Analysis Method, Device, and Computer Equipment for Predictability of Item Sales Data

Abstract

The present application relates to a method, device, computer device and storage medium for analyzing the predictability of item sales data. The method includes: detecting periodic features of a historical sales data set for predicting item sales; when there are periodic features, calling an autocorrelation function to process the historical sales data set to obtain a set of period length values; based on each period length value in the set of period lengths, decomposing the historical sales data set through a seasonal decomposition function to obtain a residual data set; determining a first index value representing the ratio of the value of the unpredictable part in the historical sales data set according to the historical sales data set and the residual data set; obtaining a set of predicted values of item sales through each naive model; verifying each set of predicted values according to a test value set to obtain a second index value; evaluating the predictability of the historical sales data set according to the first index value and the second index value. Using this method can improve the accuracy of analyzing the predictability of item sales data.

Description

Technical Field

[0001] The present application relates to the technical field of data processing, and particularly to a method, device, computer device, and storage medium for analyzing the predictability of item sales data. Background Art

[0002] With the advent of the big data era and the development of prediction technologies, more and more prediction methods are applied to the supply chain field, and the results of sales volume prediction serve as an important basis for supply chain replenishment decisions and inventory planning. However, in many industries, not all item sales data can achieve good prediction effects. A supply chain solution formulated based on a very inaccurate prediction result is inappropriate, and its final effect is not as good as that of a solution using a non-prediction method. Therefore, it is necessary to analyze the predictability of item sales data in order to formulate the most suitable supply chain solution.

[0003] However, in the prior art, the predictability of item sales data is usually determined based on the manual experience of supply chain managers and their intuitive perception of sales data, resulting in low accuracy in the analysis of the predictability of item sales data. Summary of the Invention

[0004] Based on this, in view of the above technical problems, it is necessary to provide a method, device, computer device, and storage medium for analyzing the predictability of item sales data, which can improve the accuracy of the analysis of the predictability of item sales data.

[0005] A method for predicting the predictability of item sales data, the method comprising:

[0006] Obtaining a historical sales data set for predicting item sales;

[0007] Detecting the periodic characteristics of the historical sales data set;

[0008] When the historical sales data set has the periodic characteristics, calling an autocorrelation function to process the historical sales data set to obtain a set of cycle length values;

[0009] Assigning each cycle length value in the set of cycle lengths to the seasonal factor parameter in the seasonal decomposition function, and decomposing the historical sales data set through the seasonal decomposition function to obtain a residual data set;

[0010] Determining a first index value according to the historical sales data set and the residual data set, the first index value being used to characterize the ratio of the value of the unpredictable part in the historical sales data set;

[0011] Inputting the historical sales data set into each naive model to obtain a set of predicted values corresponding to each naive model;

[0012] Verify each of the predicted value sets according to the test value set to obtain a second index value; the second index value is a measure of the predictability of the historical sales data set.

[0013] Evaluate the predictability of the historical sales data set according to the first index value and the second index value.

[0014] In one embodiment, the obtaining the historical sales data set for predicting the sales volume of an item includes:

[0015] Obtain a sample of the historical sales data set for predicting the sales volume of an item.

[0016] Fit the sample of the historical sales data set according to the target fitting function to obtain a corresponding fitted data set.

[0017] Determine the historical sales data set according to the fitted data set and the sample of the historical sales data set.

[0018] In one embodiment, the detecting the periodic characteristics of the historical sales data set includes:

[0019] Perform power spectrum processing on the historical sales data set to obtain the probability values of the historical sales data in the historical sales data set.

[0020] Calculate the entropy according to each of the probability values to obtain the entropy value of the historical sales data set.

[0021] When the entropy value is within a preset range, the historical sales data set has periodic characteristics.

[0022] In one embodiment, when the historical sales data set has the periodic characteristics, calling an autocorrelation function to process the historical sales data set to obtain a set of cycle length values includes:

[0023] Shift the historical sales data set by a preset cycle length value in sequence to obtain each shifted historical sales data set.

[0024] Call an autocorrelation function to perform correlation analysis on the historical sales data set and each of the shifted historical sales data sets to obtain each correlation value and the corresponding confidence interval value.

[0025] Obtain a set of cycle length values according to each of the correlation values and the corresponding confidence interval values.

[0026] In one embodiment, the obtaining the set of cycle length values according to each of the correlation values and the corresponding confidence interval values includes:

[0027] When the correlation value is greater than the confidence interval value and the correlation value is greater than the correlation values of a preset number of adjacent ones, determine the preset cycle length value corresponding to the correlation value as the cycle length value, and obtain a set of cycle length values.

[0028] In one embodiment, the determining the first metric value according to the historical sales data set and the residual data set includes:

[0029] Obtain the residual data with the largest value from the residual data set;

[0030] Determine the first metric value according to the mean of the historical sales data set and the residual data with the largest value.

[0031] In one embodiment, the verifying each of the predicted value sets according to the test value set to obtain a second metric value includes:

[0032] Verify each of the predicted value sets according to the test value set in the historical sales data set to obtain a set of symmetric mean absolute percentage error values;

[0033] Determine the second metric value according to the symmetric mean absolute percentage error value with the smallest value in the set of symmetric mean absolute percentage error values.

[0034] In one embodiment, the evaluating the predictability of the historical sales data set according to the first metric value and the second metric value includes:

[0035] Determine a target metric value according to the first metric value and the second metric value;

[0036] Evaluate the predictability of the historical sales data set according to the target metric value.

[0037] In one embodiment, the determining the target metric value according to the first metric value and the second metric value includes any one of the following methods:

[0038] Select the metric value with the largest value from the first metric value and the second metric value as the target metric value; or

[0039] Use the weighted average of the first metric value and the second metric value as the target metric value.

[0040] An analysis device for the predictability of item sales data, the device includes:

[0041] An acquisition module, configured to acquire a historical sales data set for predicting item sales;

[0042] A detection module, configured to detect the periodicity of the historical sales data set;

[0043] A calling module, configured to, when the periodic feature exists in the historical sales volume dataset, call an autocorrelation function to process the historical sales volume dataset to obtain a set of cycle length values;

[0044] A decomposition module, configured to assign each cycle length value in the set of cycle lengths to the seasonal factor parameter in a seasonal decomposition function, and decompose the historical sales volume dataset through the seasonal decomposition function to obtain a residual dataset;

[0045] A determination module, configured to determine a first metric value according to the historical sales volume dataset and the residual dataset, where the first metric value is used to characterize the ratio of the value of the unpredictable part in the historical sales volume dataset;

[0046] A prediction module, configured to input the historical sales volume dataset into each naive model to obtain a set of prediction values corresponding to each naive model;

[0047] A verification module, configured to verify each set of prediction values according to a set of test values to obtain a second metric value; the second metric value is a metric for measuring the predictability of the historical sales volume dataset;

[0048] An evaluation module, configured to evaluate the predictability of the historical sales volume dataset according to the first metric value and the second metric value.

[0049] A computer device, including a memory and a processor, where the memory stores a computer program, and when the processor executes the computer program, the following steps are implemented:

[0050] Obtain a historical sales volume dataset for predicting the sales volume of an item;

[0051] Detect the periodic feature of the historical sales volume dataset;

[0052] When the periodic feature exists in the historical sales volume dataset, call an autocorrelation function to process the historical sales volume dataset to obtain a set of cycle length values;

[0053] Assign each cycle length value in the set of cycle lengths to the seasonal factor parameter in a seasonal decomposition function, and decompose the historical sales volume dataset through the seasonal decomposition function to obtain a residual dataset;

[0054] Determine a first metric value according to the historical sales volume dataset and the residual dataset, where the first metric value is used to characterize the ratio of the value of the unpredictable part in the historical sales volume dataset;

[0055] Input the historical sales volume dataset into each naive model to obtain a set of prediction values corresponding to each naive model;

[0056] Verify each of the predicted value sets according to the test value set to obtain a second metric value; the second metric value is a metric for measuring the predictability of the historical sales data set;

[0057] Evaluate the predictability of the historical sales data set based on the first metric value and the second metric value.

[0058] A computer-readable storage medium having a computer program stored thereon, the computer program when executed by a processor implements the following steps:

[0059] Obtain a historical sales data set for predicting the sales volume of an item;

[0060] Detect the periodic characteristics of the historical sales data set;

[0061] When the historical sales data set has the periodic characteristics, call the autocorrelation function to process the historical sales data set to obtain a set of cycle length values;

[0062] Assign each cycle length value in the cycle length set to the seasonal factor parameter in the seasonal decomposition function, and decompose the historical sales data set through the seasonal decomposition function to obtain a residual data set;

[0063] Determine a first metric value according to the historical sales data set and the residual data set, and the first metric value is used to characterize the ratio of the values of the unpredictable part in the historical sales data set;

[0064] Input the historical sales data set into each naive model to obtain a set of predicted values corresponding to each naive model;

[0065] Verify each of the predicted value sets according to the test value set to obtain a second metric value; the second metric value is a metric for measuring the predictability of the historical sales data set;

[0066] Evaluate the predictability of the historical sales data set based on the first metric value and the second metric value.

[0067] The above-mentioned method, device, computer equipment and storage medium for analyzing the predictability of item sales data analyze the historical sales data set for predicting item sales with periodic characteristics by calling the autocorrelation function when detecting the periodic characteristics of the historical sales data set, and obtain the set of cycle length values existing in the historical sales data set; assign each cycle length value in the cycle length set to the seasonal factor parameter in the seasonal decomposition function, and decompose the historical sales data set through the seasonal decomposition function to obtain the residual data set; determine the first index value of the proportion of the unpredictable part of the historical sales data set according to the historical sales data set and the residual data set; input the historical sales data set into each naive model to obtain the set of predicted values corresponding to each naive model; verify each set of predicted values according to the set of test values of each naive model, and obtain the second index value with the best prediction effect from it; evaluate the predictability of the historical sales data set through the first index value and the second index value; that is, perform data analysis based on the periodic characteristics of the historical sales data itself, analyze the historical sales data according to the data periodic law, and obtain the evaluation result of the predictability of the data itself, avoiding the influence of factors such as manual experience, item or industry characteristics on the data predictability, and improving the accuracy of the analysis of the predictability of historical sales data. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] Figure 1 It is an application environment diagram of the method for analyzing the predictability of item sales data in an embodiment;

[0069] Figure 2 It is a flowchart of the method for analyzing the predictability of item sales data in an embodiment;

[0070] Figure 3 It is a flowchart of the method for detecting the periodic characteristics of the historical sales data set in an embodiment;

[0071] Figure 4 It is a flowchart of the method for determining the cycle length value in an embodiment;

[0072] Figure 5 It is a flowchart of the method for analyzing the predictability of item sales data in another embodiment;

[0073] Figure 6 It is a structural block diagram of the device for analyzing the predictability of item sales data in an embodiment;

[0074] Figure 7 It is a structural block diagram of the device for analyzing the predictability of item sales data in another embodiment;

[0075] Figure 8 It is an internal structure diagram of the computer equipment in an embodiment. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0076] In order to make the objectives, technical solutions and advantages of the present application more clear and understandable, the present application will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0077] The method for analyzing the predictability of item sales volume data provided by the present application can be applied to, for example, Figure 1 the application environment shown in the figure. Among them, the terminal 102 communicates with the server 104 through a network. The terminal 102 obtains a historical sales volume data set for predicting item sales volume from the server 104; detects the periodic characteristics of the historical sales volume data set; when the historical sales volume data set has periodic characteristics, calls the autocorrelation function to process the historical sales volume data set to obtain a set of cycle length values; assigns each cycle length value in the cycle length set to the seasonal factor parameter in the seasonal decomposition function, and decomposes the historical sales volume data set through the seasonal decomposition function to obtain a residual data set; determines a first index value according to the historical sales volume data set and the residual data set, and the first index value is used to represent the ratio of the value of the unpredictable part in the historical sales volume data set; inputs the historical sales volume data set into each naive model to obtain a set of predicted values corresponding to each naive model; verifies each set of predicted values according to the test value set to obtain a second index value; the second index value is a measure for measuring the predictability of the historical sales volume data set; evaluates the predictability of the historical sales volume data set according to the first index value and the second index value. Among them, the terminal 102 can be, but is not limited to, various personal computers, laptop computers, smart phones, tablet computers, and the server 104 can be implemented by an independent server or a server cluster composed of multiple servers.

[0078] In one embodiment, as Figure 2 shown, a method for analyzing the predictability of item sales volume data is provided. Taking the terminal in Figure 1 as an example, the method includes the following steps:

[0079] Step 202, obtaining a historical sales volume data set for predicting item sales volume.

[0080] Among them, the historical sales volume data set is the sales volume data of the item without trend in the past preset time period, and each historical sales volume data in the historical sales volume data set has a corresponding time attribute. For example, the historical data set of the item sales volume in the past month includes 30 historical sales volume data, and the time attribute value corresponding to each historical sales volume data can be a date, a time sequence (for example, the nth day). The historical sales volume data set can be determined according to the time period to be predicted. For example, to predict the predicted sales volume value of items during a shopping festival and obtain accurate replenishment strategy data and logistics strategy data for supply chain items, a historical sales volume data set of the shopping festival is selected.

[0081] Furthermore, the method for obtaining the historical sales dataset used to predict the item sales volume includes: obtaining a sample of the historical sales dataset used to predict the item sales volume; fitting the sample of the historical sales dataset according to the target fitting function to obtain the corresponding fitted dataset; and determining the historical sales dataset based on the fitted dataset and the sample of the historical sales dataset. The target fitting function includes a linear function and an exponential function. The expression of the linear function can be: y = ax + b, and the expression of the exponential function can be: y = x a + b. Where x represents the time attribute parameter of the historical sales data.

[0082] Respectively fit the sample of the historical sales dataset with the linear function and the exponential function, determine the parameter values of parameters a and b in the linear function and the exponential function to obtain the function expressions of the linear function and the exponential function; assign the time attribute values of each historical sales data in the sample of the historical sales dataset to the time attribute parameters of the linear function and the exponential function respectively to obtain the first fitted dataset of the linear function and the second fitted dataset of the exponential function; determine the fitted dataset from the first fitted dataset and the second fitted dataset according to the goodness-of-fit index value of the linear function and the goodness-of-fit index value of the exponential function. Based on the time attribute value of the historical sales data, subtract each historical sales data in the sample of the historical sales dataset from the corresponding fitted data in the fitted dataset to obtain the detrended historical sales dataset.

[0083] Specifically, according to the prediction time period for predicting the item sales volume, obtain a sample of the historical sales dataset of the corresponding item sales volume from the server, fit the sample of the historical sales dataset through the target function to obtain the corresponding fitted dataset, and determine the detrended historical sales dataset according to the difference between each historical sales data in the sample of the historical sales dataset and each fitted data in the fitted dataset.

[0084] Step 204, detect the periodic characteristics of the historical sales dataset.

[0085] The periodic characteristic refers to that the data is regular within a fixed time period. The periodic characteristics include weekly periodic characteristics, monthly periodic characteristics, annual periodic characteristics, seasonal characteristics, etc.

[0086] Specifically, the terminal calls the power spectrum estimation function to convert the historical sales dataset into sales frequency values, and obtains the probability values of the occurrence of each sales frequency value; determines the entropy value of the historical sales dataset according to each probability value; when the entropy value is within the preset range, it indicates that the historical sales dataset has periodic characteristics; the entropy value is used to characterize the certainty level of the historical sales dataset.

[0087] Step 206, when the historical sales dataset has periodic characteristics, call the autocorrelation function to process the historical sales dataset to obtain a set of cycle length values.

[0088] Among them, the autocorrelation function is used to determine the correlation value between the data sequence and the observed values of the data sequence at each predicted time length value.

[0089] Specifically, when there is a periodicity in the historical sales dataset, the autocorrelation function is called to sequentially obtain the observed value sets of the historical sales dataset at each preset time length value; each historical sales data in the historical sales dataset is analyzed with the observed values obtained each time to obtain the corresponding correlation value set, and there is a corresponding confidence interval value for the correlation value. The period length value set of the original sales dataset is determined according to the distribution trend of the correlation value set and the confidence interval value; the period length value set includes at least one period length value.

[0090] Step 208, assign each period length value in the period length set to the seasonal factor parameter in the seasonal decomposition function, and decompose the historical sales dataset through the seasonal decomposition function to obtain the residual dataset.

[0091] Specifically, the terminal calls the seasonal decomposition function (for example, decompose), and sequentially assigns the period length values in the period length value to the seasonal factor parameter, performs seasonal decomposition on the historical sales dataset, determines the residual data in the historical sales dataset, and obtains the residual dataset.

[0092] Step 210, determine the first index value according to the historical sales dataset and the residual dataset, and the first index value is used to represent the ratio of the unpredictable part of the values in the historical sales dataset.

[0093] Specifically, obtain the residual data with the largest value from the residual dataset, and determine the mean value of the historical sales dataset; determine the first index value according to the mean value of the historical sales dataset and the residual data with the largest value, and obtain the ratio of the unpredictable part of the values in the historical sales dataset.

[0094] Step 212, input the historical sales dataset into each naive model to obtain the corresponding prediction value sets of each naive model.

[0095] Among them, the historical sales dataset is divided into a test set and a training set according to a preset ratio. The prediction ratio can be but is not limited to 7:3, that is, 70% of the historical sales dataset is used as the training set, and 30% of the historical sales dataset is used as the test set.

[0096] Among them, the naive model accurately and quickly outputs the predicted sales volume value of the item by reducing the data processing time, and updates the quantity of items in the item supply chain and realizes the reasonable allocation of items based on the predicted sales volume value. The naive model may include Naive, Snaive, Mean prediction method (Mean), moving average, Simple Exponential Smoothing (SES), and Holt-Winters (Holt).

[0097] Among them, the prediction result obtained by Naive is the last historical data value, and the last historical data value is determined by the data dimension, which can be the last day's historical data value, the last week's historical data value, the last historical data value, etc.; it can be expressed as y n+1 = y n .

[0098] The prediction result of Snaive can be expressed as: y n+1 = y n-t+1 , that is, the prediction result is equal to the historical data value at the time point t before, and t can be preset.

[0099] The prediction result of Mean can be expressed as: That is, the prediction result of the data is equal to the mean value of the historical data. The prediction result of the moving average can be expressed as: That is, the predicted value of the data is equal to the mean value of the previous t historical data, and t can be preset.

[0100] Specifically, the training set in the historical sales data set is input into each naive model for training to obtain a trained naive model; the item sales volume is predicted through the trained naive model to obtain a set of predicted values of the item sales volume for each naive model.

[0101] Step 214, verify each set of predicted values according to the test value set to obtain a second index value; the second index value is a measure of the predictability of the historical sales data set.

[0102] Specifically, verify each set of predicted values according to the test value set in the historical sales data set to obtain a set of symmetric mean absolute percentage error values; determine the second index value according to the symmetric mean absolute percentage error value with the smallest value in the set of symmetric mean absolute percentage error values. The calculation expression of the symmetric mean absolute percentage error value (sMAPE) is:

[0103]

[0104] Among them is the predicted value, y iIt is a test value, that is, the true value of the item sales volume. The value range of sMAPE is truncated in [0, 1]. The closer the value is to 0, the smaller the error of the naive model for the item sales volume value, and the better the prediction effect. The higher the accuracy of the historical sales volume dataset used for predicting the item sales volume data.

[0105] Step 216, evaluate the predictability of the historical sales volume dataset according to the first index value and the second index value.

[0106] Among them, the predictability is used to determine whether to use the replenishment strategy data based on the prediction method. When the predictability of the historical sales volume dataset is evaluated to be high, the replenishment strategy data based on the prediction method is used; when the predictability of the historical sales volume dataset is evaluated to be low, the replenishment strategy data based on the prediction method is not used. The replenishment strategy data based on the prediction method replenishes goods according to the predicted sales volume data and the item inventory data; for example, based on the replenishment strategy of the prediction method, the predicted sales volumes for the next three days are 23 pieces, 30 pieces, and 27 pieces respectively, the total sales volume is 80 pieces, and the current inventory is 50 pieces, then 30 pieces need to be replenished. The replenishment strategy data not based on the prediction method replenishes the item according to the set replenishment strategy data; for example, the item is replenished 30 pieces each time or when the item inventory quantity is less than 30 pieces, the item inventory quantity needs to be replenished to 70 pieces.

[0107] Specifically, a target index value is determined based on a first index value and a second index value. The target index value can be, but is not limited to, the largest value among the first index value and the second index value, or the weighted average of the first index value and the second index value. The value range of the target index value is in the interval [0, 1]. The predictability of the historical sales dataset is evaluated based on the target index value. When the historical sales dataset can be used to predict the item sales volume, the historical sales dataset is input into the item sales volume prediction model to obtain the predicted sales volume value of the item. The inventory quantity of the item in the supply volume is queried according to the predicted sales volume value to generate item replenishment strategy data. And the replenishment strategy data is sent to the logistics business processing terminal to generate corresponding logistics strategy data. The item replenishment is realized based on the logistics strategy data, which improves the reasonable allocation of resources and the processing efficiency. In the above method for analyzing the predictability of item sales data, by determining the periodic characteristics of the historical sales dataset, the autocorrelation function is called to analyze the historical sales dataset, and a set of cycle length values existing in the historical sales dataset is obtained. Each cycle length value in the set of cycle lengths is assigned to the seasonal factor parameter in the seasonal decomposition function, and the historical sales dataset is decomposed by the seasonal decomposition function to obtain a residual dataset. The first index value representing the proportion of the unpredictable part of the historical sales dataset is determined based on the historical sales dataset and the residual dataset. The historical sales dataset is input into each naive model to obtain a set of predicted values corresponding to each naive model. Each set of predicted values is verified according to the test value set of each naive model, and the second index value with the best prediction effect is obtained therefrom. The predictability of the historical sales dataset is evaluated through the first index value and the second index value. That is, data analysis is performed based on the periodic characteristics of the historical sales data itself, and the historical sales data is analyzed according to the data periodic law to obtain an evaluation result on the predictability of the data itself, avoiding the influence of factors such as manual experience, item or industry characteristics on the data predictability, improving the accuracy of the analysis of the predictability of historical sales data, and generating replenishment strategy data for the supply chain to achieve reasonable allocation of items.

[0108] In one embodiment, as Figure 3 shown, a method for detecting the periodic characteristics of a historical sales dataset is provided. Taking the case where this method is applied to Figure 1 the terminal as an example, the method includes the following steps:

[0109] Step 302, perform power spectrum processing on the historical sales dataset to obtain the probability values of each historical sales data in the historical sales dataset.

[0110] Among them, the probability value can be the probability value of a frequency value appearing in the power spectrum or the probability value of a frequency interval.

[0111] Specifically, the terminal calls the power spectrum estimation function to convert the historical sales dataset into sales frequency values, obtains the sales frequency values corresponding to the historical sales dataset, determines the frequency interval values of each sales frequency value according to the characteristics of the business scenario, and obtains the probability values of each frequency interval.

[0112] Step 304, perform entropy calculation according to each probability value to obtain the entropy value of the historical sales dataset.

[0113] Among them, the entropy calculation expression for calculating entropy (H) according to each probability value can be:

[0114]

[0115] P(f) is the probability value of each sales frequency in the frequency interval obtained according to the power spectrum.

[0116] Specifically, by assigning the probability values of each frequency interval to the entropy calculation expression and performing accumulation, the entropy value characterizing the certainty of the historical sales dataset is obtained.

[0117] Step 306, determine whether the entropy value is within the preset range. If so, the historical sales dataset has periodic characteristics; otherwise, it does not have periodic characteristics.

[0118] Among them, the preset range can be but is not limited to (0, 0.7); when the entropy value is within the interval (0, 0.7), it indicates that the historical sales dataset has periodic characteristics; when the entropy value is not within the interval (0, 0.7), it indicates that the historical sales dataset does not have periodic characteristics.

[0119] In this embodiment, by calling the power spectrum estimation function to detect the periodic characteristics of the historical sales dataset, converting the historical sales data into sales frequency, analyzing the sales frequency according to the frequency characteristics in the power spectrum, and detecting the periodic characteristics of the sales historical dataset, the data processing operation is simplified, and the periodic characteristics of the historical sales dataset can be accurately detected.

[0120] In one embodiment, as Figure 4 shown, a method for determining the period length value is provided. Taking the method applied to the Figure 1 terminal as an example for illustration, it includes the following steps:

[0121] Step 402, shift the historical sales dataset in sequence with a preset time length value to obtain each shifted historical sales dataset.

[0122] Among them, the preset time length is a preset number of time units set in advance. The time unit can be, but is not limited to, days. The preset number can be integer values such as 1, 2, 3, etc. For example, the historical sales volume dataset is the historical sales volume data of an item from the 1st day to the 30th day of June. Translating the historical sales volume dataset according to the preset number 1 as the preset time length value, the obtained translated historical sales volume dataset is the historical sales volume data from the 2nd day to the 30th day. Translating the historical sales volume dataset according to the preset number 2 as the preset time length value, the obtained translated historical sales volume dataset is the historical sales volume data from the 3rd day to the 30th day.

[0123] Step 404, call the autocorrelation function to perform correlation analysis on the historical sales volume dataset and each translated historical sales volume dataset, and obtain each correlation value and the corresponding confidence interval value.

[0124] Among them, the correlation value and the confidence interval value are determined according to the historical sales volume dataset and each translated historical sales volume dataset.

[0125] Step 406, when the correlation value is greater than the confidence interval value and the correlation value is greater than the correlation value of the adjacent preset number, determine the preset time length value corresponding to the correlation value as the cycle length value, and obtain the cycle length value set.

[0126] In the above method for determining the cycle length value, the historical sales volume dataset is translated in turn by the preset time length value to obtain each translated historical sales volume dataset. The autocorrelation function is called to perform correlation analysis on the historical sales volume dataset and each translated historical sales volume dataset. Based on the strategy that the correlation value is greater than the confidence interval value and the correlation value is greater than the correlation value of the adjacent preset number, the cycle length value set is obtained. By analyzing the historical sales volume dataset through the preset time length and calling the autocorrelation function, the cycle length existing in the historical sales volume dataset can be accurately detected, improving the accuracy of data processing.

[0127] In another embodiment, as Figure 5 shown, a method for analyzing the predictability of item sales data is provided. Taking the example that this method is applied to Figure 1 the terminal in

[0128] Step 502, obtain a sample of the historical sales volume dataset for predicting the item sales volume.

[0129] Step 504, fit the sample of the historical sales volume dataset according to the target fitting function to obtain the corresponding fitted dataset.

[0130] Step 506, determine the historical sales volume dataset according to the fitted dataset and the sample of the historical sales volume dataset.

[0131] Step 508, detect the periodic characteristics of the historical sales dataset.

[0132] Step 510, when the historical sales dataset has periodic characteristics, call the autocorrelation function to process the historical sales dataset to obtain a set of cycle length values.

[0133] Step 512, assign each cycle length value in the cycle length set to the seasonal factor parameter in the seasonal decomposition function, and decompose the historical sales dataset through the seasonal decomposition function to obtain a residual dataset.

[0134] Step 514, determine a first index value according to the historical sales dataset and the residual dataset, and the first index value is used to represent the ratio of the values of the unpredictable part in the historical sales dataset.

[0135] Step 516, input the historical sales dataset into each naive model to obtain a set of predicted values corresponding to each naive model.

[0136] Step 518, verify each set of predicted values according to the test value set to obtain a second index value.

[0137] Specifically, verify each set of predicted values according to the test value set in the historical sales dataset to obtain a set of symmetric mean absolute percentage error values; determine the second index value according to the symmetric mean absolute percentage error value with the smallest value in the set of symmetric mean absolute percentage error values.

[0138] Step 520, use the weighted average of the first index value and the second index value as the target index value, and evaluate the predictability of the historical sales dataset according to the target index value.

[0139] Optionally, select the index value with the largest value from the first index value and the second index value as the target index value, and evaluate the predictability of the historical sales dataset according to the target index value.

[0140] In the above method for analyzing the predictability of item sales data, a fitted data set is obtained by fitting a historical sales data set sample for predicting item sales retrieved from a server; the historical sales data set sample is detrended based on the fitted data set to obtain a detrended historical sales data set; the historical sales data set is processed by a power spectrum estimation method, and the entropy value of the historical sales data set is determined according to the obtained probability value; the historical sales data set is detected for periodic characteristics based on the entropy value; when the historical sales data set has periodic characteristics, based on the periodic characteristics of the data, the period length value existing in the historical sales data set is determined by calling an autocorrelation function; the historical sales data is seasonally decomposed according to the periodic law of the data to obtain a residual data set; a first index value representing the ratio of the value of the unpredictable part in the historical sales data set is determined according to the mean of the historical sales data set and the largest residual value in the residual data set.

[0141] The historical sales data set is input into each naive model to obtain a set of predicted values corresponding to each naive model; each set of predicted values is verified according to the test value set in the historical sales data set to obtain a set of symmetric mean absolute percentage error values; a second index value is determined according to the smallest symmetric mean absolute percentage error value in the set of symmetric mean absolute percentage error values; the weighted average of the first index value and the second index value is used as the target index value, and the predictability of the historical sales data set is evaluated according to the target index value. That is, based on the objective historical sales data set of item sales retrieved from the server and the periodic characteristics of the data, it is determined that the historical sales data set has periodic characteristics; according to the periodic law of the data, the historical sales data set is analyzed and processed by the terminal calling an autocorrelation function, a seasonal decomposition function, and a naive model to determine the target index value for evaluating the predictability of the historical sales data set. Analyzing based on the periodic characteristics of the data itself improves the accuracy of the analysis of the predictability of historical sales data and reduces the processing time of the terminal, thereby improving the processing performance of the terminal for data.

[0142] It should be understood that although Figures 2 - 5 the steps in the flowchart of Figures 2 - 5 are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise clearly stated in this article, the execution of these steps has no strict order limit, and these steps can be executed in other orders. Moreover,

[0143] In one embodiment, as Figure 6 shown, a device for analyzing the predictability of item sales volume data is provided, including: an acquisition module 602, a detection module 604, a call module 606, a decomposition module 608, a determination module 610, a prediction module 612, a verification module 614, and an evaluation module 616, where:

[0144] The acquisition module 602 is configured to acquire a historical sales volume data set for predicting the item sales volume.

[0145] The detection module 604 is configured to detect the periodicity of the historical sales volume data set.

[0146] The call module 606 is configured to, when the historical sales volume data set has periodic characteristics, call the autocorrelation function to process the historical sales volume data set to obtain a set of cycle length values.

[0147] The decomposition module 608 is configured to assign each cycle length value in the cycle length set to the seasonal factor parameter in the seasonal decomposition function, and decompose the historical sales volume data set through the seasonal decomposition function to obtain a residual data set.

[0148] The determination module 610 is configured to determine a first index value according to the historical sales volume data set and the residual data set, and the first index value is used to characterize the ratio of the unpredictable part of the historical sales volume data set.

[0149] The prediction module 612 inputs the historical sales volume data set into each naive model to obtain a set of predicted values corresponding to each naive model.

[0150] The verification module 614 is configured to verify each set of predicted values according to the test value set to obtain a second index value; the second index value is a measure for measuring the predictability of the historical sales volume data set.

[0151] The evaluation module 616 is configured to evaluate the predictability of the historical sales volume data set according to the first index value and the second index value.

[0152] In the above-mentioned analysis device for the predictability of the sales volume data of the item, by determining the periodic characteristics of the historical sales volume data set, the autocorrelation function is called to analyze the historical sales volume data set, and the set of cycle length values existing in the historical sales volume data set is obtained; each cycle length value in the set of cycle length values is assigned to the seasonal factor parameter in the seasonal decomposition function, and the historical sales volume data set is decomposed by the seasonal decomposition function to obtain the residual data set; according to the historical sales volume data set and the residual data set, the first index value of the proportion of the unpredictable part of the historical sales volume data set is determined; the historical sales volume data set is input into each naive model to obtain the set of predicted values corresponding to each naive model; according to the test value sets of each naive model, each set of predicted values is verified, and the second index value with the best prediction effect is obtained from them; the predictability of the historical sales volume data set is evaluated through the first index value and the second index value; that is, based on the self-periodic characteristics of the historical sales volume data, data analysis is carried out, and the historical sales volume data is analyzed according to the data periodic law to obtain the evaluation result of the self-predictability of the data, avoiding the influence of factors such as manual experience, item or industry characteristics on the data predictability, improving the accuracy of the analysis of the predictability of the historical sales volume data, and generating the replenishment strategy data of the supply chain to achieve reasonable allocation of items.

[0153] In one embodiment, as Figure 7 shown, a device for analyzing the predictability of item sales volume data is provided. In addition to including the acquisition module 602, the detection module 604, the call module 606, the decomposition module 608, the determination module 610, the prediction module 612, the verification module 614, and the evaluation module 616, it further includes: a fitting module 618 and a processing module 620, where:

[0154] In one embodiment, the acquisition module 602 is further configured to acquire a sample of the historical sales volume data set for predicting the item sales volume.

[0155] The fitting module 618 is configured to fit the sample of the historical sales volume data set according to the target fitting function to obtain the corresponding fitting data set; and determine the historical sales volume data set according to the fitting data set and the sample of the historical sales volume data set.

[0156] The processing module 620 is configured to perform power spectrum processing on the historical sales volume data set to obtain the probability values of each historical sales volume data in the historical sales volume data set; and calculate the entropy value of the historical sales volume data set according to each probability value.

[0157] In one embodiment, the detection module 604 is further configured to determine that the historical sales volume data set has periodic characteristics when the entropy value is within the preset range value.

[0158] In one embodiment, the calling module 606 is further configured to shift the historical sales volume data set in sequence with a preset cycle length value to obtain each shifted historical sales volume data set; call the autocorrelation function to perform correlation analysis on the historical sales volume data set and each shifted historical sales volume data set to obtain each correlation value and the corresponding confidence interval value; and obtain a set of cycle length values according to each correlation value and the corresponding confidence interval value.

[0159] In one embodiment, the determining module 610 is further configured to determine that the preset cycle length value corresponding to the correlation value is the cycle length value when the correlation value is greater than the confidence interval value and the correlation value is greater than the correlation values of the adjacent preset quantity, so as to obtain a set of cycle length values.

[0160] In one embodiment, the determining module 610 is further configured to obtain the residual data with the largest value from the residual data set; and determine a first index value according to the mean value of the historical sales volume data set and the residual data with the largest value.

[0161] In one embodiment, the verification module 614 is further configured to verify each predicted value set according to the test value set in the historical sales volume data set to obtain a set of symmetric mean absolute percentage error values; and determine a second index value according to the symmetric mean absolute percentage error value with the smallest value in the set of symmetric mean absolute percentage error values.

[0162] In one embodiment, the evaluation module 616 is further configured to determine a target index value according to the first index value and the second index value; and evaluate the predictability of the historical sales volume data set according to the target index value.

[0163] In one embodiment, the determining module 610 is further configured to select the index value with the largest value from the first index value and the second index value as the target index value; or use the weighted average value of the first index value and the second index value as the target index value.

[0164] In one embodiment, by fitting a sample of the historical sales volume data set obtained from the server for predicting the sales volume of an item, a fitting data set is obtained; the historical sales volume data set sample is detrended according to the fitting data set to obtain a detrended historical sales volume data set; the historical sales volume data set is processed by a power spectrum estimation method, and the entropy value of the historical sales volume data set is determined according to the obtained probability value; the periodic characteristics of the historical sales volume data set are detected according to the entropy value; when the historical sales volume data set has periodic characteristics, based on the periodic characteristics of the data, the autocorrelation function is called to determine the cycle length value existing in the historical sales volume data set; the historical sales volume data is seasonally decomposed according to the periodic law of the data to obtain a residual data set; and a first index value representing the ratio of the value of the unpredictable part in the historical sales volume data set is determined according to the mean value of the historical sales volume data set and the residual data with the largest value in the residual data set.

[0165] Input the historical sales volume dataset into each naive model to obtain the corresponding predicted value sets of each naive model; verify each predicted value set according to the test value set in the historical sales volume dataset to obtain the symmetric mean absolute percentage error value set; determine the second index value according to the symmetric mean absolute percentage error value with the smallest value in the symmetric mean absolute percentage error value set; use the weighted average of the first index value and the second index value as the target index value, and evaluate the predictability of the historical sales volume dataset according to the target index value. That is, based on the objective historical sales volume dataset of the item sales obtained from the server and the periodic characteristics of the data, it is determined that the historical sales volume dataset has periodic characteristics; according to the periodic law of the data, the autocorrelation function, seasonal decomposition function and naive model are called by the terminal to analyze and process the historical sales volume dataset, and the target index value for evaluating the predictability of the historical sales volume dataset is determined. Analyze based on the periodic characteristics of the data itself to improve the accuracy of the analysis of the predictability of historical sales volume data and reduce the processing time of the terminal, thereby improving the processing performance of the terminal for data.

[0166] For the specific limitations of the item sales volume data predictability analysis device, reference can be made to the limitations of the item sales volume data predictability analysis method in the above text, which will not be elaborated here. Each module in the above item sales volume data predictability analysis device can be implemented in whole or in part by software, hardware and their combination. The above modules can be embedded in the processor of the computer device in hardware form or independent of it, or stored in the memory of the computer device in software form, so that the processor can call and execute the operations corresponding to each of the above modules.

[0167] In one embodiment, a computer device is provided. The computer device can be a terminal, and its internal structure diagram can be as Figure 8 shown. The computer device includes a processor, a memory, a communication interface, a display screen and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile 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 the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, a carrier network, NFC (Near Field Communication) or other technologies. When the computer program is executed by the processor, it implements an item sales volume data predictability analysis method. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covered on the display screen, or a button, trackball or touchpad set on the shell of the computer device, or an external keyboard, touchpad or mouse, etc.

[0168] Those skilled in the art can understand that Figure 8 the structure shown in Figure 8 is only a block diagram of some structures related to the solution of this application, and does not constitute a limitation on the computer device to which the solution of this application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine some components, or have different component arrangements.

[0169] In one embodiment, a computer device is provided, including a memory and a processor. A computer program is stored in the memory, and when the processor executes the computer program, the following steps are implemented:

[0170] Obtain a historical sales volume data set for predicting the sales volume of an item;

[0171] Detect the periodic characteristics of the historical sales volume data set;

[0172] When the historical sales volume data set has periodic characteristics, call the autocorrelation function to process the historical sales volume data set to obtain a set of cycle length values;

[0173] Assign each cycle length value in the cycle length set to the seasonal factor parameter in the seasonal decomposition function, and decompose the historical sales volume data set through the seasonal decomposition function to obtain a residual data set;

[0174] Determine a first index value according to the historical sales volume data set and the residual data set. The first index value is used to characterize the ratio of the value of the unpredictable part in the historical sales volume data set;

[0175] Input the historical sales volume data set into each naive model to obtain a set of predicted values corresponding to each naive model;

[0176] Verify each set of predicted values according to the test value set to obtain a second index value; The second index value is a measure of the predictability of the historical sales volume data set;

[0177] Evaluate the predictability of the historical sales volume data set according to the first index value and the second index value.

[0178] In one embodiment, when the processor executes the computer program, the following steps are also implemented:

[0179] Obtain a sample of the historical sales volume data set for predicting the sales volume of an item;

[0180] Fit the sample of the historical sales volume data set according to the target fitting function to obtain a corresponding fitted data set;

[0181] Determine the historical sales volume data set according to the fitted data set and the sample of the historical sales volume data set.

[0182] In one embodiment, when the processor executes the computer program, the following steps are further implemented:

[0183] Perform power spectrum processing on the historical sales volume dataset to obtain the probability values of each historical sales volume data in the historical sales volume dataset;

[0184] Calculate the entropy based on each probability value to obtain the entropy value of the historical sales volume dataset;

[0185] When the entropy value is within the preset range, the historical sales volume dataset has periodic characteristics.

[0186] In one embodiment, when the processor executes the computer program, the following steps are further implemented:

[0187] Shift the historical sales volume dataset in sequence with a preset cycle length value to obtain each shifted historical sales volume dataset;

[0188] Call the autocorrelation function to perform correlation analysis on the historical sales volume dataset and each shifted historical sales volume dataset to obtain each correlation value and the corresponding confidence interval value;

[0189] Obtain the set of cycle length values based on each correlation value and the corresponding confidence interval value.

[0190] In one embodiment, when the processor executes the computer program, the following steps are further implemented:

[0191] When the correlation value is greater than the confidence interval value and the correlation value is greater than the correlation values of the adjacent preset quantity, determine the preset cycle length value corresponding to the correlation value as the cycle length value to obtain the set of cycle length values.

[0192] In one embodiment, when the processor executes the computer program, the following steps are further implemented:

[0193] Obtain the residual data with the largest value from the residual dataset;

[0194] Determine the first index value based on the mean value of the historical sales volume dataset and the residual data with the largest value.

[0195] In one embodiment, when the processor executes the computer program, the following steps are further implemented:

[0196] Verify each predicted value set according to the test value set in the historical sales volume dataset to obtain the set of symmetric mean absolute percentage error values;

[0197] Determine the second index value according to the symmetric mean absolute percentage error value with the smallest value in the set of symmetric mean absolute percentage error values.

[0198] In one embodiment, when the processor executes the computer program, the following steps are further implemented:

[0199] Determine a target metric value based on a first metric value and a second metric value;

[0200] Evaluate the predictability of a historical sales volume data set based on the target metric value.

[0201] In one embodiment, when the processor executes a computer program, the following steps are further implemented:

[0202] Select the metric value with the largest numerical value from the first metric value and the second metric value as the target metric value; or use the weighted average of the first metric value and the second metric value as the target metric value.

[0203] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the following steps are implemented:

[0204] Obtain a historical sales volume data set for predicting the sales volume of an item;

[0205] Detect the periodic characteristics of the historical sales volume data set;

[0206] When the historical sales volume data set has periodic characteristics, call the autocorrelation function to process the historical sales volume data set to obtain a set of period length values;

[0207] Assign each period length value in the set of period lengths to the seasonal factor parameter in the seasonal decomposition function, and decompose the historical sales volume data set through the seasonal decomposition function to obtain a residual data set;

[0208] Determine a first metric value based on the historical sales volume data set and the residual data set. The first metric value is used to characterize the ratio of the value of the unpredictable part in the historical sales volume data set;

[0209] Input the historical sales volume data set into each naive model to obtain a set of predicted values corresponding to each naive model;

[0210] Verify each set of predicted values according to a set of test values to obtain a second metric value; The second metric value is a metric for measuring the predictability of the historical sales volume data set;

[0211] Evaluate the predictability of the historical sales volume data set based on the first metric value and the second metric value.

[0212] In one embodiment, when the computer program is executed by the processor, the following steps are further implemented:

[0213] Obtain a sample of the historical sales volume data set for predicting the sales volume of an item;

[0214] Fit the sample of the historical sales volume data set according to a target fitting function to obtain a corresponding fitted data set;

[0215] Determine the historical sales volume dataset based on the fitting dataset and the sample of the historical sales volume dataset.

[0216] In one embodiment, when the computer program is executed by a processor, the following steps are further implemented:

[0217] Perform power spectrum processing on the historical sales volume dataset to obtain the probability values of each historical sales volume data in the historical sales volume dataset;

[0218] Calculate the entropy based on each probability value to obtain the entropy value of the historical sales volume dataset;

[0219] When the entropy value is within the preset range, the historical sales volume dataset has periodic characteristics.

[0220] In one embodiment, when the computer program is executed by a processor, the following steps are further implemented:

[0221] Translate the historical sales volume dataset in sequence with the preset period length value to obtain each translated historical sales volume dataset;

[0222] Call the autocorrelation function to perform correlation analysis on the historical sales volume dataset and each translated historical sales volume dataset to obtain each correlation value and the corresponding confidence interval value;

[0223] Obtain the set of period length values based on each correlation value and the corresponding confidence interval value.

[0224] In one embodiment, when the computer program is executed by a processor, the following steps are further implemented:

[0225] When the correlation value is greater than the confidence interval value and the correlation value is greater than the correlation values of the adjacent preset quantity, determine the preset period length value corresponding to the correlation value as the period length value to obtain the set of period length values.

[0226] In one embodiment, when the computer program is executed by a processor, the following steps are further implemented:

[0227] Obtain the residual data with the largest value from the residual dataset;

[0228] Determine the first index value based on the mean value of the historical sales volume dataset and the residual data with the largest value.

[0229] In one embodiment, when the computer program is executed by a processor, the following steps are further implemented:

[0230] Verify each predicted value set according to the test value set in the historical sales volume dataset to obtain the set of symmetric mean absolute percentage error values;

[0231] Determine the second index value according to the symmetric mean absolute percentage error value with the smallest value in the set of symmetric mean absolute percentage error values.

[0232] In one embodiment, when the computer program is executed by a processor, the following steps are further implemented:

[0233] Determine a target metric value based on a first metric value and a second metric value;

[0234] Evaluate the predictability of a historical sales dataset based on the target metric value.

[0235] In one embodiment, when the computer program is executed by a processor, the following steps are further implemented:

[0236] Select the metric value with the largest numerical value from the first metric value and the second metric value as the target metric value; or

[0237] Use the weighted average of the first metric value and the second metric value as the target metric value.

[0238] Those of ordinary skill in the art can understand that all or part of the processes of implementing the methods in the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods. Among them, any reference to a memory, storage, database, or other medium used in the various embodiments provided in the present application can include at least one of non-volatile and volatile memories. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, or optical memory, etc. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc.

[0239] The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity of description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as the scope described in this specification.

[0240] The above-described embodiments only represent several implementation manners of the present application. Their descriptions are relatively specific and detailed, but they should not be construed as limiting the scope of the invention patent. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present application, several deformations and improvements can still be made, and these all belong to the protection scope of the present application. Therefore, the protection scope of the patent of the present application should be subject to the appended claims.

Claims

1. A method for analyzing the predictability of item sales volume data, characterized in that, the method includes: Obtain a historical sales volume data set for predicting the item sales volume; Detect the periodic characteristics of the historical sales volume data set; When the historical sales volume data set has the periodic characteristics, call the autocorrelation function to process the historical sales volume data set to obtain a set of cycle length values; Assign each cycle length value in the cycle length set to the seasonal factor parameter in the seasonal decomposition function, and perform seasonal decomposition on the historical sales volume data set through the seasonal decomposition function to determine the residual data in the historical sales volume data set, and obtain a residual data set; Determine a first index value according to the historical sales volume data set and the residual data set, and the first index value is used to represent the ratio of the unpredictable part of the numerical value in the historical sales volume data set; Input the historical sales volume data set into each naive model to obtain a set of predicted values corresponding to each naive model; Verify each set of predicted values according to the test value set to obtain a set of symmetric mean absolute percentage error values, and determine a second index value according to the symmetric mean absolute percentage error value with the smallest value in the set of symmetric mean absolute percentage error values; the second index value is a measure for measuring the predictability of the historical sales volume data set; Evaluate the predictability of the historical sales volume data set according to the first index value and the second index value.

2. The method according to claim 1, characterized in that, the obtaining of the historical sales volume data set for predicting the item sales volume includes: Obtain a sample of the historical sales volume data set for predicting the item sales volume; Fit the sample of the historical sales volume data set according to the target fitting function to obtain a corresponding fitted data set; Determine the historical sales volume data set according to the fitted data set and the sample of the historical sales volume data set.

3. The method according to claim 1, characterized in that, the detecting of the periodic characteristics of the historical sales volume data set includes: Perform power spectrum processing on the historical sales volume data set to obtain the probability values of each historical sales volume data in the historical sales volume data set; Perform entropy calculation according to each of the probability values to obtain the entropy value of the historical sales volume data set; When the entropy value is within the preset range value, the historical sales volume data set has periodic characteristics.

4. The method according to claim 1, characterized in that, when the historical sales volume data set has the periodic characteristics, calling the autocorrelation function to process the historical sales volume data set to obtain a set of cycle length values includes: Shift the historical sales volume data set in sequence with a preset time length value to obtain each shifted historical sales volume data set; Call the autocorrelation function to perform correlation analysis on the historical sales volume data set and each of the shifted historical sales volume data sets to obtain each correlation value and the corresponding confidence interval value; Obtain a set of cycle length values according to each of the correlation values and the corresponding confidence interval values.

5. The method according to claim 4, characterized in that, the obtaining of the set of cycle length values according to each of the correlation values and the corresponding confidence interval values includes: When the correlation value is greater than the confidence interval value and the correlation value is greater than the correlation values of an adjacent preset number, determine the preset time length value corresponding to the correlation value as the cycle length value, and obtain a set of cycle length values.

6. The method according to claim 1, wherein, the determining the first metric value according to the historical sales volume data set and the residual data set includes: obtaining the residual data with the largest value from the residual data set; determining the first metric value according to the mean value of the historical sales volume data set and the residual data with the largest value.

7. The method according to claim 1, wherein, the evaluating the predictability of the historical sales volume data set according to the first metric value and the second metric value includes: determining a target metric value according to the first metric value and the second metric value; evaluating the predictability of the historical sales volume data set according to the target metric value.

8. The method according to claim 7, wherein, the determining the target metric value according to the first metric value and the second metric value includes any of the following methods: selecting the metric value with the largest value from the first metric value and the second metric value as the target metric value; or taking the weighted average value of the first metric value and the second metric value as the target metric value.

9. An apparatus for analyzing the predictability of item sales data, wherein, the apparatus includes: an acquisition module for acquiring a historical sales volume data set for predicting item sales; a detection module for detecting the periodicity of the historical sales volume data set; a call module for, when the historical sales volume data set has the periodicity feature, calling an autocorrelation function to process the historical sales volume data set to obtain a set of cycle length values; a decomposition module for assigning each cycle length value of the cycle length set to a seasonal factor parameter in a seasonal decomposition function, and performing seasonal decomposition on the historical sales volume data set through the seasonal decomposition function to determine the residual data in the historical sales volume data set, and obtaining a residual data set; a determination module for determining a first metric value according to the historical sales volume data set and the residual data set, where the first metric value is used to characterize the ratio of the value of the unpredictable part in the historical sales volume data set; a prediction module for inputting the historical sales volume data set into each naive model to obtain a set of predicted values corresponding to each naive model; a verification module for verifying each set of predicted values according to a test value set to obtain a set of symmetric mean absolute percentage error values, and determining a second metric value according to the symmetric mean absolute percentage error value with the smallest value in the set of symmetric mean absolute percentage error values; the second metric value is a metric for measuring the predictability of the historical sales volume data set; an evaluation module for evaluating the predictability of the historical sales volume data set according to the first metric value and the second metric value.

10. A computer device, including a memory and a processor, where the memory stores a computer program, wherein, when the processor executes the computer program, the steps of the method according to any one of claims 1 to 8 are implemented.

11. A computer-readable storage medium having a computer program stored thereon, wherein, when the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 8.

Images