CN116579804A - Holiday commodity sales prediction method, holiday commodity sales prediction device and computer storage medium - Google Patents

Abstract

The invention discloses a holiday commodity sales prediction method, a holiday commodity sales prediction device and a computer storage medium, wherein the method comprises the following steps: acquiring historical sales volume data of all the on-sale commodities in the store to form a commodity sales volume standard data set; inputting the commodity sales standard data set into a machine learning reference model, screening out screened commodities with a prediction result lower than a preset threshold according to the prediction precision of the holiday sales of each commodity target section, and forming a screened commodity data set based on the historical sales data of the screened commodities; aiming at the screening commodity data set, carrying out holiday characteristic item expansion and commodity weight index characteristic item expansion; and inputting the screening commodity data set after the characteristic item expansion is completed into a machine learning sales volume prediction model to obtain sales volume predicted values of all shops selling commodities in the target holiday. According to the method, the goods with abnormal sales volume fluctuation in the holidays are selected through screening, and the holiday characteristic items are constructed in a targeted mode to help model prediction, so that sales volume prediction accuracy of the screened goods in the holiday period is effectively improved.

Description

Holiday commodity sales prediction method, holiday commodity sales prediction device and computer storage medium

Technical Field

The invention relates to the technical field of data analysis, in particular to a holiday commodity sales prediction method, a holiday commodity sales prediction device and a computer storage medium.

Background

Accurate sales prediction is an important premise for enterprises to make reasonable operation plans and develop efficient management of supply chains. Because commodity sales in daily periods have a certain time law, enterprises can usually predict accurately by adopting a traditional optimization algorithm. In contrast, sales regularity of commodities during special holidays is difficult to capture, and the characteristic of large fluctuation is often presented. In practice, the enterprise adopts a traditional optimization algorithm aiming at daily period to predict that a large deviation exists between sales volume during holidays and actual values. However, the amount of sales of the product that is explosive during a particular holiday can bring high profits to the retail establishment, and the accuracy of the predictions is critical to the establishment.

Retailers are typically faced with peak customer purchases before and after holidays, and less reliable holiday sales predictions can negatively impact purchasing, inventory management, production control, personnel scheduling, etc., whether the enterprise makes too optimistic or too pessimistic predictions, thereby incurring unnecessary operating costs and profit losses. For example, in order to capture sales profits during spring festival, enterprises can transport a certain amount of commodities from regional distribution centers to front distribution centers closer to retail stores in advance based on the sales profits prediction value of the spring festival, and once predicted deviation occurs, especially fresh products, the quality guarantee period is short, the fresh products are easy to consume, when warehouse extrusion problems occur, the enterprises can generate higher inventory cost and operation risk, and waste of personnel and resources is also brought. Under the condition, if the predicted sales volume is smaller than the actual sales volume of the commodities, the commodity inventory of the front-end distribution center cannot meet the actual demands, the commodity delivery time is increased, the satisfaction degree of customers is reduced, and finally the customer loss is caused.

Therefore, in order to improve the core competitiveness of retail enterprises, how to improve sales prediction accuracy of commodities during special holidays is a problem that needs to be solved. The popularity of enterprise informatization and the development of related hardware technology have made it possible to collect historical transaction data. Each commodity trade order in holidays can be sensed and stored, such as the price of the commodity, the number of sales, the number of customer orders, etc. The accumulation of these data presents opportunities for scientific predictions of sales during a particular holiday, but how to use these data to improve the accuracy of predictions is a hotspot problem for both the enterprise and academia.

In view of this, the present invention is specifically proposed.

Disclosure of Invention

In order to solve the technical problems, the invention provides an optimization processing method, a device and a storage medium suitable for big data indexes, and provides a new method for order processing, commodity screening and auxiliary sales prediction based on holiday characteristics and week weight index characteristic expansion in the prediction process.

Specifically, the following technical scheme is adopted:

a holiday commodity sales prediction method comprises the following steps:

acquiring historical sales volume data of all the on-sale commodities in the store to form a commodity sales volume standard data set;

inputting the commodity sales standard data set into a machine learning reference model, screening out screened commodities with a prediction result lower than a preset threshold according to the prediction precision of the holiday sales of each commodity target section, and forming a screened commodity data set based on the historical sales data of the screened commodities;

performing holiday feature item expansion and commodity weight index feature item expansion on the screening commodity data set;

and inputting the screening commodity data set after the characteristic item expansion is completed into a machine learning sales volume prediction model to obtain sales volume predicted values of all shops selling commodities in the target holiday.

In an alternative embodiment of the present invention, in the holiday commodity sales prediction method, inputting a commodity sales standard data set into a machine learning reference model, and screening the screened commodity with a prediction result lower than a preset threshold according to the prediction accuracy of the commodity target holiday sales includes:

inputting the commodity sales standard data set into a machine learning reference model for model training;

The machine learning reference model after the training is operated gives out sales quantity predicted values of all the sold commodities in the store on the test set;

calculating sales predicting error values of all the goods on sale based on the sales predicting values of all the goods on sale in the store in sequence;

selecting a prediction error value critical point from sales volume prediction error values of all the shops on sale, screening out the commodities with sales volume prediction error values higher than the prediction error value critical point, and summarizing the commodities into a screened commodity list;

optionally, the machine learning reference model is an XGBoost model;

optionally, date data, and/or weather data of an area to which the store belongs, and/or commodity category data, and/or commodity price data are input into the machine learning reference model to assist in the prediction.

In an optional embodiment of the present invention, in the holiday commodity sales predicting method, the step of expanding holiday feature items for the screened commodity data set includes:

counting the dates of the target holidays, respectively expanding each target holiday date forward by a first preset time interval Ts1, expanding backward by a second preset time interval Ts2 to form a complete target holiday period, and constructing corresponding holiday characteristic items according to the complete target holiday period;

Respectively numbering the date of a first preset time interval Ts1 before each target holiday, a target holiday period and a second preset time interval Ts2 after each target holiday period, so as to form a target holiday period interval coding characteristic item to identify each interval in each holiday period, and meanwhile, the numerical value of the holiday characteristic item on other dates is 0;

optionally, the first preset time interval Ts1 is equal to the second preset time interval Ts2.

In the holiday commodity sales predicting method according to the present invention, if commodity sales in a plurality of target holiday periods are predicted simultaneously in the holiday feature item expanding process for the screening commodity data set, dates of different target holiday periods are identified by adding holiday category coding feature items, and the machine learning sales predicting model is used to distinguish different target holiday periods.

In the holiday commodity sales prediction method, in the holiday feature item expansion process for the screening commodity data set, if commodity sales in a plurality of target holiday periods are predicted at the same time and different target holiday periods are partially overlapped, date overlapped by different target holiday periods is identified by adding a holiday overlapping identification feature item, and sales of different target holidays on the overlapped date can be reasonably learned and sales prediction values of corresponding target holidays can be given when the machine learning sales prediction model is used for prediction.

In the holiday commodity sales prediction method according to the present invention, if commodity sales in a plurality of target holiday periods are predicted simultaneously and there is an obvious classification in the plurality of target holidays in the holiday feature item expansion process for the screening commodity data set, the classification of different holidays is identified by adding holiday large-class codes, and the machine learning sales prediction model is used for learning the change rule of commodity sales in each large-class program mark holiday period respectively.

In an optional embodiment of the present invention, in the holiday commodity sales predicting method, the commodity weight index feature item expansion for the screened commodity data set includes:

screening data in commodity data sets, carrying out circulation record according to a circulation period T, summarizing and calculating sales of all the same circulation period to obtain average sales of all the circulation period in the circulation period, sequentially calculating average sales of all the circulation period in the circulation period, selecting one circulation period as a reference circulation period, setting a weight index of the reference circulation period as a, and setting the weight coefficients of other circulation periods as (Q/Q0) a, wherein Q is the average sales of other circulation periods, and Q0 is the average sales of the reference circulation period to obtain daily weight index characteristics of all the circulation periods in the circulation period;

Sequentially adding the daily weight index of each circulation day of each commodity to serve as the circulation period weight index characteristic of the corresponding commodity;

optionally, sequentially screening and calculating the average value of sales of all monday, the average value of sales of all tuesdays and the average value of sales of all tuesdays in a commodity data set, setting a day with the lowest average sales as a reference cycle day, setting a day weight index of the reference cycle day as a, and then using the weight coefficient of the rest 6 days as (Q/Q0) a, wherein Q is the average sales of other days, and Q0 is the average sales of the reference cycle day to obtain day weight index characteristics of monday to sunday; sequentially adding the day weight index of each commodity from monday to sunday to serve as Zhou Quanchong index characteristics of the corresponding commodity;

the daily weight index features and Zhou Quanchong index features of all commodities are incorporated into the screening commodity dataset.

In an alternative embodiment of the present invention, in a holiday commodity sales predicting method, obtaining historical sales data of all on-sale commodities in a store, forming a commodity sales standard data set includes:

performing data preprocessing steps of abnormal order processing, missing value processing, daily sales aggregation and data set division on historical sales volume data of goods sold by a store, so as to form a goods sales volume standard data set;

The abnormal order processing is to delete orders with sales quantity smaller than or equal to a preset sales quantity threshold value;

the missing value processing is to delete or fill the order of the missing sales value;

daily sales aggregation is to add sales of all sales orders according to commodities and dates so as to form a daily sales record data set of each commodity;

the data set division is to divide the processed commodity sales data set into a training set and a testing set according to the selected time interval, and the training set and the testing set are used for training and testing the machine learning reference model.

The invention also provides a holiday commodity sales predicting device, which comprises:

the data processing module is used for acquiring historical sales volume data of all the goods sold in the store and forming a goods sales volume standard data set;

the commodity data screening module inputs the commodity sales standard data set into a machine learning reference model, screens out screened commodities with a prediction result lower than a preset threshold according to the prediction precision of the holiday sales of each commodity target section, and forms a screened commodity data set based on the historical sales data of the screened commodities;

the feature item expansion module is used for expanding holiday feature items and commodity weight index feature items aiming at the screening commodity data set;

And the sales predicting module inputs the screening commodity data set after the characteristic item expansion is completed into a machine learning sales predicting model to obtain sales predicting values of all shops in the commodity sold in the target holiday.

The invention also provides a computer storage medium which stores a computer executable program, and the method for predicting the sales volume of holiday commodity is realized when the computer executable program is executed.

Compared with the prior art, the invention has the beneficial effects that:

the holiday commodity sales volume prediction method is based on the processing mode of order processing, commodity screening and characteristic expansion, and can be used for predicting sales volume of retail store commodity during holidays, and data preprocessing is carried out on store commodity sales order data, so that a sales volume data set in a standard format is formed; then calculating corresponding prediction evaluation indexes according to the operation result of the standard data set in the machine learning reference model, and screening all commodities to form a screened commodity data set; and finally, constructing a corresponding holiday characteristic item according to the target holiday period, expanding the holiday characteristic item to a screening commodity data set, and inputting the screening commodity data set into a machine learning sales prediction model to give sales prediction values of all commodity holiday periods. Therefore, the holiday commodity sales predicting method can screen commodities with abnormal fluctuation of holiday sales (commodities with sales cannot be accurately predicted by a reference model), and pertinently construct holiday characteristic items to help model prediction, so that sales predicting precision of screened commodities in a holiday period is effectively improved.

The invention provides a holiday commodity sales prediction method, which comprises the following steps of: and training and predicting all commodities by using a machine learning reference model, and judging the commodities needing to further improve the sales predicting result by using a holiday sales predicting model according to the accuracy of the predicting result, so that the machine learning sales predicting model can be used for fitting the sales fluctuation rule of the commodities in the holiday period in a targeted manner, the overall operation time is shortened, and the overall predicting accuracy of the machine learning sales predicting model is improved.

The invention provides a feature expansion method, which comprises the following steps of: according to the method, several types of characteristic items marked with key time points, intervals, orders and categories of holidays are constructed aiming at the target holiday period, so that the holiday sales prediction model taking the tree model as a main body can better learn various rules of commodity sales change in the holiday period, and the holiday sales prediction precision of the holiday sales prediction model is effectively improved.

Therefore, the holiday commodity sales predicting method of the invention guides the commodity sales plan and the replenishment plan of retail stores during the holidays by using store historical sales data to perform order processing, utilizing a reference model to perform commodity screening, performing holiday feature item expansion aiming at the target holiday and machine learning sales predicting model to give the full commodity holiday sales predicting value.

Description of the drawings:

FIG. 1 is a flow chart of a holiday commodity sales prediction method disclosed in an embodiment of the present invention;

fig. 2 is a flowchart of a prediction model establishment in a holiday commodity sales prediction method according to an embodiment of the present invention.

Detailed Description

In order to make the objects, technical solutions and advantages of the embodiments of the present invention more clear, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. It will be apparent that the described embodiments are some, but not all, embodiments of the invention.

Thus, the following detailed description of the embodiments of the invention is not intended to limit the scope of the invention, as claimed, but is merely representative of some embodiments of the invention. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.

It should be noted that, under the condition of no conflict, the embodiments of the present invention and the features and technical solutions in the embodiments may be combined with each other.

It should be noted that: like reference numerals and letters denote like items in the following figures, and thus once an item is defined in one figure, no further definition or explanation thereof is necessary in the following figures.

In the description of the present invention, it should be noted that, the terms "upper", "lower", and the like indicate an azimuth or a positional relationship based on the azimuth or the positional relationship shown in the drawings, or an azimuth or a positional relationship conventionally put in use of the inventive product, or an azimuth or a positional relationship conventionally understood by those skilled in the art, such terms are merely for convenience of describing the present invention and simplifying the description, and do not indicate or imply that the apparatus or element to be referred must have a specific azimuth, be constructed and operated in a specific azimuth, and thus should not be construed as limiting the present invention. Furthermore, the terms "first," "second," and the like, are used merely to distinguish between descriptions and should not be construed as indicating or implying relative importance.

Referring to fig. 1 and 2, a holiday commodity sales predicting method according to the present embodiment includes:

acquiring historical sales volume data of all the on-sale commodities in the store to form a commodity sales volume standard data set;

inputting the commodity sales standard data set into a machine learning reference model, screening out screened commodities with a prediction result lower than a preset threshold according to the prediction precision of the holiday sales of each commodity target section, and forming a screened commodity data set based on the historical sales data of the screened commodities;

Performing holiday feature item expansion and commodity weight index feature item expansion on the screening commodity data set;

and inputting the screening commodity data set after the characteristic item expansion is completed into a machine learning sales volume prediction model to obtain sales volume predicted values of all shops selling commodities in the target holiday.

The holiday commodity sales volume prediction method is based on the processing mode of order processing, commodity screening and characteristic expansion, and can be used for predicting sales volume of retail store commodity during holidays, and data preprocessing is carried out on store commodity sales order data, so that a sales volume data set in a standard format is formed; then calculating corresponding prediction evaluation indexes according to the operation result of the standard data set in the machine learning reference model, and screening all commodities to form a screened commodity data set; and finally, constructing a corresponding holiday characteristic item according to the target holiday period, expanding the holiday characteristic item to a screening commodity data set, and inputting the screening commodity data set into a machine learning sales prediction model to give sales prediction values of all commodity holiday periods. Therefore, the holiday commodity sales predicting method can screen commodities with abnormal fluctuation of holiday sales (commodities with sales cannot be accurately predicted by a reference model), and pertinently construct holiday characteristic items to help model prediction, so that sales predicting precision of screened commodities in a holiday period is effectively improved.

The invention provides a holiday commodity sales predicting method, which is characterized in that all commodities are trained and predicted through a machine learning reference model, and then the commodities needing to further improve sales predicting results by using a holiday sales predicting model are judged according to the accuracy of the predicting results, so that the machine learning sales predicting model can be used for fitting the sales fluctuation rule of the commodities in the holiday period in a targeted manner, the overall operation time is shortened, and the overall predicting accuracy of the machine learning sales predicting model is improved.

Therefore, the holiday commodity sales predicting method of the invention guides the commodity sales plan and the replenishment plan of retail stores during the holidays by using store historical sales data to perform order processing, utilizing a reference model to perform commodity screening, performing holiday feature item expansion aiming at the target holiday and machine learning sales predicting model to give the full commodity holiday sales predicting value.

In the holiday commodity sales prediction method of this embodiment, inputting the commodity sales standard data set into the machine learning reference model, and screening the screened commodity with the prediction result lower than the preset threshold according to the prediction accuracy of the target holiday sales of each commodity includes:

Inputting the commodity sales standard data set into a machine learning reference model for model training;

the machine learning reference model after the training is operated gives out sales quantity predicted values of all the sold commodities in the store on the test set;

calculating sales predicting error values of all the goods on sale based on the sales predicting values of all the goods on sale in the store in sequence;

and selecting a prediction error value critical point from sales volume prediction error values of all shops on sale, screening out the commodities with sales volume prediction error values higher than the prediction error value critical point, and summarizing the commodities into a screened commodity list.

Optionally, the machine learning reference model is constructed by taking the XGBoost model as a main model. The XGBoost (eXtreme Gradient Boosting) model is an open source framework for gradient lifting (Gradient Boosting) created by doctor Chen Tianji of the university of washington, and is an efficient, flexible and portable machine learning algorithm. The model is an addition model formed by a plurality of weak learners, the deviation of the prediction results of all the weak learners is corrected by the newly added weak learners, and the model uses the information of the second derivative when solving the optimization target on the basis of GBDT objective function, so that the definition of the optimization target is more accurate, and the training speed is accelerated; meanwhile, a regular term is added in the optimization objective function to limit the complexity of the model and prevent overfitting. Overall, the XGBoost model has very excellent performance in algorithm performance while the overall running speed is fast.

Optionally, date data, and/or weather data of an area to which the store belongs, and/or commodity category data, and/or commodity price data are input into the machine learning reference model to assist in the prediction.

The feature expansion method is provided by the holiday commodity sales prediction method, and the feature items of key time points, intervals, orders and categories of holidays are marked according to the target holiday period, so that the holiday sales prediction model taking the tree model as a main body can learn various rules of commodity sales change in the holiday period better, and the holiday sales prediction precision of the holiday sales prediction model is improved effectively.

Further, the present embodiment is directed to a retail store full commodity screening process comprising:

the date data are converted into characteristic items and are combined into a commodity sales standard data set, wherein the characteristic items comprise a year number, a quarter number, a month number, a week number and a day number, and the characteristic items are converted into an integer data format for storage so as to be used by directly inputting an XGBoost model subsequently.

The weather data are converted into characteristic items and are combined into a commodity sales standard data set, wherein the characteristic items comprise weather states, day average temperature, night average temperature, wind power and air quality, and the weather data are converted into integer data formats for storage, so that the weather data can be directly input into the XGBoost model for use.

Optionally, the weather data information may be expanded, and the precipitation, barometric pressure, humidity, wind direction, and extreme weather warning information may be converted into an integer data format for incorporation as a feature item into a commodity sales standard data set.

Converting the commodity class information into characteristic items, incorporating the characteristic items into a commodity sales standard data set, including a part number, a major class number, a middle class number, a minor class number and a commodity number, and converting the commodity class information into an integer data format for storage so as to directly input an XGBoost model for use.

And converting the commodity unit price information into characteristic items, merging the characteristic items into a commodity sales standard data set, and converting the characteristic items into a floating point data format for storage so as to be directly input into an XGBoost model for use.

Alternatively, various price information of a commodity such as standard price, sales price, etc. may be converted into a floating point data format and incorporated as a feature item into a commodity sales volume standard data set.

Inputting the data set which is integrated with the four major types of characteristic items into a reference model which takes the XGBoost model as a main body for training, and giving out sales prediction values of all commodities in a target holiday period to form a commodity target holiday prediction sales data set.

As an alternative embodiment of the invention, parameter optimization of the XGBoost model is performed on the whole training data set, and the XGBoost model parameters are optimized by grid search.

And storing the optimal parameters, and calling the group of parameters to perform model construction and prediction when the sales are predicted.

The important parameters to be optimally adjusted in the XGBoost model training include booster, objective, learning _rate, max_depth, n_evators and n_ jobs, subsample, colsample _ bytree, colsample _byevel, which correspond to a model solving mode, a loss function, a model learning rate, the maximum depth of a tree, the number of submodels, the number of parallel threads, a training subsampled ratio, the proportion of feature samples when the tree is built and the proportion of feature samples when the tree nodes are split.

Classifying the screened commodity data set according to commodities, sequentially calculating errors between the sales predicted value and the true value in the holiday period of each commodity target section, and calculating by adopting a WMAPE formula.

And calculating the average value of all the commodity prediction errors WMAPEs, and screening commodities by taking the selected multiple of the average value as a threshold value/critical value, namely screening commodities with the prediction error WMAPEs larger than the threshold value/critical value, so as to form a screening commodity list.

In the holiday commodity sales prediction method of this embodiment, the step of expanding holiday feature items for the screened commodity data set includes:

counting the dates of the target holidays, respectively expanding each target holiday date forward by a first preset time interval Ts1, expanding backward by a second preset time interval Ts2 to form a complete target holiday period, and constructing corresponding holiday characteristic items according to the complete target holiday period;

the dates of the three intervals are respectively numbered respectively according to a first preset time interval Ts1 before each target holiday, a target holiday period and a second preset time interval Ts2 after each target holiday period, so that a target holiday period interval coding characteristic item is formed to identify each interval in each holiday period, and meanwhile, the numerical value of the holiday characteristic item on other dates is 0.

Optionally, the first preset time interval Ts1 is equal to the second preset time interval Ts2.

Specifically, as an implementation manner of the present embodiment, when the first preset time interval Ts1 and the second preset time interval Ts2 are the same and equal to 7 days (one week), the dates of the three sections of the week before, during and the week after each holiday are respectively numbered (sequentially encoded in {1,2,3 }), so as to form a holiday period section encoding feature term to identify each section in each holiday period, and the numerical value of the feature term on other dates is 0;

Further, the date of the previous week of each holiday is encoded with { -7, -6, -5, -4, -3, -2, -1} in sequence, and then the date of the holiday period and the date of the next week are encoded with {1,2, 3.}, in sequence, whereby the order of dates within each holiday period is noted, while the value of the feature item on other dates is 0.

It should be understood by those skilled in the art that the first preset time interval Ts1 and the second preset time interval Ts2 in this embodiment may be the same or different, and may be specifically preset according to the holiday of the target festival.

Optionally, in the holiday commodity sales predicting method of the present embodiment, in the process of expanding the holiday feature item for the screened commodity data set, if commodity sales in multiple target holiday periods are predicted at the same time, dates of different target holiday periods are identified by adding holiday category coding feature items, and the machine learning sales predicting model is used for distinguishing different target holiday periods.

Optionally, in the holiday commodity sales predicting method according to this embodiment, in the process of expanding the holiday feature item for the screened commodity data set, if commodity sales in multiple target holiday periods are predicted at the same time and there is a situation that different target holiday periods partially overlap, date overlapping different target holiday periods is identified by adding a holiday overlapping identification feature item, and the machine learning sales predicting model is used for reasonably learning sales of different target holidays on overlapping dates and providing sales predicting values of corresponding target holidays when predicting.

Optionally, in the holiday commodity sales predicting method according to this embodiment, in the process of expanding the holiday feature item for the screened commodity data set, if commodity sales in a plurality of target holiday periods are predicted at the same time, and obvious classifications (such as common holidays and legal holidays) exist in the plurality of target holidays, the classification of different holidays is identified by adding holiday large class codes, and the machine learning sales predicting model is used for respectively learning the change rule of commodity sales in each large class program label holiday period.

In the holiday commodity sales predicting method of this embodiment, the commodity weight index feature item expansion for the screened commodity data set includes:

screening data in commodity data sets, carrying out circulation record according to a circulation period T, summarizing and calculating sales of all the same circulation period to obtain average sales of all the circulation period in the circulation period, sequentially calculating average sales of all the circulation period in the circulation period, selecting one circulation period as a reference circulation period, setting a weight index of the reference circulation period as a, and setting the weight coefficients of other circulation periods as (Q/Q0) a, wherein Q is the average sales of other circulation periods, and Q0 is the average sales of the reference circulation period to obtain daily weight index characteristics of all the circulation periods in the circulation period;

And sequentially adding the daily weight index of each commodity on each cycle day to serve as the cycle period weight index characteristic of the corresponding commodity.

Optionally, sequentially screening and calculating the average value of sales of all monday, the average value of sales of all tuesdays and the average value of sales of all tuesdays in a commodity data set, setting a day with the lowest average sales as a reference cycle day, setting a day weight index of the reference cycle day as a, and then using the weight coefficient of the rest 6 days as (Q/Q0) a, wherein Q is the average sales of other days, and Q0 is the average sales of the reference cycle day to obtain day weight index characteristics of monday to sunday; sequentially adding the day weight index from monday to sunday (preset period Tm) of each commodity to serve as Zhou Quanchong index characteristics of the corresponding commodity;

the daily weight index features and Zhou Quanchong index features of all commodities are incorporated into the screening commodity dataset.

The daily weight index feature and Zhou Quanchong index feature of the commodity are calculated in a cycle period of about 7 days, and the present embodiment may also be calculated in other cycle periods, for example, in a cycle period of about 10 days.

Further, the training and predicting process of the machine learning sales prediction model of the present embodiment includes:

Inputting the screened commodity sales data set which is integrated with the holiday characteristic items, the Zhou Quanchong index and the daily weight index characteristic items into a holiday sales prediction model which takes the XGBoost model as a main body for training, and giving sales prediction values of screened commodities in spring festival.

Alternatively, the overall prediction effect of the machine-learned sales prediction model may be analyzed by calculating a corresponding prediction error WMAPE value from the sales prediction value.

As an optional implementation mode of the invention, parameter optimization of the XGBoost model is carried out aiming at the whole training data set, and the parameters of the XGBoost model are optimized by utilizing grid search;

saving the optimal parameters, and calling the group of parameters to perform model construction and prediction when the sales quantity is predicted;

the important parameters to be optimally adjusted in the XGBoost model training include booster, objective, learning _rate, max_depth, n_evators and n_ jobs, subsample, colsample _ bytree, colsample _byevel, which correspond to a model solving mode, a loss function, a model learning rate, the maximum depth of a tree, the number of submodels, the number of parallel threads, a training subsampled ratio, the proportion of feature samples when the tree is built and the proportion of feature samples when the tree nodes are split.

The method for predicting sales volume of commodities on holidays in this embodiment, obtaining historical sales volume data of all commodities on sale in a store, and forming a commodity sales volume standard data set includes:

performing data preprocessing steps of abnormal order processing, missing value processing, daily sales aggregation and data set division on historical sales volume data of goods sold by a store, so as to form a goods sales volume standard data set;

the abnormal order processing is to delete orders with sales quantity smaller than or equal to a preset sales quantity threshold, and optionally, the preset sales quantity threshold is selected to be 0;

the missing value processing is to delete or fill the order of the missing sales value;

daily sales aggregation is to add sales of all sales orders according to commodities and dates so as to form a daily sales record data set of each commodity;

the data set division is to divide the processed commodity sales data set into a training set and a testing set according to the selected time interval, and the training set and the testing set are used for training and testing the machine learning reference model.

Specifically, the data set after the processing is divided into a training data set and a test data set according to a set proportion, wherein the daily sales record number proportion of each commodity in the two data sets is about 8:2, and meanwhile, the daily sales record data of the corresponding commodity in at least one complete spring festival period is contained in the test set.

And carrying out data format unified processing on the data set and checking each column of data missing value to form a commodity sales standard data set for the machine learning reference model.

Optionally, sales of all commodities in the data set in all target holiday periods can be supplemented completely, wherein dates without sales records are filled with 0, so that a subsequent model can capture the complete sales change rule of the commodities in the target holiday periods.

Alternatively, the abnormal value inspection and processing may be performed on all of the daily sales data of each commodity in turn, and the daily sales data of abnormally high or abnormally low may be removed or smoothed.

Optionally, the sales order data set may be eliminated from any sales records or very few records of merchandise within the target holiday period, simplifying the overall data set size.

The invention also provides a holiday commodity sales predicting device, which comprises:

the data processing module is used for acquiring historical sales volume data of all the goods sold in the store and forming a goods sales volume standard data set;

the commodity data screening module inputs the commodity sales standard data set into a machine learning reference model, screens out screened commodities with a prediction result lower than a preset threshold according to the prediction precision of the holiday sales of each commodity target section, and forms a screened commodity data set based on the historical sales data of the screened commodities;

The feature item expansion module is used for expanding holiday feature items and commodity weight index feature items aiming at the screening commodity data set;

and the sales predicting module inputs the screening commodity data set after the characteristic item expansion is completed into a machine learning sales predicting model to obtain sales predicting values of all shops in the commodity sold in the target holiday.

The present embodiment also provides a computer storage medium storing a computer executable program which, when executed, implements a holiday commodity sales prediction method as described above.

The computer storage medium of this embodiment may include a data signal propagated in baseband or as part of a carrier wave with readable program code embodied therein. Such a propagated data signal may take any of a variety of forms, including, but not limited to, electro-magnetic, optical, or any suitable combination of the foregoing. A computer readable storage medium may also be any readable medium that can communicate, propagate, or transport a program for use by or in connection with an instruction execution system, apparatus, or device. Program code embodied on a computer readable storage medium may be transmitted using any appropriate medium, including but not limited to wireless, wireline, optical fiber cable, RF, etc., or any suitable combination of the foregoing.

The embodiment also provides an electronic device, which comprises a processor and a memory, wherein the memory is used for storing a computer executable program, and when the computer program is executed by the processor, the processor executes the holiday commodity sales prediction method.

The electronic device is in the form of a general purpose computing device. The processor may be one or a plurality of processors and work cooperatively. The invention does not exclude that the distributed processing is performed, i.e. the processor may be distributed among different physical devices. The electronic device of the present invention is not limited to a single entity, but may be a sum of a plurality of entity devices.

The memory stores a computer executable program, typically machine readable code. The computer readable program may be executable by the processor to enable an electronic device to perform the method, or at least some of the steps of the method, of the present invention.

The memory includes volatile memory, such as Random Access Memory (RAM) and/or cache memory, and may be non-volatile memory, such as Read Only Memory (ROM).

It should be understood that elements or components not shown in the above examples may also be included in the electronic device of the present invention. For example, some electronic devices further include a display unit such as a display screen, and some electronic devices further include a man-machine interaction element such as a button, a keyboard, and the like. The electronic device may be considered as covered by the invention as long as the electronic device is capable of executing a computer readable program in a memory for carrying out the method or at least part of the steps of the method.

From the above description of embodiments, those skilled in the art will readily appreciate that the present invention may be implemented by hardware capable of executing a specific computer program, such as the system of the present invention, as well as electronic processing units, servers, clients, handsets, control units, processors, etc. included in the system. The invention may also be implemented by computer software executing the method of the invention, e.g. by control software executed by a microprocessor, an electronic control unit, a client, a server, etc. It should be noted, however, that the computer software for performing the method of the present invention is not limited to being executed by one or a specific hardware entity, but may also be implemented in a distributed manner by unspecified specific hardware. For computer software, the software product may be stored on a computer readable storage medium (which may be a CD-ROM, a usb disk, a removable hard disk, etc.), or may be stored distributed over a network, as long as it enables the electronic device to perform the method according to the invention.

Example 1

The embodiment is constructed for the problem of predicting sales of goods by retail stores during spring festival;

referring to fig. 2, the order processing procedure of the retail store sales data according to the present embodiment includes:

The sales quantity of sales orders of retail stores is extracted completely, and the sales orders with the sales quantity less than or equal to zero are removed;

the sales quantity of sales orders of retail stores is extracted completely, and the sales orders with the missing sales quantity or the wrong data format are filled with data or corrected in the data format;

the sales order data subjected to abnormal order processing and missing value filling are aggregated according to the days, so that daily sales volume data of all commodities in retail stores are obtained, and the commodities and the dates are sequentially ordered;

dividing the processed data set into a training data set and a test data set according to a set proportion, wherein the daily sales record number proportion of each commodity in the two data sets is about 8:2, and simultaneously ensuring that the test set contains daily sales record data of the corresponding commodity in at least one complete spring festival period;

and carrying out data format unified processing on the data set and checking each column of data missing value to form a commodity sales standard data set for a subsequent model.

Optionally, sales of all commodities in the data set in all spring festival periods can be supplemented completely, wherein the date without sales record is filled with 0, so that a subsequent model can capture the complete sales change rule of the commodities in the spring festival period;

Alternatively, the abnormal value inspection and processing can be sequentially performed on all daily sales data of each commodity, and the daily sales data with abnormally high or abnormally low can be removed or smoothed;

optionally, the sales order data set may be eliminated without any sales records or with very few records in the spring festival period, simplifying the overall data set size.

Further, the present embodiment is directed to a retail store full commodity screening process comprising:

converting date data into characteristic items, incorporating the characteristic items into a commodity sales standard data set, wherein the characteristic items comprise a year number, a quarter number, a month number, a week number and a day number, and converting the characteristic items into an integer data format for storage so as to be convenient for the subsequent direct input of an XGBoost model for use;

converting weather data into characteristic items, and integrating the characteristic items into a commodity sales standard data set, wherein the characteristic items comprise weather states, day average temperature, night average temperature, wind power and air quality, and the weather data are converted into integer data formats for storage so as to be directly input into an XGBoost model for use;

optionally, the weather data information can be expanded, the precipitation, the atmospheric pressure, the humidity, the wind direction and the extreme weather early warning information are converted into integer data formats, and the integer data formats are used as characteristic items to be combined into a commodity sales standard data set;

Converting the category information of each level of commodity into characteristic items, and incorporating the characteristic items into a commodity sales standard data set, wherein the characteristic items comprise a part number, a major class number, a middle class number, a minor class number and a commodity number, and converting the characteristic items into an integer data format for storage so as to facilitate the subsequent direct input of an XGBoost model for use;

converting commodity unit price information into characteristic items, merging the characteristic items into a commodity sales standard data set, and converting the characteristic items into a floating point data format for storage so as to directly input an XGBoost model for use later;

alternatively, various price information such as standard price, sales price and the like of the commodity can be converted into a floating point data format and can be used as characteristic items to be incorporated into a commodity sales volume standard data set;

inputting the data set which is integrated with the four major types of characteristic items into a reference model which takes the XGBoost model as a main body for training, and giving sales prediction values of all commodities in spring festival period to form a commodity spring festival prediction sales data set;

as an optional implementation mode of the invention, parameter optimization of the XGBoost model is carried out aiming at the whole training data set, and the parameters of the XGBoost model are optimized by utilizing grid search;

saving the optimal parameters, and calling the group of parameters to perform model construction and prediction when the sales quantity is predicted;

Important parameters to be optimally adjusted in XGBoost model training include booster, objective, learning _rate, max_depth, n_evators and n_ jobs, subsample, colsample _ bytree, colsample _byevel, which correspond to a model solving mode, a loss function, a model learning rate, the maximum depth of a tree, the number of submodels, the number of parallel threads, a training subsamples duty ratio, the proportion of feature samples when the tree is built and the proportion of feature samples when the tree nodes are split;

classifying the commodity spring festival predicted sales volume data set according to commodities, sequentially calculating errors between sales volume predicted values and true values in the period of each commodity spring festival, and calculating by adopting a WMAPE formula;

calculating the average value of all commodity prediction errors WMAPEs, and screening commodities by taking a selected multiple of the average value as a threshold value/critical value, namely screening commodities with the prediction errors WMAPEs larger than the threshold value/critical value to form a screening commodity list;

integrating daily sales data of the screened commodity and corresponding characteristic item (date characteristic, weather characteristic, commodity coding characteristic and commodity price characteristic) data into a screened commodity data set according to the screened commodity list;

dividing the screening commodity data set into a training data set and a test data set according to the selected demarcation point, wherein the daily sales volume record number proportion of each commodity in the two data sets is about 8:2, and simultaneously ensuring that the test set contains daily sales volume record data of the corresponding commodity in at least one complete spring festival period;

Alternatively, an appropriate threshold selection range may be selected by plotting the prediction error WMAPE of all the commodities, and sequentially testing the selected range for a plurality of thresholds to select an optimal threshold.

Further, the feature term expansion of the screening commodity sales data set according to the embodiment includes three items of spring festival period feature term, zhou Quanchong index and daily weight index feature term, and the processing procedure includes:

the construction process of the spring festival period characteristic item comprises the following steps:

integrating and extracting the date of legal seven-day holidays in spring festival and the date of each week before and after in the screening commodity sales volume data set to form a complete spring festival sales period each year;

the dates of the three sections of the week before the spring festival, the period of the spring festival and the week after the spring festival are respectively numbered (coded in sequence of {1,2,3 }), so that a spring festival period section coding characteristic item is formed to identify each section in the spring festival period each year, and meanwhile, the numerical value of the characteristic item on other dates is 0;

further, the date of the previous week of the spring festival is encoded with { -7, -6, -5, -4, -3, -2, -1} in sequence, then the date of the period of the spring festival and the date of the subsequent week is encoded with {1,2, 3.}, thereby marking the date order in the period of the spring festival each year, while the value of the feature item on the other date is 0;

Sequentially merging the spring festival period characteristic items constructed above into a screening commodity data set;

optionally, if commodity sales in multiple holiday periods are predicted at the same time, a holiday category encoding feature item can be added to identify dates of different holiday periods, so that a subsequent model can distinguish different holiday periods;

optionally, if sales of commodities in multiple holiday periods are predicted at the same time and different holiday periods are partially overlapped, a holiday overlapping identification feature item can be added to identify dates on which different holiday periods overlap, so that a subsequent model can reasonably learn sales of overlapping dates and give corresponding sales prediction values when predicting;

optionally, if sales of goods in multiple holidays are predicted at the same time, and there are obvious classifications (such as common holidays and legal holidays) in multiple holidays, a holiday major class code can be added to identify the classes of different holidays, so that the subsequent model can learn the change rule of sales of goods in each major class holiday period respectively.

The construction process of the characteristic items of the week weight index and the day weight index comprises the following steps:

Calculating average sales of each commodity from monday to sunday in the screened commodity data set in sequence, setting the daily weight index of the day with the lowest average sales as 1.0, and dividing the average sales of the rest 6 days by the lowest value to obtain the daily weight index from monday to sunday;

sequentially adding the day weight index of each commodity from monday to sunday to serve as Zhou Quanchong index of the corresponding commodity;

the daily weight index and the weekly weight index of all commodities are incorporated as feature items into the screening commodity data set.

Further, the training and predicting process of the holiday sales predicting model of the present embodiment includes:

inputting the screening commodity sales data set which is integrated with the spring festival period characteristic item, the Zhou Quanchong index and the daily weight index characteristic item into a holiday sales prediction model taking the XGBoost model as a main body for training, and giving a sales prediction value of the screening commodity in the spring festival;

optionally, calculating a corresponding prediction error WMAPE value according to the commodity sales prediction value to analyze the overall prediction effect of the holiday sales prediction model;

as an optional implementation mode of the invention, parameter optimization of the XGBoost model is carried out aiming at the whole training data set, and the parameters of the XGBoost model are optimized by utilizing grid search;

Saving the optimal parameters, and calling the group of parameters to perform model construction and prediction when the sales quantity is predicted;

the important parameters to be optimally adjusted in the XGBoost model training include booster, objective, learning _rate, max_depth, n_evators and n_ jobs, subsample, colsample _ bytree, colsample _byevel, which correspond to a model solving mode, a loss function, a model learning rate, the maximum depth of a tree, the number of submodels, the number of parallel threads, a training subsampled ratio, the proportion of feature samples when the tree is built and the proportion of feature samples when the tree nodes are split.

The holiday commodity sales predicting method based on order processing, commodity screening and feature expansion gives sales predicting values during commodity spring festival, and the sales and replenishment plans of retail stores during spring festival can be guided after summarized processing.

The embodiment of the invention provides an implementation process of a holiday commodity sales predicting method based on order processing, commodity screening and feature expansion, wherein retail store sales order data are processed into a data set in a standard format in the order processing process, a predicting result obtained after a reference model is input is screened out of commodity sets needing further improving the holiday sales predicting effect, and then corresponding feature items are constructed for the screened commodity sets and target holidays to expand the data sets, so that the holiday sales predicting model can effectively predict sales values of commodities in the holiday period, and model parameter optimizing strategies can guarantee generalization capacity of model overall predicting performance.

The above embodiments are only for illustrating the present invention and not for limiting the technical solutions described in the present invention, and although the present invention has been described in detail in the present specification with reference to the above embodiments, the present invention is not limited to the above specific embodiments, and thus any modifications or equivalent substitutions are made to the present invention; all technical solutions and modifications thereof that do not depart from the spirit and scope of the invention are intended to be included in the scope of the appended claims.

Claims (10)

1. The holiday commodity sales prediction method is characterized by comprising the following steps of:

acquiring historical sales volume data of all the on-sale commodities in the store to form a commodity sales volume standard data set;

inputting the commodity sales standard data set into a machine learning reference model, screening out screened commodities with a prediction result lower than a preset threshold according to the prediction precision of the holiday sales of each commodity target section, and forming a screened commodity data set based on the historical sales data of the screened commodities;

performing holiday feature item expansion and commodity weight index feature item expansion on the screening commodity data set;

and inputting the screening commodity data set after the characteristic item expansion is completed into a machine learning sales volume prediction model to obtain sales volume predicted values of all shops selling commodities in the target holiday.

2. The holiday commodity sales prediction method according to claim 1, wherein inputting the commodity sales standard data set into a machine learning reference model, and screening the screened commodity with the prediction result lower than the preset threshold according to the prediction accuracy of the commodity target holiday sales comprises:

inputting the commodity sales standard data set into a machine learning reference model for model training;

the machine learning reference model after the training is operated gives out sales quantity predicted values of all the sold commodities in the store on the test set;

calculating sales predicting error values of all the goods on sale based on the sales predicting values of all the goods on sale in the store in sequence;

selecting a prediction error value critical point from sales volume prediction error values of all the shops on sale, screening out the commodities with sales volume prediction error values higher than the prediction error value critical point, and summarizing the commodities into a screened commodity list;

optionally, the machine learning reference model is constructed by taking the XGBoost model as a main model;

optionally, date data, and/or weather data of an area to which the store belongs, and/or commodity category data, and/or commodity price data are input into the machine learning reference model to assist in the prediction.

3. A holiday commodity sales prediction method according to claim 1 or 2, wherein said holiday feature term expansion for said screening commodity data set comprises:

counting the dates of the target holidays, respectively expanding each target holiday date forward by a first preset time interval Ts1, expanding backward by a second preset time interval Ts2 to form a complete target holiday period, and constructing corresponding holiday characteristic items according to the complete target holiday period;

respectively numbering the date of a first preset time interval Ts1 before each target holiday, a target holiday period and a second preset time interval Ts2 after each target holiday period, so as to form a target holiday period interval coding characteristic item to identify each interval in each holiday period, and meanwhile, the numerical value of the holiday characteristic item on other dates is 0;

optionally, the first preset time interval Ts1 is equal to the second preset time interval Ts2.

4. A holiday commodity sales prediction method according to claim 3, wherein in the holiday feature item expansion process for the screened commodity data set, if commodity sales in a plurality of target holiday periods are predicted at the same time, dates of different target holiday periods are identified by adding holiday category coding feature items, and the machine learning sales prediction model is used for distinguishing different target holiday periods.

5. The holiday commodity sales prediction method according to claim 3, wherein in the holiday feature item expansion process for the screened commodity data set, if commodity sales in a plurality of target holiday periods are predicted at the same time and different target holiday periods are partially overlapped, date of overlapping of different target holiday periods is identified by adding a holiday overlapping identification feature item, and sales of different target holidays on the overlapping date can be reasonably learned and sales prediction values of corresponding target holidays can be given when the machine learning sales prediction model predicts.

6. The holiday commodity sales prediction method according to claim 3, wherein in the holiday feature item expansion process for the screened commodity data set, if commodity sales in a plurality of target holiday periods are predicted simultaneously and obvious classifications exist in the plurality of target holidays, classification of different holidays is identified by adding holiday large class codes, and the machine learning sales prediction model is used for learning the change rule of commodity sales in each large class program label holiday period respectively.

7. The holiday commodity sales prediction method according to claim 1, wherein said commodity weight index feature term expansion for said screening commodity data set comprises:

screening data in commodity data sets, carrying out circulation record according to a circulation period T, summarizing and calculating sales of all the same circulation period to obtain average sales of all the circulation period in the circulation period, sequentially calculating average sales of all the circulation period in the circulation period, selecting one circulation period as a reference circulation period, setting a weight index of the reference circulation period as a, and setting the weight coefficients of other circulation periods as (Q/Q0) a, wherein Q is the average sales of other circulation periods, and Q0 is the average sales of the reference circulation period to obtain daily weight index characteristics of all the circulation periods in the circulation period;

sequentially adding the daily weight index of each circulation day of each commodity to serve as the circulation period weight index characteristic of the corresponding commodity;

optionally, sequentially screening and calculating the average value of sales of all monday, the average value of sales of all tuesdays and the average value of sales of all tuesdays in a commodity data set, setting a day with the lowest average sales as a reference cycle day, setting a day weight index of the reference cycle day as a, and then using the weight coefficient of the rest 6 days as (Q/Q0) a, wherein Q is the average sales of other days, and Q0 is the average sales of the reference cycle day to obtain day weight index characteristics of monday to sunday;

Sequentially adding the day weight index from monday to sunday (preset period Tm) of each commodity to serve as Zhou Quanchong index characteristics of the corresponding commodity;

the daily weight index features and Zhou Quanchong index features of all commodities are incorporated into the screening commodity dataset.

8. The holiday commodity sales predicting method according to claim 1, wherein obtaining historical sales data of all on-sale commodities in a store, forming a commodity sales standard data set comprises:

performing data preprocessing steps of abnormal order processing, missing value processing, daily sales aggregation and data set division on historical sales volume data of goods sold by a store, so as to form a goods sales volume standard data set;

the abnormal order processing is to delete orders with sales quantity smaller than or equal to a preset sales quantity threshold value;

the missing value processing is to delete or fill the order of the missing sales value;

daily sales aggregation is to add sales of all sales orders according to commodities and dates so as to form a daily sales record data set of each commodity;

the data set division is to divide the processed commodity sales data set into a training set and a testing set according to the selected time interval, and the training set and the testing set are used for training and testing the machine learning reference model.

9. A holiday commodity sales predicting apparatus, comprising:

the data processing module is used for acquiring historical sales volume data of all the goods sold in the store and forming a goods sales volume standard data set;

the commodity data screening module inputs the commodity sales standard data set into a machine learning reference model, screens out screened commodities with a prediction result lower than a preset threshold according to the prediction precision of the holiday sales of each commodity target section, and forms a screened commodity data set based on the historical sales data of the screened commodities;

the feature item expansion module is used for expanding holiday feature items and commodity weight index feature items aiming at the screening commodity data set;

and the sales predicting module inputs the screening commodity data set after the characteristic item expansion is completed into a machine learning sales predicting model to obtain sales predicting values of all shops in the commodity sold in the target holiday.

10. A computer storage medium storing a computer executable program, wherein the computer executable program when executed implements a holiday commodity sales prediction method according to any one of claims 1-8.