Hot search information pushing method based on big data

Abstract

The invention relates to the technical field of hot search pushing, and discloses a hot search information pushing method based on big data, which screens user effective behavior data through a multi-dimensional data cleaning strategy, constructs a user interest weight model in combination with a time attenuation factor, and dynamically balances the diversity and correlation of pushed contents by using information entropy. The method specifically comprises the following steps: cleaning historical browsing data of a user, removing invalid behaviors, and generating a second heat collection set; calculating interest weights of the user for different types of hot search based on the time decay factor, and optimizing weight distribution through normalization processing; pushing strategies are designed for the new users and the old users respectively, the new users adopt global information entropy to screen high-diversity content, the old users dynamically adjust a pushing list in combination with weighted information entropy, it is ensured that the content conforms to interests, and information cocoon rooms are avoided.

Description

Technical Field

[0001] The present invention relates to the technical field of hot search pushing, and specifically provides a method for pushing hot search information based on big data. Background Art

[0002] With the explosive growth of Internet information, the hot search pushing system has become an important way for users to obtain hot information. Traditional pushing methods mainly rely on users' historical behaviors (such as clicks and browsing durations) to build interest models, but there are the following problems:

[0003] Invalid browsing data is not effectively filtered, resulting in bias in the interest model. For example, content that a user accidentally clicks but does not read in depth may be misjudged as an interest point. Over-reliance on historical behaviors leads to homogenization of pushed content. Users are exposed to information in a single field for a long time, which restricts their horizons and exacerbates group polarization. Existing technologies do not introduce a time decay factor and cannot reflect the dynamic changes of users' interests. For example, a user's recent interest in technology content may be much higher than their historical preference for entertainment. When there is a lack of sufficient behavioral data, the pushing strategy is single, usually filling with popular lists, which cannot meet personalized needs. In existing technologies, information entropy is mostly used to evaluate the global content diversity, but it is not combined with users' interests. Existing technologies use information entropy to screen high-diversity content, but do not consider users' personalized needs during pushing, resulting in an imbalance between content relevance and diversity. In addition, the strategies for new and old users are not differentiated: the pushing for new users lacks diversity guidance, while the pushing for old users may fall into an information cocoon due to excessive personalization.

[0004] In summary, there is an urgent need for a method for pushing hot search information based on big data to improve the pushing accuracy and diversity and enhance user stickiness. Summary of the Invention

[0005] The present invention provides a method for pushing hot search information based on big data, which helps to solve the problems mentioned in the above background art.

[0006] The present invention provides the following technical solution: A method for pushing hot search information based on big data, specifically including:

[0007] Manually classify the hot search information, and divide all hot search information into different types;

[0008] The types include: politics, economy, technology, entertainment, and sports;

[0009] Obtain the users for whom hot search information is to be pushed, denoted as target users;

[0010] Obtain all the hot search information browsed by the target users historically to form a first hot search set;

[0011] Determine whether the browsing of the hot search information in the first hot search set by the target user is effective according to the data cleaning strategy, specifically including:

[0012] Arbitrarily select a piece of hot search information from the first hot search set and record it as the target hot search;

[0013] Obtain the number of times the target user browses the target hot search, and record it as the browsing times;

[0014] If the browsing times = 1, then obtain the browsing duration of the target user for the target hot search, and record it as the browsing duration;

[0015] Set the first duration threshold and the second duration threshold;

[0016] The first duration threshold and the second duration threshold are used to determine whether the browsing of the target hot search by the target user is effective; among them, the second duration threshold is greater than the first duration threshold;

[0017] If the browsing duration ≤ the first duration threshold;

[0018] Then it is determined that the browsing of the target hot search by the target user is ineffective, and at the same time, the target hot search is removed from the first hot search set;

[0019] If the first duration threshold < browsing duration < the second duration threshold;

[0020] Then it is determined that the browsing of the target hot search by the target user is effective, and at the same time, the target hot search is retained in the first hot search set;

[0021] If the browsing duration ≥ the second duration threshold, then obtain the interaction depth of the target user for the target hot search;

[0022] The interaction depth specifically includes 0 and 1;

[0023] If the target user has the behavior of liking, commenting or sharing during the browsing of the target hot search, then it is determined that the interaction depth of the target user for the target hot search is 1;

[0024] If the target user does not have the behavior of liking, commenting or sharing during the browsing of the target hot search, then it is determined that the interaction depth of the target user for the target hot search is 0;

[0025] If the interaction depth of the target user for the target hot search is 1, then it is determined that the browsing of the target hot search by the target user is effective, and at the same time, the target hot search is retained in the first hot search set;

[0026] If the interaction depth of the target user for the target hot search is 0, then it is determined that the browsing of the target hot search by the target user is ineffective, and at the same time, the target hot search is removed from the first hot search set.

[0027] Optionally, determining whether the browsing of the hot search information in the first hot search set by the target user is effective according to the data cleaning strategy specifically includes:

[0028] If the number of views ≠ 1;

[0029] Obtain the time interval between the end time of the target user's first view of the target hot search and the start time of the second view of the target hot search, denoted as the first time interval, obtain the time interval between the end time of the target user's second view of the target hot search and the start time of the third view of the target hot search, denoted as the second time interval... Obtain the time interval between the end time of the target user's penultimate view of the target hot search and the start time of the last view of the target hot search, denoted as the A-th time interval;

[0030] Where A = the number of views - 1;

[0031] Calculate (the first time interval + the second time interval +... + the A-th time interval) ÷ A, and denote the result as the average time interval;

[0032] Set a time interval threshold, which is used to determine whether the browsing of the target hot search by the target user is effective;

[0033] If the average time interval ≥ the time interval threshold, it is determined that the browsing of the target hot search by the target user is effective, and at the same time, the target hot search is retained in the first hot search set;

[0034] If the average time interval < the time interval threshold, it is determined that the browsing of the target hot search by the target user is ineffective, and at the same time, the target hot search is removed from the first hot search set;

[0035] Judge whether the browsing of all hot search information in the first hot search set is effective through the above method;

[0036] Obtain all the hot search information retained in the first hot search set to form a second hot search set.

[0037] Optionally, it further includes:

[0038] Calculate the interest weight of the target user for each type of hot search information, specifically including:

[0039] The hot search information of the political type in the second hot search set constitutes the political type set, the hot search information of the economic type in the second hot search set constitutes the economic type set, the hot search information of the technology type in the second hot search set constitutes the technology type set, the hot search information of the entertainment type in the second hot search set constitutes the entertainment type set, and the hot search information of the sports type in the second hot search set constitutes the sports type set;

[0040] Obtain the moments when the target user browses the hot searches in each different type of collection respectively, and calculate the time difference from the current time;

[0041] Set a time decay factor α, which is used to weight the historical browsing data of the target user during the process of calculating the interest weight of the target user for each type of hot search information to achieve the effect of paying more attention to recent behaviors;

[0042] Calculate the interest weight of the target user for each type of hot search information through the following formula:

[0043]

[0044] Where:

[0045] w i is the interest weight of the target user for the i-th type of hot search information; when i = 1, it represents the interest weight for political type hot search information, when i = 2, it represents the interest weight for economic type hot search information, when i = 3, it represents the interest weight for science and technology type hot search information, when i = 4, it represents the interest weight for entertainment type hot search information, and when i = 5, it represents the interest weight for sports type hot search information;

[0046] t ij is the time difference from the current time of the j-th browsing behavior of the target user for the i-th type of hot search information;

[0047] α is the time decay factor;

[0048] b ij is the contribution value of the j-th browsing behavior of the target user to the i-th type of hot search information; if the target user only browses, the contribution value is recorded as 1, if the target user likes the content during browsing, the contribution value is recorded as 2, if the target user comments during browsing, the contribution value is recorded as 3, if the target user shares the content during browsing, the contribution value is recorded as 4; if the user has multiple behaviors during browsing, select the contribution value corresponding to the behavior with the largest contribution value as the contribution value;

[0049] Calculate the interest weights w1, w2, w3, w4 and w5 through the above method.

[0050] Optionally, calculating the interest weight of the target user for each type of hot search information specifically includes:

[0051] Normalize the interest weight of the target user so that the sum is 1. Specifically, normalize the interest weight of the target user through the following formula;

[0052]

[0053] Obtain \(w1'\), \(w2'\), \(w3'\), \(w4'\) and \(w5'\) through the above calculation method.

[0054] Optionally, it further includes:

[0055] Set a first quantity threshold, which is used to determine whether the target user is a new user;

[0056] Obtain the quantity of hot search information in the second hot search set, denoted as the judgment quantity;

[0057] If the judgment quantity ≤ the first quantity threshold, it is determined that the target user is a new user;

[0058] If the judgment quantity > the first quantity threshold, it is determined that the target user is an old user;

[0059] If the target user is a new user, push hot search information to the target user according to the first strategy;

[0060] If the target user is an old user, push hot search information to the target user according to the second strategy.

[0061] Optionally, when the target user is a new user, pushing hot search information to the target user according to the first strategy specifically includes:

[0062] Obtain all current hot search information to form a hot search information pool;

[0063] Set the push quantity, the first information entropy, and the second information entropy;

[0064] The push quantity, the first information entropy, and the second information entropy are used to select hot search information for pushing to the target user from the hot search information pool, where the first information entropy < the second information entropy;

[0065] Arbitrarily select the push quantity of hot search information from the hot search information pool to form a target push set;

[0066] Classify the hot search information in the target push set according to its type, and obtain the proportion of each type of hot search information to all hot search information in the target push set, denoted as \(P(x1)\), \(P(x2)\), \(P(x3)\), \(P(x4)\) and \(P(x5)\) respectively;

[0067] Calculate the information entropy of the hot search information in the target push set through the following formula:

[0068]

[0069] If the first information entropy < \(H(ω)\) < the second information entropy, then push the hot search information in the target push set to the target user;

[0070] If \(H(\omega)\leq\) the first information entropy or \(H(\omega)\geq\) the second information entropy, then randomly select the number of hot search information to be pushed from the hot search information pool to form a target push set until the first information entropy \(<\) the information entropy of the hot search information in the target push set \(<\) the second information entropy.

[0071] Optionally, if the target user is an old user, then push hot search information to the target user according to the second strategy, which specifically includes:

[0072] Obtain all current hot search information to form a hot search information pool;

[0073] Set the number of pushes, the first information entropy, and the second information entropy;

[0074] The number of pushes, the first information entropy, and the second information entropy are used to select hot search information for pushing hot search information to the target user from the hot search information pool, where the first information entropy \(<\) the second information entropy;

[0075] Randomly select the number of hot search information to be pushed from the hot search information pool to form a target push set;

[0076] Classify the hot search information in the target push set according to its type, and obtain the ratio of each type of hot search information to all hot search information in the target push set, denoted as \(P(x1)\), \(P(x2)\), \(P(x3)\), \(P(x4)\), and \(P(x5)\) respectively;

[0077] Calculate the weighted information entropy of the hot search information in the target push set through the following formula:

[0078]

[0079] If the first information entropy \(< H(\omega)'\ <\) the second information entropy, then push the hot search information in the target push set to the target user;

[0080] If \(H(\omega)'\leq\) the first information entropy or \(H(\omega)'\geq\) the second information entropy, then randomly select the number of hot search information to be pushed from the hot search information pool to form a target push set until the first information entropy \(<\) the weighted information entropy of the hot search information in the target push set \(<\) the second information entropy.

[0081] The present invention has the following beneficial effects:

[0082] 1. This method for pushing hot search information based on big data classifies hot search information manually to ensure that different types of hot search information can accurately correspond to the needs of target users. For example, classifications such as politics, economy, technology, entertainment, and sports enable the system to push precise content according to the interests of different users, improving the relevance of the push. The push of classified information avoids information overload, and users can receive updated hot search information within their areas of interest, thus enhancing the user experience. Personalized push after classification can provide more data support for the optimization of subsequent algorithms, making the hot search recommendation system more intelligent.

[0083] 2. This method for pushing hot search information based on big data determines the attention of users to hot search information by judging the browsing duration and interaction depth, and can clearly distinguish whether the interest of users in certain information is real. It effectively removes invalid information, making the pushed content more in line with user needs. For hot search information with a browsing count of 1, by setting a first duration threshold and a second duration threshold and comparing them with the browsing duration of the target user, the interference of spam information can be effectively reduced. If the browsing duration ≤ the first duration threshold, it indicates that the browsing duration of the target user for this hot search information is very short, which may be a short browsing caused by accidental touch, so it is determined that the browsing of the target user for this target hot search is invalid. If the browsing duration ≥ the second duration threshold, at this time, the interaction depth of the target user for the target hot search is combined to judge whether the browsing of the target user for the target hot search is valid. If the interaction depth is 0, it means that although the user has browsed this hot search information for a long time, no interaction behavior has occurred, which may be the situation of staying on this hot search information due to accidental touch, so it is determined that the browsing of the target user for this target hot search is invalid; in addition, for hot search information with a browsing count not equal to 1, by setting a time interval threshold and obtaining the average time interval, it is judged whether the browsing of the target user for the target hot search is valid; if the average time interval < the time interval threshold, it means that the target user has browsed the same hot search information multiple times in a short period, with the suspicion of swiping hot searches, so it is determined that the browsing of the target user for this target hot search is invalid; in this way, effective and invalid behaviors can be accurately distinguished, providing an accurate data basis for the subsequent push of hot search information.

[0084] 3. This method for pushing hot search information based on big data can deeply explore user preferences by calculating the interest weights of target users for different types of hot search information, and then push more personalized hot search information, enhancing the acceptance of users for the pushed content. The calculation of interest weights is not only based on the historical behavior of users, but also combines a time decay factor, making the recommended content more in line with the current immediate interests of users, rather than simply relying on historical data. By using the time decay factor, the dynamic changes of user interests can be updated in a timely manner, especially when user interests change, the system can quickly adjust the recommendation strategy.

[0085] 4. The method for pushing hot search information based on big data determines whether the target user is a new user or an old user and adopts different pushing strategies. If the judged quantity ≤ the first quantity threshold, it indicates that the browsing records of the target user are few, and at this time, the target user is recognized as a new user. If the judged quantity > the first quantity threshold, it indicates that the browsing records of the target user are many, and then the target user is recognized as an old user. For new users, randomly select the pushing quantity of hot search information from the hot search information pool, and calculate the information entropy of the selected hot search information. The larger the information entropy, the more extensive the types of the selected hot search information are, and the smaller the information entropy, the more narrow the types of the selected hot search information are. Ensure that the information entropy of the selected hot search information is between the first information entropy and the second information entropy, which can ensure that the hot search information pushed to the target user is neither too extensive nor too narrow, avoiding the existence of information cocoons. For old users, randomly select the pushing quantity of hot search information from the hot search information pool, and calculate the weighted information entropy of the selected hot search information by combining the interest weights of the target user for each type of hot search information. Ensure that the weighted information entropy of the selected hot search information is between the first information entropy and the second information entropy, which can ensure that the hot search information pushed to the target user not only conforms to the interests of the target user but also avoids information cocoons. BRIEF DESCRIPTION OF THE DRAWINGS

[0086] Figure 1 It is a schematic flowchart of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0087] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0088] Embodiment, referring to Figure 1 , a method for pushing hot search information based on big data, specifically including:

[0089] Manually classify the hot search information, and divide all hot search information into different types;

[0090] The types include: politics, economy, technology, entertainment, and sports;

[0091] The method for pushing hot search information based on big data classifies hot search information manually to ensure that different types of hot search information can accurately correspond to the needs of target users. For example, classifications such as politics, economy, technology, entertainment, and sports enable the system to push accurate content according to the interests of different users, improving the relevance of the push. The classified information push avoids information overload, and users can receive updated hot search information within their areas of interest, thereby enhancing the user experience. Performing personalized push after classification can provide more data support for the optimization of subsequent algorithms, making the hot search recommendation system more intelligent;

[0092] Obtain the user who will receive the pushed hot search information, denoted as the target user;

[0093] Obtain all the hot search information browsed by the target user historically to form a first hot search set;

[0094] Judge whether the browsing of the hot search information in the first hot search set by the target user is effective according to the data cleaning strategy, specifically including:

[0095] Arbitrarily select a piece of hot search information from the first hot search set, denoted as the target hot search;

[0096] Obtain the number of times the target user browses the target hot search, denoted as the browsing times;

[0097] If the browsing times = 1, then obtain the browsing duration of the target user for the target hot search, denoted as the browsing duration;

[0098] Set a first duration threshold and a second duration threshold;

[0099] The first duration threshold and the second duration threshold are used to judge whether the browsing of the target hot search by the target user is effective; among them, the second duration threshold is greater than the first duration threshold;

[0100] If the browsing duration ≤ the first duration threshold;

[0101] Then it is determined that the browsing of the target hot search by the target user is invalid, and at the same time, the target hot search is removed from the first hot search set;

[0102] If the first duration threshold < the browsing duration < the second duration threshold;

[0103] Then it is determined that the browsing of the target hot search by the target user is effective, and at the same time, the target hot search is retained in the first hot search set;

[0104] If the browsing duration ≥ the second duration threshold, then obtain the interaction depth of the target user for the target hot search;

[0105] The interaction depth specifically includes 0 and 1;

[0106] If the target user likes, comments on, or shares the target hot search during the browsing process of the target hot search, it is determined that the interaction depth of the target user with the target hot search is 1;

[0107] If the target user does not like, comment on, or share the target hot search during the browsing process of the target hot search, it is determined that the interaction depth of the target user with the target hot search is 0;

[0108] If the interaction depth of the target user with the target hot search is 1, it is determined that the browsing of the target user for the target hot search is effective, and at the same time, the target hot search is retained in the first hot search set;

[0109] If the interaction depth of the target user with the target hot search is 0, it is determined that the browsing of the target user for the target hot search is ineffective, and at the same time, the target hot search is removed from the first hot search set;

[0110] This method for pushing hot search information based on big data determines the attention of users to hot search information by judging the browsing duration and interaction depth, and can clearly distinguish whether the interest of users in certain information is real. It effectively removes invalid information, making the pushed content more in line with user needs. For hot search information with a browsing count of 1, by setting the first duration threshold and the second duration threshold and comparing them with the browsing duration of the target user, the interference of spam information can be effectively reduced. If the browsing duration ≤ the first duration threshold, it means that the browsing duration of the target user for this hot search information is very short, which may be a short browsing caused by accidental touch. Therefore, it is determined that the browsing of the target user for this target hot search is ineffective. If the browsing duration ≥ the second duration threshold, at this time, the interaction depth of the target user with the target hot search is combined to judge whether the browsing of the target user for the target hot search is effective. If the interaction depth is 0, it means that although the user has browsed this hot search information for a long time, no interaction behavior has occurred, which may be the situation of staying on this hot search information due to accidental touch. Therefore, it is determined that the browsing of the target user for this target hot search is ineffective.

[0111] The determination of whether the browsing of the target user for the hot search information in the first hot search set is effective according to the data cleaning strategy specifically includes:

[0112] If the browsing count ≠ 1;

[0113] Obtain the time interval between the end moment of the target user's first browsing of the target hot search and the start moment of the second browsing of the target hot search, denoted as the first time interval, obtain the time interval between the end moment of the target user's second browsing of the target hot search and the start moment of the third browsing of the target hot search, denoted as the second time interval... Obtain the time interval between the end moment of the target user's penultimate browsing of the target hot search and the start moment of the last browsing of the target hot search, denoted as the A-th time interval;

[0114] where A = browsing count - 1;

[0115] Calculate (the first time interval + the second time interval +... + the A-th time interval) ÷ A, and record the result as the average time interval;

[0116] Set a time interval threshold, which is used to determine whether the browsing of the target hot search by the target user is effective;

[0117] If the average time interval ≥ the time interval threshold, it is determined that the browsing of the target hot search by the target user is effective, and at the same time, the target hot search is retained in the first hot search set;

[0118] If the average time interval < the time interval threshold, it is determined that the browsing of the target hot search by the target user is ineffective, and at the same time, the target hot search is removed from the first hot search set;

[0119] Judge whether the browsing of all hot search information in the first hot search set is effective through the above method;

[0120] Obtain all the hot search information retained in the first hot search set to form a second hot search set;

[0121] In addition, for the hot search information with the browsing times not equal to 1, by setting the time interval threshold and obtaining the average time interval, it is judged whether the browsing of the target hot search by the target user is effective; if the average time interval < the time interval threshold, it means that the target user browses the same hot search information multiple times in a short period of time, and there is a suspicion of swiping the hot search, so it is determined that the browsing of the target user for this target hot search is ineffective; in this way, effective and ineffective behaviors can be accurately distinguished, providing an accurate data basis for the subsequent push of hot search information.

[0122] It also includes:

[0123] Calculate the interest weight of the target user for each type of hot search information, specifically including:

[0124] The hot search information of the political type in the second hot search set constitutes a political type set, the hot search information of the economic type in the second hot search set constitutes an economic type set, the hot search information of the technology type in the second hot search set constitutes a technology type set, the hot search information of the entertainment type in the second hot search set constitutes an entertainment type set, and the hot search information of the sports type in the second hot search set constitutes a sports type set;

[0125] Respectively obtain the moments when the target user browses the hot searches in each different type set, and calculate the time difference from the current time;

[0126] Set a time decay factor α, which is used to weight the historical browsing data of the target user during the process of calculating the interest weight of the target user for each type of hot search information to achieve the effect of paying more attention to recent behaviors;

[0127] Calculate the interest weight of the target user for each type of hot search information through the following formula:

[0128]

[0129] Where:

[0130] w i is the interest weight of the target user for the i-th type of hot search information; when i = 1, it represents the interest weight for political type hot search information, when i = 2, it represents the interest weight for economic type hot search information, when i = 3, it represents the interest weight for science and technology type hot search information, when i = 4, it represents the interest weight for entertainment type hot search information, when i = 5, it represents the interest weight for sports type hot search information;

[0131] t ij is the time difference from the j-th browsing behavior of the target user for the i-th type of hot search information to the current time;

[0132] α is the time decay factor;

[0133] b ij is the contribution value of the j-th browsing behavior of the target user to the i-th type of hot search information; if the target user only browses, the contribution value is recorded as 1, if the target user likes during the browsing process, the contribution value is recorded as 2, if the target user comments during the browsing process, the contribution value is recorded as 3, if the target user shares during the browsing process, the contribution value is recorded as 4; if the user has multiple behaviors during the browsing process, select the contribution value corresponding to the behavior with the largest contribution value as the contribution value;

[0134] Calculate the interest weights q1, q2, w3, w4, and w5 through the above method;

[0135] This hot search information push method based on big data can deeply explore user preferences by calculating the interest weights of the target user for different types of hot search information, and then push more personalized hot search information to improve the acceptance of the pushed content by users. The calculation of interest weights is not only based on the user's historical behavior but also combines the time decay factor, making the recommended content more in line with the current user's immediate interests rather than simply relying on historical data. By using the time decay factor, the dynamic changes of user interests can be updated in a timely manner, especially when the user's interests change, the system can quickly adjust the recommendation strategy.

[0136] The calculation of the interest weight of the target user for each type of hot search information specifically includes:

[0137] Normalize the interest weights of the target users so that their sum is 1. Specifically, normalize the interest weights of the target users through the following formula;

[0138]

[0139] Obtain w1′, w2′, w3′, w4′, and w5′ through the above calculation method.

[0140] It also includes:

[0141] Set a first quantity threshold, which is used to determine whether the target user is a new user;

[0142] Obtain the number of hot search information in the second hot search set, denoted as the judgment quantity;

[0143] If the judgment quantity ≤ the first quantity threshold, it is determined that the target user is a new user;

[0144] If the judgment quantity > the first quantity threshold, it is determined that the target user is an old user;

[0145] If the target user is a new user, push hot search information to the target user according to the first strategy;

[0146] If the target user is an old user, push hot search information to the target user according to the second strategy;

[0147] This method for pushing hot search information based on big data adopts different pushing strategies by determining whether the target user is a new user or an old user; if the judgment quantity ≤ the first quantity threshold, it indicates that the browsing record of the target user is less, and at this time, the target user is determined to be a new user; if the judgment quantity > the first quantity threshold, it indicates that the browsing record of the target user is more, and then the target user is determined to be an old user;

[0148] If the target user is a new user, pushing hot search information to the target user according to the first strategy specifically includes:

[0149] Obtain all current hot search information to form a hot search information pool;

[0150] Set the push quantity, the first information entropy, and the second information entropy;

[0151] The push quantity, the first information entropy, and the second information entropy are used to select hot search information for pushing hot search information to the target user from the hot search information pool, where the first information entropy < the second information entropy;

[0152] Arbitrarily select the push quantity of hot search information from the hot search information pool to form a target push set;

[0153] Classify the hot search information in the target push set according to its type, and obtain the ratio of each type of hot search information to all the hot search information in the target push set, denoted as P(x1), P(x2), P(x3), P(x4), and P(x5) respectively;

[0154] Calculate the information entropy of the hot search information in the target push set through the following formula:

[0155]

[0156] If the first information entropy < H(ω) < the second information entropy, then push the hot search information in the target push set to the target user;

[0157] If H(ω) ≤ the first information entropy or H(ω) ≥ the second information entropy, then randomly select the number of push hot search information from the hot search information pool to form the target push set until the first information entropy < the information entropy of the hot search information in the target push set < the second information entropy;

[0158] For new users, randomly select the number of push hot search information from the hot search information pool, and calculate the information entropy of the selected hot search information. The larger the information entropy, the more extensive the types of the selected hot search information are, and the smaller the information entropy, the more narrow the types of the selected hot search information are. Ensuring that the information entropy of the selected hot search information is between the first information entropy and the second information entropy can ensure that the hot search information pushed to the target user is neither too extensive nor too narrow, avoiding the existence of information cocoons;

[0159] If the target user is an old user, then push the hot search information to the target user according to the second strategy, specifically including:

[0160] Obtain all the current hot search information to form the hot search information pool;

[0161] Set the number of pushes, the first information entropy, and the second information entropy;

[0162] The number of pushes, the first information entropy, and the second information entropy are used to select the hot search information for pushing the hot search information to the target user from the hot search information pool, where the first information entropy < the second information entropy;

[0163] Randomly select the number of push hot search information from the hot search information pool to form the target push set;

[0164] Classify the hot search information in the target push set according to its type, and obtain the ratio of each type of hot search information to all the hot search information in the target push set, denoted as P(x1), P(x2), P(x3), P(x4), and P(x5) respectively;

[0165] Calculate the weighted information entropy of the hot search information in the target push set through the following formula:

[0166]

[0167] If the first information entropy < H(ω)' < the second information entropy, then push the hot search information in the target push set to the target user;

[0168] If H(ω)' ≤ the first information entropy or H(ω)' ≥ the second information entropy, then randomly select the number of push hot search information from the hot search information pool again to form the target push set until the first information entropy < the weighted information entropy of the hot search information in the target push set < the second information entropy;

[0169] For old users, randomly select the number of push hot search information from the hot search information pool, and calculate the weighted information entropy of the selected hot search information in combination with the interest weight of the target user for each type of hot search information, ensuring that the weighted information entropy of the selected hot search information is between the first information entropy and the second information entropy, which can ensure that the hot search information pushed to the target user not only meets the interests of the target user but also avoids the information cocoon.

[0170] It should be noted that in this article, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device comprising a series of elements not only includes those elements, but also includes other elements not expressly listed, or also includes elements inherent to such process, method, article or device.

[0171] The above are only the preferred embodiments of the present invention. It should be pointed out that for those of ordinary skill in the art, without departing from the technical principle of the present invention, several improvements and refinements can be made, and these improvements and refinements should also be regarded as the protection scope of the present invention.

Claims

1. A method for pushing hot search information based on big data, characterized in that: Specifically include: Manually classify the hot search information and divide all the hot search information into different types; The genres mentioned include: politics, economics, technology, entertainment, and sports; Get the user who pushes the hot search information and record it as the target user; Obtain all hot search information that the target user has browsed in the past to form the first hot search set; According to the data cleaning strategy, it is determined whether the target user's browsing of the hot search information in the first hot search set is effective, specifically including: Randomly select a hot search information from the first hot search set and record it as the target hot search; Get the number of times the target user browses the target hot search, recorded as the number of views; If the number of views = 1, then obtain the time the target user spends viewing the target hot search, and record it as the viewing time; Setting a first duration threshold and a second duration threshold; The first duration threshold and the second duration threshold are used to determine whether the target user's browsing of the target hot search is valid; wherein the second duration threshold is greater than the first duration threshold; If the browsing time is less than or equal to the first time threshold; The target user's browsing of the target hot search is deemed invalid, and the target hot search is removed from the first hot search set; If the first duration threshold < browsing duration < second duration threshold; The target user's browsing of the target hot search is deemed valid, and the target hot search is retained in the first hot search set; If the browsing time is greater than or equal to the second time threshold, the interaction depth of the target user on the target hot search is obtained; The interaction depth specifically includes 0 and 1; If the target user clicks "like", "comments" or "shares" during the browsing process of the target hot search, the target user's interaction depth for the target hot search is considered to be 1; If the target user does not click "like", "comment" or "share" during the browsing process of the target hot search, the target user's interaction depth for the target hot search is considered to be 0; If the target user's interaction depth for the target hot search is 1, the target user's browsing of the target hot search is deemed valid, and the target hot search is retained in the first hot search set; If the target user's interaction depth for the target hot search is 0, it is determined that the target user's browsing of the target hot search is invalid, and the target hot search is removed from the first hot search set.

2. A method for pushing hot search information based on big data according to claim 1, characterized in that: The step of judging whether the target user's browsing of the hot search information in the first hot search set is effective according to the data cleaning strategy specifically includes: If the number of views ≠ 1; Obtain the time interval between the end time of the target user's first browsing of the target hot search and the start time of the second browsing of the target hot search, recorded as the first time interval; obtain the time interval between the end time of the target user's second browsing of the target hot search and the start time of the third browsing of the target hot search, recorded as the second time interval... obtain the time interval between the end time of the target user's second-to-last browsing of the target hot search and the start time of the last browsing of the target hot search, recorded as the Ath time interval; Where A = number of views - 1; Calculate (first time interval + second time interval + ... + Ath time interval) ÷ ​​A, and record the result as the average time interval; Setting a time interval threshold, wherein the time interval threshold is used to determine whether the target user's browsing of the target hot search is valid; If the average time interval is ≥ the time interval threshold, the target user's browsing of the target hot search is deemed valid, and the target hot search is retained in the first hot search set; If the average time interval is less than the time interval threshold, the target user's browsing of the target hot search is deemed invalid, and the target hot search is removed from the first hot search set; Determine whether browsing all hot search information in the first hot search set is valid by using the above method; All hot search information retained in the first hot search set is obtained to form a second hot search set.

3. A method for pushing hot search information based on big data according to claim 2, characterized in that: Also includes: Calculate the target user's interest weight for each type of hot search information, including: The hot search information of the type of politics in the second hot search set constitutes a political type set, the hot search information of the type of economy in the second hot search set constitutes an economic type set, the hot search information of the type of technology in the second hot search set constitutes a technology type set, the hot search information of the type of entertainment in the second hot search set constitutes an entertainment type set, and the hot search information of the type of sports in the second hot search set constitutes a sports type set; Get the time when the target user browses the hot searches in each type of collection respectively, and calculate the time difference with the current time; Setting a time decay factor α, which is used to weight the target user's historical browsing data in the process of calculating the target user's interest weight for each type of hot search information, so as to achieve the effect of paying more attention to recent behavior; The target user's interest weight for each type of hot search information is calculated using the following formula: in: w i is the interest weight of the target user in the i-th type of hot search information; when i=1, it represents the interest weight in political hot search information, when i=2, it represents the interest weight in economic hot search information, when i=3, it represents the interest weight in technology hot search information, when i=4, it represents the interest weight in entertainment hot search information, when i=5, it represents the interest weight in sports hot search information; t ij is the time difference between the target user’s j-th browsing behavior of the i-th type of hot search information and the current time; α is the time decay factor; b ij is the contribution value of the target user's j-th browsing behavior to the i-th type of hot search information; if the target user only browses, the contribution value is recorded as 1; if the target user likes during the browsing process, the contribution value is recorded as 2; if the target user comments during the browsing process, the contribution value is recorded as 3; if the target user shares during the browsing process, the contribution value is recorded as 4; if the user has multiple behaviors during the browsing process, the contribution value corresponding to the behavior with the largest contribution value is selected as the contribution value; The interest weights w1, w2, w3, w4 and w5 are calculated in the above manner.

4. A method for pushing hot search information based on big data according to claim 3, characterized in that: The calculation of the interest weight of the target user for each type of hot search information specifically includes: The target user interest weights are normalized so that their sum is 1. Specifically, the target user interest weights are normalized using the following formula; w1′, w2′, w3′, w4′ and w5′ are calculated and obtained in the above manner.

5. A method for pushing hot search information based on big data according to claim 4, characterized in that: Also includes: Setting a first quantity threshold, where the first number threshold is used to determine whether the target user is a new user; Obtain the number of hot search information in the second hot search set, and record it as the judgment number; If the judgment quantity is ≤ the first quantity threshold, the target user is determined to be a new user; If the judgment quantity is greater than the first quantity threshold, the target user is determined to be an old user; If the target user is a new user, push hot search information to the target user according to the first strategy; If the target user is an old user, the hot search information is pushed to the target user according to the second strategy.

6. A method for pushing hot search information based on big data according to claim 5, characterized in that: If the target user is a new user, the hot search information is pushed to the target user according to the first strategy, specifically including: Obtain all current hot search information to form a hot search information pool; Set the push quantity, first information entropy and second information entropy; The push quantity, the first information entropy and the second information entropy are used to select hot search information from the hot search information pool for hot search information push to the target user, wherein the first information entropy is less than the second information entropy; Randomly select a number of hot search information to be pushed from the hot search information pool to form a target push set; Classify the hot search information in the target push set according to its type, and obtain the ratio of each type of hot search information to all hot search information in the target push set, which are recorded as P(x1), P(x2), P(x3), P(x4) and P(x5) respectively; The information entropy of hot search information in the target push set is calculated using the following formula: If the first information entropy < H)ω) < the second information entropy, the hot search information in the target push set is pushed to the target user; If H(ω)≤the first information entropy or H(ω)≥the second information entropy, then a random number of hot search information are selected from the hot search information pool to form a target push set, until the first information entropy < the information entropy of the hot search information in the target push set < the second information entropy.

7. A method for pushing hot search information based on big data according to claim 5, characterized in that: If the target user is an old user, the hot search information is pushed to the target user according to the second strategy, which specifically includes: Obtain all current hot search information to form a hot search information pool; Set the push quantity, first information entropy and second information entropy; The push quantity, the first information entropy and the second information entropy are used to select hot search information from the hot search information pool for hot search information push to the target user, wherein the first information entropy is less than the second information entropy; Randomly select a number of hot search information to be pushed from the hot search information pool to form a target push set; Classify the hot search information in the target push set according to its type, and obtain the ratio of each type of hot search information to all hot search information in the target push set, which are recorded as P(x1), P(x2), P(x3), P(x4) and P(x5) respectively; The weighted information entropy of hot search information in the target push set is calculated using the following formula: If the first information entropy < H(ω)′ < the second information entropy, the hot search information in the target push set is pushed to the target user; If H(ω)′≤the first information entropy or H(ω)′≥the second information entropy, then any number of hot search information to be pushed is selected from the hot search information pool to form a target push set, until the first information entropy <the weighted information entropy of the hot search information in the target push set <the second information entropy.

Images