CN113837783A - Time series model parameter optimization method and device and computer equipment - Google Patents
Time series model parameter optimization method and device and computer equipment Download PDFInfo
- Publication number
- CN113837783A CN113837783A CN202010588655.0A CN202010588655A CN113837783A CN 113837783 A CN113837783 A CN 113837783A CN 202010588655 A CN202010588655 A CN 202010588655A CN 113837783 A CN113837783 A CN 113837783A
- Authority
- CN
- China
- Prior art keywords
- time sequence
- parameter
- time series
- determining
- value
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0201—Market modelling; Market analysis; Collecting market data
- G06Q30/0202—Market predictions or forecasting for commercial activities
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0201—Market modelling; Market analysis; Collecting market data
Landscapes
- Business, Economics & Management (AREA)
- Strategic Management (AREA)
- Engineering & Computer Science (AREA)
- Accounting & Taxation (AREA)
- Development Economics (AREA)
- Finance (AREA)
- Entrepreneurship & Innovation (AREA)
- Game Theory and Decision Science (AREA)
- Data Mining & Analysis (AREA)
- Economics (AREA)
- Marketing (AREA)
- Physics & Mathematics (AREA)
- General Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The application relates to a time series model parameter optimization method, a time series model parameter optimization device, computer equipment and a storage medium. The method comprises the following steps: acquiring a training time sequence data set and a verification time sequence data set representing sales volumes of various articles; determining an optimizable periodic parameter set according to the time sequence duration of the training time sequence data set and the time sequence duration range of the periodic parameters which can be optimized; determining the periodic parameter value of the optimizable periodic parameter according to the evaluation index value and the verification time sequence data set; determining a preset effect value of a holiday type according to the training time sequence data set to optimize a holiday item parameter to obtain a holiday item parameter value; determining the variable point number supported by the time series model according to the time sequence duration, optimizing the trend item parameter according to the variable point number to obtain a trend item parameter value, determining the optimal parameter value group of the time series model, and obtaining the optimized time series model. By adopting the method, the accuracy of the time series model in predicting the sales of different types of articles can be improved.
Description
Technical Field
The application relates to the technical field of product sales prediction, in particular to a time series model parameter optimization method, a time series model parameter optimization device, computer equipment and a storage medium.
Background
With the advancement of informatization and datamation, the intelligent supply operation construction of enterprises is becoming more and more important. The forecast of sales volume, as the "first line of defense" in modern supply chain management, has profound effects on the production, restocking, and sales plans of enterprises. How to accurately predict the sales volume of the classification of the SKUs with different sales expressions and various classifications is an important basic stone for the operation of the enterprise supply chain. At present, a time series prophet model which is applied to a time series model of sales prediction and can be a generalized model based on a time series is widely accepted in the industry and applied to various manufacturing scenes due to the advantages of flexibility, strong interpretability, wide application scenes and the like.
When the Prophet model is used for performing classification prediction, a user needs to set the model parameters of each classification one by one, and the difficulty is high; if the product class characteristics are neglected and the uniform model parameters are used, the accuracy of the sales prediction result is low.
Disclosure of Invention
In view of the above, it is necessary to provide a time series model parameter optimization method, apparatus, computer device and storage medium capable of improving accuracy of sales prediction for different types of articles.
A method of time series model parameter optimization, the method comprising:
acquiring a training time sequence data set and a verification time sequence data set representing sales volumes of various articles;
determining an optimizable periodic parameter set when the timing duration of the training timing data set is within an optimizable timing duration range of periodic parameters in a time series model;
determining the periodic parameter value of each periodic parameter in the optimizable periodic parameter set according to the evaluation index value and the verification time sequence data set;
determining a festival type corresponding to the training time sequence data set;
optimizing the holiday item parameters in the time series model according to the preset effect value of each holiday type to obtain holiday item parameter values;
determining the variable point quantity supported by the time series model according to the time sequence duration, and optimizing a trend item parameter in the time series model according to the variable point quantity to obtain a trend item parameter value;
and obtaining a target parameter value set corresponding to each category identification according to each periodic parameter value, each holiday parameter value and each trend parameter value.
In one embodiment, the acquiring a training time-series data set and a verification time-series data set representing sales of the items of the respective categories includes:
acquiring a time series data sample representing the sales volume of the article;
classifying the time sequence data samples according to category identification to obtain time sequence data samples of sales volume of various categories of articles;
determining training set duration and verification set duration according to the time sequence duration and a preset time sequence duration range of the time sequence data samples of the sales volume of each category article;
and determining a training time sequence data set and a verification time sequence data set corresponding to each class identification from the time sequence data samples according to the training set duration and the verification set duration.
In one embodiment, before determining the training set duration and the verification set duration according to the time sequence duration and the preset time sequence duration range of the time sequence data sample of the sales volume of each article, the method further comprises:
performing missing value processing on the time series data samples of the sales volume of each article, and detecting whether abnormal data exist in the preprocessed time series data samples;
and when abnormal data exist in the processed time series data samples, smoothing the abnormal data.
In one embodiment, before the determining the optimizable set of periodic parameters when the timing duration of the training timing data set is within an optimizable timing duration range of periodic parameters in the time series model, the method further comprises:
receiving a cycle parameter optimization instruction;
and optimizing the cycle type in the time series model according to the cycle parameter optimization instruction, and determining the cycle parameters in the time series model.
In one embodiment, the preset effect value comprises a pre-nodal effect value and a post-nodal effect value; optimizing the holiday term parameters in the time series model according to the preset effect values of the holiday types to obtain holiday term parameter values, wherein the holiday term parameter values comprise:
and assigning the pre-festival effect value and the post-festival effect value to a virtual variable of a festival and holiday item in the time series model to obtain a parameter value of the festival and holiday item.
In one embodiment, the determining, according to the time-series duration, the variable point number supported by the trend item, and determining, according to the variable point number, an optimized trend item parameter value includes:
determining the variable point quantity supported by the trend item according to the time sequence duration;
segmenting the training time sequence data set according to the variable point quantity to obtain a time sequence data segment set;
and calculating the growth rate value of each time sequence data segment in the time sequence data segment set, determining a full value of a predicted growth rate according to each growth rate value, and taking the predicted growth rate value as a trend item parameter value.
In one embodiment, the method further comprises:
acquiring time series data for predicting the sales volume of the goods from a server;
and inputting the time sequence data into an optimized time sequence model, classifying the time sequence data according to the category identification to obtain a sales prediction value of each item identified by the category in a given time period.
An apparatus for time series model parameter optimization, the apparatus comprising:
the acquisition module is used for acquiring a training time sequence data set and a verification time sequence data set corresponding to each category identification;
the judging module is used for determining an optimizable periodic parameter set when the time sequence duration of the training time sequence data set is within the time sequence duration range which can be optimized by the periodic parameters in the time sequence model;
the period parameter optimization module is used for periodically determining the period parameter value of each periodic parameter in the optimizable periodic parameter set according to the evaluation index value and the verification time sequence data set;
the holiday parameter optimization module is used for determining a holiday type corresponding to the training time sequence data set; optimizing the holiday item parameters in the time series model according to the preset effect value of each holiday type to obtain holiday item parameter values;
the trend parameter optimization module is used for determining the variable point number supported by the time sequence model according to the time sequence duration, and optimizing the trend item parameter in the time sequence model according to the variable point number to obtain a trend item parameter value;
and the parameter combination module is used for obtaining a target parameter value group type corresponding to each category identification according to each periodic parameter value, each holiday parameter value and each trend item parameter value.
A computer device comprising a memory and a processor, the memory storing a computer program, the processor implementing the following steps when executing the computer program:
acquiring a training time sequence data set and a verification time sequence data set corresponding to each category identification;
determining an optimizable periodic parameter set when the timing duration of the training timing data set is within an optimizable timing duration range of periodic parameters in a time series model;
determining the periodic parameter value of each periodic parameter in the optimizable periodic parameter set according to the evaluation index value and the verification time sequence data set;
determining a festival type corresponding to the training time sequence data set;
optimizing the holiday item parameters in the time series model according to the preset effect value of each holiday type to obtain holiday item parameter values;
determining the variable point quantity supported by the time series model according to the time sequence duration, and optimizing a trend item parameter in the time series model according to the variable point quantity to obtain a trend item parameter value;
and obtaining a target parameter value set corresponding to each category identification according to each periodic parameter value, each holiday parameter value and each trend parameter value.
A computer-readable storage medium, on which a computer program is stored which, when executed by a processor, carries out the steps of:
acquiring a training time sequence data set and a verification time sequence data set corresponding to each category identification;
determining an optimizable periodic parameter set when the timing duration of the training timing data set is within an optimizable timing duration range of periodic parameters in a time series model;
determining the periodic parameter value of each periodic parameter in the optimizable periodic parameter set according to the evaluation index value and the verification time sequence data set;
determining a festival type corresponding to the training time sequence data set;
optimizing the holiday item parameters in the time series model according to the preset effect value of each holiday type to obtain holiday item parameter values;
determining the variable point quantity supported by the time series model according to the time sequence duration, and optimizing a trend item parameter in the time series model according to the variable point quantity to obtain a trend item parameter value;
and obtaining a target parameter value set corresponding to each category identification according to each periodic parameter value, each holiday parameter value and each trend parameter value.
The time series model parameter method, the time series model parameter device, the computer equipment and the storage medium are characterized in that a training time series data set and a verification time series data set representing sales volumes of various articles are obtained; determining an optimizable periodic parameter set in the time sequence model according to the time sequence duration of the training time sequence data set and the optimizable time sequence duration range of the periodic parameter; determining the periodic parameter value of each periodic parameter in the optimizable periodic parameter set according to the evaluation index value and the verification time sequence data set; optimizing a holiday item parameter in the time sequence model according to a preset effect value of a holiday type corresponding to the training time sequence data set to obtain a holiday item parameter value; determining the variable point number supported by the time sequence model according to the time sequence duration, and optimizing the trend item parameter in the time sequence model according to the variable point number to obtain a trend item parameter value; the optimal parameter combination of the time series model is obtained, the optimal time series model is determined, the parameters of the time series model are optimized according to the training time series data set and the verification time series data set of the sales volume of each class of articles, the periodic parameters, the holiday item parameters and the trend items of each class are optimized, the optimal parameter combination is obtained, and the accuracy of the time series model in predicting the sales volume of different classes of articles is improved.
Drawings
FIG. 1 is a diagram of an exemplary embodiment of a time series model parameter optimization method;
FIG. 2 is a schematic flow chart diagram of a method for time series model parameter optimization in one embodiment;
FIG. 3 is a flow diagram illustrating a method for partitioning a training time series data set and a validation time series data set, according to one embodiment;
FIG. 4 is a schematic flow chart of a time series model parameter optimization method in another embodiment;
FIG. 5 is a block diagram showing an exemplary embodiment of a time series model parameter optimizing apparatus;
FIG. 6 is a block diagram showing the construction of a time series model parameter optimizing apparatus according to another embodiment;
FIG. 7 is a diagram illustrating an internal structure of a computer device according to an embodiment.
Detailed Description
In order to make the objects, technical solutions and advantages of the present application more apparent, the present application is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the present application and are not intended to limit the present application.
The time series model parameter optimization method provided by the application can be applied to the application environment shown in fig. 1. Wherein the terminal 102 communicates with the server 104 via a network. The terminal 102 acquires a training time sequence data set and a verification time sequence data set representing sales volumes of various articles from the server 104; determining an optimizable periodic parameter set when the time sequence duration of the training time sequence data set is within an optimizable time sequence duration range of periodic parameters in the time sequence model; determining the periodic parameter value of each periodic parameter in the optimizable periodic parameter set according to the evaluation index value and the verification time sequence data set; determining a festival type corresponding to a training time sequence data set; optimizing a holiday item parameter in the time sequence model according to the preset effect value of each holiday type to obtain a holiday item parameter value; determining the variable point number supported by the time sequence model according to the time sequence duration, and optimizing the trend item parameter in the time sequence model according to the variable point number to obtain a trend item parameter value; and obtaining a target parameter value set corresponding to each category identification according to each periodic parameter value, each holiday parameter value and each trend parameter value to obtain an optimized time series model. The terminal 102 may be, but is not limited to, various personal computers and notebook computers, and the server 104 may be implemented by an independent server or a server cluster composed of a plurality of servers.
In an embodiment, as shown in fig. 2, a time series model parameter optimization method is provided, which is described by taking the method as an example applied to the terminal in fig. 1, and includes the following steps:
step 202, a training time sequence data set and a verification time sequence data set representing sales volume of each article are obtained.
Wherein the training time series data set is used for training a time series model; and the verification time sequence data set is used for verifying the trained time sequence model. Each time series data in the time series data set comprises a time series representing the article sale time and a numerical value representing the article sale quantity.
The time sequence model comprises a period term, a holiday term and a trend term; the time series model may be a Prophet model, which may be expressed as:
y(t)＝g(t)+s(t)+11(t)+∈
wherein s (t) is a period term, g (t) is a trend term, h (t) is a holiday term, and epsilon is an error term.
Specifically, time series data samples representing the sales volume of the article are obtained from a database of a server, each time series data in the time series data samples carries an article type identifier, the time series data samples are input into a time series model, classifying the time series data samples according to the category identification to obtain time series data samples representing the sales volume of each category article, for example, a time series data sample 1 in the goods production supply chain is obtained from a database, the time series data sample comprises sales data of class A goods, sales data of class B goods and sales data of class C goods, classifying the time series data sample 1 according to the type identification corresponding to each of the type A articles, the type B articles and the type C articles to obtain a time series data sample A, a time series data sample B and a time series data sample C; determining the data granularity of the time series data samples, and determining the time length of the time series data samples according to the data granularity; and determining a training time sequence data set and a verification time sequence data set of sales volume of each class of articles according to the time length. Wherein, the data granularity comprises daily granularity, weekly granularity, monthly granularity and annual granularity.
The time sequence duration refers to the time sequence duration of a training time sequence data set determined by taking the degree of day as a dimension; for example, the training time-series data set includes 500 items of type a sales data, and the data points corresponding to the 500 items of type a sales data are determined to be 60 data points with the degree of day as the dimension. The time sequence duration may also be determined by taking the dimensions of the week, month and year, which is not limited herein. The time sequence durations corresponding to the parameters of the different types of periods are different, as shown in table 1, when the time sequence duration of the training time sequence data set is within the range of (0,21) days, it is indicated that the training time sequence data set does not support time sequence model operation, the accuracy of the time sequence model prediction after optimization is low, simple time sequence model operation is supported, and the prediction cost value of the article in the period of 7 is obtained; when the time sequence duration of the training time sequence data set is in the range of [21,45) days, parameters of a periodic cycle and a seasonal mode can be optimized, and the parameter optimization of a monthly cycle and a yearly cycle is not supported; when the time sequence duration of the training time sequence data set is in the range of [45,400) days, parameters of a week period, a month period and a seasonal mode can be optimized, and parameter optimization of a year period is not supported; parameters for the weekly, monthly, yearly and seasonal patterns may be optimized when the timing duration of the training timing dataset is in the [400, +) day range. The periodic parameter set comprises at least one periodic parameter of a week period, a month period and an year period.
Specifically, the terminal receives a period parameter optimization instruction and optimizes the period type in the time series model; the time sequence model default period is a year period, a month period and a day period, the time sequence model default period is optimized according to a period parameter optimization instruction, the period type of the time sequence model is determined to be the year period, the month period and the year period, and periodic parameters in the time sequence model are determined; an optimizable periodic parameter set is determined when a timing duration of the training timing data set is within an optimizable timing duration range of periodic parameters in the time series model. The optimization cycle type carried by the cycle parameter optimization instruction is determined according to an application scenario of preset data, for example, in an application scenario of a manufacturing industry, a sales value needs to be predicted according to the monthly degree, and reasonable distribution management of the inventory quantity and the sales quantity on the supply quantity can be performed, for example, a monthly cycle needs to be added in a time series model when the sales volume is rushed at the end of a month.
Table 1:
and step 206, determining the periodic parameter value of each periodic parameter in the optimizable periodic parameter set according to the evaluation index value and the verification time sequence data set.
The evaluation index may include any one of a Mean Absolute Error (MAE), a Mean Square Error (MSE), a Root Mean Square Error (RMSE), and a Mean Absolute Percentage Error (MAPE), and is determined according to an actual demand of the commodity.
Time series data models exist in weekly, monthly, yearly and seasonal patterns. Whether the period items in the time series model have period types of a week period, a month period and an annual period can be represented by Boolean parameter values 'True' and 'False', wherein each period type has two conditions of 'True' and 'False', wherein True represents the existence of the period, and False represents the nonexistence of the period; there may be eight periodic models for the periodic terms, i.e., there are corresponding eight sets of periodic parameters: { (True ), (False, False), (True, False), (True, False, True, False), (False, True), (False, True), (False, True) }; for example, "True, True" may represent a weekly, monthly and yearly period of presence and "True, False" represents a weekly, non-monthly and yearly period of presence. The corresponding parameter value of the seasonal pattern is any one of Multiplicative and additive.
Specifically, training is carried out according to the periodic parameter group according to a training time sequence data set, so that a well-trained periodic model can be obtained; inputting a verification time sequence data set into the trained eight periodic model to obtain a verification result; and performing index calculation on the verification result according to the evaluation index, determining the periodic parameter value of the periodic parameter, and determining the periodic parameter value of each periodic parameter in the optimizable periodic parameter set according to the periodic parameter value. For example, if the periodic parameter value is True, the periodic parameter value corresponding to the preset periodic type is acquired. The selectable periodic patterns are:
wherein N represents a periodic parameter, i.e. the fourier order; p represents a period length; a isn、bnRepresenting an estimated parameter; the cycle length is preset; for example, the cycle length of the weekly cycle is 7 days, the cycle length of the monthly cycle is 30.5 days, and the cycle length of the annual cycle is 365.25 days. And when the periodic parameter value is True, namely the periodic parameter value is a value with a preset value.
And step 208, determining the festival type corresponding to the training time sequence data set.
The time series model is internally provided with a preset festival type, a festival name and a preset effect value, wherein the preset effect value comprises a pre-festival effect value and a post-festival effect value. As shown in table 2, the holiday types include traditional chinese holidays, western holidays, sales festivals, and the like; traditional Chinese festivals include the New year's day, the spring festival, the Qingming festival, the labor festival, the mid-autumn festival, the national celebration festival, and the like; western festivals including valentine's day, women's day, Christmas, etc.; the sales sections comprise 6.18 promotion sections, twenty-one promotion sections, twenty-two promotion sections and the like, and corresponding preset effect values exist in different festivals.
Table 2:
and step 210, optimizing the holiday item parameters in the time series model according to the preset effect value of each holiday type to obtain holiday item parameter values.
Specifically, according to the festival name in the festival type determined by the training time sequence data set, a pre-festival effect value and a post-festival effect value of the festival name are obtained, and the pre-festival effect value and the post-festival effect value are input into a festival and holiday item model to obtain a festival and holiday item parameter value.
And 212, determining the variable point number supported by the time sequence model according to the time sequence duration, and optimizing the trend item parameter in the time sequence model according to the variable point number to obtain a trend item parameter value.
And 214, obtaining a target parameter value set corresponding to each category identification according to each periodic parameter value, each holiday parameter value and each trend item parameter value.
In the time series model parameter optimization method, a training time series data set and a verification time series data set representing sales volume of various articles are obtained from a server; determining the corresponding optimizable periodic parameters of the training time sequence data set of each article according to the time sequence duration of the training time sequence data set, the optimizable periodic parameters and the optimizable time sequence duration range; determining the periodic parameter value of each periodic parameter in the optimizable periodic parameter set through the evaluation index value and the verification time sequence data set based on grid search; the method comprises the steps that holiday item parameter values are obtained through configuring holiday information built in a time series model and holiday item parameters in the time series model; determining the variable point number supported by the time sequence model according to the time sequence duration, and optimizing the trend item parameter in the time sequence model according to the variable point number to obtain a trend item parameter value; and determining an optimized time series model according to the periodic parameter values, the holiday term parameter values and the trend term parameter values obtained by optimization. The periodic parameters and the trend parameters are optimized according to the time sequence duration of the training time sequence data set, the holiday item parameters are optimized by configuring holiday information, holiday item parameter values are determined, the optimized time sequence model is determined according to the obtained parameter combinations, sales of different types of articles can be simultaneously predicted, manual adjustment of model parameters is not needed, and accuracy of model prediction is improved.
In one embodiment, as shown in fig. 3, a training time-series data set and verification time-series data set method for determining sales volume of various items is provided, which is described by taking the method as an example applied to the terminal in fig. 1, and includes the following steps:
at step 302, a time series data sample characterizing the sales volume of the item is obtained.
And step 304, classifying the time series data samples according to the category identification to obtain the time series data samples of the sales volume of each category of articles.
And step 306, performing missing value processing on the time series data samples of the sales volume of each article, and detecting whether abnormal data exists in the preprocessed time series data samples.
And 308, when abnormal data exists in the processed time series data samples, smoothing the abnormal data.
Specifically, when a missing value exists in a time sequence data sample of the sales volume of each type of article, zero filling processing is carried out on the missing item; and when data with the numerical value being more than 3 times of the standard deviation of the mean value of the time series data samples exists in the processed time series data samples, the data is abnormal data, and the numerical value of the abnormal data is updated to be the numerical value with the standard deviation being 3 times of the mean value of the time series data samples. By preprocessing the time series data, the low accuracy of time series model prediction caused by data loss is avoided.
And 310, determining the training set duration and the verification set duration according to the time sequence duration and the preset time sequence duration range of the time sequence data samples of the sales volume of each article.
Specifically, the time series data are segmented according to the time length of the time series data samples, and a training time series data set and a verification time series data set used for optimizing the period parameters are obtained. Optionally, the segmenting the time series data according to the time length to obtain a training set and a verification set for optimizing the period parameter includes: determining the segmentation ratio of the training set and the verification set according to a preset total time interval corresponding to the time length; and segmenting the time series data according to the segmentation ratio to obtain a training set and a verification set for optimizing the period item parameters. As shown in table 3, the time length T of the time series data samples in the present scheme needs to be satisfied, and when the time length is in the preset total time length interval of [21,45 ] days, the time series data sample corresponding to the time length of T-7 is used as the training time series data set, and the time series data sample corresponding to seven days after the time length is used as the verification time series data set; when the duration is in a total duration interval preset in [45,400) days, selecting a time series data sample corresponding to the duration with the maximum value from 70% T duration and (T-30) days as a training time series data set, and taking a time series data sample corresponding to the duration of seven days later as a verification time series data set; and when the time length is in a preset total time length interval of [400, +) days, taking the time sequence data sample corresponding to the time length of T-30 as a training time sequence data set, and taking the time sequence data sample corresponding to the time length thirty days after the time length as a verification time sequence data set.
Table 3:
in the embodiment, time series data samples of sales volume of various articles are obtained by classifying time series data samples obtained from a server according to article identifications carried by the time series data samples, and missing value zero padding processing and data smoothing processing are performed on the data before the time series data samples for representing the sales volume of various articles are divided into a training time series data set and a verification time series data set; dividing data according to a preset time sequence duration range in which the time sequence duration of the time sequence data sample is located to obtain a corresponding training time sequence data set and a corresponding verification time sequence data set; the integrity of the time series data sample can be determined, and the accuracy of time series model prediction can be improved.
In another embodiment, as shown in fig. 4, a time series model parameter optimization method is provided, which is described by taking the method as an example applied to the terminal in fig. 1, and includes the following steps:
At step 404, a cycle parameter optimization instruction is received.
In particular, the cycle parameter optimization instructions are used to optimize the built-in cycle types in the time series model. For example, the built-in cycle types in the time series cycle model include year cycle, week cycle, and day cycle; the daily cycle is suitable for predicting the fluctuation situation in one day, for example, the 10 o 'clock at the station is more, the 18 o' clock is less, and the like; the data volume obtained by the daily period prediction is small and the data utilization rate is low in actual manufacturing industry requirements; and optimizing the daily period in the time period sequence model to deal with the situation of 'monthly end sales rush amount', increasing the monthly period types and deleting the daily period.
And 406, optimizing the cycle type in the time series model according to the cycle parameter optimization instruction, and determining the cycle parameter in the time series model.
In step 408, an optimizable periodic parameter set is determined when the timing length of the training timing data set is within an optimizable timing length range of the periodic parameters in the time series model.
And step 410, determining the periodic parameter value of each periodic parameter in the optimizable periodic parameter set according to the evaluation index value and the verification time sequence data set.
And step 414, assigning the pre-festival effect value and the post-festival effect value of the festival type to the virtual variable of the festival and holiday item in the time series model to obtain the parameter value of the festival and holiday item.
And 418, calculating a growth rate value of each time sequence data segment in the time sequence data segment set, determining a predicted growth rate value according to each growth rate value, and taking the predicted growth rate value as a trend item parameter value.
Specifically, determining the time sequence duration of a training time sequence data set by taking days as granularity; determining a corresponding preset time sequence duration range according to the time sequence duration, and determining the number of variable points supportable by a trend item by taking the number of days of a preset training time sequence data set as a period; and segmenting the variable point pairs to obtain a time sequence data segment set. As shown in table 4, when the time sequence duration is within the range of (0,21) days, taking preset training time sequence data set days 3 as a cycle, and determining the number of variable points supportable by the trend term as an integer value obtained by dividing the time sequence duration by 3; when the time sequence duration is in the range of [21, 100) days, taking preset training time sequence data set days 5 as a period, and determining the number of variable points supportable by a trend item as an integral value obtained by dividing the time sequence duration by 5; when the time sequence duration is in the range of [100, 200) days, the number of the variable points which can be supported by the trend item is 25; when the time sequence duration is in the range of [200, + inf) days, taking the number of days 7 of a preset training time sequence data set as a period, and determining the number of variable points which can be supported by the trend term as an integer value obtained by dividing the time sequence duration by 7.
Table 4:
for example, the time sequence duration is 270 days, when the time sequence duration is in the range of [200, + inf), the number of variable points which can be supported by the trend item is determined to be 30 by taking the preset training time sequence data set number of days 7 as a cycle, the training time sequence data set is divided into 30 time sequence data sections, and the slope value, namely the growth rate value, of each section is calculated; and obtaining a predicted growth rate value in a given time period, namely a trend term parameter value according to the growth rate value and the Laplace distribution.
And step 420, obtaining a target parameter value set corresponding to each category identification according to each periodic parameter value, each holiday parameter value and each trend item parameter value.
Optionally, in one embodiment, time series data for predicting sales of the item is acquired from a server; and inputting the time sequence data into the optimized time sequence model, and classifying the time sequence data according to the category identification to obtain a sales prediction value of the article identified by each category in a given time period. Optionally, acquiring time-series data for predicting the sales amount of the article from a database of the server; the time sequence data comprises sales data of articles marked by three types of identifiers, namely '01', '02' and '03', the time sequence data is input into an optimized time sequence model, the time sequence data is distinguished according to the type identifiers, time sequence data corresponding to the three types of identifiers is obtained, corresponding parameter value groups are determined, the predicted sales of the articles corresponding to the types of identifiers in a given time period are predicted through the optimized time sequence model, manual parameter adjustment is not needed, reasonable distribution of nodes on a supply chain can be achieved according to the predicted sales, reasonable distribution of resources is improved, and the universality of the time sequence model is improved.
In the time model parameter optimization method, before optimizing time series model parameters, a cycle parameter optimization instruction is generated according to the triggering of a prediction scene to optimize the time series model, and a cycle type corresponding to the prediction scene is determined; determining an optimizable periodic parameter set according to the time sequence duration of the training time sequence data set, and determining periodic parameter values of periodic parameters in the optimizable periodic parameter set according to the training time sequence data set, the verification time sequence data set and the evaluation index value which represent the sales volume of various articles; the method comprises the steps that holiday item parameter values are obtained through configuring holiday information built in a time series model and holiday item parameters in the time series model; determining the number of variable points supported by a time sequence model according to the time sequence duration, dividing a training time sequence data set according to the number of variable points to obtain a time sequence data segment set, determining a predicted growth rate value based on the growth rate value of each time sequence data segment and obeying Laplace distribution, and obtaining a trend item parameter value; and determining an optimized time series model according to the periodic parameter values, the holiday term parameter values and the trend term parameter values obtained by optimization. By optimizing the periodic parameter, the holiday term parameter and the trend term parameter, the accuracy of time series model prediction and the model universality can be improved.
It should be understood that although the various steps in the flow charts of fig. 2-4 are shown in order as indicated by the arrows, the steps are not necessarily performed in order as indicated by the arrows. The steps are not performed in the exact order shown and described, and may be performed in other orders, unless explicitly stated otherwise. Moreover, at least some of the steps in fig. 2-4 may include multiple steps or multiple stages, which are not necessarily performed at the same time, but may be performed at different times, which are not necessarily performed in sequence, but may be performed in turn or alternately with other steps or at least some of the other steps.
In one embodiment, as shown in fig. 5, there is provided a time series model parameter optimization apparatus, including: an obtaining module 502, a judging module 504, a period parameter optimizing module 506, a holiday parameter optimizing module 508, a trend parameter optimizing module 510 and a parameter combining module 512, wherein:
an obtaining module 502 is configured to obtain a training time sequence data set and a verification time sequence data set representing sales volumes of various items.
The determining module 504 is configured to determine an optimizable periodic parameter set when a time sequence duration of the training time sequence data set is within an optimizable time sequence duration range of the periodic parameter in the time sequence model.
And a period parameter optimization module 506, configured to periodically determine, according to the evaluation index value and the verification time series data set, a period parameter value of each periodic parameter in the optimizable periodic parameter set.
A holiday parameter optimization module 508 for determining a holiday type corresponding to the training time series data set; and optimizing the holiday item parameters in the time sequence model according to the preset effect values of the holiday types to obtain holiday item parameter values, wherein the preset effect values comprise a pre-holiday effect value and a post-holiday effect value.
And the trend parameter optimization module 510 is configured to determine the number of variable points supported by the time series model according to the time sequence duration, and optimize the trend item parameter in the time series model according to the number of variable points to obtain a trend item parameter value.
And the parameter combination module 512 is configured to obtain a target parameter value set corresponding to each category identifier according to each periodic parameter value, each holiday parameter value, and each trend item parameter value.
In the time series model parameter optimization device, a training time sequence data set and a verification time sequence data set representing sales volume of various articles are obtained from a server; determining the corresponding optimizable periodic parameters of the training time sequence data set of each article according to the time sequence duration of the training time sequence data set, the optimizable periodic parameters and the optimizable time sequence duration range; determining the periodic parameter value of each periodic parameter in the optimizable periodic parameter set through the evaluation index value and the verification time sequence data set based on grid search; the method comprises the steps that holiday item parameter values are obtained through configuring holiday information built in a time series model and holiday item parameters in the time series model; determining the variable point number supported by the time sequence model according to the time sequence duration, and optimizing the trend item parameter in the time sequence model according to the variable point number to obtain a trend item parameter value; and determining an optimized time series model according to the periodic parameter values, the holiday term parameter values and the trend term parameter values obtained by optimization. The periodic parameters and the trend parameters are optimized according to the time sequence duration of the training time sequence data set, the holiday item parameters are optimized by configuring holiday information, holiday item parameter values are determined, the optimized time sequence model is determined according to the obtained parameter combinations, sales of different types of articles can be simultaneously predicted, manual adjustment of model parameters is not needed, and accuracy of model prediction is improved.
In another embodiment, as shown in fig. 6, there is provided a time series model parameter optimization apparatus, comprising: the acquiring module 502, the determining module 504, the period parameter optimizing module 506, the holiday parameter optimizing module 508, the trend parameter optimizing module 510, and the parameter combining module 512, further include: wherein:
in one embodiment, the acquisition module 502 is further configured to acquire a time series data sample indicative of an amount of sales of the item.
And the classification module 514 is configured to classify the time-series data samples according to the category identifiers to obtain time-series data samples of sales volumes of the various categories of articles.
The determining module 516 is configured to determine a training set duration and a verification set duration according to the time sequence duration of the time sequence data sample of the sales volume of each category of articles and a preset time sequence duration range; and determining a training time sequence data set and a verification time sequence data set corresponding to each category identification from the time sequence data samples according to the training set time length and the verification set time length.
The preprocessing module 518 is configured to perform missing value processing on the time series data samples of the sales volume of each category of articles, and detect whether abnormal data exists in the preprocessed time series data samples; and when abnormal data exists in the processed time series data samples, smoothing the abnormal data.
In one embodiment, the cycle parameter optimization module 506 is further configured to receive a cycle parameter optimization instruction; and optimizing the cycle type in the time series model according to the cycle parameter optimization instruction, and determining the cycle parameter in the time series model.
In one embodiment, the holiday parameter optimization module 508 is further configured to assign the pre-holiday effect value and the post-holiday effect value to a virtual variable of a holiday term in the time series model to obtain a holiday term parameter value.
In one embodiment, the trend parameter optimization module 510 is further configured to determine the number of change points supported by the trend item according to the time sequence duration; segmenting the training time sequence data set according to the variable point number to obtain a time sequence data segment set; and calculating the growth rate value of each time sequence data segment in the time sequence data segment set, determining a predicted growth rate value according to each growth rate value, and taking the predicted growth rate value as a trend item parameter value.
A prediction module 520, configured to obtain time-series data for predicting sales of the item from the server; and inputting the time sequence data into the optimized time sequence model, and classifying the time sequence data according to the category identification to obtain a sales prediction value of the article identified by each category in a given time period.
In one embodiment, time series data samples of sales volume of various articles are obtained by classifying time series data samples obtained from a server according to article identifications carried by the time series data samples, and missing value zero padding processing and data smoothing processing are carried out on data before the time series data samples for representing the sales volume of various articles are divided into a training time series data set and a verification time series data set; and dividing the data according to a preset time sequence duration range in which the time sequence duration of the time sequence data sample is located to obtain a corresponding training time sequence data set and a corresponding verification time sequence data set.
Determining the corresponding optimizable periodic parameters of the training time sequence data set of each article according to the time sequence duration of the training time sequence data set, the optimizable periodic parameters and the optimizable time sequence duration range; determining the periodic parameter value of each periodic parameter in the optimizable periodic parameter set through the evaluation index value and the verification time sequence data set based on grid search; the method comprises the steps that holiday item parameter values are obtained through configuring holiday information built in a time series model and holiday item parameters in the time series model; determining the variable point number supported by the time sequence model according to the time sequence duration, and optimizing the trend item parameter in the time sequence model according to the variable point number to obtain a trend item parameter value; and determining an optimized time series model according to the periodic parameter values, the holiday term parameter values and the trend term parameter values obtained by optimization.
For specific limitations of the time series model parameter optimization device, reference may be made to the above limitations of the time series model parameter optimization method, which are not described herein again. The modules in the time series model parameter optimization device can be wholly or partially realized by software, hardware and a combination thereof. The modules can be embedded in a hardware form or independent from a processor in the computer device, and can also be stored in a memory in the computer device in a software form, so that the processor can call and execute operations corresponding to the modules.
In one embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as shown in fig. 7. The computer device includes a processor, a memory, a communication interface, a display screen, and an input device connected by a system bus. Wherein the processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device comprises a nonvolatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of an operating system and computer programs in the non-volatile storage medium. The communication interface of the computer device is used for carrying out wired or wireless communication with an external terminal, and the wireless communication can be realized through WIFI, an operator network, NFC (near field communication) or other technologies. The computer program is executed by a processor to implement a time series model parameter optimization method. The display screen of the computer equipment can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer equipment can be a touch layer covered on the display screen, a key, a track ball or a touch pad arranged on the shell of the computer equipment, an external keyboard, a touch pad or a mouse and the like.
Those skilled in the art will appreciate that the architecture shown in fig. 7 is merely a block diagram of some of the structures associated with the disclosed aspects and is not intended to limit the computing devices to which the disclosed aspects apply, as particular computing devices may include more or less components than those shown, or may combine certain components, or have a different arrangement of components.
In one embodiment, a computer device is provided, comprising a memory and a processor, the memory having a computer program stored therein, the processor implementing the following steps when executing the computer program:
acquiring a training time sequence data set and a verification time sequence data set representing sales volumes of various articles;
determining an optimizable periodic parameter set when the time sequence duration of the training time sequence data set is within an optimizable time sequence duration range of periodic parameters in the time sequence model;
determining the periodic parameter value of each periodic parameter in the optimizable periodic parameter set according to the evaluation index value and the verification time sequence data set;
determining a festival type corresponding to a training time sequence data set;
optimizing a holiday item parameter in the time sequence model according to the preset effect value of each holiday type to obtain a holiday item parameter value;
determining the variable point number supported by the time sequence model according to the time sequence duration, and optimizing the trend item parameter in the time sequence model according to the variable point number to obtain a trend item parameter value;
and obtaining a target parameter value set corresponding to each category identification according to each periodic parameter value, each holiday parameter value and each trend item parameter value.
In one embodiment, the processor, when executing the computer program, further performs the steps of:
acquiring a time series data sample representing the sales volume of the article;
classifying the time series data samples according to the category identification to obtain the time series data samples of the sales volume of each category of articles;
determining the training set duration and the verification set duration according to the time sequence duration of the time sequence data samples of the sales volume of each type of article and a preset time sequence duration range;
and determining a training time sequence data set and a verification time sequence data set corresponding to each category identification from the time sequence data samples according to the training set time length and the verification set time length.
In one embodiment, the processor, when executing the computer program, further performs the following steps;
performing missing value processing on the time series data samples of the sales volume of each type of article, and detecting whether abnormal data exist in the preprocessed time series data samples;
and when abnormal data exists in the processed time series data samples, smoothing the abnormal data.
In one embodiment, the processor, when executing the computer program, further performs the steps of:
receiving a cycle parameter optimization instruction;
and optimizing the cycle type in the time series model according to the cycle parameter optimization instruction, and determining the cycle parameter in the time series model.
In one embodiment, the processor, when executing the computer program, further performs the steps of:
and assigning the pre-festival effect value and the post-festival effect value to the virtual variable of the festival and holiday item in the time sequence model to obtain the parameter value of the festival and holiday item.
In one embodiment, the processor, when executing the computer program, further performs the following steps;
determining the variable point number supported by the trend item according to the time sequence duration;
segmenting the training time sequence data set according to the variable point number to obtain a time sequence data segment set;
and calculating the growth rate value of each time sequence data segment in the time sequence data segment set, determining the full value of the predicted growth rate according to each growth rate value, and taking the predicted growth rate value as the parameter value of the trend item.
In one embodiment, the processor, when executing the computer program, further performs the steps of:
acquiring time series data for predicting the sales volume of the goods from a server;
and inputting the time sequence data into the optimized time sequence model, and classifying the time sequence data according to the category identification to obtain a sales prediction value of the article identified by each category in a given time period.
In one embodiment, a computer-readable storage medium is provided, having a computer program stored thereon, which when executed by a processor, performs the steps of:
acquiring a training time sequence data set and a verification time sequence data set representing sales volumes of various articles;
determining an optimizable periodic parameter set when the time sequence duration of the training time sequence data set is within an optimizable time sequence duration range of periodic parameters in the time sequence model;
determining the periodic parameter value of each periodic parameter in the optimizable periodic parameter set according to the evaluation index value and the verification time sequence data set;
determining a festival type corresponding to a training time sequence data set;
optimizing a holiday item parameter in the time sequence model according to the preset effect value of each holiday type to obtain a holiday item parameter value;
determining the variable point number supported by the time sequence model according to the time sequence duration, and optimizing the trend item parameter in the time sequence model according to the variable point number to obtain a trend item parameter value;
and obtaining a target parameter value set corresponding to each category identification according to each periodic parameter value, each holiday parameter value and each trend item parameter value.
In one embodiment, the computer program when executed by the processor further performs the steps of:
acquiring a time series data sample representing the sales volume of the article;
classifying the time series data samples according to the category identification to obtain the time series data samples of the sales volume of each category of articles;
determining the training set duration and the verification set duration according to the time sequence duration of the time sequence data samples of the sales volume of each type of article and a preset time sequence duration range;
and determining a training time sequence data set and a verification time sequence data set corresponding to each category identification from the time sequence data samples according to the training set time length and the verification set time length.
In one embodiment, the computer program when executed by the processor further performs the steps of:
performing missing value processing on the time series data samples of the sales volume of each type of article, and detecting whether abnormal data exist in the preprocessed time series data samples;
and when abnormal data exists in the processed time series data samples, smoothing the abnormal data.
In one embodiment, the computer program when executed by the processor further performs the steps of:
receiving a cycle parameter optimization instruction;
and optimizing the cycle type in the time series model according to the cycle parameter optimization instruction, and determining the cycle parameter in the time series model.
In one embodiment, the computer program when executed by the processor further performs the steps of:
and assigning the pre-festival effect value and the post-festival effect value to the virtual variable of the festival and holiday item in the time sequence model to obtain the parameter value of the festival and holiday item.
In one embodiment, the computer program when executed by the processor further performs the steps of:
determining the variable point number supported by the trend item according to the time sequence duration;
segmenting the training time sequence data set according to the variable point number to obtain a time sequence data segment set;
and calculating the growth rate value of each time sequence data segment in the time sequence data segment set, determining the full value of the predicted growth rate according to each growth rate value, and taking the predicted growth rate value as the parameter value of the trend item.
In one embodiment, the computer program when executed by the processor further performs the steps of:
acquiring time series data for predicting the sales volume of the goods from a server;
and inputting the time sequence data into the optimized time sequence model, and classifying the time sequence data according to the category identification to obtain a sales prediction value of the article identified by each category in a given time period.
It will be understood by those skilled in the art that all or part of the processes of the methods of the embodiments described above can be implemented by hardware related to instructions of a computer program, which can be stored in a non-volatile computer-readable storage medium, and when executed, can include the processes of the embodiments of the methods described above. Any reference to memory, storage, database or other medium used in the embodiments provided herein can include at least one of non-volatile and volatile memory. Non-volatile Memory may include Read-Only Memory (ROM), magnetic tape, floppy disk, flash Memory, optical storage, or the like. Volatile Memory can include Random Access Memory (RAM) or external cache Memory. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM), among others.
The technical features of the above embodiments can be arbitrarily combined, and for the sake of brevity, all possible combinations of the technical features in the above embodiments are not described, but should be considered as the scope of the present specification as long as there is no contradiction between the combinations of the technical features.
The above examples only express several embodiments of the present application, and the description thereof is more specific and detailed, but not construed as limiting the scope of the invention. It should be noted that, for a person skilled in the art, several variations and modifications can be made without departing from the concept of the present application, which falls within the scope of protection of the present application. Therefore, the protection scope of the present patent shall be subject to the appended claims.
Claims (10)
1. A method for time series model parameter optimization, the method comprising:
acquiring a training time sequence data set and a verification time sequence data set representing sales volumes of various articles;
determining an optimizable periodic parameter set when the timing duration of the training timing data set is within an optimizable timing duration range of periodic parameters in a time series model;
determining the periodic parameter value of each periodic parameter in the optimizable periodic parameter set according to the evaluation index value and the verification time sequence data set;
determining a festival type corresponding to the training time sequence data set;
optimizing the holiday item parameters in the time series model according to the preset effect value of each holiday type to obtain holiday item parameter values;
determining the variable point quantity supported by the time series model according to the time sequence duration, and optimizing a trend item parameter in the time series model according to the variable point quantity to obtain a trend item parameter value;
and obtaining a target parameter value set corresponding to each category identification according to each periodic parameter value, each holiday parameter value and each trend parameter value.
2. The method of claim 1, wherein the obtaining a training time series data set and a validation time series data set characterizing sales of the respective item class comprises:
acquiring a time series data sample representing the sales volume of the article;
classifying the time sequence data samples according to category identification to obtain time sequence data samples of sales volume of various categories of articles;
determining training set duration and verification set duration according to the time sequence duration and a preset time sequence duration range of the time sequence data samples of the sales volume of each category article;
and determining a training time sequence data set and a verification time sequence data set corresponding to each class identification from the time sequence data samples according to the training set duration and the verification set duration.
3. The method of claim 2, wherein prior to determining a training set duration and a validation set duration based on the time series data samples of the sales volume for each of the items and the predetermined time series duration range, the method further comprises:
performing missing value processing on the time series data samples of the sales volume of each article, and detecting whether abnormal data exist in the preprocessed time series data samples;
and when abnormal data exist in the processed time series data samples, smoothing the abnormal data.
4. The method of claim 1, wherein prior to determining the optimizable set of periodic parameters when the timing duration of the training timing data set is within an optimizable range of timing durations of periodic parameters in a time series model, the method further comprises:
receiving a cycle parameter optimization instruction;
and optimizing the cycle type in the time series model according to the cycle parameter optimization instruction, and determining the cycle parameters in the time series model.
5. The method of claim 1, wherein the preset effect values comprise pre-nodal effect values and post-nodal effect values; optimizing the holiday term parameters in the time series model according to the preset effect values of the holiday types to obtain holiday term parameter values, wherein the holiday term parameter values comprise:
and assigning the pre-festival effect value and the post-festival effect value to a virtual variable of a festival and holiday item in the time series model to obtain a parameter value of the festival and holiday item.
6. The method of claim 1, wherein the determining the number of change points supported by the trend item according to the time sequence duration and the determining the optimized trend item parameter value according to the number of change points comprise:
determining the variable point quantity supported by the trend item according to the time sequence duration;
segmenting the training time sequence data set according to the variable point quantity to obtain a time sequence data segment set;
and calculating the growth rate value of each time sequence data segment in the time sequence data segment set, determining a predicted growth rate value according to each growth rate value, and taking the predicted growth rate value as a trend item parameter value.
7. The method of claim 1, further comprising:
acquiring time series data for predicting the sales volume of the goods from a server;
and inputting the time sequence data into an optimized time sequence model, classifying the time sequence data according to the category identification to obtain a sales prediction value of each item identified by the category in a given time period.
8. An apparatus for time series model parameter optimization, the apparatus comprising:
the acquisition module is used for acquiring a training time sequence data set and a verification time sequence data set which represent sales volumes of various articles;
the judging module is used for determining an optimizable periodic parameter set when the time sequence duration of the training time sequence data set is within the time sequence duration range which can be optimized by the periodic parameters in the time sequence model;
the period parameter optimization module is used for periodically determining the period parameter value of each periodic parameter in the optimizable periodic parameter set according to the evaluation index value and the verification time sequence data set;
the holiday parameter optimization module is used for determining a holiday type corresponding to the training time sequence data set; optimizing the holiday item parameters in the time series model according to the preset effect value of each holiday type to obtain holiday item parameter values;
the trend parameter optimization module is used for determining the variable point number supported by the time sequence model according to the time sequence duration, and optimizing the trend item parameter in the time sequence model according to the variable point number to obtain a trend item parameter value;
and the parameter combination module is used for obtaining a target parameter value set corresponding to each type identifier according to each periodic parameter value, each holiday parameter value and each trend parameter value.
9. A computer device comprising a memory and a processor, the memory storing a computer program, characterized in that the processor, when executing the computer program, implements the steps of the method of any of claims 1 to 7.
10. A computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, carries out the steps of the method of any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010588655.0A CN113837783A (en) | 2020-06-24 | 2020-06-24 | Time series model parameter optimization method and device and computer equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010588655.0A CN113837783A (en) | 2020-06-24 | 2020-06-24 | Time series model parameter optimization method and device and computer equipment |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113837783A true CN113837783A (en) | 2021-12-24 |
Family
ID=78964578
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010588655.0A Pending CN113837783A (en) | 2020-06-24 | 2020-06-24 | Time series model parameter optimization method and device and computer equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113837783A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115660565A (en) * | 2022-12-27 | 2023-01-31 | 北京京东乾石科技有限公司 | Exit quantity determining method and device, electronic equipment and computer readable medium |
| CN117786417A (en) * | 2024-02-28 | 2024-03-29 | 之江实验室 | Model training method, transient source identification method and device and electronic equipment |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20090292662A1 (en) * | 2008-05-26 | 2009-11-26 | Kabushiki Kaisha Toshiba | Time-series data analyzing apparatus, time-series data analyzing method, and computer program product |
| CN109447716A (en) * | 2018-11-09 | 2019-03-08 | 四川长虹电器股份有限公司 | Method for Sales Forecast method and server based on Recognition with Recurrent Neural Network |
| CN109493151A (en) * | 2019-01-10 | 2019-03-19 | 哈步数据科技(上海)有限公司 | Method for Sales Forecast method and system |
| CN110135876A (en) * | 2018-02-09 | 2019-08-16 | 北京京东尚科信息技术有限公司 | The method and device of Method for Sales Forecast |
| CN110555578A (en) * | 2018-06-01 | 2019-12-10 | 北京京东尚科信息技术有限公司 | sales prediction method and device |
| CN111274531A (en) * | 2020-01-19 | 2020-06-12 | 苏宁云计算有限公司 | Commodity sales amount prediction method, commodity sales amount prediction device, computer equipment and storage medium |
-
2020
- 2020-06-24 CN CN202010588655.0A patent/CN113837783A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20090292662A1 (en) * | 2008-05-26 | 2009-11-26 | Kabushiki Kaisha Toshiba | Time-series data analyzing apparatus, time-series data analyzing method, and computer program product |
| CN110135876A (en) * | 2018-02-09 | 2019-08-16 | 北京京东尚科信息技术有限公司 | The method and device of Method for Sales Forecast |
| CN110555578A (en) * | 2018-06-01 | 2019-12-10 | 北京京东尚科信息技术有限公司 | sales prediction method and device |
| CN109447716A (en) * | 2018-11-09 | 2019-03-08 | 四川长虹电器股份有限公司 | Method for Sales Forecast method and server based on Recognition with Recurrent Neural Network |
| CN109493151A (en) * | 2019-01-10 | 2019-03-19 | 哈步数据科技(上海)有限公司 | Method for Sales Forecast method and system |
| CN111274531A (en) * | 2020-01-19 | 2020-06-12 | 苏宁云计算有限公司 | Commodity sales amount prediction method, commodity sales amount prediction device, computer equipment and storage medium |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115660565A (en) * | 2022-12-27 | 2023-01-31 | 北京京东乾石科技有限公司 | Exit quantity determining method and device, electronic equipment and computer readable medium |
| CN117786417A (en) * | 2024-02-28 | 2024-03-29 | 之江实验室 | Model training method, transient source identification method and device and electronic equipment |
| CN117786417B (en) * | 2024-02-28 | 2024-05-10 | 之江实验室 | Model training method, transient source identification method and device and electronic equipment |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US10659542B2 (en) | System and methods for optimal allocation of multi-tenant platform infrastructure resources | |
| CN107818344A (en) | The method and system that user behavior is classified and predicted | |
| WO2019109790A1 (en) | Sales volume prediction method and device, and computer-readable storage medium | |
| CN111401801B (en) | Processing system and method for determining decentralized storage of target object warehouse | |
| CN114219412B (en) | Automatic replenishment method and replenishment system based on intelligent commodity system sales prediction | |
| CN107622326B (en) | User classification and available resource prediction method, device and equipment | |
| CN110706015B (en) | Feature selection method for advertisement click rate prediction | |
| CN110555578A (en) | sales prediction method and device | |
| CA3007940A1 (en) | Systems and methods of utilizing multiple forecast models in forecasting customer demands for products at retail facilities | |
| CN113837783A (en) | Time series model parameter optimization method and device and computer equipment | |
| CN109522923A (en) | Customer address polymerization, device and computer readable storage medium | |
| CN112288163A (en) | Target factor prediction method of target object and related equipment | |
| CN114926305A (en) | Electric power material management and control method and device, computer equipment and storage medium | |
| CN111274531A (en) | Commodity sales amount prediction method, commodity sales amount prediction device, computer equipment and storage medium | |
| CN111178830A (en) | Cost accounting method and device, computer equipment and storage medium | |
| CN110825526A (en) | Distributed scheduling method and device based on ER relationship, equipment and storage medium | |
| Hung et al. | Capacity rationing decision procedures with order profit as a continuous random variable | |
| KR20210103834A (en) | Method and Apparatus for Demand Expectation Based on Machine Learning | |
| CN115729687A (en) | Task scheduling method and device, computer equipment and storage medium | |
| CN113837782B (en) | Periodic term parameter optimization method and device of time sequence model and computer equipment | |
| CN107844496A (en) | Statistical information output intent and device | |
| CN109767031A (en) | Model classifiers method for building up, device, computer equipment and storage medium | |
| CN115511562A (en) | Virtual product recommendation method and device, computer equipment and storage medium | |
| CN110033292A (en) | Information output method and device | |
| CN113780611A (en) | Inventory management method and device |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |