Recommendation degree algorithm for deep processing between short video platform arrival data

Abstract

The invention belongs to the technical field of short video arrival person data processing and analysis, and particularly relates to a recommendation degree algorithm for deep processing between short video platform arrival person data, which comprises the following steps: selecting a target arrival person; sorting and cleaning data distribution of eight groups of target persons and fan ages; adding the arrival person data in the arrival person library, performing comparative analysis on the arrival persons in the arrival person library and the target arrival persons, and comprehensively processing the data between the arrival persons and the target arrival persons so as to obtain the recommendation degree; compared with the prior art, the method has the advantages that according to a result obtained by the method, when a person who can sell a commodity is found for the commodity, the existing popular person of the commodity is taken as a target, and a person similar to the popular person can be evaluated to the recommendation degree, so that the time spent on BD connection establishment of the person is greatly shortened, the capacity of the found person for carrying out the commodity is relatively stable, and the service life of the person is prolonged. And the problem that the sales condition is not stable in the process of manually establishing the connection and searching can be avoided.

Description

Technical Field

[0001] The present invention belongs to the technical field of short video talent data processing and analysis, and in particular to a recommendation algorithm for deep processing of talent data on a short video platform. Background Art

[0002] With the rapid rise and popularity of social media platforms, especially short video platforms like TikTok, a large number of users and content creators have been attracted. These content creators, also known as "influencers", have accumulated a large number of fans and exerted a wide influence on the platform. As a result, companies and brands have begun to seek to promote their products and services through these influencers.

[0003] In this context, the analysis and profiling of influencers has become an important research area. Commercial entities and scientific research institutions are seeking to develop advanced analytical tools and algorithms to better understand the influence of influencers and the characteristics of their fan groups, so as to make product recommendations and marketing activities more accurate and effective. For example, Douyin has provided eight major profiles of user influencers for business use.

[0004] At present, it is a common practice in the industry to analyze user behavior on social media through data mining technology, which includes understanding user interaction patterns, content preferences, active time, etc. However, these methods often focus on general user data rather than specifically targeting expert groups, and the two cannot be well matched together.

[0005] In daily operations, operators are often required to manually match people. The current social media portrait data cannot accurately describe the user portraits of influencers, especially in identifying the characteristics of specific user groups. Social media does not provide a comparison method between influencers and influencers, or between influencers and products. When companies or influencer association BDs find influencers that match specific products in the company's influencer database, they may encounter problems such as low matching efficiency and poor association results, which leads to unsatisfactory marketing results and resource utilization.

[0006] Based on this, this case is brought forward. Summary of the invention

[0007] The purpose of the present invention is to provide a recommendation algorithm for deep processing of expert data on a short video platform to solve the above-mentioned defects.

[0008] In order to achieve the above object, the technical solution of the present invention is as follows:

[0009] A recommendation algorithm for deep processing of data of experts on a short video platform includes the following steps:

[0010] S10. Select target talent;

[0011] S20. Sort and clean the data distribution of the eight major groups and the age of fans of the target talents, and obtain the weighted data groups (d1, d2, d3) of the top three groups among the eight major groups of target talents and the weighted data groups (d1', d2', d3') of the top three groups among the age data of fans of the target talents;

[0012] S30. Add the expert data in the expert database;

[0013] S40. Select three groups of data corresponding to the first three groups of eight groups of experts in any expert database and the target expert, and three groups of data corresponding to the first three groups of age data of fans of the target expert;

[0014] S50. Calculate the initial data difference between the expert in the expert library and the target expert in the selected project;

[0015] S60. Calculate the difference weights between the experts in the expert database and the target experts in the selected items, and obtain the difference weight data set (f1, f2, f3) of the items selected by the eight groups and the difference weight data set (f1', f2', f3') of the items selected by the fans' age data;

[0016] S70. Calculate the reliability and obtain the reliability data set (g1, g2):

[0017] g1=f1*d1+f2*d2+f3*d3;

[0018] g2=f1'*d1'+f2'*d2'+f3'*d3';

[0019] S80. Obtain the recommendation degree by scaling the data density of the reliability data group.

[0020] Furthermore, in step S20, the eight groups of target influencers are sorted and cleaned, including the following steps:

[0021] S21. Select the top three population data of the eight population data of the target expert to obtain data group A;

[0022] S22. Allocate the obtained data group A by percentage to obtain data group B;

[0023] S23. Redistribute data group B according to the coefficient group, wherein the coefficient group is (x, y, z), and x+y+z=1, x≥y≥z, and the coefficient group (x, y, z) is an empirical value, that is, the population data with the highest proportion in data group B is *x, the data with the second highest proportion is *y, and the data with the third highest proportion is *z, to obtain data group C;

[0024] S24. Allocate the data group C by percentage to obtain the weighted data group (d1, d2, d3).

[0025] Furthermore, in step S20, the fan age data of the target influencer is sorted and cleaned, including the following steps:

[0026] S25. Select the top three target influencer fan age data to obtain data set A';

[0027] S26. Allocate the obtained data group A' by percentage to obtain data group B';

[0028] S27. Redistribute the data set B' according to the coefficient group, wherein the coefficient group is (x, y, z), and x+y+z=1, x≥y≥z, and the coefficient group (x, y, z) is an empirical value, that is, the population data with the highest proportion in the data set B' is *x, the data with the second highest proportion is *y, and the data with the third highest proportion is *z, to obtain the data set C';

[0029] S28. Allocate the data group C' by percentage to obtain the weighted data group (d1', d2', d3').

[0030] Furthermore, the coefficient group takes values ​​of (0.5, 0.3, 0.2).

[0031] Furthermore, the recommendation degree = ((g1+1) / 2)^0.5+0.2*(((g2+1) / 2)^0.5).

[0032] Compared with the prior art, the advantages of the present invention are: through the results obtained by the above-mentioned calculation method, when looking for experts who can sell products for the products, the existing hot-selling experts of the products are taken as the target, and the experts similar to the hot-selling experts can be evaluated and recommended, thereby greatly reducing the time spent on expert BD establishment, and the ability of the found experts to bring products is relatively stable, and the problem of unstable expert sales that occurs in the process of manual establishment of links will not occur (the source of the problem is that the standards for manual establishment of links are different each time). BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 It is a flowchart of a recommendation algorithm for deep processing of expert data on a short video platform in an embodiment. DETAILED DESCRIPTION

[0034] The present invention is further described in detail below with reference to the embodiments.

[0035] The present invention is further described in detail below in conjunction with the embodiments. It should be understood that the directions or positional relationships indicated by the terms "upper", "lower", "front", "back", "left", "right", "top", "bottom", "inside" and "outside" herein are based on the directions or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific direction, be constructed and operated in a specific direction, and therefore cannot be understood as limiting the present invention.

[0036] This embodiment proposes a recommendation algorithm for deep processing of short video platform expert data, such as Figure 1 As shown, the following steps are included:

[0037] S10. Select a target expert. The purpose of matching is to find other expert users who are suitable or similar to the target expert based on the target expert;

[0038] S20. Arrange and clean the data distribution of the eight major groups and the age of fans of the target influencer. The arrangement and cleaning process of the eight major groups of data of the target influencer is shown in steps S21-S24, and the arrangement and cleaning process of the age data of the fans of the target influencer is shown in steps S25-S28;

[0039] S21. Select the top three population data of the eight population data of the target expert to obtain data group A;

[0040] S22. Allocate the obtained data group A by percentage to obtain data group B;

[0041] S23. Redistribute data group B according to the coefficient group, wherein the coefficient group is (x, y, z), and x+y+z=1, x≥y≥z, and the coefficient group (x, y, z) is an empirical value, that is, the population data with the highest proportion in data group B is *x, the data with the second highest proportion is *y, and the data with the third highest proportion is *z, to obtain data group C;

[0042] S24. Allocate the proportion of data group C by percentage to obtain the weighted data group (d1, d2, d3);

[0043] S25. Select the top three target influencer fan age data to obtain data set A';

[0044] S26. Allocate the obtained data group A' by percentage to obtain data group B';

[0045] S27. Redistribute the data set B' according to the coefficient group, wherein the coefficient group is (x, y, z), and x+y+z=1, x≥y≥z, and the coefficient group (x, y, z) is an empirical value, that is, the population data with the highest proportion in the data set B' is *x, the data with the second highest proportion is *y, and the data with the third highest proportion is *z, to obtain the data set C';

[0046] S28. Allocate the data group C' by percentage to obtain the weighted data group (d1', d2', d3').

[0047] S30. Add the expert data in the expert database (compare and analyze the experts in the expert database with the target experts and comprehensively process the data between the two to obtain the recommendation degree);

[0048] S40. Select three groups of data corresponding to the first three groups of eight groups of experts in any expert database and the target expert, and three groups of data corresponding to the first three groups of age data of fans of the target expert;

[0049] S50. Calculate the initial data difference between the expert in the expert library and the target expert in the selected project;

[0050] S60. Calculate the difference weights between the experts in the expert database and the target experts in the selected items, and obtain the difference weight data set (f1, f2, f3) of the items selected by the eight groups and the difference weight data set (f1', f2', f3') of the items selected by the fans' age data;

[0051] S70. Calculate the reliability and obtain the reliability data set (g1, g2):

[0052] g1=f1*d1+f2*d2+f3*d3;

[0053] g2=f1'*d1'+f2'*d2'+f3'*d3';

[0054] S80. The recommendation degree is obtained by scaling the data density of the reliability data group, where the recommendation degree = ((g1+1) / 2)^0.5+0.2*(((g2+1) / 2)^0.5).

[0055] The following is a demonstration based on actual cases.

[0056] Table 1 below shows the eight groups of target influencers:

[0057]

[0058]

[0059] Table 1: Data of eight groups of selected target influencers

[0060] The top three population data of the eight target groups were selected, namely 33.5% of the middle-aged and elderly in small towns, 22% of the elderly in cities, and 21.8% of the senior middle class. The rest of the data were discarded, and data group A (33.5%, 22%, 21.8%) was obtained.

[0061] The obtained data group A is distributed according to percentage:

[0062] Middle-aged and elderly people in small towns: 33.5% / (33.5%+22%+21.8%)=0.4334;

[0063] Urban silver-haired: 22% / (33.5%+22%+21.8%)=0.2846;

[0064] Senior middle class: 21.8% / (33.5%+22%+21.8%)=0.2820;

[0065] The data set B (0.4334, 0.2846, 0.2820) was obtained.

[0066] The data set B is redistributed according to the coefficient group, which is (0.5, 0.3, 0.2):

[0067] Middle-aged and elderly people in small towns: 0.4334*0.5=0.2167;

[0068] Urban silver hair: 0.2846*0.3=0.0854;

[0069] Senior middle class: 0.2820*0.2=0.0564;

[0070] The data set C (0.2167, 0.0854, 0.0564) was obtained.

[0071] When doing a portrait analysis for an expert, the top three items in the expert portrait grouping have a deeper degree of priority / weight. For example, the portrait proportion of the first-ranked label is actually much greater than the proportion of the second-ranked label. When used to recommend matching products or other experts, we will redistribute them through coefficient groups to raise or lower their weights. Simply put, if an expert ranks first among young people in small towns at 35%, then our operations will pay more attention to young people in small towns, and we will also pay attention to the second and third ones, but the intensity is not that great. Specific coefficient group values, such as (05, 0.3, 02) in this embodiment, are empirical data obtained through a large amount of data and examples.

[0072] Allocate the percentage of data group C:

[0073] Middle-aged and elderly people in small towns: 0.2167 / (0.2167+0.0854+0.0564)=0.6045;

[0074] Urban silver-haired: 0.0854 / (0.2167+0.0854+0.0564)=0.2382;

[0075] Senior middle class: 0.0564 / (0.2167+0.0854+0.0564)=0.1573;

[0076] The data set D (0.6045, 0.2382, 0.1573) was obtained.

[0077] Table 2 below shows the age data of fans of selected target influencers:

[0078] 31-40 41-50 24-30 18-23 Over 50 years old 36％ 23％ 19％ 5％ 16％

[0079] Table 2: Age data of fans of selected target influencers

[0080] The top three age groups of fans of the target influencer are selected, namely, '31-40' 36%, '41-50' 23%, '24-30' 19%, and the rest are discarded to obtain data group A' (36%, 23%, 19%).

[0081] The obtained data group A' is distributed according to percentage:

[0082] '31-40': 36% / (36%+23%+19%)=0.4615;

[0083] '41-50': 23% / (36%+23%+19%)=0.2949;

[0084] '24-30': 19% / (36%+23%+19%)=0.2436;

[0085] The data set B' (0.4615, 0.2949, 0.2436) was obtained.

[0086] The data set B' is redistributed according to the coefficient group, which is (0.5, 0.3, 0.2):

[0087] '31-40': 0.4615*0.5=0.2308;

[0088] '41-50': 0.2949*0.3=0.0885;

[0089] '24-30': 0.2436*0.2=0.0487;

[0090] The data set C' (0.2308, 0.0885, 0.0487) is obtained;

[0091] Allocate the data group C' by percentage:

[0092] '31-40': 0.2308 / (0.2308+0.0885+0.0487)=0.6272;

[0093] '41-50': 0.0885 / (0.2308+0.0885+0.0487)=0.2404;

[0094] '24-30': 0.0487 / (0.2308+0.0885+0.0487)=0.1324;

[0095] The data set D (0.6272, 0.2404, 0.1324) was obtained.

[0096] Add the expert data in the expert database. Table 3 below shows the eight major groups of data for any expert in the expert database:

[0097]

[0098] Table 3: Data of eight major groups of any talent in the talent database

[0099] Select the data corresponding to the first three groups of the eight target experts, namely, 0.4905 for the middle-aged and elderly in small towns, 0.0232 for the silver-haired in cities, and 0.0086 for the senior middle class, and calculate the initial data difference between the experts in the expert library and the target experts in the selected projects:

[0100] Middle-aged and elderly people in small towns: 0.4905-0.335=0.1555;

[0101] Urban silver hair: 0.0232-0.22 = -0.1968;

[0102] Senior middle class: 0.0086-0.218=-0.2094;

[0103] The data set E (0.1555, -0.1968, -0.2094) was obtained.

[0104] Calculate the difference weights between the experts in the expert library and the target experts in the selected projects, and obtain the difference weight data set of the selected projects of the eight major groups:

[0105] Middle-aged and elderly people in small towns: 0.1555 / 0.335=0.4642;

[0106] Urban silver hair: -0.1968 / 0.22=-0.8945;

[0107] Senior middle class: -0.2094 / 0.218=-0.9606;

[0108] The data set F (0.4642, -0.8945, -0.9606) was obtained.

[0109] Table 4 below shows the age data of fans of any talent in the talent database:

[0110] 31-40 41-50 24-30 18-23 Over 50 years old 42.23％ 21.24％ 19.56％ 8.22％ 8.73％

[0111] Table 4: Age data of fans of any talent in the talent database

[0112] Select the data corresponding to the first three groups in the target expert's fan age data, i.e. '31-40' 0.4223, '41-50' 0.2124, '24-30' 0.1956, and calculate the initial data difference between the expert in the expert library and the target expert in the selected items:

[0113] '31-40': 0.4223-0.36=0.0623;

[0114] '41-50': 0.2124-0.23 = -0.0176;

[0115] '24-30': 0.1956-0.19=0.0056;

[0116] The data set E' (0.0623, -0.0176, 0.0056) was obtained.

[0117] Calculate the difference weights between the experts in the expert library and the target experts in the selected items, and obtain the difference weight data set of the selected items of the fan age data:

[0118] '31-40': 0.0623 / 0.36=0.1731;

[0119] '41-50': -0.0176 / 0.23 = -0.0765;

[0120] '24-30': 0.0056 / 0.19=0.0295;

[0121] The data set F' (0.1731, -0.0765, 0.0295) was obtained.

[0122] Calculation reliability:

[0123] Data of eight major groups = 0.4642*0.6045+-0.8945*0.2382+-0.9606*0.1573=-0.0836;

[0124] Fan age data = 0.1731*0.6272+-0.0765*0.2404+0.0295*0.1324=0.0940;

[0125] The data set G (-0.0836, 0.0940) was obtained.

[0126] The data density of data group G is scaled to obtain the recommended degree:

[0127] Recommendation degree = ((-0.0836+1) / 2)^0.5+0.2*(((0.0940+1) / 2)^0.5) = 0.8248 = 82.48%.

[0128] The above embodiments are only used to explain the concept of the present invention, rather than to limit the protection of the rights of the present invention. Any non-substantial changes to the present invention using this concept shall fall within the protection scope of the present invention.

Claims

1. A recommendation algorithm for deep processing of data between experts on short video platforms, characterized in that: The following steps are involved: S10. Select target talent; S20. Sort and clean the data distribution of the eight major groups and the age of fans of the target talents, and obtain the weighted data groups (d1, d2, d3) of the top three groups among the eight major groups of target talents and the weighted data groups (d1', d2', d3') of the top three groups among the age data of fans of the target talents; S30. Add the expert data in the expert database; S40. Select three groups of data corresponding to the first three groups of eight groups of experts in any expert database and the target expert, and three groups of data corresponding to the first three groups of age data of fans of the target expert; S50. Calculate the initial data difference between the expert in the expert library and the target expert in the selected project; S60. Calculate the difference weights between the experts in the expert database and the target experts in the selected items, and obtain the difference weight data set (f1, f2, f3) of the items selected by the eight groups and the difference weight data set (f1', f2', f3') of the items selected by the fans' age data; S70. Calculate the reliability and obtain the reliability data set (g1, g2): g1=f1*d1+f2*d2+f3*d3; g2=f1'*d1'+f2'*d2'+f3'*d3'; S80. Obtain the recommendation degree by scaling the data density of the reliability data group.

2. A recommendation algorithm for deep processing of short video platform expert data as claimed in claim 1, characterized in that: In step S20, the eight groups of target influencers are sorted and cleaned, including the following steps: S21. Select the top three population data of the eight population data of the target expert to obtain data group A; S22. Allocate the obtained data group A by percentage to obtain data group B; S23. Redistribute data group B according to the coefficient group, wherein the coefficient group is (x, y, z), and x+y+z=1, x≥y≥z, and the coefficient group (x, y, z) is an empirical value, that is, the population data with the highest proportion in data group B is *x, the data with the second highest proportion is *y, and the data with the third highest proportion is *z, to obtain data group C; S24. Allocate the data group C by percentage to obtain the weighted data group (d1, d2, d3).

3. The recommendation algorithm for deep processing of short video platform expert data as claimed in claim 1, characterized in that: In step S20, the fan age data of the target influencer is sorted and cleaned, including the following steps: S25. Select the top three target influencer fan age data to obtain data set A'; S26. Allocate the obtained data group A' by percentage to obtain data group B'; S27. Redistribute the data set B' according to the coefficient group, wherein the coefficient group is (x, y, z), and x+y+z=1, x≥y≥z, and the coefficient group (x, y, z) is an empirical value, that is, the population data with the highest proportion in the data set B' is *x, the data with the second highest proportion is *y, and the data with the third highest proportion is *z, to obtain the data set C'; S28. Allocate the data group C' by percentage to obtain the weighted data group (d1', d2', d3').

4. A recommendation algorithm for deep processing of short video platform expert data as described in claim 2 or 3, characterized in that: The coefficient group takes values ​​of (0.5, 0.3, 0.2).

5. The recommendation algorithm for deep processing of short video platform expert data as claimed in claim 1, characterized in that: The recommendation degree = ((g1+1) / 2)^0.5+0.2*(((g2+1) / 2)^0.5).

Images