A data analysis method for optimizing search engine recommended content

Abstract

The present application relates to the field of electric digital data processing, in particular to a data analysis method for optimizing search engine recommended content, user data, information data and user interaction with information data are collected; a concern degree coefficient is obtained according to the user's stay time, interaction and recommended browsing time of information; an information data vector of each information data is constructed according to the concern degree coefficient of each user to each information data; the push weight of the user to the search material content is obtained according to the relationship between the information data vector of the information data and the search material content and the concern degree coefficient of the user to each information data; the search material content is accurately pushed to each user according to the push weight of the user to the search material content. According to different users and different interactive behaviors, the present application determines the push weight of the search material content to different users, and realizes accurate recommendation of the search material content.

Description

Technical Field

[0001] This application relates to the field of electronic digital data processing, specifically to a data analysis method for optimizing search engine recommended content. Background Technology

[0002] With the continuous advancement of technology, the way information is delivered online has evolved from the simple and direct methods of the past to more intelligent and precise information push. Big data technology, including data collection, storage, processing, and analysis, has become a key tool for achieving precise information push, enabling a better understanding of user behavior and interests, and thus providing highly personalized content and product recommendations based on individual differences.

[0003] Precision recommendation methods refer to using advanced algorithms to recommend highly relevant products or content to users based on their preferences and interests. This is typically based on technologies such as machine learning, data mining, and artificial intelligence to analyze and understand user behavior and provide targeted suggestions. In this way, internet companies can improve user satisfaction, increase user engagement, and boost sales conversion rates.

[0004] Collaborative filtering is an algorithm used in recommendation systems that analyzes user behavior and interests to discover content that users might be interested in. Traditional collaborative filtering algorithms calculate a user's potential interest based on their ratings of content. However, in today's big data environment of the internet age, users are unlikely to spend time rating every piece of data they encounter, and the algorithm doesn't take into account the differences between users. Summary of the Invention

[0005] To address the aforementioned technical problems, this invention provides a data analysis method for optimizing search engine recommended content, thereby resolving existing issues.

[0006] The data analysis method for optimizing search engine recommended content according to the present invention adopts the following technical solution:

[0007] One embodiment of the present invention provides a data analysis method for optimizing search engine recommended content, the method comprising the following steps:

[0008] Collect user interaction data on various information data, including: likes, favorites, and comments;

[0009] For each user's information data, a level of interest coefficient is obtained based on the user's dwell time on the information, likes, favorites, number of comments, and recommended browsing time of the information; an information data vector is constructed for each information data based on the level of interest coefficient of each user; and a similarity is obtained based on the relationship between the information data and the information data vector of the search results.

[0010] Based on the similarity between the information data vectors of each piece of information data and the information data content of the search results, and the user's degree of interest coefficient for each piece of information data, we obtain the user's estimated degree of interest coefficient for the search results content; based on the user's estimated degree of interest coefficient for the search results content and the user's degree of interest coefficient for all information data, we obtain the user's push weight for the search results content.

[0011] The system accurately pushes search results to each user based on the weight they assign to the search results.

[0012] Preferably, the step of obtaining the degree of interest coefficient based on the user's dwell time on the information, likes, favorites, number of comments, and recommended browsing time of the information includes:

[0013] The dwell time ratio is calculated based on the user's dwell time on the information and the recommended browsing time for the information.

[0014] Interaction weight is determined by the number of times a user likes, saves, and comments on a message.

[0015] The product of dwell time ratio and interaction weight is used as the attention coefficient.

[0016] Preferably, the step of obtaining the dwell time ratio based on the user's dwell time on the information and the recommended browsing time of the information includes:

[0017] Calculate the ratio of the user's dwell time on the information to the recommended browsing time of the information. When the ratio is greater than or equal to t, the dwell time ratio is set to 1; when the ratio is less than t, the dwell time ratio is set to the ratio, where t is a preset dwell threshold.

[0018] Preferably, the step of obtaining the interaction weight based on the number of times a user likes, favorites, and comments on information includes:

[0019] Calculate the product of the like operation and the first ratio adjustment parameter, calculate the product of the favorite operation and the second ratio adjustment parameter, and calculate the product of the number of comments and the third ratio adjustment parameter. The sum of the above three products is used as the interaction weight; where the first, second, and third ratio adjustment parameters are preset parameters.

[0020] Preferably, the step of constructing the information data vector for each piece of information data based on the user's level of interest in each piece of information data includes:

[0021] For each piece of information data, the information data vector is composed of the coefficients representing the degree of interest each user has in the information data.

[0022] Preferably, the step of obtaining similarity based on the relationship between information data and information data vectors of search results includes:

[0023] The mean of the information data vector obtained from acquiring information data and the information data content searched;

[0024] The mean of the information data and the search results content is used as the corresponding mean in the calculation of the Pearson coefficient. The Pearson coefficient of the information data vector of the information data and the search results content is calculated.

[0025] Calculate the cosine similarity coefficient between the information data vectors of the information data and the information data vectors of the searched materials;

[0026] The mean of the Pearson coefficient and the cosine similarity coefficient is used as the similarity.

[0027] Preferably, the mean of the information data vectors obtained from the information data and the searched data content includes:

[0028] For information data and search results, calculate the mean of all non-zero elements in the information data vector.

[0029] Preferably, the step of obtaining the user's estimated interest coefficient in the search results based on the similarity between each piece of information data and the information data vector of the search results, as well as the user's interest coefficient in each piece of information data, includes:

[0030] The similarity between each piece of information data and the information data vector of the search content is sorted from largest to smallest to obtain a similarity set. The number of pieces of information data most similar to the search content is obtained based on the distribution of the element values ​​in the similarity set.

[0031] For the first H pieces of information data in the similarity set, where H is the number of pieces of information data most similar to the search content, calculate the product of the similarity between the information data and the information data vector of the search content and the coefficient of user interest in the information data; calculate the sum of the products of all information data.

[0032] Calculate the sum of the similarity between all information data and the information data vectors of the search results;

[0033] The ratio of the sum to the sum of similarities is used as a coefficient to estimate the user's level of interest in the search results.

[0034] Preferably, obtaining the number of information data most similar to the searched content based on the numerical distribution of elements in the similarity set includes:

[0035] Remove elements with values ​​less than or equal to 0 from the similarity set, and calculate the difference in similarity between the next element and the previous element in the similarity set as the derivative of the previous element.

[0036] Starting from the first element in the similarity set, if the derivative of an element is less than the judgment threshold, then the index position of the element in the similarity set is taken as the number of information data that are most similar to the content of the searched material; where the judgment threshold is a preset parameter.

[0037] If no element with a derivative less than a preset threshold is found after iterating through all elements in the similarity set, then the number of elements filtered will be used as the number of information data that is most similar to the content of the searched material; where the number of elements filtered is a preset parameter.

[0038] Preferably, the step of obtaining the user's push weight for the search results based on the user's estimated level of interest in the search results and the user's level of interest in all information data includes:

[0039] Calculate the mean of the coefficients of user interest in all information data, and use the difference between the user's estimated coefficient of interest in the search results and the mean as the push weight for the search results.

[0040] The present invention has at least the following beneficial effects:

[0041] This invention first obtains a coefficient of user concern for information data based on user dwell time and user interaction with information data, effectively evaluating the relationship between users and data;

[0042] Based on the degree of interest that each piece of information data and search results receive from all users, the similarity between each piece of information data and search results is further calculated using cosine similarity and Pearson correlation coefficient. Information data related to search results is then analyzed in detail to increase the acceptability of the pushed content to users.

[0043] By combining the relevance of various information data with the search results, a weighted average method is used to estimate the expected level of interest coefficient of users who have not viewed the search results but may receive the pushed search results, so as to accurately match the pushed search results with the content that users expect to be pushed.

[0044] This invention takes into account the different data and information habits of different users, that is, the differences in interaction behavior between users, and combines the user's own characteristics to determine the push weight of search information content to different users, so as to achieve accurate recommendation of search information content. Attached Figure Description

[0045] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0046] Figure 1 A flowchart of a data analysis method for optimizing search engine recommended content provided by the present invention;

[0047] Figure 2 A diagram illustrating the degree of interest between various user data and various information data;

[0048] Figure 3 For information data A schematic diagram of information data vectors. Detailed Implementation

[0049] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of a data analysis method for optimizing search engine recommendation content proposed according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0050] Unless otherwise defined, 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 pertains.

[0051] The following description, in conjunction with the accompanying drawings, details a specific scheme for a data analysis method for optimizing search engine recommended content provided by the present invention.

[0052] One embodiment of the present invention provides a data analysis method for optimizing search engine recommended content.

[0053] Specifically, the following data analysis method for optimizing search engine recommended content is provided; please refer to [link / reference]. Figure 1 The method includes the following steps:

[0054] Step S001: Collect user interaction data on information data.

[0055] A large amount of user and information data is collected online, and a database is built to record the interaction between user data and information data. The database records each user's interaction with each piece of information data, such as likes, favorites, comments, and browsing time.

[0056] In this embodiment, any data representing a specific meaning, such as a piece of text, a video, or a product, is referred to as information data. Search results are placed in a marketplace and randomly distributed to different users, recording the interaction between users and the search results. This embodiment sets the interaction between users and the psychology course to include likes, favorites, and comments; implementers can set these according to actual needs. Assume the database contains a total of... User data Individual information data.

[0057] Step S002: Based on all available data, define the user's level of interest in the data and the similarity between data, and predict users who may be interested in the search results.

[0058] Traditional collaborative filtering algorithms involve users rating information data. The rating indicates how much a user likes the data. However, in practice, users cannot rate every piece of information they see, which undoubtedly increases their time cost.

[0059] In practice, users express their level of interest in a particular piece of information through actions such as liking, sharing, and commenting. Therefore, we define a user interest coefficient for information; this example uses user i's interest in information j as an example:

[0060]

[0061]

[0062]

[0063] In the formula, Indicates user Information The degree of concern coefficient As the base coefficient, this embodiment sets an empirical value. The implementer can set it up themselves. User Information The length of stay is compared to Indicates user Information Interaction weight;

[0064] formula In this context, the calculation method is as follows: , The time a user spends on this information. For information Recommended browsing time, The dwell time threshold is set at an empirical value of 0.7 in this embodiment, but implementers can set it themselves. The recommended browsing time refers to the time required to fully understand the information. The recommended viewing time is calculated in several ways. For example, the recommended viewing time for a video is its length, measured in minutes. The recommended viewing time for a text message is calculated as the number of words divided by the average reading speed, assuming a reading speed of 200 words per minute. When a user spends more than 70% of the recommended viewing time on the information, it is considered that the user has absorbed the information. However, due to various reasons, users cannot absorb information at a constant speed; video information may be viewed at accelerated speeds, and text information may be skipped or read abruptly.

[0065] formula middle, These are the first, second, and third proportional adjustment parameters, respectively. This embodiment uses the formula... The three proportional adjustment parameters These are set to constants 1, 3, and 5 respectively, and implementers can set them according to their actual situation. On behalf of users Information The "like" action, when a user Information Like, will Set it to 1, otherwise set it to 0. Indicates user Information The "favorite" option is set to 1 if it's "favorite" and 0 otherwise. This indicates the user Information The number of comments.

[0066] Obviously, for users Spending on information The more time spent online The larger the value, the richer the user's interaction with the information, the more frequent the interactions, and the higher the interaction weight. The larger the value, the more likely the user is to be affected. Information The more concerned one is, the higher their level of concern will be. The larger the base coefficient. Used to correct situations where the user has received the information but has not interacted with it. The case where the value is 0. After the user finishes browsing... information A value of 1 indicates that the user is quite interested in the information; otherwise, there is no interaction. It equals 0 at this time. An equal value of 0 is illogical; the base coefficient... This is used to correct the situation.

[0067] To facilitate data calculation, this embodiment uniformly uses the degree of concern coefficient. Scale to between [0, 10]:

[0068]

[0069] In the formula, For users scaling to [0, 10] Information The degree of concern coefficient For users Information The degree of concern coefficient It is an exponential function with the natural constant e as the base.

[0070] For all users and all information data, the correlation coefficient between each user's data and each piece of information data is as follows: Figure 2 As shown:

[0071] Figure 2 The first column of letters Indicates each user, first line of letters This represents various information data. A single user cannot interact with all information data; therefore, a large portion of the data in the above diagram is 0. The information data includes the search results to be pushed to the user.

[0072] Assuming to use Specifically, to precisely target search results, the first step is to find data similar to psychology course materials. If a user cares about data But I haven't seen it yet. can Recommended to this user. The similarity between information data is evaluated using the vectors of their corresponding two information data points. For a given information data point... The information data vector is as follows Figure 3 As shown, the information data vector is an n-dimensional vector.

[0073] If the information data vectors of two pieces of information are similar, it means that the degree of concern of all users for the two pieces of information tends to be consistent. That is, in most cases, different users care about both pieces of information or do not care about them at all, and the content expressed by the two pieces of information is more likely to be similar.

[0074] Define an information data vector for any piece of information data. Information data vectors related to search content Similarity:

[0075]

[0076]

[0077] In the formula, For information data Related search results The similarity between information data vectors, The cosine similarity coefficient represents the similarity between two information data vectors. The Pearson correlation coefficient, cosine similarity coefficient, and Pearson correlation coefficient represent two information data vectors. Both are well-known techniques and will not be elaborated further in this embodiment.

[0078] It should be noted that the mean of the information data vector is required in the calculation of the Pearson correlation coefficient. ,in, The number of non-zero coefficients representing the degree of concern. Information viewed users The average of the concern level coefficients. A concern level coefficient of 0 is not included in the calculation because it indicates the psychology course data has not yet been viewed. The calculation of the average concern level coefficient for other data is similar.

[0079] Both the cosine similarity coefficient and the Pearson correlation coefficient are numbers between -1 and 1. The average of the similarity obtained by the two methods is used as a standard to measure the similarity between information data. The larger the value, the more similar the two information data vectors are, indicating that most users have a similar level of reaction and acceptance towards the two information data contents.

[0080] A search result for content that was not found Users who browse (Right now ), for the content of the search results Estimated level of concern coefficient for:

[0081]

[0082] In the formula, User Search results The estimated level of concern coefficient It is the number of pieces of information most similar to the content of the search query. If H=2, then... Is related to the content of the search results The two information data with the highest similarity between the information data vectors are selected, and in this embodiment, the empirical value H=2 is set, which can be set by the implementer. It is the content of the search. With information and data The similarity of information data vectors, User For information data The similarity coefficient calculated in the previous step is used as a weight to predict the user's level of interest. The degree of interest in the content of the searched information.

[0083] Prerequisite: Provide information to the estimated users. right The degree of concern, user First, you need to browse the information data. .Right now Non-zero.

[0084] Regarding the parameter of the number of most similar numbers Definition: Each data point is related to The data points exhibit similarity, with a value range of [-1, 1]. Positive values ​​indicate a positive correlation, while negative values ​​indicate a negative correlation. Two negatively correlated data points represent different data contents. The greater the similarity, the more similar the data points are.

[0085] Search results (m-1) similarity scores are calculated with other information data. Data with negative similarity scores is related to the search results. Clearly, these are not the same type of data; therefore, remove all similarities less than or equal to 0. Let's assume that at the end... We obtain a similarity set by sorting all similarities in descending order. .

[0086] Treating each similarity value as a scatter point in a two-dimensional plane, the derivative of each data point is obtained using the backward differencing method. If the derivative of a data point changes significantly, it indicates a sharp drop in data similarity, suggesting a substantial difference between subsequent data and that data. A threshold T is set to define the degree of change in the derivative as follows: In this embodiment, an empirical value is used for T. The implementer can set it themselves, that is, to assume that subsequent data and search information content are consistent. The content differs significantly.

[0087] about The selection specifically includes the following two situations:

[0088] 1. When the elements in the similarity set show a slow decreasing trend overall, and all derivatives in the similarity set... The differences are not significant, and there is no difference. For elements that are selected, set H to the filter quantity. , In this embodiment, the empirical value is set to 10. The implementer can set it himself. Since the overall decrease of elements in the set is relatively slow, it indicates that the differences between elements are small. Therefore, a small number of similarities is usually sufficient for the subsequent analysis.

[0089] 2. Starting from the first element in the similarity set, when a similar data point exists... The first data point with a large change in derivative, i.e. Then the set containing that data point The index position in the middle is used as .

[0090] Considering the varying levels of user activity, for example, some users like, save, and comment multiple times on information they care about, while others are more reserved, only liking and rarely commenting on information they care about.

[0091] Therefore, the same level of concern will result in different coefficients for users with different personality traits and behaviors. Thus, it is necessary to consider the overall behavioral interaction of users with all information data to construct a user profile. Search results Push weight :

[0092]

[0093] In the formula, For users Search results Push weight, User Search results The estimated level of concern coefficient For users The mean of the level of concern coefficient, where the mean is calculated by the user The average value of the level of interest in all viewed information data.

[0094] To illustrate this formula: Consider an active user, 'a', who interacts with all the data they've viewed, exhibiting a high level of interest in that data. Conversely, consider an inactive user, 'b', who rarely interacts with the data, resulting in a low level of interest. If the estimated... Greater than This does not necessarily mean that user A agrees with this. More interested because user A showed a high level of interest in almost all data. It is only relatively small compared to user a, and its size is uncertain compared to user b's other data.

[0095] To avoid this situation, we introduce the average level of user interest. This is based on the user's level of interest in the search results. When the attention coefficient of a given data item is greater than the average attention coefficient of other data items, the push weight is adjusted. A positive value indicates that... More interested. User Push weight The larger the value, the more satisfied the user is with the search results. The more interested you are.

[0096] Step S003: Accurately push search results to users based on the user's weight in the search results.

[0097] At this point, for other information data that has been viewed related to the search results but has not been seen, Each user has a push weight calculated using the above method. By arranging all users' push weights from high to low, a user push list can be obtained.

[0098] Set minimum push threshold Search results are presented in order of priority based on user recommendations. Push to all push weights Greater than Users.

[0099] Thus, this embodiment completes the accurate recommendation of search results.

[0100] The embodiments of the present invention first obtain the degree coefficient of user concern for information data based on user dwell time and user interaction with information data, which effectively evaluates the relationship between users and data;

[0101] Based on the degree of interest that each piece of information data and search results receive from all users, the similarity between each piece of information data and search results is further calculated using cosine similarity and Pearson correlation coefficient. Information data related to search results is then analyzed in detail to increase the acceptability of the pushed content to users.

[0102] By combining the relevance of various information data with the search results, a weighted average method is used to estimate the expected level of interest coefficient of users who have not viewed the search results but may receive the pushed search results, so as to accurately match the pushed search results with the content that users expect to be pushed.

[0103] This invention takes into account the different data and information habits of different users, that is, the differences in interaction behavior between users, and combines the user's own characteristics to determine the push weight of search information content for different users, so as to achieve accurate recommendation of search information content.

[0104] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this specification. Additionally, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some embodiments, multitasking and parallel processing are possible or may be advantageous.

[0105] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

[0106] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them; modifications to the technical solutions described in the foregoing embodiments, or equivalent substitutions of some of the technical features, do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. A data analysis method for optimizing content recommendation by a search engine, characterized by, The method comprises the following steps: Collecting interaction data of each user on each information data, the interaction data comprising: a like operation, a collection operation and a comment number; For each user and each information data, obtaining a concern degree coefficient according to a user's stay time on information, a like operation, a collection operation, a comment number and a recommended browsing time of information; constructing an information data vector of each information data according to the concern degree coefficients of each user on each information data; and obtaining similarity according to the relationship between the information data vectors of each information data and the search material content; Obtaining a user's estimated concern degree coefficient for the search material content according to the similarity of the information data vectors of each information data and the search material content and the concern degree coefficients of the user for each information data; and obtaining a push weight of the user for the search material content according to the user's estimated concern degree coefficient for the search material content and the concern degree coefficients of the user for all information data; According to the push weight of the user for the search material content, the search material content is accurately pushed to each user; The construction of the information data vector of each information data according to the concern degree coefficients of each user on each information data comprises: for each information data, the concern degree coefficients of each user on the information data are used to form the information data vector of each information data; The obtaining of the push weight of the user for the search material content according to the user's estimated concern degree coefficient for the search material content and the concern degree coefficients of the user for all information data comprises: calculating the mean value of the concern degree coefficients of the user for all information data, and using the difference between the user's estimated concern degree coefficient for the search material content and the mean value as the push weight of the user for the search material content.

2. The data analysis method for optimizing the content recommendation of a search engine according to claim 1, characterized in that, The obtaining of the concern degree coefficient according to the user's stay time on information, a like operation, a collection operation, a comment number and a recommended browsing time of information comprises: Obtaining a stay time ratio according to the user's stay time on information and the recommended browsing time of information; Obtaining an interaction weight according to the like operation, the collection operation and the comment number of the user on information; The product of the stay time ratio and the interaction weight is used as the concern degree coefficient.

3. The data analysis method for optimizing the content recommendation of a search engine according to claim 2, characterized in that, The obtaining of the stay time ratio according to the user's stay time on information and the recommended browsing time of information comprises: Calculating the ratio of the user's stay time on information to the recommended browsing time of information, when the ratio is greater than or equal to t, the stay time ratio is set to 1; when the ratio is less than t, the stay time ratio is set to the ratio, wherein t is a preset stay threshold.

4. The data analysis method for optimizing the content recommendation of a search engine according to claim 2, wherein, The obtaining of the interaction weight according to the like operation, the collection operation and the comment number of the user on information comprises: The sum of the products of the like operation and a first proportional adjustment parameter, the products of the collection operation and a second proportional adjustment parameter, and the products of the comment number and a third proportional adjustment parameter is used as the interaction weight; wherein the first, second and third proportional adjustment parameters are preset parameters.

5. The data analysis method for optimizing the content recommendation of a search engine according to claim 2, wherein, The obtaining of the similarity according to the relationship between the information data vectors of each information data and the search material content comprises: Obtaining the mean value of the information data vectors of the information data and the search material content; The mean value of the information data and the search material content is taken as a corresponding mean value in a Pearson coefficient calculation process, and the Pearson coefficient of the information data vector of the information data and the search material content is calculated; The cosine similarity coefficient of the information data vector of the information data and the search material content is calculated; The mean value of the Pearson coefficient and the cosine similarity coefficient is taken as the similarity.

6. The data analysis method for optimizing the content recommendation of a search engine according to claim 5, characterized in that, The mean value of the information data vector of the information data and the search material content includes: For the information data vector of the information data and the search material content, the mean value of all non-zero elements in the information data vector is calculated.

7. The data analysis method for optimizing the content recommendation of a search engine according to claim 1, wherein, The estimated attention coefficient of the user to the search material content is obtained according to the similarity of the information data vector of each information data and the search material content and the attention coefficient of the user to each information data, and includes: The similarities of the information data vector of each information data and the search material content are sorted from large to small to obtain a similarity set, and the number of information data most similar to the search material content is obtained according to the element value distribution in the similarity set; For the first H information data in the similarity set, where H is the number of information data most similar to the search material content, the product of the similarity of the information data vector of the information data and the search material content and the attention coefficient of the user to the information data is calculated, and the sum of the products of all information data is calculated; The sum of the similarities of the information data vector of all information data and the search material content is calculated; The ratio of the sum value to the sum of the similarities is taken as the estimated attention coefficient of the user to the search material content.

8. The data analysis method for optimizing the content recommendation of a search engine according to claim 7, characterized in that, The number of information data most similar to the search material content is obtained according to the element value distribution in the similarity set, and includes: The elements in the similarity set whose values are less than or equal to 0 are removed, the similarity difference between the latter element and the former element in the similarity set is calculated as the derivative of the former element; Starting from the first element in the similarity set, if the derivative of the element is less than the judgment threshold, the index position of the element in the similarity set is taken as the number of information data most similar to the search material content; wherein the judgment threshold is a preset parameter; If all elements in the similarity set are traversed and no element with a derivative less than the preset threshold is found, the screening number is taken as the number of information data most similar to the search material content; wherein the screening number is a preset parameter.

Images