Risk prediction method based on multi-mode e-commerce big data and real-time perception

Abstract

The invention relates to a risk prediction method based on multi-modal e-commerce big data and real-time perception, and the method comprises the steps: predetermining a selling cycle as a total time period, and carrying out the equal time interval division in the total time period, and obtaining a plurality of sub time periods; the exposure frequency and the selling number of the target commodity in each time period are obtained, and unified dimension processing is carried out and recorded as lt; i, i, i , i < data > 1 and i < data > 2 < gt >; wherein the idata1 represents the exposure frequency after the dimension is unified, and the idata2 represents the selling quantity after the dimension is unified; subtracting the idata1 from the idata2 to calculate a competitive power index of the target commodity of the ti, and judging whether the competitive power index of the ti is lower than a first threshold value or not, and if yes, performing first risk early warning; performing exposure frequency recessive growth screening on all the sub-time periods in the total time period to obtain screened sub-time periods; and carrying out merging analysis, judging whether a first risk exists, and if yes, carrying out first risk early warning. According to the method, the situation of insufficient competitiveness can be found out as much as possible, and richer data support is provided for related planners.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of e-commerce big data analysis, and particularly relates to a risk prediction method based on multi-modal e-commerce big data and real-time perception. BACKGROUND

[0002] When selling e-commerce goods, various methods such as advertising are often used to expose the goods for sale to increase the number of users browsing the goods, thereby increasing the number of goods sold. Generally speaking, the increase in the number of goods browsed should be proportional to the number of goods sold. If the number of goods browsed increases but the number of goods sold does not keep up, it is equivalent to high input and low output, and the competitiveness of the goods is obviously insufficient. Therefore, the present application provides a risk prediction method to correlate the number of goods browsed and the number of goods sold, and to predict the risk of the competitiveness of the goods in a timely manner to inform relevant planners to adjust the selling strategy in the next cycle. SUMMARY

[0003] The present application aims to at least solve one of the problems of the prior art, and provides a risk prediction method based on multi-modal e-commerce big data and real-time perception.

[0004] To achieve the above-mentioned purpose, the present application adopts the following technical solutions:

[0005] Specifically, the present application provides a risk prediction method based on multi-modal e-commerce big data and real-time perception, which includes the following:

[0006] A total time period is determined as a selling period, and the total time period is divided into a plurality of sub-time periods at equal time intervals, and the sub-time periods are recorded in chronological order as t_i, i∈[1, P], P being the number of sub-time periods;

[0007] The exposure frequency and the number of goods sold of the target goods in each sub-time period t_i are obtained and processed in a unified dimension, and are recorded as <t_i, i_data_1, i_data_2>, wherein i_data_1 represents the exposure frequency after unified dimensioning, and i_data_2 represents the number of goods sold after unified dimensioning;

[0008] The competitiveness index of the target goods in t_i is calculated by subtracting i_data_1 from i_data_2, and it is determined whether the competitiveness index of t_i is lower than a first threshold value. If so, a first risk warning is performed;

[0009] The exposure frequency of all sub-time periods in the total time period is screened for implicit growth to obtain screened sub-time periods;

[0010] The screened sub-time periods are analyzed and combined to determine whether there is a first risk. If so, a first risk warning is also performed.

[0011] Furthermore, specifically, the implicit increase in exposure frequency is filtered across all sub-time periods within the total time period to obtain the filtered sub-time periods, including:

[0012] The exposure frequencies within the total time period are arranged in the order of acquisition time to form the first data sequence {1_data_1, 2_data_1, ..., P_data_1};

[0013] The first data sequence compares adjacent data. When the exposure frequency acquired relatively later is smaller than the exposure frequency acquired relatively earlier, the exposure frequency acquired relatively later is marked as a removed exposure frequency and removed from the first data sequence. Then, the exposure frequency of the next adjacent frequency is compared with the exposure frequency acquired relatively earlier. If the exposure frequency of the next adjacent frequency is smaller than the exposure frequency acquired relatively earlier, the exposure frequency of the next adjacent frequency is marked as a removed exposure frequency and removed from the first data sequence. The above operation is repeated until there is a time segment corresponding to an exposure frequency larger than the exposure frequency acquired relatively earlier. This time segment is then used as the new exposure frequency acquired relatively earlier for analysis.

[0014] Repeat the above filtering method until the first data sequence has been processed, at which point a second data sequence is obtained after removing all exposure frequencies.

[0015] At this point, the number of data points in the second data sequence is denoted as α, and the number of all removed exposure frequencies is denoted as β. A judgment is made: if α / β is greater than the second threshold, it is determined that the exposure frequency in the total time period has a significant upward trend. At this point, the time period corresponding to the data in the second data sequence is recorded as the filtered time period. If α / β is not greater than the second threshold, it is determined that the exposure frequency in the total time period does not have a significant upward trend. At this point, the average value Q of all removed exposure frequencies is calculated, and the removed exposure frequencies higher than Q*γ are added to the second data sequence in the original order to obtain the updated second data sequence. At this point, the time period corresponding to the data in the updated second data sequence is recorded as the filtered time period.

[0016] Where γ is a preset scaling factor set by the user.

[0017] Furthermore, specifically, the filtered time periods are merged and analyzed to determine if the primary risk exists. If so, a primary risk warning is issued, including...

[0018] The average exposure frequency within the filtered time period is recorded as Avg_data_1, and the sales quantity within the filtered time period is recorded as Avg_data_2.

[0019] At this point, the competitiveness index of the target products in the selected time period is calculated by subtracting Avg_data_1 from Avg_data_2, and it is determined whether the competitiveness index of the selected time period is lower than the first threshold. If so, the first risk warning is issued.

[0020] Furthermore, the method also includes,

[0021] By performing a migration analysis on the overall competitiveness index of the previous total time period and the overall competitiveness index of the current total time period, it is determined whether there is a competitiveness fluctuation in the overall competitiveness of the current total time period, and the competitiveness fluctuation analysis results are output.

[0022] Furthermore, specifically, the process of performing a transfer analysis of the overall competitiveness indicators from the previous total time period to the overall competitiveness indicators for the current total time period includes:

[0023] For the previous total time period, calculate the competitiveness index of the target product corresponding to each sub-time period. At this time, with the sub-time period number as the horizontal axis and the competitiveness index as the vertical axis, P first data points will be formed in the two-dimensional plane coordinate system. Then, curve fitting is performed on the P first data points to obtain the competitiveness index curve of the previous total time period.

[0024] For the current total time period, calculate the competitiveness index of the target product for each sub-time period. This will generate P second data points in the second plane coordinate system. Using the competitiveness index curve of the previous total time period as a reference curve, perform spatiotemporal migration analysis on the P second data points as follows:

[0025] Calculate the shortest distance from each of the P second data points to the reference curve, and count the number of points less than the third threshold, denoted as M. If the value of M / P is greater than the fourth threshold, it is determined that there is no fluctuation in the overall competitiveness during the current total time period.

[0026] If the M / P value is not greater than the fourth threshold, it is determined that there is a fluctuation in competitiveness at this time. Then, the average value of the competitiveness index in the current total time period is calculated and recorded as the first value, and the average value of the competitiveness index in the previous total time period is recorded as the second value.

[0027] If the first value is greater than the second value, the overall competitiveness fluctuation trend of the current total time period is determined to be a positive trend; if the first value is not greater than the second value, the overall competitiveness fluctuation trend of the current total time period is determined to be a negative trend.

[0028] Furthermore, the method also includes,

[0029] The risk prediction results are visualized and an analysis report is generated. The relevant data is packaged and sent to the corresponding IP address of the relevant planning personnel.

[0030] This invention also proposes a risk prediction system based on multimodal e-commerce big data and real-time perception, including the following:

[0031] The time period division module is used to predetermine the sales cycle as the total time period, divide the total time period into multiple sub-time periods with equal time intervals, and record them as t_i in chronological order, i∈[1,P], where P is the number of sub-time periods;

[0032] The data acquisition module is used to obtain the exposure frequency and sales quantity of the target product within each time period t_i and perform unified dimension processing, denoted as .<t_i,i_data_1,i_data_2> , where i_data_1 represents the exposure frequency after unification of dimensions, and i_data_2 represents the sales quantity after unification of dimensions;

[0033] The first risk warning module is used to calculate the competitiveness index of the target product t_i by subtracting i_data_1 from i_data_2, and to determine whether the competitiveness index of t_i is lower than the first threshold. If so, the first risk warning is issued.

[0034] The interval filtering module is used to filter all sub-time periods within the total time period by implicitly increasing the exposure frequency, resulting in filtered sub-time periods.

[0035] The merge analysis module is used to merge and analyze the filtered time periods to determine whether there is a primary risk, and if so, to issue a primary risk warning.

[0036] The beneficial effects of this invention are as follows:

[0037] This invention proposes a risk prediction method based on multimodal e-commerce big data and real-time perception. It divides a predetermined sales cycle into multiple time periods and performs correlation analysis on the exposure frequency and sales volume of target products within each time period to determine if there are any instances of insufficient competitiveness. Then, considering the correlation between the exposure frequency and sales volume of target products, it selects some time periods with implicit increases in exposure frequency as typical analysis objects. These typical analysis objects are then combined for competitiveness analysis to identify as many instances of insufficient competitiveness as possible. Furthermore, this invention proposes a method to determine whether there are fluctuations in overall competitiveness in the current time period by performing a migration analysis on the overall competitiveness index of the previous total time period, providing richer data support for relevant planners. Attached Figure Description

[0038] The above and other features of this disclosure will become more apparent from the detailed description of the embodiments illustrated in conjunction with the accompanying drawings. In the accompanying drawings, the same reference numerals denote the same or similar elements. Obviously, the drawings described below are merely some embodiments of this disclosure. For those skilled in the art, other drawings can be obtained from these drawings without any creative effort. In the drawings:

[0039] Figure 1 The flowchart shown is a risk prediction method based on multimodal e-commerce big data and real-time perception according to the present invention. Detailed Implementation

[0040] The following will provide a clear and complete description of the concept, specific structure, and technical effects of the present invention in conjunction with embodiments and accompanying drawings, so as to fully understand the purpose, solution, and effects of the present invention. It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. The same reference numerals used throughout the accompanying drawings indicate the same or similar parts.

[0041] Example 1, referring to Figure 1 This invention proposes a risk prediction method based on multimodal e-commerce big data and real-time perception, including the following:

[0042] Step 110: Predetermine the sales cycle as the total time period, divide the total time period into multiple sub-time periods with equal time intervals, and record them as t_i in chronological order, i∈[1,P], where P is the number of sub-time periods;

[0043] Step 120: Obtain the exposure frequency and sales quantity of the target product within each time period t_i, and perform unified dimension processing, denoted as .<t_i,i_data_1,i_data_2> Where i_data_1 represents the exposure frequency after unification of dimensions, and i_data_2 represents the sales quantity after unification of dimensions; the unification of dimensions operation can establish the correspondence between exposure frequency and sales quantity through machine learning and other methods, unifying the two to a standard under which they can be directly compared in terms of magnitude. This is a relatively mature technology in this field, so it will not be elaborated on.

[0044] Step 130: Calculate the competitiveness index of the target product t_i by subtracting i_data_1 from i_data_2, and determine whether the competitiveness index of t_i is lower than the first threshold. If so, issue the first risk warning.

[0045] Step 140: Perform implicit growth filtering on exposure frequency for all time segments within the total time period to obtain the filtered time segments.

[0046] Step 150: Perform a combined analysis on the filtered time periods to determine if there is a primary risk. If so, issue a primary risk warning.

[0047] In this embodiment 1, the predetermined sales cycle is divided into multiple time periods, and a correlation analysis is performed on the exposure frequency and sales volume of the target product within each time period to determine whether there is a lack of competitiveness. Then, considering the correlation between the exposure frequency and sales volume of the target product, some time periods with implicit growth in exposure frequency are selected as typical analysis objects. Subsequently, a combined competitiveness analysis is performed on the typical analysis objects to identify as many cases of insufficient competitiveness as possible.

[0048] In a preferred embodiment of the present invention, specifically, the exposure frequency implicit growth screening is performed on all sub-time periods within the total time period to obtain the screened sub-time periods, including...

[0049] The exposure frequencies within the total time period are arranged in the order of acquisition time to form the first data sequence {1_data_1, 2_data_1, ..., P_data_1};

[0050] The first data sequence compares adjacent data. When the exposure frequency acquired relatively later is smaller than the exposure frequency acquired relatively earlier, the exposure frequency acquired relatively later is marked as a removed exposure frequency and removed from the first data sequence. Then, the exposure frequency of the next adjacent frequency is compared with the exposure frequency acquired relatively earlier. If the exposure frequency of the next adjacent frequency is smaller than the exposure frequency acquired relatively earlier, the exposure frequency of the next adjacent frequency is marked as a removed exposure frequency and removed from the first data sequence. The above operation is repeated until there is a time segment corresponding to an exposure frequency larger than the exposure frequency acquired relatively earlier. This time segment is then used as the new exposure frequency acquired relatively earlier for analysis.

[0051] Repeat the above filtering method until the first data sequence has been processed, at which point a second data sequence is obtained after removing all exposure frequencies.

[0052] The above process can be illustrated by an example. Suppose the first data sequence is {A,B,C,D,E,F,G}, where B is less than A, so B is marked and removed. C is also less than A, so C is marked and removed. D is greater than A, so it is retained. At this time, D is used as the exposure frequency relative to the previously acquired one. Continue to compare the size relationship between E and D. If E is less than D, D is marked and removed. If F is greater than D, it is retained. At this time, F is used as the exposure frequency relative to the previously acquired one, and so on.

[0053] At this point, the number of data points in the second data sequence is denoted as α, and the number of all removed exposure frequencies is denoted as β. A judgment is made: if α / β is greater than the second threshold, it is determined that the exposure frequency in the total time period has a significant upward trend. At this point, the time period corresponding to the data in the second data sequence is recorded as the filtered time period. If α / β is not greater than the second threshold, it is determined that the exposure frequency in the total time period does not have a significant upward trend. At this point, the average value Q of all removed exposure frequencies is calculated, and the removed exposure frequencies higher than Q*γ are added to the second data sequence in the original order to obtain the updated second data sequence. At this point, the time period corresponding to the data in the updated second data sequence is recorded as the filtered time period.

[0054] Where γ is a preset scaling factor set by the user.

[0055] In this preferred method, considering that even if the exposure frequency shows an upward trend in practical applications, it is impossible for it to show an upward trend in every adjacent time segment, the above method is used to perform implicit analysis on the upward trend of the exposure frequency in the time segment and adjust the content of the second data sequence accordingly to more accurately find the filtered time segments. At this time, the data of the filtered time segments should be more sensitive than that of any individual time segment. By using the filtered time segments to conduct competitive analysis of the target product, the first risk warning event outside the analysis level of any individual time segment can be found.

[0056] In a preferred embodiment of the present invention, specifically, the filtered time periods are merged and analyzed to determine whether a first risk exists. If so, a first risk warning is also issued, including...

[0057] The average exposure frequency within the filtered time period is denoted as Avg_data_1, and the sales quantity within the filtered time period is denoted as Avg_data_2.

[0058] At this point, the competitiveness index of the target products in the selected time period is calculated by subtracting Avg_data_1 from Avg_data_2, and it is determined whether the competitiveness index of the selected time period is lower than the first threshold. If so, the first risk warning is issued.

[0059] In a preferred embodiment of the present invention, the method further includes,

[0060] By performing a migration analysis on the overall competitiveness index of the previous total time period and the overall competitiveness index of the current total time period, it is determined whether there is a competitiveness fluctuation in the overall competitiveness of the current total time period, and the competitiveness fluctuation analysis results are output.

[0061] As a preferred embodiment of the present invention, specifically, the process of performing a migration analysis of the overall competitiveness index of the current total time period on the overall competitiveness index of the previous total time period includes,

[0062] For the previous total time period, calculate the competitiveness index of the target product corresponding to each sub-time period. At this time, with the sub-time period number as the horizontal axis and the competitiveness index as the vertical axis, P first data points will be formed in the two-dimensional plane coordinate system. Then, curve fitting is performed on the P first data points to obtain the competitiveness index curve of the previous total time period.

[0063] For the current total time period, calculate the competitiveness index of the target product for each sub-time period. This will generate P second data points in the second plane coordinate system. Using the competitiveness index curve of the previous total time period as a reference curve, perform spatiotemporal migration analysis on the P second data points as follows:

[0064] Calculate the shortest distance from each of the P second data points to the reference curve, and count the number of points less than the third threshold, denoted as M. If the value of M / P is greater than the fourth threshold, it is determined that there is no fluctuation in the overall competitiveness during the current total time period.

[0065] If the M / P value is not greater than the fourth threshold, it is determined that there is a fluctuation in competitiveness at this time. Then, the average value of the competitiveness index in the current total time period is calculated and recorded as the first value, and the average value of the competitiveness index in the previous total time period is recorded as the second value.

[0066] If the first value is greater than the second value, the overall competitiveness fluctuation trend of the current total time period is determined to be a positive trend; if the first value is not greater than the second value, the overall competitiveness fluctuation trend of the current total time period is determined to be a negative trend.

[0067] In this preferred embodiment, by performing a migration analysis on the overall competitiveness index of the previous total time period and the overall competitiveness index of the current total time period, and using the fitting result of the overall competitiveness of the previous total time period as an approximate standard for the overall competitiveness index of the current total time period, a spatiotemporal migration analysis of the overall competitiveness index of the current total time period can be performed. This can yield a more accurate analysis result of the competitiveness fluctuation trend, thereby providing relevant planners with a practically guiding reference for planning in the next total time period.

[0068] In a preferred embodiment of the present invention, the method further includes,

[0069] The risk prediction results are visualized and an analysis report is generated. The relevant data is packaged and sent to the corresponding IP address of the relevant planning personnel.

[0070] In this preferred embodiment, to facilitate the work of relevant planning personnel, the relevant analysis process is visualized and the relevant data is packaged and sent to the corresponding IP address of the relevant planning personnel.

[0071] Example 2: This invention also proposes a risk prediction system based on multimodal e-commerce big data and real-time perception, including the following:

[0072] The time period division module is used to predetermine the sales cycle as the total time period, divide the total time period into multiple sub-time periods with equal time intervals, and record them as t_i in chronological order, i∈[1,P], where P is the number of sub-time periods;

[0073] The data acquisition module is used to obtain the exposure frequency and sales quantity of the target product within each time period t_i and perform unified dimension processing, denoted as .<t_i,i_data_1,i_data_2> , where i_data_1 represents the exposure frequency after unification of dimensions, and i_data_2 represents the sales quantity after unification of dimensions;

[0074] The first risk warning module is used to calculate the competitiveness index of the target product t_i by subtracting i_data_1 from i_data_2, and to determine whether the competitiveness index of t_i is lower than the first threshold. If so, the first risk warning is issued.

[0075] The interval filtering module is used to filter all sub-time periods within the total time period by implicitly increasing the exposure frequency, resulting in filtered sub-time periods.

[0076] The merge analysis module is used to merge and analyze the filtered time periods to determine whether there is a primary risk, and if so, to issue a primary risk warning.

[0077] Furthermore, the functional modules in the various embodiments of the present invention can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module. The integrated modules described above can be implemented in hardware or as software functional modules.

[0078] If the integrated module is implemented as a software functional module and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the above embodiments of the present invention can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or system capable of carrying the computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc.

[0079] Although the description of the invention has been quite detailed and particularly of several described embodiments, it is not intended to limit it to any of these details or embodiments or any particular embodiment, but should be considered as providing a broad possible interpretation of the claims by referring to the appended claims and taking into account the prior art, thereby effectively covering the intended scope of the invention. Furthermore, the invention has been described above with respect to embodiments foreseeable by the inventors in order to provide a useful description, and non-substantial modifications to the invention that have not yet been foreseen may still represent equivalent modifications.

[0080] The above description is merely a preferred embodiment of the present invention. The present invention is not limited to the above-described embodiments. Any embodiment that achieves the technical effects of the present invention using the same means should fall within the protection scope of the present invention. Within the protection scope of the present invention, various modifications and variations can be made to the technical solutions and / or implementation methods.

Claims

1. A risk prediction method based on multimodal e-commerce big data and real-time perception, characterized in that, Including the following: The sales cycle is predetermined as the total time period. The total time period is divided into multiple sub-time periods by equal time intervals, and they are t_i in chronological order, i∈[1,P], where P is the number of sub-time periods. Obtain the exposure frequency and sales quantity of the target product within each time period t_i, and perform unit-standardization processing, denoted as .<t_i,i_data_1,i_data_2> , where i_data_1 represents the exposure frequency after unification of dimensions, and i_data_2 represents the sales quantity after unification of dimensions; Calculate the competitiveness index of the target product t_i by subtracting i_data_1 from i_data_2, and determine whether the competitiveness index of t_i is lower than the first threshold. If so, issue the first risk warning. The filtered time periods are obtained by performing implicit growth filtering on the exposure frequency of all time periods within the total time period. The selected time periods are merged and analyzed to determine whether the primary risk exists. If so, a primary risk warning is issued.

2. The risk prediction method based on multimodal e-commerce big data and real-time perception according to claim 1, characterized in that, Specifically, the implicit increase in exposure frequency is filtered across all time segments within the total time period to obtain the filtered time segments, including: The exposure frequencies within the total time period are arranged in the order of acquisition time to form the first data sequence {1_data_1, 2_data_1, ..., P_data_1}; The first data sequence compares adjacent data. When the exposure frequency acquired relatively later is smaller than the exposure frequency acquired relatively earlier, the exposure frequency acquired relatively later is marked as a removed exposure frequency and removed from the first data sequence. Then, the exposure frequency of the next adjacent frequency is compared with the exposure frequency acquired relatively earlier. If the exposure frequency of the next adjacent frequency is smaller than the exposure frequency acquired relatively earlier, the exposure frequency of the next adjacent frequency is marked as a removed exposure frequency and removed from the first data sequence. The above operation is repeated until there is a time segment corresponding to an exposure frequency larger than the exposure frequency acquired relatively earlier. This time segment is then used as the new exposure frequency acquired relatively earlier for analysis. Repeat the above filtering method until the first data sequence has been processed, at which point a second data sequence is obtained after removing all exposure frequencies. At this point, the number of data points in the second data sequence is denoted as α, and the number of all removed exposure frequencies is denoted as β. A judgment is made: if α / β is greater than the second threshold, it is determined that the exposure frequency in the total time period has a significant upward trend. At this point, the time period corresponding to the data in the second data sequence is recorded as the filtered time period. If α / β is not greater than the second threshold, it is determined that the exposure frequency in the total time period does not have a significant upward trend. At this point, the average value Q of all removed exposure frequencies is calculated, and the removed exposure frequencies higher than Q*γ are added to the second data sequence in the original order to obtain the updated second data sequence. At this point, the time period corresponding to the data in the updated second data sequence is recorded as the filtered time period. Where γ is a preset scaling factor set by the user.

3. The risk prediction method based on multimodal e-commerce big data and real-time perception according to claim 2, characterized in that, Specifically, the filtered time periods are merged and analyzed to determine if the primary risk exists. If so, a primary risk warning is issued, including... The average exposure frequency within the filtered time period is recorded as Avg_data_1, and the sales quantity within the filtered time period is recorded as Avg_data_2. At this point, the competitiveness index of the target products in the selected time period is calculated by subtracting Avg_data_1 from Avg_data_2, and it is determined whether the competitiveness index of the selected time period is lower than the first threshold. If so, the first risk warning is issued.

4. The risk prediction method based on multimodal e-commerce big data and real-time perception according to claim 1, characterized in that, The method also includes, By performing a migration analysis on the overall competitiveness index of the previous total time period and the overall competitiveness index of the current total time period, it is determined whether there is a competitiveness fluctuation in the overall competitiveness of the current total time period, and the competitiveness fluctuation analysis results are output.

5. The risk prediction method based on multimodal e-commerce big data and real-time perception according to claim 4, characterized in that, Specifically, the process of performing a transfer analysis of the overall competitiveness indicators from the previous total time period to the overall competitiveness indicators for the current total time period includes: For the previous total time period, calculate the competitiveness index of the target product corresponding to each sub-time period. At this time, with the sub-time period number as the horizontal axis and the competitiveness index as the vertical axis, P first data points will be formed in the two-dimensional plane coordinate system. Then, curve fitting is performed on the P first data points to obtain the competitiveness index curve of the previous total time period. For the current total time period, calculate the competitiveness index of the target product for each sub-time period. This will generate P second data points in the second plane coordinate system. Using the competitiveness index curve of the previous total time period as a reference curve, perform spatiotemporal migration analysis on the P second data points as follows: Calculate the shortest distance from each of the P second data points to the reference curve, and count the number of points less than the third threshold, denoted as M. If the value of M / P is greater than the fourth threshold, it is determined that there is no fluctuation in the overall competitiveness during the current total time period. If the M / P value is not greater than the fourth threshold, it is determined that there is a fluctuation in competitiveness at this time. Then, the average value of the competitiveness index in the current total time period is calculated and recorded as the first value, and the average value of the competitiveness index in the previous total time period is recorded as the second value. If the first value is greater than the second value, the overall competitiveness fluctuation trend of the current total time period is determined to be a positive trend; if the first value is not greater than the second value, the overall competitiveness fluctuation trend of the current total time period is determined to be a negative trend.

6. The risk prediction method based on multimodal e-commerce big data and real-time perception according to any one of claims 1-5, characterized in that, The method also includes, The risk prediction results are visualized and an analysis report is generated. The relevant data is packaged and sent to the corresponding IP address of the relevant planning personnel.

7. A risk prediction system based on multimodal e-commerce big data and real-time perception, characterized in that: Including the following: The time period division module is used to predetermine the sales cycle as the total time period, divide the total time period into multiple sub-time periods with equal time intervals, and record them as t_i in chronological order, i∈[1,P], where P is the number of sub-time periods; The data acquisition module is used to obtain the exposure frequency and sales quantity of the target product within each time period t_i and perform unified dimension processing, denoted as .<t_i,i_data_1,i_data_2> , where i_data_1 represents the exposure frequency after unification of dimensions, and i_data_2 represents the sales quantity after unification of dimensions; The first risk warning module is used to calculate the competitiveness index of the target product t_i by subtracting i_data_1 from i_data_2, and to determine whether the competitiveness index of t_i is lower than the first threshold. If so, the first risk warning is issued. The interval filtering module is used to filter all sub-time periods within the total time period by implicitly increasing the exposure frequency, resulting in filtered sub-time periods. The merge analysis module is used to merge and analyze the filtered time periods to determine whether there is a primary risk, and if so, to issue a primary risk warning.

Images