Method and device for determining recommendation group, equipment and storage medium

Abstract

The invention provides a method, device and equipment for determining a recommendation group and a storage medium, and the method comprises the steps: determining user attention information based on the operation information of a user on a commodity within a preset time; for any one to-be-selected recommendation group in the at least one to-be-selected recommendation group, determining first recommendation information based on each recommended commodity in the to-be-selected recommendation group, and calculating a first correlation between the first recommendation information and the user attention information and a second correlation between the first recommendation information and any second recommendation information; the second recommendation information is determined according to each recommended commodity in any recommendation group meeting a preset condition in the recommendation groups for the user; determining the recommendation degree of the to-be-selected recommendation group according to the first correlation and the second correlation of the to-be-selected recommendation group; wherein the higher the first correlation is and the smaller the second correlation is, the higher the recommendation degree is; and determining the to-be-selected recommendation group with the maximum recommendation degree from the at least one to-be-selected recommendation group, and taking the to-be-selected recommendation group as a newly-added recommendation group for the user.

Description

Technical Field

[0001] This application relates to the field of computer technology, and in particular, to a method, apparatus, device, and storage medium for determining a recommended group. Background Art

[0002] In the e-commerce scenario, various lists are usually inserted to attract users' attention. Users are often influenced by the recommendations of the products on the lists, thus increasing the possibility of purchase. The lists can also serve as a purchase guide for consumers to help users make choices among numerous products. However, numerous lists involve the problem of list sorting. Neither a fixed order nor a random order can well meet users' preferences, resulting in a waste of display traffic. For example, a fixed order may lead to lists that users are not interested in being ranked in the front, and a random order may lead to lists that do not quite meet users' preferences. Summary of the Invention

[0003] This application provides a method, apparatus, device, and storage medium for determining a recommended group, so that the determined recommended group has both diversity and relevance.

[0004] In a first aspect, this application provides a method for determining a recommended group, and the method includes:

[0005] Determine user attention information based on the operation information of the user on products within a preset time;

[0006] For any one of at least one recommended group to be selected, determine first recommended information based on the recommended products in the recommended group to be selected, and calculate the first relevance between the first recommended information and the user attention information and the second relevance between the first recommended information and any one of the second recommended information; the second recommended information is determined based on the recommended products in any one of the recommended groups that meet the preset conditions in the recommended groups for the user;

[0007] For any one of at least one recommended group to be selected, determine the recommendation degree of the recommended group to be selected according to the first relevance and the second relevance of the recommended group to be selected; wherein, the greater the first relevance and the smaller the second relevance, the greater the recommendation degree;

[0008] Determine the recommended group to be selected with the greatest recommendation degree from the at least one recommended group to be selected as the newly added recommended group for the user.

[0009] The first relevance calculated based on the first relevant information of the recommended group to be selected and the user's concerned information can reflect whether there is a high relevance between the recommended group to be selected and the user's concerned information. In other words, the first relevance can reflect whether the recommended group to be selected meets the user's concerns or preferences. Since the second recommended information is determined based on the recommended group for the user, the second relevance calculated based on the first recommended information and the second recommended information of the recommended group to be selected can reflect the relevance between the recommended group to be selected and the recommended group for the user. Then, the recommendation degree of the recommended group to be selected is determined based on the first relevance and the second relevance. The greater the first relevance and the smaller the second relevance, the greater the recommendation degree. In this way, it is ensured that the recommended group to be selected added to the recommended group for the user not only meets the user's concerns or preferences but also avoids the recommended group to be selected with a high relevance to the recommended group for the user from being added to the recommended group for the user again, improving the diversity of the recommended group for the user.

[0010] In a possible design, determining the user's concerned information based on the user's operation information on the commodity within a preset time includes:

[0011] Based on the user's operation information on the commodity within the preset time, determine the commodity categories of the commodities with the set operation information at different commodity category levels.

[0012] Based on the commodity categories of the commodities with the set operation information at different commodity category levels, determine the user's concerned vector as the user's concerned information.

[0013] Determining the first recommended information based on each recommended commodity in the recommended group to be selected includes:

[0014] Based on the commodity categories of each recommended commodity in the recommended group to be selected at different commodity category levels, determine the first recommended vector as the first recommended information.

[0015] In a possible design, determining the user's concerned vector as the user's concerned information based on the commodity categories of the commodities with the set operation information at different commodity category levels includes:

[0016] For any one of the different commodity category levels, determine the user preference weight values corresponding to each commodity category of the commodities with the set operation information at the commodity category level, and determine the first commodity category level vector corresponding to the commodity category level according to the user preference weight values corresponding to each commodity category and the set vectors corresponding to each commodity category.

[0017] Determine the user attention vector based on the first product category level vectors corresponding to the different product category levels respectively and the product category level weight values corresponding to the different product category levels respectively.

[0018] In a possible design, the setting operation information is an order placement operation;

[0019] Determining the user preference weight values corresponding to the respective product categories at the product category level for the product with the setting operation information includes:

[0020] Determine the N product categories with the most order placements at the product category level; N is a positive integer;

[0021] Determine the order placement preferences of the user for each of the N product categories according to the order placement times corresponding to the N product categories respectively;

[0022] Take the result of performing exponentialization and normalization operations on the N order placement preferences as the user preference weight values corresponding to the respective product categories.

[0023] In a possible design, determining the first recommendation vector serving as the first recommendation information based on the product categories of the respective recommended products in the to-be-selected recommendation group at different product category levels includes:

[0024] For any one of the different product category levels, determine the second product category level vector corresponding to the product category level according to the set vectors corresponding to the respective product categories at the product category level;

[0025] Determine the first recommendation vector based on the second product category level vectors corresponding to the different product category levels respectively and the product category level weight values corresponding to the different product category levels respectively.

[0026] In a possible design, any one of the recommendation groups that meet the preset conditions in the recommendation group for the user is determined in the following manner:

[0027] When the length of the recommendation group for the user is less than or equal to the preset window size, any one of the recommendation groups in the recommendation group for the user meets the preset conditions;

[0028] When the length of the recommendation group for the user is greater than the preset window size, the M recommendation groups newly added to the recommendation group for the user all meet the preset conditions; where M is equal to the preset window size.

[0029] When the length of the recommended group for a user is large, there is no need to calculate the second correlation between the recommended group to be selected and each recommended group in the recommended group for the user. In this application, conditions are designed for the recommended group for the user, and it is only necessary to calculate the second correlation between the recommended group to be selected and each recommended group that meets the preset conditions in the recommended group for the user, saving computing resources and improving computing efficiency.

[0030] In a possible design, determining the recommendation degree of the recommended group to be selected according to the first correlation and the second correlation of the recommended group to be selected includes:

[0031] Determining the recommendation degree of the recommended group to be selected according to a preset correlation weight value, the first correlation, and the second correlation.

[0032] In a possible design, the method further includes:

[0033] If there is no recommended group for the user, determining the recommendation degree of each recommended group to be selected according to the first correlation of each recommended group to be selected in the at least one recommended group to be selected; wherein, the greater the first correlation, the greater the recommendation degree;

[0034] Taking the recommended group to be selected with the largest recommendation degree as the newly added recommended group for the user.

[0035] In a second aspect, this application also provides a device for determining a recommended group. The device includes: a determination unit and a calculation unit;

[0036] The determination unit is used to determine user attention information based on the operation information of the user on the commodity within a preset time;

[0037] The determination unit is used to, for any one of the at least one recommended group to be selected, determine first recommendation information based on the recommended commodities in the recommended group to be selected. The calculation unit is used to calculate the first correlation between the first recommendation information and the user attention information and the second correlation between the first recommendation information and any second recommendation information; the second recommendation information is determined based on the recommended commodities in any one of the recommended groups that meet the preset conditions in the recommended group for the user;

[0038] The determination unit is used to, for any one of the at least one recommended group to be selected, determine the recommendation degree of the recommended group to be selected according to the first correlation and the second correlation of the recommended group to be selected; wherein, the greater the first correlation and the smaller the second correlation, the greater the recommendation degree;

[0039] The determining unit is further configured to determine, from the at least one recommended group to be selected, the recommended group to be selected with the highest recommendation degree as the newly added recommended group for the user.

[0040] In a possible design, when the determining unit is configured to determine user attention information based on the operation information of the user on the goods within a preset time, it is specifically configured to: based on the operation information of the user on the goods within the preset time, determine the product categories of the goods with the set operation information at different product category levels; based on the product categories of the goods with the set operation information at different product category levels, determine the user attention vector as the user attention information; when the determining unit is configured to determine the first recommendation information based on each recommended product in the recommended group to be selected, it is specifically configured to: based on the product categories of each recommended product in the recommended group to be selected at different product category levels, determine the first recommendation vector as the first recommendation information.

[0041] In a possible design, when the determining unit is configured to determine the user attention vector as the user attention information based on the product categories of the goods with the set operation information at different product category levels, it is specifically configured to: for any one of the different product category levels, determine the user preference weight values corresponding to each product category of the goods with the set operation information at the product category level, and determine the first product category level vector corresponding to the product category level according to the user preference weight values corresponding to each product category and the set vectors corresponding to each product category; determine the user attention vector according to the first product category level vectors corresponding to the different product category levels and the product category level weight values corresponding to the different product category levels.

[0042] In a possible design, the set operation information is the order placement operation; when the determining unit is configured to determine the user preference weight values corresponding to each product category of the goods with the set operation information at the product category level, it is specifically configured to: determine the N product categories with the most order placement times at the product category level; N is a positive integer; according to the order placement times corresponding to the N product categories, determine the order placement preferences of the user for each of the N product categories; use the result of exponentiating and normalizing the N order placement preferences as the user preference weight values corresponding to each product category.

[0043] In a possible design, when the determining unit is used to determine the first recommendation vector as the first recommendation information based on the product categories of each recommended product in the to-be-selected recommendation group at different product category levels, it is specifically used for: for any one of the different product category levels, determining the second product category level vector corresponding to the product category level according to the set vectors respectively corresponding to the product categories under the product category level; and determining the first recommendation vector according to the second product category level vectors respectively corresponding to the different product category levels and the product category level weight values respectively corresponding to the different product category levels.

[0044] In a possible design, any one of the recommendation groups that meet the preset conditions in the recommendation group for the user is determined by the following method:

[0045] When the length of the recommendation group for the user is less than or equal to the preset window size, any one of the recommendation groups in the recommendation group for the user meets the preset conditions;

[0046] When the length of the recommendation group for the user is greater than the preset window size, the M recommendation groups newly added to the recommendation group for the user all meet the preset conditions; where M is equal to the preset window size.

[0047] In a possible design, when the determining unit is used to determine the recommendation degree of the to-be-selected recommendation group according to the first relevance and the second relevance of the to-be-selected recommendation group, it is specifically used for: determining the recommendation degree of the to-be-selected recommendation group according to the preset relevance weight value, the first relevance and the second relevance.

[0048] In a possible design, the determining unit is further used to, if there is no recommendation group for the user, determine the recommendation degree of each to-be-selected recommendation group according to the first relevance of each to-be-selected recommendation group in the at least one to-be-selected recommendation group; where the greater the first relevance, the greater the recommendation degree; and taking the to-be-selected recommendation group with the greatest recommendation degree as the newly added recommendation group for the user.

[0049] In a third aspect, the present application further provides a device for determining a recommendation group, and the device includes: a processor, and a memory communicatively connected to the processor;

[0050] The memory stores computer execution instructions;

[0051] The processor executes the computer execution instructions stored in the memory to implement the method described in the first aspect above.

[0052] Fourthly, the present application also provides a computer-readable storage medium. The readable storage medium includes a program which, when executed on a device, causes the device to execute the method according to any one of the above first aspect.

[0053] Fifthly, the present application also provides a computer program product. The computer program product includes a computer program which, when executed by a processor, implements the method according to the first aspect described above. Description of the Drawings

[0054] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following briefly introduces the drawings required for the description of the embodiments. Obviously, the drawings in the following description are only some embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0055] Figure 1 It is a flowchart showing the method for determining a recommended group provided by an embodiment of the present application;

[0056] Figure 2 It is a structural schematic diagram of the device for determining a recommended group provided by an embodiment of the present application Figure 1 ;

[0057] Figure 3 It is a structural schematic diagram of the device for determining a recommended group provided by an embodiment of the present application Figure 2 。 Detailed Embodiments

[0058] In order to make the objectives, technical solutions and advantages of the present application clearer, the present application will be further described in detail below with reference to the drawings. Obviously, the described embodiments are only some embodiments of the present application, rather than all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present application.

[0059] The application scenarios described in the embodiments of the present application are for more clearly illustrating the technical solutions of the embodiments of the present application, and do not constitute a limitation on the technical solutions provided by the embodiments of the present application. Those of ordinary skill in the art know that with the emergence of new application scenarios, the technical solutions provided by the embodiments of the present application are equally applicable to similar technical problems. Among them, in the description of the present application, unless otherwise specified, "a plurality of" means two or more.

[0060] At present, when determining recommendation groups for users, for example, when recommending product lists for users, we often only consider the user's preference for the products, resulting in similar product lists recommended to users. This can easily lead to user fatigue, reduce user experience, and even cause users to no longer purchase related products. It also squeezes out exposure opportunities for other product lists, which will lead to a decline in overall product conversions and affect business results.

[0061] To this end, this application proposes Figure 1 The method for determining the recommended group shown in the figure can be performed by a server, a chip in the server, or a functional module in the server, and this application does not limit this. The following description takes the server as the execution subject as an example. The method specifically includes:

[0062] Step 101: Determine user attention information based on user operation information on the product within a preset time.

[0063] Exemplarily, based on the user's operation information on the product within a preset time, the product category of the product with the set operation information at different product category levels is determined; then, based on the product category of the product with the set operation information at different product category levels, a user attention vector as the user attention information is determined; when determining the user attention vector, for any product category level in the different product category levels, first determine the user preference weight value corresponding to each product category at the product category level for the product with the set operation information, and determine the first product category level vector corresponding to the product category level according to the user preference weight value corresponding to each product category and the set vector corresponding to each product category; then determine the user attention vector according to the first product category level vectors corresponding to different product category levels and the product category level weight values ​​corresponding to different product category levels.

[0064] Among them, the operation information is set as an order operation. In the process of determining the user preference weight values ​​corresponding to each commodity category at the commodity category level for the commodity with the set operation information, first determine the N commodity categories (N is a positive integer) with the most orders at the commodity category level; then determine the user's order preference for each commodity category in the N commodity categories according to the number of orders corresponding to the N commodity categories; and then use the results of indexing and normalizing the N order preferences as the user preference weight values ​​corresponding to each commodity category. The commodity category level generally includes the first-level commodity category, the second-level commodity category, the third-level commodity category..., the first-level commodity category includes the second-level commodity category, the second-level commodity category includes the third-level commodity category, and the inclusion relationship between the remaining commodity categories is similar, which will not be repeated here.

[0065] For example, assume that the preset time is one week, N is 4, and taking the first-level commodity category as an example, assume that the situation of the first-level commodity category with the most order placement times by the user within one week is shown in Table 1 as follows:

[0066] First-level commodity category name Snacks Baked pastries Instant food Wine and beverages Number of orders 7 2 1 1

[0067] Table 1

[0068] According to Table 1, it can be seen that the user placed 7 orders for commodities belonging to the casual snacks commodity category within one week, 2 orders for commodities belonging to the baked pastries commodity category, 1 order for commodities belonging to the instant food commodity category, and 1 order for commodities belonging to the alcoholic beverages and drinks commodity category. Next, determine the order placement preferences of the user for each commodity category among the 4 commodity categories shown in Table 1. The order placement preferences of the user can be represented by calculating the proportion of the order placement times of each commodity category level; for example, according to Table 1, it can be seen that the user placed a total of 11 orders (7 + 2 + 1 + 1) for commodities belonging to the first-level commodity category within one week. Then, the proportion of the order placement times for the casual snacks commodity category = 7 / 11 ≈ 0.64. Similarly, the proportion of the order placement times for the baked pastries commodity category = 2 / 11 ≈ 0.18, the proportion of the order placement times for the instant food commodity category = 1 / 11 ≈ 0.09, and the proportion of the order placement times for the alcoholic beverages and drinks commodity category = 1 / 11 ≈ 0.09. Therefore, the order placement preferences of the user for each commodity category under the 4 commodity category levels shown in Table 1 are shown in Table 2 as follows:

[0069] First-level commodity category name Snacks Baked pastries Instant food Wine and beverages Order preference 0.64 0.18 0.09 0.09

[0070] Table 2

[0071] Then, perform exponentialization and normalization operations on the order placement preferences, and use the results of the exponentialization and normalization operations as the user preference weight values corresponding to each commodity category respectively. Among them, the exponentialization result can be calculated using the formula e x where x is the order placement preference; for example, according to Table 2, the order placement preference for the casual snacks commodity category is 0.64. Therefore, the corresponding exponentialization result = e 0.64 ≈ 1.889; similarly, the exponentialization result corresponding to the baked pastries commodity category = e 0.18 ≈ 1.199, the exponentialization result corresponding to the instant food commodity category = e 0.09 ≈ 1.09, and the exponentialization result corresponding to the alcoholic beverages and drinks commodity category = e 0.09 ≈ 1.09. Therefore, the exponentialization results corresponding to the 4 commodity categories of casual snacks, baked pastries, instant food, and alcoholic beverages and drinks in Table 2 are shown in Table 3 as follows:

[0072] First-level commodity category name Snacks Baked pastries Instant food Wine and beverages Indexing result 1.889 1.199 1.09 1.09

[0073] Table 3

[0074] The normalization operation result is obtained as follows: calculate the sum of the indexed results of each commodity category, and take the result of dividing the indexed result of each commodity category by the sum of the indexed results of each commodity category as the user preference weight value corresponding to each commodity category. For example, taking Table 2 as an example, the sum of the indexed results of the 4 commodity categories of casual snacks, baked pastries, instant foods, and alcoholic and non-alcoholic beverages in Table 2 = 1.889 + 1.199 + 1.09 + 1.09 = 5.268. Then, the normalization result corresponding to the casual snacks commodity category, that is, the user preference weight value corresponding to the casual snacks commodity category = 1.889 / 5.268 ≈ 0.36; the user preference weight value corresponding to the baked pastries commodity category = 1.199 / 5.268 ≈ 0.22; the user preference weight value corresponding to the instant foods commodity category = 1.09 / 5.268 ≈ 0.21; the user preference weight value corresponding to the alcoholic and non-alcoholic beverages category = 1.09 / 5.268 ≈ 0.21; Therefore, the user preference weight values corresponding to the 4 commodity categories of casual snacks, baked pastries, instant foods, and alcoholic and non-alcoholic beverages shown in Table 3 are shown in Table 4 as follows:

[0075] First-level commodity category name Snacks Baked pastries Instant food Wine and beverages User preference weight value 0.36 0.22 0.21 0.21

[0076] Table 4

[0077] Exemplarily, after determining the user preference weight values corresponding to each commodity category under the first-level commodity category, the first commodity category level vector corresponding to the commodity category level can be determined according to the user preference weight values corresponding to each commodity category and the set vectors corresponding to each commodity category. For example, assume that the set vector corresponding to the casual snacks commodity category is E 休闲零食 , the set vector corresponding to the baked pastries commodity category is E 烘焙糕点 , the set vector corresponding to the instant foods commodity category is E 方便速食 , and the set vector corresponding to the alcoholic and non-alcoholic beverages commodity category is E 酒水饮料 . Then, the first commodity category level vector corresponding to the first-level commodity category level determined according to the user preference weight values corresponding to each commodity category shown in Table 4 and the set vectors corresponding to each commodity category = 0.36 × E 休闲零食 + 0.22 × E 烘焙糕点 + 0.21 × E 方便速食 + 0.21 × E 酒水饮料 , denoted as E cate1 .

[0078] The determination process of the first commodity category level vectors corresponding to the second-level commodity category level, the third-level commodity category level, etc. is similar to the determination process of the first commodity category level vector corresponding to the first-level commodity category level, and will not be elaborated here.

[0079] Further exemplarily, after determining the first commodity category level vectors corresponding to different commodity category levels respectively, the user attention vector can be determined according to the first commodity category level vectors corresponding to different commodity category levels respectively and the commodity category level weight values corresponding to different commodity category levels respectively. For example, assume that the first commodity category level vector corresponding to the secondary commodity category level is E cate2 , the first commodity category level vector corresponding to the tertiary commodity category level is E cate3 ; the weight value of the primary commodity category level is 0.2, the weight value of the secondary commodity category level is 0.3, and the weight value of the tertiary commodity category level is 0.5 (the weight values of each commodity category level can be set according to actual business requirements, and this application does not limit this, and only this is an example here), then the user attention vector E user = 0.2 × E cate1 + 0.3 × E cate2 + 0.5 × E cate3 .

[0080] Step 102: For any one of the to-be-selected recommendation groups among at least one to-be-selected recommendation group, determine the first recommendation information based on the recommended commodities in the to-be-selected recommendation group, and calculate the first correlation between the first recommendation information and the user attention information and the second correlation between the first recommendation information and any one of the second recommendation information; the second recommendation information is determined according to the recommended commodities in any one of the recommendation groups that meet the preset conditions among the recommendation groups for the user.

[0081] Exemplarily, when determining the first recommendation vector serving as the first recommendation information based on the commodity categories of the recommended commodities in the to-be-selected recommendation group at different commodity category levels: for any one of the commodity category levels among different commodity category levels, determine the second commodity category level vector corresponding to the commodity category level according to the set vectors corresponding to the respective commodity categories at the commodity category level; determine the first recommendation vector according to the second commodity category level vectors corresponding to different commodity category levels respectively and the commodity category level weight values corresponding to different commodity category levels respectively; wherein, calculating the first correlation between the first recommendation information and the user attention information is equivalent to calculating the first correlation between the first recommendation vector and the user attention vector.

[0082] Assume that the primary commodity category included in a to-be-selected recommendation group is leisure snacks, the secondary commodity categories are roasted nuts and dried fruits, and the tertiary commodity categories are plum products, bean snacks, and almonds; wherein, the set vector corresponding to the leisure snacks commodity is E 休闲零食 , the set vectors corresponding to the roasted nuts and dried fruits commodity categories are E 炒货坚果 and E 果脯蜜饯 , and the set vectors corresponding to the plum products, bean snacks, and almonds commodity categories are E梅类制品 , E 豆类零食 and E 巴旦木 ; then the second product category level vector corresponding to the first-level product category included in the to-be-selected recommendation group = E 休闲零食 , the second product category level vector corresponding to the second-level product category = E 炒货坚果 + E 果脯蜜饯 , the second product category level vector corresponding to the third-level product category = E 梅类制品 + E 豆类零食 + E 巴旦木 ; assuming that the product category level weight values corresponding to the first-level product category to the third-level product category are still 0.2, 0.3, and 0.5 respectively, then the first recommendation vector E item = 0.2 × E 休闲零食 + 0.3 × (E 炒货坚果 + E 果脯蜜饯 ) + 0.5 × (E 梅类制品 + E 豆类零食 + E 巴旦木 ).

[0083] In addition, there can be multiple product category levels such as the first-level product category and the second-level product category included in the to-be-selected recommendation group. When calculating the second product category level vector corresponding to each product category level, just add the set vectors corresponding to each product category. For example, assuming that the first-level product categories included in a to-be-selected recommendation group are leisure snacks and instant foods, then the second product category level vector corresponding to the first-level product category included in this to-be-selected recommendation group = E 休闲零食 + E 方便速食 .

[0084] Exemplarily, after determining the first recommendation vector corresponding to each to-be-selected recommendation group, the first correlation between the first recommendation vector and the user attention vector can be calculated. The present application does not limit the calculation formula of the first correlation. For example, the cosine similarity calculation formula can be adopted to calculate the similarity score between the first recommendation vector and the user attention vector to represent the first correlation.

[0085] Exemplarily, since the second recommendation information is determined according to each recommended product in any one of the recommendation groups that meet the preset conditions for the user, therefore, the determination process of the second recommendation vector for each recommendation group in the recommendation groups for the user is similar to the determination process of the first recommendation vector for each to-be-selected recommendation group in the to-be-selected recommendation groups, which will not be elaborated here. Calculating the second correlation between the first recommendation information and any one of the second recommendation information is equivalent to calculating the second correlation between the first recommendation vector and any one of the second recommendation vectors.

[0086] It should be noted that, in order to save computing resources and improve computing efficiency, when calculating the second correlation between the first recommendation vector and the second recommendation vectors, instead of calculating the second correlation between the first recommendation vector and all the second recommendation vectors in the recommendation group for the user, a preset condition is set, and the second correlation is only calculated between the first recommendation vector and the second recommendation vectors corresponding to any one of the recommendation groups that meet the preset condition in the recommendation group for the user.

[0087] Any one of the recommendation groups that meet the preset condition in the recommendation group for the user is determined in the following manner:

[0088] When the length of the recommendation group for the user is less than or equal to the preset window size, any one of the recommendation groups in the recommendation group for the user meets the preset condition; that is to say, when the length of the recommendation group for the user is less than or equal to the preset window size, for any one of the to-be-selected recommendation groups, the second correlation between it and each recommendation group in the recommendation group for the user should be calculated; among them, the first preset window size is usually set to a value within the range of 3 to 5.

[0089] When the length of the recommendation group for the user is greater than the preset window size, the M recommendation groups newly added to the recommendation group for the user all meet the preset condition; that is to say, at this time, only the second correlation between each recommendation group newly added to the recommendation group for the user needs to be calculated; among them, M is equal to the preset window size.

[0090] Exemplarily, if there is no recommendation group for the user, the recommendation degree of each to-be-selected recommendation group is determined according to the first correlation of each to-be-selected recommendation group in at least one to-be-selected recommendation group; among them, the greater the first correlation, the greater the recommendation degree; the to-be-selected recommendation group with the greatest recommendation degree is used as the newly added recommendation group for the user. In other words, the to-be-selected recommendation group with the greatest recommendation degree is used as the first recommendation group in the recommendation group for the user.

[0091] For example, assume that there are 3 to-be-selected recommendation groups, denoted as to-be-selected recommendation group ① to to-be-selected recommendation group ③ respectively. At this time, there is no recommendation group for the user. The first correlation between the first recommendation vector corresponding to to-be-selected recommendation group ① and the user attention vector is 0.4, the first correlation between the first recommendation vector corresponding to to-be-selected recommendation group ② and the user attention vector is 0.65, and the first correlation between the first recommendation vector corresponding to to-be-selected recommendation group ③ and the user attention vector is 0.42; since the greater the first correlation, the greater the recommendation degree, therefore, the recommendation degree of to-be-selected recommendation group ② is the greatest, and to-be-selected recommendation group ② is used as the first to-be-selected recommendation group added to the recommendation group for the user.

[0092] Step 103: For any one of the at least one recommended group to be selected, determine the recommendation degree of the recommended group to be selected according to the first relevance and the second relevance of the recommended group to be selected; wherein, the greater the first relevance and the smaller the second relevance, the greater the recommendation degree.

[0093] Exemplarily, if the recommended group for the user is not empty, the recommendation degree of the recommended group to be selected should be determined according to the preset relevance weight value, the first relevance, and the second relevance, rather than only according to the first relevance. Specifically, calculate the second relevance between the first recommended vector corresponding to each recommended group to be selected and each recommended group that meets the preset conditions in the recommended group for the user, and record the maximum value of the second relevance as maxSim; record the first relevance between the first recommended vector corresponding to the currently traversed recommended group to be selected and the user attention vector as p, then the recommendation degree of the recommended group to be selected can be calculated by the following formula:

[0094] Recommendation degree = w×p - (1 - w)×maxSim

[0095] where, w is the preset relevance weight value;

[0096] Alternatively, the recommendation degree of the recommended group to be selected can also be calculated by the following formula:

[0097]

[0098] Since the greater the first relevance and the smaller the second relevance, the greater the recommendation degree; on the contrary, the smaller the first relevance and the greater the second relevance, the smaller the recommendation degree; in this way, it is ensured that the recommended group to be selected newly added to the recommended group for the user not only conforms to the user's attention or user preference, but also avoids the recommended group to be selected with a relatively high relevance to the recommended group in the recommended group for the user from being newly added to the recommended group for the user again, improving the diversity of the recommended group for the user.

[0099] Step 104: Determine the recommended group to be selected with the maximum recommendation degree from the at least one recommended group to be selected as the newly added recommended group for the user.

[0100] The length of the recommended group for the user can also be restricted. When the length of the recommended group for the user reaches the length limit, the addition of recommended groups to the recommended group for the user can be stopped.

[0101] Figure 2 and Figure 3 FIG. is a schematic structural diagram of a possible apparatus for determining a recommended group provided by an embodiment of the present application. These apparatuses for determining a recommended group can be used to implement the functions of the server in the above method embodiments, and thus can also achieve the beneficial effects possessed by the above method embodiments.

[0102] As Figure 2 shown, the device 200 for determining a recommended group includes a determination unit 210 and a calculation unit 220. The device 200 for determining a recommended group is used to implement the functions of the server in the method embodiments described above Figure 1 shown.

[0103] When the device 200 for determining a recommended group is used to implement the functions of the server in the method embodiments described above Figure 1 shown:

[0104] The determination unit 210 is configured to determine user attention information based on the operation information of the user on the commodity within a preset time;

[0105] The determination unit 210 is configured to, for any one of at least one recommended group to be selected, determine first recommendation information based on the recommended commodities in the recommended group to be selected. The calculation unit 220 is configured to calculate a first correlation between the first recommendation information and the user attention information and a second correlation between the first recommendation information and any second recommendation information. The second recommendation information is determined based on the recommended commodities in any one of the recommended groups that meet the preset conditions in the recommended group for the user;

[0106] The determination unit 210 is configured to, for any one of at least one recommended group to be selected, determine the recommendation degree of the recommended group to be selected according to the first correlation and the second correlation of the recommended group to be selected. Wherein, the greater the first correlation and the smaller the second correlation, the greater the recommendation degree;

[0107] The determination unit 210 is further configured to determine, from the at least one recommended group to be selected, the recommended group to be selected with the greatest recommendation degree as the newly added recommended group for the user.

[0108] In a possible design, when the determination unit 210 is configured to determine user attention information based on the operation information of the user on the commodity within a preset time, it is specifically configured to: determine the commodity categories of the commodities with set operation information at different commodity category levels based on the operation information of the user on the commodity within the preset time; determine the user attention vector as the user attention information based on the commodity categories of the commodities with set operation information at different commodity category levels. When the determination unit 210 is configured to determine first recommendation information based on the recommended commodities in the recommended group to be selected, it is specifically configured to: determine the first recommendation vector as the first recommendation information based on the commodity categories of the recommended commodities in the recommended group to be selected at different commodity category levels.

[0109] In a possible design, when the determining unit 210 is used to determine a user attention vector as the user attention information based on the product categories of products with set operation information at different product category levels, it is specifically configured to: for any one of the different product category levels, determine the user preference weight values corresponding to the respective product categories of the products with set operation information at the product category level, and determine a first product category level vector corresponding to the product category level according to the user preference weight values corresponding to the respective product categories and the set vectors corresponding to the respective product categories; determine the user attention vector according to the first product category level vectors corresponding to the different product category levels and the product category level weight values corresponding to the different product category levels.

[0110] In a possible design, the set operation information is an order placement operation; when the determining unit 210 is used to determine the user preference weight values corresponding to the respective product categories of the products with set operation information at the product category level, it is specifically configured to: determine the N product categories with the most order placement times at the product category level; N is a positive integer; determine the order placement preference of the user for each of the N product categories according to the order placement times corresponding to the N product categories; use the result of exponentiating and normalizing the N order placement preferences as the user preference weight values corresponding to the respective product categories.

[0111] In a possible design, when the determining unit 210 is used to determine a first recommendation vector as the first recommendation information based on the product categories of the respective recommended products in the to-be-selected recommendation group at different product category levels, it is specifically configured to: for any one of the different product category levels, determine a second product category level vector corresponding to the product category level according to the set vectors corresponding to the respective product categories at the product category level; determine the first recommendation vector according to the second product category level vectors corresponding to the different product category levels and the product category level weight values corresponding to the different product category levels.

[0112] In a possible design, any one of the recommendation groups that meet the preset conditions in the recommendation group for the user is determined by the following method:

[0113] When the length of the recommendation group for the user is less than or equal to the preset window size, any one of the recommendation groups in the recommendation group for the user meets the preset conditions;

[0114] When the length of the recommendation group for the user is greater than the preset window size, the M recommendation groups newly added to the recommendation group for the user all meet the preset conditions; where M is equal to the preset window size.

[0115] In a possible design, when the determining unit 210 is used to determine the recommendation degree of the to-be-selected recommendation group according to the first relevance and the second relevance of the to-be-selected recommendation group, it is specifically used to: determine the recommendation degree of the to-be-selected recommendation group according to the preset relevance weight value, the first relevance, and the second relevance.

[0116] In a possible design, the determining unit 210 is further used to, if there is no recommendation group for the user, determine the recommendation degree of each to-be-selected recommendation group according to the first relevance of each to-be-selected recommendation group in the at least one to-be-selected recommendation group; where the greater the first relevance, the greater the recommendation degree; and use the to-be-selected recommendation group with the highest recommendation degree as the newly added recommendation group for the user.

[0117] For a more detailed description of the above determining unit 210 and calculating unit 220, reference can be directly made to Figure 1 the relevant descriptions in the method embodiments shown, which will not be elaborated here.

[0118] As Figure 3 shown, the apparatus 300 for determining a recommendation group includes a processor 310 and an interface circuit 320. The processor 310 and the interface circuit 320 are coupled to each other. It can be understood that the interface circuit 320 can be a transceiver or an input / output interface. Optionally, the apparatus 300 for determining a recommendation group may further include a memory 330, which is used to store the instructions executed by the processor 310 or the input data required for the processor 310 to run the instructions or the data generated after the processor 310 runs the instructions.

[0119] When the apparatus 300 for determining a recommendation group is used to implement Figure 1 the method shown, the processor 310 is used to implement the function of the above determining unit 210, and the interface circuit 320 is used to implement the function of the above calculating unit 220.

[0120] The division of units in the embodiments of the present application is illustrative. It is only a logical function division. In actual implementation, there may be other division methods. In addition, in each embodiment of the present application, the functional units can be integrated in one processor, or can exist separately physically, or two or more units can be integrated in one unit. The above integrated units can be implemented in the form of hardware or in the form of software functional units.

[0121] Although the preferred embodiments of the present application have been described, additional changes and modifications can be made to these embodiments by those skilled in the art once they learn the basic creative concept. Therefore, the appended claims are intended to be construed to include the preferred embodiments as well as all changes and modifications falling within the scope of the present application.

[0122] Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the claims of the present application and their equivalent technologies, the present application is also intended to include these modifications and variations.

Claims

1. A method for determining a recommendation group, characterized in that: The method includes: Determine the user's attention information based on the user's operation information on the product within a preset time; For any one of at least one recommendation group to be selected, determine first recommendation information based on each recommended commodity in the recommendation group to be selected, and calculate a first correlation between the first recommendation information and the information concerned by the user and a second correlation between the first recommendation information and any second recommendation information; the second recommendation information is determined based on each recommended commodity in any one of the recommendation groups for the user that meets a preset condition; For any one of the at least one recommendation group to be selected, determining the recommendation degree of the recommendation group to be selected according to the first correlation of the recommendation group to be selected and the second correlation of the recommendation group to be selected; wherein the greater the first correlation and the smaller the second correlation, the greater the recommendation degree; A to-be-selected recommendation group with the greatest recommendation degree is determined from the at least one to-be-selected recommendation group as a newly added recommendation group for the user.

2. The method according to claim 1, characterized in that The determining of the user's attention information based on the user's operation information on the product within a preset time includes: Based on the operation information of the user on the commodity within the preset time, determining the commodity category of the commodity with the set operation information at different commodity category levels; determining a user attention vector as the user attention information based on the commodity categories at different commodity category levels of the commodity with the set operation information; The determining the first recommendation information based on each recommended product in the recommendation group to be selected includes: Based on the commodity categories of each recommended commodity in the to-be-selected recommendation group at different commodity category levels, a first recommendation vector serving as the first recommendation information is determined.

3. The method according to claim 2, characterized in that The determining, based on the commodity categories of the commodities with the set operation information at different commodity category levels, a user attention vector as the user attention information comprises: For any one of the different commodity category levels, determine the user preference weight values ​​corresponding to each commodity category under the commodity category level for the commodity with the set operation information, and determine the first commodity category level vector corresponding to the commodity category level according to the user preference weight values ​​corresponding to each commodity category and the setting vectors corresponding to each commodity category; The user attention vector is determined according to the first commodity category level vectors respectively corresponding to the different commodity category levels and the commodity category level weight values ​​respectively corresponding to the different commodity category levels.

4. The method according to claim 3, characterized in that The setting operation information is an order placement operation; The determining of the user preference weight values ​​corresponding to each commodity category at the commodity category level for the commodity with the set operation information includes: Determine N product categories with the most orders at the product category level; N is a positive integer; Determining the user's order preference for each of the N product categories according to the number of orders respectively corresponding to the N product categories; The results of indexing and normalizing the N order preferences are used as the user preference weight values ​​corresponding to each product category.

5. The method according to claim 2, characterized in that The determining, based on the commodity categories of each recommended commodity in the to-be-selected recommendation group at different commodity category levels, a first recommendation vector as the first recommendation information comprises: For any one of the different commodity category levels, determine a second commodity category level vector corresponding to the commodity category level according to the setting vectors respectively corresponding to each commodity category under the commodity category level; The first recommendation vector is determined according to the second commodity category level vectors respectively corresponding to the different commodity category levels and the commodity category level weight values ​​respectively corresponding to the different commodity category levels.

6. The method according to any one of claims 1 to 5, characterized in that Any one of the recommendation groups for the user that meets the preset condition is determined in the following manner: When the length of the recommendation group for the user is less than or equal to the preset window size, any one of the recommendation groups for the user satisfies the preset condition; When the length of the recommendation group for the user is greater than the preset window size, the M recommendation groups newly added to the recommendation group for the user all meet the preset condition; wherein M is equal to the preset window size.

7. The method according to any one of claims 1 to 5, characterized in that The determining the recommendation degree of the to-be-selected recommendation group according to the first correlation of the to-be-selected recommendation group and the second correlation of the to-be-selected recommendation group includes: The recommendation degree of the to-be-selected recommendation group is determined according to a preset correlation weight value, the first correlation, and the second correlation.

8. The method according to any one of claims 1 to 5, characterized in that The method further comprises: If there is no recommendation group for the user, determining the recommendation degree of each recommendation group to be selected according to the first correlation of each recommendation group to be selected in the at least one recommendation group to be selected; wherein the greater the first correlation, the greater the recommendation degree; The recommended group to be selected with the largest recommendation degree is used as a newly added recommended group for the user.

9. A device for determining a recommendation group, characterized in that: The device comprises: a determination unit and a calculation unit; The determining unit is used to determine the user's attention information based on the user's operation information on the product within a preset time; The determining unit is used to determine the first recommendation information for any one of the at least one recommendation groups to be selected based on each recommended commodity in the recommendation group to be selected, and the calculating unit is used to calculate the first correlation between the first recommendation information and the user's attention information and the second correlation between the first recommendation information and any second recommendation information; the second recommendation information is determined based on each recommended commodity in any one of the recommendation groups for the user that meets the preset conditions; The determining unit is used to determine, for any one of the at least one recommendation group to be selected, a recommendation degree of the recommendation group to be selected according to a first correlation of the recommendation group to be selected and a second correlation of the recommendation group to be selected; wherein the greater the first correlation and the smaller the second correlation, the greater the recommendation degree; The determining unit is further configured to determine a to-be-selected recommendation group with the greatest recommendation degree from the at least one to-be-selected recommendation group as a newly added recommendation group for the user.

10. A device for determining a combination of objects, characterized in that include: A processor, and a memory communicatively connected to the processor; The memory stores computer-executable instructions; The processor executes the computer-executable instructions stored in the memory to implement the method according to any one of claims 1 to 8.

11. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-executable instructions, which are used to implement the method according to any one of claims 1 to 8 when executed by a processor.

12. A computer program product, characterized in that The invention comprises a computer program, which implements the method according to any one of claims 1 to 8 when being executed by a processor.

Images