Intelligent sales data management platform for enterprise service

Abstract

The invention relates to the technical field of sales data management, in particular to an intelligent sales data management platform for enterprise services. Obtaining suspected association nodes between the sales volume and the influence factors according to the sales volume characteristic time sequence and the influence characteristic time sequence; obtaining an initial association node and association according to the change association characteristics of the suspected association node in the neighbor time period; obtaining an influence coefficient according to the difference characteristics of the sales characteristic time sequence and the influence characteristic time sequence in the neighborhood data change of the initial association node; according to the difference characteristic of the influence coefficient and the influence lag characteristic of the initial association node, obtaining an influence contribution degree; and obtaining a final influence weight according to the relevance and the influence coefficients and the influence contribution degrees of all the initial relevance nodes. According to the invention, prediction is carried out according to the final influence weight, the influence feature time sequence and the sales volume feature time sequence of all the influence factors, the predicted sales volume is obtained, and the accuracy of sales volume prediction is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of sales data management, and in particular to a sales data intelligent management platform for enterprise services. Background Art

[0002] The sales data intelligent management platform for enterprise services is a comprehensive system that integrates multi-dimensional data processing and analysis. The platform can obtain various relevant data sources for sales analysis and forecasting. Through sales analysis and forecasting, the company's operating strategy can be formulated in advance, allowing the company to develop healthily. Traditional sales data analysis and forecasting mainly rely on historical data statistics or simple regression analysis models. Such methods are more effective when the data scale is small and the market environment is relatively stable. However, with the increase of internal and external influencing factors, such as the public evaluation status of the company, commodity price fluctuations and weather changes, sales will fluctuate; changes in various influencing factors have led to low accuracy of sales forecasts under traditional methods, affecting the rationality of the formulation of corporate operating strategies. Summary of the invention

[0003] In order to solve the technical problem that the changes in the above-mentioned multiple influencing factors lead to inaccurate sales forecast results in the traditional way, the purpose of the present invention is to provide a sales data intelligent management platform for enterprise services. The technical solution adopted is as follows: The data acquisition module is used to obtain the time series of product sales history characteristics and the time series of influence characteristics of different sales influencing factors; The sales volume influence judgment module is used to obtain the suspected association node between the sales volume and the influencing factor according to the change trend characteristics of the sales volume feature time series and the influencing feature time series; obtain the initial association node and the correlation between the sales volume and the influencing factor according to the change association characteristics of the sales volume feature time series and the influencing feature time series in the time period adjacent to the suspected association node; The sales volume impact analysis module is used to obtain the impact coefficient of the initial associated node according to the difference characteristics of the sales volume feature time series and the impact feature time series in the neighborhood data of the initial associated node; obtain the impact contribution of the influencing factor at the initial associated node according to the difference characteristics of the impact coefficient and the impact coefficients of other initial associated nodes and the impact lag characteristics of the initial associated node; obtain the final impact weight of the influencing factor according to the correlation, the impact coefficients and the impact contribution of all initial associated nodes; The sales forecasting module is used to make forecasts based on the final impact weights of all influencing factors, the impact feature time series, and the sales feature time series to obtain the predicted sales for future dates.

[0004] Furthermore, the step of obtaining a suspected correlation node between sales volume and influencing factors according to the change trend characteristics of the sales volume feature time series and the influencing feature time series includes: All sales inflection points in the sales feature time series and all impact inflection points in the impact feature time series are obtained through the PELT algorithm; any sales inflection point and the impact inflection point closest to the any sales inflection point are taken as a group of suspected association nodes between the sales volume and the impact factor.

[0005] Furthermore, the step of obtaining the initial associated node and the correlation between the sales volume and the influencing factor according to the change associated features of the sales volume feature time series and the influencing feature time series in the adjacent time period of the suspected associated node includes: Determine whether the data change direction from the sales inflection point to the adjacent sales inflection point in any suspected associated node is the same as the data change direction from the influence inflection point to the adjacent influence inflection point. If they are the same, treat the any suspected associated node as a positively correlated node; if they are not the same, treat the any suspected associated node as a negatively correlated node. Compare the number of positively correlated nodes and the number of negatively correlated nodes. If the number of positively correlated nodes exceeds the number of negatively correlated nodes, the correlation between sales volume and the influencing factor is positively correlated, and the positively correlated nodes are used as the initial correlated nodes. If the number of negatively correlated nodes exceeds the number of positively correlated nodes, the correlation between sales volume and the influencing factor is negatively correlated, and the negatively correlated nodes are used as the initial correlated nodes.

[0006] Furthermore, the step of obtaining the influence coefficient of the initial associated node according to the difference characteristics of the sales feature time series and the influence feature time series in the neighborhood data change of the initial associated node includes: The data segment between the sales inflection point and the adjacent sales inflection point is used as the sales data segment of the sales inflection point, and the absolute value of the difference between any data and the adjacent data in the sales data segment is calculated to obtain the sales difference; the data segment between the influence inflection point and the adjacent influence inflection point is used as the influence feature data segment of the influence inflection point, and the absolute value of the difference between any data and the adjacent data in the influence feature data segment is calculated to obtain the influence difference; the sales data segment and the influence feature data segment are aligned based on the timestamps of the sales inflection point and the influence inflection point in the same initial associated node, and the average value of the ratio of the sales difference to the influence feature difference of all timestamps after alignment is calculated and normalized to obtain the first influence index; Calculate the absolute value of the difference between the average value of the sales feature time series and the sales inflection point to obtain the first difference; calculate the absolute value of the difference between the average value of the impact feature time series and the impact inflection point to obtain the second difference; calculate the ratio of the first difference of the sales inflection point in the initial associated node to the second difference of the impact inflection point and normalize them to obtain the second impact index; calculate the average value of the first impact index and the second impact index to obtain the impact coefficient of the initial associated node.

[0007] Furthermore, the step of obtaining the influence contribution of the influencing factor at the initial associated node according to the difference characteristics between the influence coefficient and the influence coefficients of other initial associated nodes and the influence hysteresis characteristics of the initial associated node includes: The median of the influence coefficients of all initial associated nodes corresponding to the influencing factor is used as the reference coefficient; the time difference between the influence inflection point and the sales inflection point in all initial associated nodes corresponding to the influencing factor is calculated to obtain different lag degrees, and the median of the lag degree is used as the reference degree; the absolute value of the difference between the influence coefficient of the initial associated node and the reference coefficient is calculated and normalized to obtain the influence difference degree; the absolute value of the difference between the lag degree of the initial associated node and the reference degree is calculated and normalized to obtain the lag difference degree; the product of the influence difference degree and the lag difference degree is calculated and negatively correlated to map them to obtain the influence contribution degree of the initial associated node.

[0008] Furthermore, the step of obtaining the final influence weight of the influence factor according to the correlation, the influence coefficient and the influence contribution of all the initial associated nodes includes: Calculate the sum of the products of the influence coefficients and the influence contributions of all initial associated nodes to obtain the sum of weighted influence coefficients; calculate the sum of the influence contributions of all initial associated nodes to obtain the sum of contributions; calculate the ratio of the sum of the weighted influence coefficients to the sum of the contributions to obtain a first value; when the correlation is positively correlated, use the first value as the final influence weight of the influencing factor; when the correlation is negatively correlated, use the opposite of the first value as the final influence weight of the influencing factor.

[0009] Furthermore, the step of performing prediction based on the final impact weights of all impact factors, the impact feature time series and the sales feature time series to obtain the predicted sales volume for a future date includes: All influencing factors are taken as exogenous variables in the ARIMAX model, and the final impact weights corresponding to the influencing factors are taken as the regression coefficients of the exogenous variables; the ARIMAX model is used to make predictions based on the impact characteristic time series and sales characteristic time series of all influencing factors to obtain the predicted sales volume for future dates.

[0010] Furthermore, the time series of the impact characteristics of the different sales influencing factors include: The time series of weather temperature impact characteristics, price impact characteristics and public evaluation score impact characteristics.

[0011] The present invention has the following beneficial effects: In the present invention, obtaining suspected associated nodes can determine the time nodes at which changes in influencing factors may affect sales; obtaining initial associated nodes can determine the time nodes at which influencing factors actually affect sales changes, thereby determining the time range for analyzing the degree of influence of influencing factors on sales; obtaining correlation can determine the positive and negative trends of the influence of influencing factors on sales, thereby preliminarily improving the accuracy of sales forecasts. Obtaining the influence coefficient can determine the degree of influence of influencing factors on sales based on the data change characteristics of influencing factors and sales; since the influence coefficients of influencing factors at different initial associated nodes are different, obtaining the influence contribution can characterize the credibility of different influence coefficients, thereby improving the accuracy of the process of calculating the final influence weight. Obtaining the final influence weight can characterize the true degree of influence of the influencing factor on sales, thereby improving the accuracy of sales forecasts. Finally, according to the final influence weights, influence feature time series and sales feature time series of all influencing factors, forecasts are made to obtain the forecast sales of future dates. Compared with forecasts based only on historical sales data, combining the degree of influence of multi-dimensional influencing factors on sales can improve the accuracy of sales forecasts, allowing enterprises to formulate more reasonable operating strategies. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0013] Figure 1 A block diagram of a sales data intelligent management platform for enterprise services provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0014] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the following is a detailed description of the specific implementation method, structure, features and effects of a sales data intelligent management platform for enterprise services proposed by the present invention in combination with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics in one or more embodiments may be combined in any suitable form.

[0015] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0016] The following is a detailed description of a specific solution of a sales data intelligent management platform for enterprise services provided by the present invention in conjunction with the accompanying drawings.

[0017] See also Figure 1 , which shows a block diagram of a sales data intelligent management platform for enterprise services provided by an embodiment of the present invention, and the block diagram includes the following modules: The data acquisition module S1 is used to obtain the sales feature time series of the product history and the influence feature time series of different sales influencing factors.

[0018] When predicting sales data, it is necessary to first consider the historical trend of sales, so the historical sales feature time series of the product before the current date is obtained, and the sales data of the day is collected once a day. Secondly, the change in sales will be affected by many factors. In the embodiment of the present invention, the analysis product is clothing. The change in sales of clothing is not only affected by the product price, but also by the seasonal temperature. For example, the sales of summer clothing will increase significantly after the temperature rises, and the sales will also decrease when the temperature drops. At the same time, the product sales of the enterprise will be affected by public evaluation factors. When the enterprise has a good reputation in the network, it will have a positive impact on sales; when the enterprise has negative news in the network, it will have a negative impact on sales. Therefore, in the embodiment of the present invention, the influence feature time series of different sales influencing factors includes: the influence feature time series of weather temperature, the influence feature time series of price, and the influence feature time series of public evaluation points; the collection frequency of each influence feature time series is consistent with the collection frequency and collection time range of the sales feature time series. In an embodiment of the present invention, the process of obtaining the public evaluation score is as follows: first, through crawler technology or by retrieving text data related to the company's products from a third party, the text is segmented and preprocessed using a Chinese word segmentation tool, and finally the positive and negative emotional words in the text are matched and scored using a pre-built sentiment dictionary to obtain public evaluation scores for different time periods. The higher the public evaluation score, the better the reputation. It should be noted that in order to avoid the dimensional influence of different features during the analysis process, the sales feature time series and the different impact feature time series are all subjected to data normalization processing; in order to avoid special circumstances causing fluctuations in local data in the sales feature time series and the impact feature time series and affecting the analysis of the overall data change trend, the sales feature time series and the different impact feature time series need to be filtered to filter out the influence of special circumstances and noise data on the overall time series data; the implementer can determine the frequency and time range of time series acquisition according to the implementation scenario, and determine different influencing factors according to the products sold.

[0019] The sales impact judgment module S2 is used to obtain the suspected association nodes between sales and influencing factors based on the change trend characteristics of the sales feature time series and the influence feature time series; and obtain the initial association nodes and correlation between sales and influencing factors based on the change association characteristics of the sales feature time series and the influence feature time series in the adjacent time periods of the suspected association nodes.

[0020] In order to improve the accuracy of sales forecasting, it is necessary to analyze the degree of influence of different influencing factors on sales changes; therefore, it is necessary to first determine the time node of sales changes caused by influencing factors. When the influencing factors affecting sales are relatively stable, sales will not fluctuate significantly. When the influencing factors change, such as a decrease in sales prices, sales will tend to rise in the short term. Therefore, when the data of influencing factors change, sales data will change accordingly. Then, the suspected association nodes between sales and influencing factors are obtained according to the change trend characteristics of the sales feature time series and the influencing feature time series; preferably, in an embodiment of the present invention, the step of obtaining the suspected association nodes includes: obtaining all sales inflection points in the sales feature time series and all inflection points in the influencing feature time series by the PELT algorithm; it should be noted that the PELT algorithm is an efficient change point detection method based on dynamic programming, which is mainly used for the precise positioning of inflection points in time series. The algorithm belongs to the prior art, and the specific steps are not repeated; the sales inflection point is the time node in the sales feature time series that represents a significant change in sales, such as a position where sales significantly rise or fall, and the inflection point is the time node in the influencing feature time series where the influencing factors significantly change, such as a position where the temperature significantly rises or falls. Any sales inflection point and the most recent inflection point before any sales inflection point are taken as a set of suspected association nodes between sales and influencing factors. Since changes in sales have a lag, changes in influencing factors will lead to changes in sales. Therefore, the sales inflection point and the most recent inflection point before it are taken as a set of suspected association nodes between sales and influencing factors.

[0021] Furthermore, since sales are affected by multiple factors, for example, sales increases may be affected by factors such as increased public evaluation scores and temperature changes, the impact of early price changes on sales may be small at this time; and each sales inflection point will obtain a set of suspected associated nodes with the inflection points of different influencing factors, so not every set of suspected associated nodes can truly reflect the correlation characteristics between the influencing factors and sales; therefore, the initial associated nodes and correlations between sales and influencing factors are obtained based on the change correlation characteristics of the sales feature time series and the influencing feature time series in the adjacent time periods of the suspected associated nodes.

[0022] Preferably, in an embodiment of the present invention, the step of obtaining the initial associated node and the correlation includes: determining whether the data change direction from the sales inflection point to the adjacent sales inflection point in any suspected associated node is the same as the data change direction from the impact inflection point to the adjacent impact inflection point. If they are the same, any suspected associated node is regarded as a positively correlated node; if they are not the same, any suspected associated node is regarded as a negatively correlated node. When the data of two sequences are both rising or falling, the data change direction is consistent, which means positive correlation, for example, the public evaluation score rises and the sales volume rises; when the data of two sequences rise and fall respectively, it means negative correlation, for example, the price falls and the sales volume rises. Under normal circumstances, the correlation between the influencing factor and sales volume is fixed. If it is negatively correlated, changes in the influencing factor will only lead to opposite changes in sales volume, such as price decreases and sales volume increases; price increases and sales volume decreases; when all suspected correlation nodes between the influencing factor and sales volume have both positive correlation nodes and negative correlation nodes, it means that a certain type of correlation is caused by other influencing factors. For example, when the price increases, the sales volume increases, and the reason for the increase in sales volume is more likely to be caused by temperature and corporate reputation, indicating that the sales inflection point is not related to the price influencing factor, and the corresponding suspected correlation node is not a real correlation node. Further, compare the number of positive correlation nodes and negative correlation nodes. If the number of positive correlation nodes exceeds the number of negative correlation nodes, the correlation between sales volume and the influencing factor is positively correlated, and the positive correlation node is used as the initial correlation node; the initial correlation node represents that there is a correlation between sales volume and the influencing factor at this node, while the non-initial correlation node means that there is no correlation between the two. If the number of negative correlation nodes exceeds the number of positive correlation nodes, the correlation between sales volume and the influencing factor is negatively correlated, and the negative correlation node is used as the initial correlation node. The correlation between sales volume and influencing factors is used to determine the positivity or negativity of the regression coefficient of the exogenous variable in the final forecast.

[0023] The sales impact analysis module S3 is used to obtain the influence coefficient of the initial associated node based on the difference characteristics of the sales feature time series and the influence feature time series in the neighborhood data of the initial associated node; obtain the influence contribution of the influencing factor at the initial associated node based on the difference characteristics of the influence coefficient and the influence coefficients of other initial associated nodes and the influence lag characteristics of the initial associated node; and obtain the final influence weight of the influencing factor based on the correlation, the influence coefficients and the influence contribution of all initial associated nodes.

[0024] After obtaining the initial association node between sales and each influencing factor, the degree of influence of each change of the influencing factor on sales can be analyzed to improve the accuracy of sales forecast results. When sales change significantly due to changes in influencing factors, it means that the degree of influence of the influencing factor on sales is high; the weight in the forecast is high. When the influencing factor changes significantly and sales only change slightly, it means that the degree of influence of the influencing factor on sales is low, and the weight in the forecast process is low. Therefore, the influence coefficient of the initial association node is obtained based on the difference characteristics of the changes in the sales feature time series and the influencing feature time series in the neighborhood data of the initial association node.

[0025] Preferably, in an embodiment of the present invention, the step of obtaining the influence coefficient includes: taking the data segment from the sales inflection point to the adjacent sales inflection point as the sales data segment of the sales inflection point, calculating the absolute value of the difference between any data and the adjacent data in the sales data segment, and obtaining the sales difference; in an embodiment of the present invention, the adjacent data is the next data of the arbitrary data, and the sales difference represents the difference in sales between two adjacent dates. Taking the data segment from the impact inflection point to the adjacent impact inflection point as the impact feature data segment of the impact inflection point, calculating the absolute value of the difference between any data and the adjacent data in the impact feature data segment, and obtaining the impact difference; the impact difference represents the difference in the numerical value of the influencing factor between two adjacent dates. Aligning the sales data segment and the impact feature data segment based on the timestamps of the sales inflection point and the impact inflection point in the same initial associated node, because the influence of the influencing factor on sales has a lag, the purpose of aligning the sales data segment and the impact feature data segment based on the timestamps of the sales inflection point and the impact inflection point is to analyze the degree of influence of the change of the influencing factor on the change of sales. Calculate the average of the ratio of the sales volume difference to the impact feature difference of all timestamps after alignment and normalize them to obtain the first impact index; when the sales volume difference at the corresponding moment after the timestamp alignment is larger and the impact feature difference is smaller, the first impact index is larger, which means that the impact factor has a greater impact on sales volume, and only a small change in the impact factor can lead to a large change in sales volume. When the first impact index is smaller, it means that the impact factor has a smaller impact on sales volume. It should be noted that the data length of the sales data segment and the impact feature data segment may be different, and the shortest data range of the two is taken to calculate the first impact index.

[0026] Further, the absolute value of the difference between the average value of the sales feature time series and the sales inflection point is calculated to obtain the first difference; the larger the first difference, the greater the difference between the sales inflection point and the average sales. The absolute value of the difference between the average value of the impact feature time series and the impact inflection point is calculated to obtain the second difference; the larger the second difference, the greater the difference between the impact inflection point and the average impact factor value. The ratio of the first difference of the sales inflection point in the initial association node to the second difference of the impact inflection point is calculated and normalized to obtain the second impact index; when the first difference is larger and the second difference is smaller, the larger the second impact index is, which means that the impact factor has a greater impact on sales, and a smaller change in the impact factor can lead to a larger change in sales. Calculate the average of the first impact index and the second impact index to obtain the impact coefficient of the initial association node; when the impact coefficient is larger, it means that at the initial association node, the impact factor has a greater impact on sales. The formula for obtaining the impact coefficient includes: In the formula, R represents the influence coefficient of any initial associated node, represents normalization, M represents the minimum number of data in the sales data segment and the impact feature data segment. After aligning the sales data segment and the impact feature data segment based on the timestamps of the sales inflection point and the impact inflection point, represents the mth sales data, Indicates Sales data, represents the mth influencing factor data, Indicates Influencing factors data, represents the sales volume difference, Indicates the impact difference, represents the first impact index, H represents the first difference, K represents the second difference, Represents the second impact index.

[0027] Under normal circumstances, the degree of influence of each influencing factor on sales is relatively stable, and only a stable degree of influence can improve the accuracy of sales forecasting. After obtaining the influence coefficients of the influencing factors at all initial associated nodes, due to the superposition of multiple influencing factors in the product sales process, the influence coefficients of any sales factor at different initial associated nodes may be different. Due to the existence of other influencing factors, some of the influence coefficients corresponding to the influencing factor are too large or too small. It is necessary to further analyze all the influencing coefficients to determine the specific degree of influence of the influencing factors on sales. Therefore, the influence contribution of the influencing factors at the initial associated nodes is obtained based on the difference characteristics between the influence coefficients and the influence coefficients of other initial associated nodes and the influence lag characteristics of the initial associated nodes.

[0028] Preferably, in an embodiment of the present invention, the step of obtaining the influence contribution includes: taking the median of the influence coefficients of all initial associated nodes corresponding to the influencing factor as a reference coefficient; since other influencing factors may cause the influence coefficient of the influencing factor in some initial associated nodes to be too large or too small, the median of the influence coefficient can better represent the real influence degree of the influencing factor on sales. Calculate the time difference between the influence inflection point and the sales inflection point in all initial associated nodes corresponding to the influencing factor to obtain different lag degrees, and the lag degree represents the time interval when the sales volume changes significantly when the sales are affected by the change of the influencing factor; take the median of the lag degree as a reference degree; since the superposition of other influencing factors may cause the lag degree of the influencing factor in different initial associated nodes to be too large or too small, the median of the lag degree can better represent the real lag degree of the influencing factor on sales. Calculate the absolute value of the difference between the influence coefficient of the initial associated node and the reference coefficient and normalize them to obtain the influence difference degree; when the influence difference degree is smaller, it means that the influence coefficient is closer to the real influence degree, and the contribution degree of the influence coefficient in calculating the comprehensive influence weight is greater. Calculate the absolute value of the difference between the lag degree and the reference degree of the initial association node and normalize them to obtain the lag difference degree; when the lag difference degree is smaller, it means that the sales volume at the initial association node is less affected by other influencing factors, then the credibility of the influence coefficient of the influencing factor at the initial association node is higher, and the contribution degree in calculating the comprehensive influence weight is greater. Calculate the product of the influence difference degree and the lag difference degree and negatively map them to obtain the influence contribution degree of the initial association node; when the influence difference degree and the lag difference degree at the initial association node are both smaller, it means that the sales volume at this location is less affected by other influencing factors, then the influence coefficient of the influencing factor at the initial association node is more accurate, and the influence contribution degree is greater.

[0029] Further, after obtaining the influence coefficient and influence contribution of the influencing factor at all initial associated nodes, the final influence weight of the influencing factor can be obtained according to the correlation, the influence coefficient and influence contribution of all initial associated nodes; preferably, in an embodiment of the present invention, the step of obtaining the final influence weight includes: calculating the sum of the products of the influence coefficients and influence contributions of all initial associated nodes to obtain the sum of weighted influence coefficients; calculating the sum of the influence contributions of all initial associated nodes to obtain the sum of contributions; calculating the ratio of the sum of weighted influence coefficients to the sum of contributions to obtain a first value; the first value represents the weighted average of the influence coefficients, and the larger the first value, the greater the influence of the influencing factor on sales. When the correlation is positively correlated, the sales increase as the value of the influencing factor increases, so the first value is used as the final influence weight of the influencing factor; when the correlation is negatively correlated, the sales decrease as the value of the influencing factor increases, so the opposite of the first value is used as the final influence weight of the influencing factor. The formula for obtaining the final influence weight includes: In the formula, W represents the final impact weight of the influencing factor, represents the sign function, X represents the correlation, when it is positively correlated, Output 1, when it is negatively correlated, Output -1, N represents the number of initial associated nodes corresponding to the influencing factor, Indicates the influence contribution of the nth initial associated node, represents the influence coefficient of the nth initial associated node, represents the total contribution, represents the sum of weighted influence coefficients, Indicates the first value.

[0030] The sales forecasting module S4 is used to make forecasts based on the final impact weights of all impact factors, the impact feature time series and the sales feature time series to obtain the forecasted sales volume for future dates.

[0031] After obtaining the final impact weights of all sales, the sales forecasting model of the product can be constructed. Therefore, the forecast is made according to the final impact weights of all influencing factors, the impact feature time series and the sales feature time series to obtain the forecast sales of the future date; preferably, in the embodiment of the present invention, the step of obtaining the forecast sales of the future date includes: taking all influencing factors as exogenous variables in the ARIMAX model, and taking the final impact weights corresponding to the influencing factors as the regression coefficients of the exogenous variables; it should be noted that the ARIMAX multivariate time series forecasting model belongs to the prior art, and the model is an extension of the ARIMA model. Exogenous variables are added to improve the forecast accuracy. The final impact weights are taken as the regression coefficients of the exogenous variables to determine the degree of influence of the exogenous variables on the sales. According to the impact feature time series and the sales feature time series of all influencing factors, the ARIMAX model is used to forecast and obtain the forecast sales of the future date; the ARIMAX model belongs to the prior art, and the specific forecasting steps are not repeated. Compared with the efficiency forecast based on the historical sales data alone, the forecast accuracy of the sales can be improved by combining the impact degree of multidimensional influencing factors. After obtaining the forecast sales of the future date, the enterprise can formulate a reasonable operation strategy based on the forecast results so that the enterprise can develop healthily.

[0032] In summary, the embodiments of the present invention provide a sales data intelligent management platform for enterprise services; obtain suspected association nodes between sales and influencing factors according to the sales feature time series and the influence feature time series; obtain initial association nodes and correlation according to the change association characteristics of the suspected association node's neighboring time period; obtain the influence coefficient according to the difference characteristics of the sales feature time series and the influence feature time series in the neighborhood data of the initial association node; obtain the influence contribution according to the difference characteristics of the influence coefficient and the influence lag characteristics of the initial association node; obtain the final influence weight according to the correlation, the influence coefficient and the influence contribution of all initial association nodes. The present invention predicts according to the final influence weights of all influencing factors, the influence feature time series and the sales feature time series to obtain the predicted sales volume, thereby improving the accuracy of sales prediction.

[0033] It should be noted that the sequence of the above embodiments of the present invention is only for description and does not represent the advantages and disadvantages of the embodiments. The processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0034] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referenced to each other, and each embodiment focuses on the differences from other embodiments.

Claims

1. A sales data intelligent management platform for enterprise services, characterized in that: The platform includes the following modules: The data acquisition module is used to obtain the time series of product sales history characteristics and the time series of influence characteristics of different sales influencing factors; The sales volume influence judgment module is used to obtain the suspected association node between the sales volume and the influencing factor according to the change trend characteristics of the sales volume feature time series and the influencing feature time series; obtain the initial association node and the correlation between the sales volume and the influencing factor according to the change association characteristics of the sales volume feature time series and the influencing feature time series in the time period adjacent to the suspected association node; The sales volume impact analysis module is used to obtain the impact coefficient of the initial associated node according to the difference characteristics of the sales volume feature time series and the impact feature time series in the neighborhood data of the initial associated node; obtain the impact contribution of the influencing factor at the initial associated node according to the difference characteristics of the impact coefficient and the impact coefficients of other initial associated nodes and the impact lag characteristics of the initial associated node; obtain the final impact weight of the influencing factor according to the correlation, the impact coefficients and the impact contribution of all initial associated nodes; The sales forecasting module is used to make forecasts based on the final impact weights of all influencing factors, the impact feature time series, and the sales feature time series to obtain the predicted sales for future dates.

2. The sales data intelligent management platform for enterprise services according to claim 1, characterized in that: The step of obtaining suspected association nodes between sales volume and influencing factors according to the change trend characteristics of the sales volume feature time series and the influencing feature time series comprises: All sales inflection points in the sales feature time series and all impact inflection points in the impact feature time series are obtained through the PELT algorithm; any sales inflection point and the impact inflection point closest to the any sales inflection point are taken as a group of suspected association nodes between the sales volume and the impact factor.

3. The sales data intelligent management platform for enterprise services according to claim 2, characterized in that: The step of obtaining the initial associated nodes and the association between sales volume and influencing factors according to the change associated features of the sales volume feature time series and the influencing feature time series in the time period adjacent to the suspected associated node comprises: Determine whether the data change direction from the sales inflection point to the adjacent sales inflection point in any suspected associated node is the same as the data change direction from the influence inflection point to the adjacent influence inflection point. If they are the same, treat the any suspected associated node as a positively correlated node; if they are not the same, treat the any suspected associated node as a negatively correlated node. Compare the number of positively correlated nodes and the number of negatively correlated nodes. If the number of positively correlated nodes exceeds the number of negatively correlated nodes, the correlation between sales volume and the influencing factor is positively correlated, and the positively correlated nodes are used as the initial correlated nodes. If the number of negatively correlated nodes exceeds the number of positively correlated nodes, the correlation between sales volume and the influencing factor is negatively correlated, and the negatively correlated nodes are used as the initial correlated nodes.

4. The sales data intelligent management platform for enterprise services according to claim 2, characterized in that: The step of obtaining the influence coefficient of the initial associated node according to the difference characteristics of the sales feature time series and the influence feature time series in the neighborhood data of the initial associated node includes: The data segment between the sales inflection point and the adjacent sales inflection point is used as the sales data segment of the sales inflection point, and the absolute value of the difference between any data and the adjacent data in the sales data segment is calculated to obtain the sales difference; the data segment between the influence inflection point and the adjacent influence inflection point is used as the influence feature data segment of the influence inflection point, and the absolute value of the difference between any data and the adjacent data in the influence feature data segment is calculated to obtain the influence difference; the sales data segment and the influence feature data segment are aligned based on the timestamps of the sales inflection point and the influence inflection point in the same initial associated node, and the average value of the ratio of the sales difference to the influence feature difference of all timestamps after alignment is calculated and normalized to obtain the first influence index; Calculate the absolute value of the difference between the average value of the sales feature time series and the sales inflection point to obtain the first difference; calculate the absolute value of the difference between the average value of the impact feature time series and the impact inflection point to obtain the second difference; calculate the ratio of the first difference of the sales inflection point in the initial associated node to the second difference of the impact inflection point and normalize them to obtain the second impact index; calculate the average value of the first impact index and the second impact index to obtain the impact coefficient of the initial associated node.

5. The sales data intelligent management platform for enterprise services according to claim 2, characterized in that: The step of obtaining the influence contribution of the influence factor on the initial associated node according to the difference characteristics of the influence coefficient and the influence coefficients of other initial associated nodes and the influence hysteresis characteristics of the initial associated node comprises: The median of the influence coefficients of all initial associated nodes corresponding to the influencing factor is used as the reference coefficient; the time difference between the influence inflection point and the sales inflection point in all initial associated nodes corresponding to the influencing factor is calculated to obtain different lag degrees, and the median of the lag degree is used as the reference degree; the absolute value of the difference between the influence coefficient of the initial associated node and the reference coefficient is calculated and normalized to obtain the influence difference degree; the absolute value of the difference between the lag degree of the initial associated node and the reference degree is calculated and normalized to obtain the lag difference degree; the product of the influence difference degree and the lag difference degree is calculated and negatively correlated to map them to obtain the influence contribution degree of the initial associated node.

6. The sales data intelligent management platform for enterprise services according to claim 3, characterized in that: The step of obtaining the final influence weight of the influence factor according to the association, the influence coefficient and the influence contribution of all the initial associated nodes comprises: Calculate the sum of the products of the influence coefficients and the influence contributions of all initial associated nodes to obtain the sum of weighted influence coefficients; calculate the sum of the influence contributions of all initial associated nodes to obtain the sum of contributions; calculate the ratio of the sum of the weighted influence coefficients to the sum of the contributions to obtain a first value; when the correlation is positively correlated, use the first value as the final influence weight of the influencing factor; when the correlation is negatively correlated, use the opposite of the first value as the final influence weight of the influencing factor.

7. The sales data intelligent management platform for enterprise services according to claim 1, characterized in that: The step of performing prediction based on the final impact weights of all impact factors, the impact feature time series, and the sales feature time series to obtain the predicted sales volume for a future date includes: All influencing factors are taken as exogenous variables in the ARIMAX model, and the final impact weights corresponding to the influencing factors are taken as the regression coefficients of the exogenous variables; the ARIMAX model is used to make predictions based on the impact characteristic time series and sales characteristic time series of all influencing factors to obtain the predicted sales volume for future dates.

8. The sales data intelligent management platform for enterprise services according to claim 1, characterized in that: The time series of the impact characteristics of different sales influencing factors include: The time series of weather temperature impact characteristics, price impact characteristics and public evaluation score impact characteristics.

Images