activity period, a first time period earlier than the target time period, and a second time period later than the target time period; determining the target keyword corresponding to the item according to the item identification; determining the basic demand corresponding to the item according to the target keyword; predicting the target demand of the item in the time period to be predicted according to the time period type and the basic demand.

[0181] According to the technical solution of the embodiment of the present invention, by receiving a forecast request for the demand of an item, the forecast request indicates the item ID of the item and the time period to be predicted; determining the target demand increase activity corresponding to the time period to be predicted, and the time interval between the activity period corresponding to the target demand increase activity and the time period to be predicted is not greater than a preset threshold; determining the time period type corresponding to the time period to be predicted according to the time relationship between the activity period and the time period to be predicted; the time period type includes one or more of a target time period belonging to the activity period, a first time period earlier than the target time period, and a second time period later than the target time period; determining the target keyword corresponding to the item according to the item ID; determining the basic demand corresponding to the item according to the target keyword; and predicting the target demand of the item in the time period to be predicted according to the time period type and the basic demand. Thus, by combining the time period type of the time period to be predicted on the basis of the basic demand, the target demand of the item in the time period to be predicted is predicted, so that the accuracy of the prediction result is improved, thereby reducing the phenomenon of insufficient supply of items caused by low predicted demand and less stock, or inventory backlog caused by excessively high predicted demand and more stock.

[0182] The above specific implementations do not constitute a limitation on the protection scope of the present invention. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions may occur depending on design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for predicting the demand for goods, characterized in that: include: receiving a forecast request for item demand, the forecast request indicating an item identifier of the item and a time period to be forecasted; Determine a target demand quantity improvement activity corresponding to the time period to be predicted, wherein the time interval between the activity period corresponding to the target demand quantity improvement activity and the time period to be predicted is not greater than a preset threshold; Determining a time period type corresponding to the time period to be predicted according to a time relationship between the activity time period and the time period to be predicted; The time period type includes one or more of a target time period belonging to the activity time period, a first time period earlier than the target time period, and a second time period later than the target time period; Determining a target keyword corresponding to the item according to the item identifier; Determine the basic demand corresponding to the item according to the target keyword; According to the time period type and the basic demand, a target demand for the item in the time period to be predicted is predicted.

2. The method according to claim 1, characterized in that The step of predicting the target demand for the item in the to-be-predicted time period according to the time period type and the basic demand includes: In the case where the time period type includes the target time period, determining a target activity factor corresponding to the item according to the historical data of the item; Determining a first target demand corresponding to the target time period according to the basic demand and the target activity factor; The target demand is predicted according to the first target demand and the basic demand.

3. The method according to claim 2, characterized in that Determining the target activity factor corresponding to the item according to the historical data of the item includes: Determine, based on the historical data, a first actual demand corresponding to the item in the historical demand increase activity, first click data in an out-of-stock state, second click data in a non-out-of-stock state, and a demand conversion rate corresponding to the second click data; Determining an ideal demand corresponding to the historical demand enhancement activity according to the first actual demand, the first click data, and the demand conversion rate; Determining, based on the historical data, a second actual demand within a preset period before the historical demand increase activity; The target activity factor is determined according to the ideal demand and the second actual demand.

4. The method according to claim 2, characterized in that: The determining, according to the basic demand and the target activity factor, a first target demand corresponding to the target time period includes: Determining whether the item identifier corresponds to a target price change identifier; the target price change identifier corresponds to the target demand increase activity; In the case where there is a corresponding target price change identifier, determining a target price factor corresponding to the item according to the target price change identifier; The first target demand is determined according to the basic demand, the target activity factor and the target price factor.

5. The method according to claim 4, characterized in that Determining the target price factor corresponding to the item according to the target price change identifier includes: Determine, based on the historical data, a historical price change indicator corresponding to the item in the historical demand increase activity; Determine a historical price factor according to a first actual demand in the historical demand enhancement activity and a second actual demand in a preset period before the historical demand enhancement activity; Determining the relationship between the historical price change indicator and the historical price factor; The target price factor is determined according to the target price change identifier and the relationship.

6. The method according to claim 5, characterized in that The determining the relationship between the historical price change identifier and the historical price factor; and determining the target price factor according to the target price change identifier and the relationship, comprises: Training a price factor prediction model using the historical price change identifier and the historical price factor, so that the price factor prediction model learns the relationship between the historical price change identifier and the historical price factor; The target price change identifier is input into the price factor prediction model to obtain the target price factor.

7. The method according to claim 2, characterized in that The predicting the target demand according to the first target demand and the basic demand includes: determining whether the time period type includes the first time period and / or the second time period other than the target time period; In the case where the first time period and / or the second time period are included, determining a second target demand corresponding to the first time period and / or the second time period according to the basic demand; The target demand quantity is predicted according to the first target demand quantity and the second target demand quantity.

8. The method according to claim 1, characterized in that Determining the basic demand corresponding to the item according to the target keyword includes: Acquire historical data corresponding to the target keyword; According to the historical data, determine the historical demand through data smoothing; Determine the reference demand corresponding to the period to be predicted based on the sales data in the preset period adjacent to the period to be predicted; The reference demand is input into a basic demand prediction model to obtain a basic demand for the item output by the basic demand prediction model; the basic demand prediction model is trained using the historical demand.

9. The method according to claim 1, characterized in that: Determining the target keyword corresponding to the item according to the item identifier includes: Using a word segmentation model to segment the item title corresponding to the item identifier to obtain initial keywords; Screening the initial keywords to obtain candidate keywords; The target keyword is determined according to the candidate keywords.

10. The method according to claim 9, characterized in that Determining the target keyword according to the candidate keyword includes: Determine the time keyword set; combining the time keyword set with the item title as a target item title; Calculating the difference between each keyword in the first candidate keyword and other keywords, and the relevance between each keyword and the target item title; The target keyword is determined from the first candidate keywords according to the difference and the correlation; the target keyword includes a target time keyword related to the commodity in the time keyword set.

11. The method according to claim 10, characterized in that Determining the target keyword from the first candidate keywords according to the difference and the correlation includes: Calculating the maximum marginal correlation coefficient between each keyword and the target item title according to the difference and the correlation; According to the maximum marginal correlation coefficient, it is determined whether each keyword is a target keyword.

12. A device for predicting demand for goods, characterized in that: include: A receiving module, configured to receive a forecast request for item demand, wherein the forecast request indicates an item identifier of the item and a time period to be forecasted; A time period type determination module, used to determine a target demand quantity improvement activity corresponding to the time period to be predicted, wherein the time interval between the activity period corresponding to the target demand quantity improvement activity and the time period to be predicted is not greater than a preset threshold; and determining the time period type corresponding to the time period to be predicted according to the time relationship between the activity time period and the time period to be predicted; The time period type includes one or more of a target time period belonging to the activity time period, a first time period earlier than the target time period, and a second time period later than the target time period; A keyword mining module, used to determine the target keyword corresponding to the item according to the item identifier; An analysis module, used to determine the basic demand corresponding to the item according to the target keyword; The prediction module is used to predict the target demand of the item in the time period to be predicted according to the time period type and the basic demand.

13. An electronic device for predicting the demand for goods, characterized in that: include: one or more processors; a storage device for storing one or more programs, When the one or more programs are executed by the one or more processors, the one or more processors implement the method according to any one of claims 1 to 11.

14. A computer readable medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the method according to any one of claims 1 to 11 is implemented.

Images