A marketing promotion method based on object
By collecting and processing user, marketing products and operation time data, calculating time decay factors, generating feature vectors, and dividing product types, the problem of ignoring time factors in the existing technology is solved, and marketing effects and user interest recognition capabilities are improved.
Patent Information
- Application Number
- CN202411274863.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-12
- Publication Date
- 2024-11-26
- Estimated Expiration
- 2044-09-12
AI Technical Summary
The prior art ignores the time factor of product purchase and changes in user interests and hobbies in intelligent marketing, resulting in a decrease in marketing effect.
By collecting user, marketing products and operation time data, calculating time decay factors, generating user and marketing product feature vectors, modifying user portraits, dividing long-term and short-term products, forming user timetables, and setting time-sharing marketing strategies to improve precise marketing Effect.
It realizes time-sensitive recognition of user interests and hobbies and product attribute weights, improves the accuracy of marketing recommendations and user interest recognition capabilities, and enhances marketing effects.
Smart Images

Figure CN118798995B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of big data processing, and in particular to an object-based marketing promotion method. Background Art
[0002] Online marketing has become an important means for enterprises to enhance brand awareness, attract potential customers, and increase sales. In this context, object-based marketing methods have emerged, which emphasize the precise positioning of target audiences and the formulation of personalized promotion strategies to maximize marketing effects. At present, when conducting precision marketing for users, the intelligent marketing methods only consider the preference of some users who click on the marketing products or similar products, and ignore the preference of some users who do not click on the marketing products, resulting in the loss of some users of the marketing products, which affects the accurate judgment of the users needed for the marketing products and reduces the precision marketing effect.
[0003] The Chinese invention patent with application number 202311653896.9 provides a precision marketing method and system based on user portraits, which collects user data and marketing product data; obtains user feature vectors and marketing product feature vectors, predicts the user's interests in marketing products that have not yet been purchased through the products that the user has purchased and the product attributes, calculates the similarity between all the products purchased by the user and the marketing products, and arranges them in descending order of similarity, and delivers marketing product advertisements to users with higher rankings to improve the effectiveness of precision marketing.
[0004] However, in the above-mentioned prior art, the same weight is set for the goods purchased by the user to calculate the similarity, ignoring the time factor of the purchase of the goods and the changes in the interests and hobbies of the user at different times. If the product to be marketed is very similar to the product purchased by the user in the early stage, but the user has not purchased it recently, it is impossible to judge the changing trend of the user's interests and hobbies in real time, resulting in reduced marketing effect. Summary of the invention
[0005] The present application solves the problem that the time factor of commodity purchase and the changing interests and hobbies of users at different times are ignored in the prior art by providing an object-based marketing and promotion method, thereby achieving the technical effect of obtaining user interests and hobbies in real time based on the time factor of commodity purchase and thus improving precision marketing.
[0006] The present application provides an object-based marketing promotion method, the method comprising:
[0007] S100: Collecting user data, marketing product data and operation time data, wherein the operation time data refers to the time nodes and browsing duration corresponding to the user's click browsing and purchase behaviors of the product;
[0008] S200: Preliminarily sorting the user data, the marketing product data and the operation time data, and calculating the time decay factor of each product according to the operation time data ; Calculated according to the time attenuation factor and , and generate user feature vectors and marketing product feature vectors;
[0009] S300: Calculating the product to be marketed based on the marketing product feature vector The similarity between all products that the user has taken actions on , calculate the user's interest preference for marketing products based on the user feature vector and the similarity , and obtain the corrected user portrait feature vector ; Calculate each user portrait feature vector The similarity between the two and the marketing product feature vector G is used to sort in descending order, and the top 50% of users are obtained to form a preliminary user list;
[0010] S400: Generate a user timetable for each user according to the preliminary user list, divide commodities into long-term commodities and short-term commodities, wherein the long-term commodities refer to commodities with a long usage period and a low purchase frequency, and the short-term commodities refer to commodities with a short usage period and a high purchase frequency; mark commodities purchased by the user according to the long-term commodities and the short-term commodities to form a commodity time vector; generate a user feature report according to the commodity time vector and the user timetable; generate a primary user list and a secondary user list according to the user feature report;
[0011] S500: Setting a marketing promotion strategy according to the primary user list and the secondary user list.
[0012] Furthermore, the calculation formula of the time attenuation factor is:
[0013]
[0014] in, represents the time decay factor, Represents a weight parameter, which is used to measure the influence of the user's browsing time node and the current time node in time decay. and They represent the decay rate based on the purchase time node and the decay rate based on the browsing time node respectively. Indicates the current time node. Indicates the purchase time node, Indicates the browsing time node.
[0015] Furthermore, if the purchase time node is before the browsing time node, the time decay factor is calculated based on the purchase time node. If the browsing time node exists, the weight parameter needs to be set according to the time interval and browsing duration. The specific formula is as follows:
[0016]
[0017] in, represents the browsing time duration of the user at the browsing time node; k represents the parameter affected by the browsing time, which is used to adjust the browsing time to the weight parameter The parameters that affect the market are pre-set based on historical data and historical experience.
[0018] Furthermore, the user feature vector is: ; The marketing product feature vector is: ;
[0019] in, Indicates the nth product on which the user has taken action. Indicates the user's interest preference for the nth item after time decay. Represents the nth attribute feature of the product, Indicates the value of the weight of the nth attribute after time decay; use the TF-IDF algorithm to and Calculate the value of It shows the changes in users’ interest in products at different times; It shows the changes of attribute weights of commodities at different times.
[0020] Furthermore, the Indicates the proportion of products with the nth attribute feature. The specific calculation formula is as follows:
[0021]
[0022] in, Indicates the number of items sold with the nth attribute feature. Indicates the total quantity of the item sold.
[0023] Furthermore, the user portrait feature vector .
[0024] Furthermore, the user schedule includes the user's shopping time periods and shopping preferences, the shopping time periods include active time periods, stable time periods and slow time periods, and the shopping preferences refer to whether the user has special preferences for commodity attributes when shopping.
[0025] Furthermore, the primary user list refers to users whose similarity between the products to be marketed and the products previously purchased by the user is not less than the similarity threshold and the time difference between the purchase time node and the current time node is within the corresponding time difference threshold range; the secondary user list refers to users whose similarity between the products to be marketed and the products previously purchased by the user is less than the similarity threshold and the time difference between the purchase time node and the current time node is greater than the corresponding time difference high threshold or less than the corresponding time difference low threshold; the similarity threshold is pre-set based on historical data and is used to measure the similarity between the products to be marketed and the products purchased by the user; the time difference threshold range is pre-set based on the user's historical purchase frequency data, and the time difference threshold ranges corresponding to the long-term products and the short-term products are different.
[0026] Furthermore, the commodity level corresponding to the user is obtained according to the primary user list and the secondary user list, the commodity level including high level, medium level and low level, the commodity level is divided according to a preset purchase rate threshold range, the high level refers to commodities with a purchase rate not less than the upper threshold in the purchase rate threshold range, the medium level refers to commodities with a purchase probability within the purchase rate threshold range, and the low level refers to commodities with a purchase probability less than the lower threshold in the purchase rate threshold range; active time periods, stable time periods and slow time periods are obtained according to the user schedule, and marketing promotion is carried out according to the time-sharing marketing strategy.
[0027] Furthermore, push frequency data is obtained based on the operation time data and the user schedule, the push frequency data is normalized and the user's push frequency is predicted using a linear regression method; the user's preference for long-term and short-term products is obtained based on the product time vector, and a push ratio is generated based on the preference; and an individual time-sharing marketing strategy is generated based on the push frequency and the push ratio.
[0028] One or more technical solutions provided in this application have at least the following technical effects or advantages:
[0029] By comprehensively considering the user's historical behavior, time decay factors and product attributes, we can achieve the effect of obtaining more accurate user portrait feature vectors and marketing product feature vectors, improve the time sensitivity of user interest preferences and product attribute weights, and make recommendations more in line with users' current needs and preferences. Through similarity calculation and user screening, we can more effectively identify user groups interested in specific marketing products. BRIEF DESCRIPTION OF THE DRAWINGS
[0030] Figure 1 The present invention is a flowchart of an object-based marketing promotion method according to an embodiment of the present invention. DETAILED DESCRIPTION
[0031] To facilitate the understanding of the present invention, the present application will be described more comprehensively below with reference to the relevant drawings; the drawings show preferred embodiments of the present invention, but the present invention can be implemented in many different forms and is not limited to the embodiments described herein; on the contrary, the purpose of providing these embodiments is to enable a more thorough and comprehensive understanding of the disclosed content of the present invention.
[0032] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by technicians in the technical field to which the present invention belongs; the terms used in the specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention; the term "and / or" used herein includes any and all combinations of one or more related listed items.
[0033] Embodiment 1: Figure 1 As shown, an object-based marketing promotion method, the method comprising:
[0034] S100: Collecting user data, marketing product data and operation time data, wherein the operation time data refers to the time nodes and browsing duration corresponding to the user's click browsing and purchase behaviors of the product;
[0035] In some embodiments, user data includes: collecting user shopping habits and purchasing preferences, through which the user's shopping habits can be used to understand what kind of goods the user likes to buy, including what platform they like to buy on; collecting user evaluations of products, including evaluation content and evaluation time, through which the user's evaluation of the products after purchase can reflect the user's post-purchase status of the products.
[0036] Marketing product data: collects information such as the attributes, prices, inventory, and images of marketing products; log information: collects website access logs, including user access time, visited pages, and other information; product information is the consumer's understanding of the characteristics and specifications of the product, which is the premise for users to make purchasing decisions.
[0037] Operation time data: collects the time when users click and browse the products, the time when they purchase the products, and the duration of the users’ browsing.
[0038] S200: Preliminarily sorting the user data, the marketing product data and the operation time data, calculating the time decay factor of each product according to the operation time data; calculating the time decay factor according to the time decay factor and , and generate user feature vectors and marketing product feature vectors.
[0039] In some embodiments, the user data, the marketing product data and the operation time data are preprocessed, and the preprocessing includes data cleaning, data conversion and data integration; the data cleaning refers to preprocessing and screening the original data to remove duplicate data, missing data, erroneous data, etc. Common data cleaning methods include removing duplicates, filling missing values, correcting erroneous data, etc.; the data conversion refers to converting the original data into a form more suitable for analysis and modeling, which includes normalization, standardization, encoding, discretization and other processing of the data. Common data conversion methods include feature scaling, independent encoding, label encoding, binning, etc.; the data integration refers to merging data from multiple different sources into a unified data set, which involves operations such as data table merging, field matching, data connection, etc. Common data integration methods include merging, connecting, appending, etc. Through the above data sorting method, the collected user data, marketing product data and operation time data are calculated.
[0040] In some embodiments, the user feature vector is: ; The marketing product feature vector is: ;
[0041] in, Indicates the nth product on which the user has taken action. Indicates the user's interest preference for the nth item after time decay. Represents the nth attribute feature of the product, Indicates the value of the weight of the nth attribute after time decay; use the TF-IDF algorithm to and Calculate the value of It shows the changes in users’ interest in products at different times; It shows the changes of the attribute weights of the goods at different times; the attributes of the goods in this application are not objective attributes of the goods, but subjective attributes of the goods. For example, the objective attribute of a cabinet belongs to furniture, but its style attribute is the style, style and color of its design; at different times, the subjective attributes popular for the same goods are different, and users' preferences for the attributes of goods are also different. The weights of the attributes of goods change with the passage of time, ensuring that the weights of the attributes of goods more accurately reflect the current actual situation rather than outdated information; when adjusting the weights of the attributes of goods, it is necessary to set the weight of a certain attribute of the goods according to a large number of user purchase records and the proportion of the total number of goods sold. Indicates the proportion of products with the nth attribute feature. The specific calculation formula is as follows:
[0042]
[0043] in, Indicates the number of products with the nth attribute feature purchased by all users, Represents the total number of goods purchased by all users; by calculating the proportion of the number of goods sold of a certain attribute feature, the popularity and popularity of the attribute feature can be reflected.
[0044] In some embodiments, according to It can capture the user's current interest preferences in real time to achieve more accurate judgment and recommendation; It can obtain the specific attribute characteristics of users' preferences for products in real time and capture the changing trend of users' preferences for product attributes over time.
[0045] In some embodiments, TF-IDF (Term-Frequency-Inverse-Document-Frequency) is term frequency-inverse document frequency, which is a feature vectorization method widely used in text mining. It reflects the importance of a word to a text collection or a document in a corpus. The importance of a word increases in direct proportion to the number of times it appears in a document, but at the same time decreases in inverse proportion to the frequency of its appearance in the corpus. If a word or phrase appears frequently in an article and rarely appears in other articles, it is considered that this word or phrase has good category discrimination ability. TF (Term-Frequency), that is, term frequency, indicates the frequency of occurrence of a term (keyword) in a text; IDF (Inverse-Document-Frequency), that is, inverse document frequency, is a measure of the general importance of a word. TF-IDF combines the two to calculate the importance of a word in a document. The calculation formula of TF-IDF is Here, TF-IDF is used to measure the importance of a certain type of product to a user and the importance of the attribute characteristics of the product category to the product.
[0046] In some embodiments, represents the user's initial interest preference for the nth item, The value of is calculated using the TF-IDF algorithm, and the calculation formula is as follows:
[0047]
[0048] Wherein, TF1=the number of times the nth product appears among the products that the user has operated / the number of times all products that the user has operated appear, and IDF1=the total number of users / (the total number of users who have operated the nth product + 1).
[0049] Indicates the value of the initial weight of the nth attribute. The value of is calculated using the TF-IDF algorithm.
[0050]
[0051] Among them, TF2 = the number of times all users searched category n when performing operations on the product / the number of times all users searched for all product category attributes, IDF2 = the total number of operations performed by all users on the product / (the total number of times category n was searched to obtain the product + 1).
[0052] In some embodiments, the user's last purchase time node for a certain product and the browsing time node corresponding to the last click browsing behavior are obtained. If the purchase time node is before the browsing time node, the time decay factor is calculated based on the purchase time node; if the purchase time node is after the browsing time node, the browsing time is recorded and the time decay factor is calculated comprehensively based on the purchase time node and the browsing time node.
[0053] For example, for product A, the last time a user purchased A, that is, the purchase time node is June 2, and there is no click and browse record of product A between the purchase time node and the current time node, that is, the purchase time node is before the browse time node, and the time decay factor is calculated based on the purchase time node; if there is a browse time node for product A between the purchase time node and the current time node, that is, the purchase time node is after the browse time node, then the browse time is recorded and the time decay factor is calculated based on the purchase time node and the browse time node. The purchase time node and the browse time node mentioned above are both the time nodes when the user last purchased or browsed the product.
[0054] The calculation formula of the time decay factor is:
[0055]
[0056] in, represents the time decay factor, Represents a weight parameter, which is used to measure the influence of the user's browsing time node and the current time node in time decay. and They represent the decay rate based on the purchase time node and the decay rate based on the browsing time node respectively. Indicates the current time node. Indicates the purchase time node, Indicates the browsing time node.
[0057] In some embodiments, the decay rate needs to be pre-set based on historical data and actual experience. Preferably, an initial decay rate is set. According to experience, the monthly decay rate of the user's influence after purchasing a product is 25%, which means that if a user purchases a product at the beginning of a month, then by the end of the month, the willingness to purchase the same product again will be reduced to 75% of the beginning of the month (i.e., 1-25%=75%); the time decay factor is to consider the influence of time factors when calculating the similarity between users and products; when the purchase time and the current time are close, the time decay factor is close to 1, indicating that the influence of the interaction is large; when the purchase time and the current time are not close, the time decay factor is close to 0, indicating that the influence of the interaction is small.
[0058] If the purchase time node is before the browsing time node, the time decay factor is calculated based on the purchase time node. ; If there are browsing time nodes, you need to set the weight parameters according to the time interval and browsing duration, specifically:
[0059]
[0060] in, Indicates the browsing time length of the user at the browsing time node; k represents the parameter affected by the browsing time, which is used to adjust the browsing time to the weight parameter The parameters that affect the user are pre-set according to historical data and historical experience. Preferably, the parameters are set comprehensively according to the user's behavior pattern and historical experience. The user's behavior pattern requires monitoring the user's behavior before and after purchase, analyzing the user's historical behavior data, and understanding the user's browsing time distribution before purchase or after clicking. If the user browses for a long time before purchase, then k should be set larger.
[0061] S300: Calculating the product to be marketed based on the marketing product feature vector The similarity between all products that the user has taken actions on , calculate the user's interest preference for marketing products based on the user feature vector and the similarity , and obtain the corrected user portrait feature vector ;
[0062] ;
[0063] represents all the products that the user likes, j is equal to 1, 2, 3...K, K represents the number of all the products that the user likes, Display products and The similarity of the products in Represents the user's interest preference for the jth product after time decay;
[0064] Among them, the formula for calculating the similarity between any items is as follows:
[0065] ;
[0066] represents the similarity between item i and item j, , Respectively represent the number of users who have operated on item i and item j, Indicates the number of users who have operated on both item i and item j;
[0067] Calculate each user portrait feature vector The similarity between the two and the marketing product feature vector G is calculated, and the similarity is used to sort in descending order to obtain the top 50% of users to form a preliminary user list.
[0068] In some embodiments, based on each user portrait feature vector The similarity is calculated with the marketing product feature vector G, and the Pearson-Correlation-Coefficient is used to calculate the feature vector of each user portrait The similarity between the two and the marketing product feature vector G; the Pearson correlation coefficient (Pearson-Correlation-Coefficient) is a statistic that measures the degree of linear correlation between two continuous variables. It is used to measure the strength and direction of the linear relationship between two variables. The value range of the Pearson correlation coefficient is between -1 and +1. When the coefficient is +1, it means that the two variables are completely positively correlated. When the coefficient is -1, it means that the two variables are completely negatively correlated. When the coefficient is close to 0, it means that there is no linear relationship between the two variables. The role of the Pearson correlation coefficient is mainly reflected in the following aspects: Measuring the linear relationship between variables: The Pearson correlation coefficient can help determine whether there is a linear relationship between two variables and the strength of the linear relationship. When the coefficient is close to +1 or -1, it is considered that there is a strong linear relationship between the two variables; Verifying research hypotheses: The Pearson correlation coefficient can be used to verify research questions or hypotheses, such as determining whether a factor is related to another factor; Feature selection and variable screening: In feature selection and variable screening, the Pearson correlation coefficient is often used to evaluate the correlation between variables and target variables, thereby determining which variables have a significant impact on the target variable; Data exploration and visualization: The Pearson correlation coefficient can be used as a data exploration tool to help understand the relationship between variables in a data set, so as to more intuitively observe the correlation between variables.
[0069] The technical solutions in the above embodiments of the present application have at least the following technical effects or advantages:
[0070] This application achieves the effect of obtaining more accurate user portrait feature vectors and marketing product feature vectors by comprehensively considering the user's historical behavior, time decay factors and product attributes, improving the time sensitivity of user interest preferences and product attribute weights, making recommendations more in line with the user's current needs and preferences, and through similarity calculation and user screening, it can more effectively identify user groups interested in specific marketing products.
[0071] Embodiment 2: In embodiment 1, by comprehensively considering the historical behavior of users and time factors, it is possible to obtain marketing user groups more accurately. This embodiment makes further limitations on the basis of the above embodiment.
[0072] S400: Generate a user timetable for each user according to the preliminary user list, divide commodities into long-term commodities and short-term commodities, wherein the long-term commodities refer to commodities with a long usage period and a low purchase frequency, and the short-term commodities refer to commodities with a short usage period and a high purchase frequency; mark commodities purchased by the user according to the long-term commodities and the short-term commodities to form a commodity time vector; generate a user feature report according to the commodity time vector and the user timetable; generate a primary user list and a secondary user list according to the user feature report;
[0073] S500: Setting a marketing promotion strategy according to the primary user list and the secondary user list.
[0074] The user schedule includes the user's shopping time period and shopping preferences. The shopping time period includes an active time period, a stable time period and a slow time period. The shopping preferences refer to whether the user has special preferences for commodity attributes when shopping.
[0075] In some embodiments, a statistical analysis is performed based on the purchase time node of each commodity in the user feature vector, including the commodity name and commodity attribute information, and the purchase frequency information of the commodity is calculated based on the different times of purchase of the same commodity. The user's shopping time period and shopping preferences in a day or a week are obtained based on the analysis of the user's purchase time node information. The shopping time period includes an active time period, a stable time period and a slow time period, which are divided according to the number of commodities purchased by the user in each time period. The active time period is a time period with a large purchase volume by the user, the stable time period is a time period with purchase records but a small number of times and a small purchase volume, and the slow time period is a time period with few purchases. The shopping preference refers to whether the user has special preferences for commodity attribute characteristics when shopping, including brand preference and price preference. , regional preferences, etc., such as being fond of a certain brand, or liking high-priced goods, etc., combining the user's historical purchase records and the product time vector to identify the time pattern of the user's purchase of long-term goods and short-term goods; the product time vector is obtained by dividing the goods purchased by the user according to a certain rule, where the certain rule refers to the period of use of the goods and the frequency of purchase. Preferably, it can also be comprehensively limited according to the shelf life of the goods and the frequency of product updates. According to the period of use and the frequency of purchase, the goods are divided into long-term goods and short-term goods. The long-term goods have a long use cycle and a low purchase frequency, such as furniture and home appliances. After purchasing once, they will not be purchased again in a short time, and are consumed slowly; the short-term goods have a short use cycle and a high purchase frequency, such as daily necessities, food, cosmetics, etc., which are consumed quickly and require long-term and stable purchases.
[0076] In the above embodiment, similarity is used to analyze users who have not clicked on the marketed product, select users who are likely to purchase for marketing promotion, analyze the products that the users have purchased in the preliminary user list, determine the user's purchase frequency of short-term and long-term products, and record them to form a user feature report, which includes all the contents of the product time vector and the user timetable, and partitions the corresponding content for subsequent analysis.
[0077] The primary user list refers to users whose similarity between the products to be marketed and the products previously purchased by the user is not less than the similarity threshold and the time difference between the purchase time node and the current time node is within the corresponding time difference threshold range; the secondary user list refers to users whose similarity between the products to be marketed and the products previously purchased by the user is less than the similarity threshold and the time difference between the purchase time node and the current time node is greater than the corresponding time difference high threshold or less than the corresponding time difference low threshold; the similarity threshold is pre-set based on historical data and is used to measure the similarity between the products to be marketed and the products purchased by the user; the time difference threshold range is pre-set based on the user's historical purchase frequency data, and the time difference threshold ranges corresponding to the long-term products and the short-term products are different.
[0078] In some embodiments, a similarity threshold and a time difference threshold range are preset. First, it is determined whether the product to be marketed is a short-term product or a long-term product. Based on the determination result, a similarity analysis is performed with products of the same category (short-term or long-term) that the user has already purchased. The similarity analysis method provided in the above embodiments can be used. If the similarity is less than the similarity threshold, the user is added to the secondary user list. If the similarity is not less than the similarity threshold, the purchase time node of the user's most recent purchase of a similar product is obtained, and the time difference with the current time node is calculated. The time difference is compared with the time difference threshold range. If the time difference is within the time difference threshold range, it meets the user's purchase frequency and is added to the primary user list. Otherwise, it is added to the secondary user list. user list; the time difference threshold range is pre-set according to the user's historical purchase frequency data, the time difference threshold ranges corresponding to the long-term product and the short-term product are different, and the time difference threshold indicates whether the current product meets the user's purchase frequency; for example, in the historical data, user A buys cosmetics once a month on average, the product to be marketed and the cosmetics previously purchased by user A are highly similar and both are short-term products, the user A's most recent purchase record is exactly 28 days, that is, the time difference is 28 days, and the pre-set time difference threshold range of the corresponding short-term product is [25, 33], 28 days is within the time difference threshold range, which meets the user's purchase frequency, and at this time user A is the primary user for promoting the product to be marketed.
[0079] In some embodiments, the time difference threshold range and the similarity threshold are both pre-set based on the user's personal historical data. The similarity threshold is obtained by analyzing the name, function, design, etc. of the purchased goods in the historical data. For example, if the promoted product is a certain brand of cosmetics, and the brand of cosmetics in the purchase record of user A is the same, then its similarity threshold should be set relatively high because the user pays more attention to the brand. When promoting cosmetics, user A needs to be added to the secondary user list; the frequency of users' purchases of goods is analyzed. It is easier to analyze the pattern of the frequency of short-term goods, and then set a more accurate time difference threshold range. When analyzing long-term goods, the set time difference threshold range should be synchronously analyzed based on the historical data of other users, especially for goods with a long usage period, and then the time difference threshold range is improved.
[0080] Preferably, when marking the commodities as long-term commodities and short-term commodities, auxiliary marking is performed according to seasonal, non-seasonal, daily, special holidays, etc., and targeted analysis and promotion are performed on such commodities during marketing promotion. When creating a vector for each commodity, the commodity type and auxiliary characteristics are included. For example, the vector of a heating device used in winter is [long-term commodity, seasonal commodity (winter)].
[0081] In some embodiments, based on the characteristics of the goods, that is, the distinction between long-term goods and short-term goods, the above embodiments can be further limited and supplemented. In the above embodiments, a time factor is set, and different goods are distinguished under the premise of considering the time factor. For long-term goods, because the usage period is longer, the user's purchase or browsing records are older, resulting in a lower weight. By calculating the time difference and judging whether the product is a long-term product, the problem of low weight caused by time attenuation and affecting user screening can be solved.
[0082] The technical solutions in the above embodiments of the present application have at least the following technical effects or advantages:
[0083] This application analyzes the time patterns of users' purchases of long-term and short-term goods to more accurately identify users' purchasing preferences and shopping time periods; by setting similarity thresholds and time difference threshold ranges, primary users who are more in line with marketing promotion conditions are screened out to improve the accuracy of marketing promotion; by auxiliary marking of goods such as seasonal, non-seasonal, daily, and specific holidays, more specific analysis and promotion strategy formulation can be carried out for different types of goods during marketing promotion; and the effect of more accurately identifying marketing promotion users and being able to select marketing push based on the time vector of goods and users is achieved.
[0084] Embodiment 3: In the above embodiment, by analyzing the time pattern of users' purchase of long-term and short-term commodities, the user's purchase preference and shopping time period are more accurately identified. This embodiment makes further improvements on the basis of the above embodiment.
[0085] The commodity level corresponding to the user is obtained according to the primary user list and the secondary user list, the commodity level including high level, medium level and low level, and the commodity level is divided according to a preset purchase rate threshold range, the high level refers to commodities with a purchase rate not less than a high threshold in the purchase rate threshold range, the medium level refers to commodities with a purchase probability within the purchase rate threshold range, and the low level refers to commodities with a purchase probability less than a low threshold in the purchase rate threshold range; the active time period, the stable time period and the slow time period are obtained according to the user schedule, and marketing promotion is carried out according to the time-sharing marketing strategy.
[0086] In some embodiments, a user places different list positions for different commodities, and the success rate of the user's commodity purchase is predicted based on the different list positions, which is expressed as the purchase rate. The high-level commodities are commodities with high similarity and completely consistent with the user's purchase frequency, and it is estimated that more than 90% of the users will purchase them; the middle-level commodities are commodities with high similarity or relatively consistent purchase frequency, and it is estimated that the user has a certain probability of purchasing them; the low-level commodities are commodities with low similarity or not quite consistent purchase frequency, and it is estimated that the user has a low probability of purchasing them; specific adjustments need to be made according to the user's analysis. For example, user A may place commodity X in the primary user list, and the purchase rate may be adjusted accordingly. User list, user A's purchase rate for product X is 90%, user A is in the secondary user list for product Y, then further determine the reason for being divided into the secondary user list, set different purchase rates according to different reasons, if the time difference does not meet the time difference threshold range, that is, does not meet the user's purchase frequency, then set to 70%, if it does not meet the user's shopping preferences and other conditions, then set to 60%, the pre-set purchase rate threshold range is [65%, 85%], the purchase rate is not less than 85% of the goods are high-level goods, the purchase rate is within the purchase rate threshold range is a mid-level goods, the purchase rate is less than 65% of the goods is a low-level goods. Preferably, when determining the purchase rate, the user's soft conditions and hard conditions should be combined, not meeting the user's purchase frequency is a soft condition, if not meeting the user's shopping preferences and other conditions are hard conditions, after a series of experiments, for goods that do not meet the soft conditions, the purchase rate is basically stable at a medium level; for goods that do not meet the hard conditions, the purchase rate is basically low.
[0087] In some embodiments, the time-sharing marketing strategy is to promote the product at different levels according to different time periods of the user's schedule. The time-sharing marketing strategy is to select the stable time period and slow time period of the user for high-level products, and monitor the push effect in real time and analyze it. If the actual benefit in the slow time period is not lower than the expected effect, the proportion of push in the slow time period will be gradually increased; for mid-level and low-level products, select a part to be pushed in the user's active time period to increase exposure opportunities and purchase probability, and push the remaining part in the user's stable time period. It is also necessary to monitor the push results in real time to ensure that the number and types of products pushed in the same time period are not too many, so as to avoid information overload and maximize economic benefits.
[0088] The time-sharing marketing strategy is a refined marketing strategy based on user purchasing behavior and time habits. By analyzing the user's purchase frequency and time period preference, different levels of goods (high-level, medium-level, and low-level) are pushed to users in the most suitable time period; it can effectively reduce information overload, improve user experience, increase product exposure opportunities, and increase purchase probability. Through precise time management and push strategies, the time-sharing marketing strategy can more effectively attract users' attention and promote purchasing behavior, thereby maximizing marketing economic benefits. By reducing unnecessary pushes and avoiding information overload, more valuable information is provided, thereby improving user experience; pushing during time periods when users have a strong desire to shop can increase product exposure opportunities and purchase probability; more accurately reach target users, and improve marketing effects and conversion rates.
[0089] The technical solutions in the above embodiments of the present application have at least the following technical effects or advantages:
[0090] This application sets up a time-sharing marketing strategy through user schedules, primary user lists and secondary user lists, and pushes different levels of products (high-level, mid-level, and low-level) to users in the most suitable time periods, effectively reducing information overload, improving user experience, increasing product exposure opportunities and increasing purchase probability.
[0091] Embodiment 4: This embodiment is a further improvement on the basis of the above embodiment.
[0092] In the above embodiment, a time-sharing marketing strategy is set to segment the user's purchase time and set the product level, so as to conduct time-sharing marketing, reduce information overload, and maximize economic benefits. However, when recommending the time-sharing marketing strategy, it is impossible to control the push frequency and push ratio according to the differences between users, thereby affecting the marketing promotion effect.
[0093] The push frequency data is obtained according to the operation time data and the user schedule, the push frequency data is normalized and the user's push frequency is predicted using a linear regression method; the user's preference for long-term and short-term products is obtained according to the product time vector, and a push ratio is generated according to the preference; an individual time-sharing marketing strategy is generated according to the push frequency and the push ratio.
[0094] In some embodiments, the push frequency data refers to the total purchase frequency, average purchase interval, login frequency, and average browsing time, and the push frequency data is normalized to ensure the comparability between different features and avoid adverse effects on calculations due to excessive differences in feature dimensions or numerical ranges. The Min-Max normalization method is used to scale the feature index to the interval [0, 1]. The integrity and accuracy of the data must also be ensured, and records containing missing values or outliers must be deleted to ensure the integrity and accuracy of the data. Machine learning is used to define a linear regression model, with the push frequency data as the input variable and the push frequency as the output variable. The linear regression model is a technology well known to people in this field and is not specifically limited here.
[0095] In some embodiments, the user's preference for long-term and short-term products is obtained according to the product time vector, and a push ratio is generated according to the preference in combination with the shopping time period, and the number of long-term products and the number of short-term products in all user purchase records are obtained, and the corresponding average purchase quantity is generated. The purchase volume of long-term and short-term products of a single user is compared with the average respectively. If the number of long-term products of a single user is greater than the average of the purchase volume of long-term products of all users, the user is judged to have a long-term preference; if the number of short-term products of a single user is greater than the average of the purchase volume of short-term products of all users, the user is judged to have a short-term preference; according to the user's preference characteristics, a pre-set push rule is output to adjust the push ratio. An individual time-sharing marketing strategy is formulated according to the push frequency and push ratio corresponding to the user, and marketing promotion is carried out for the user according to the individual time-sharing marketing strategy.
[0096] In some embodiments, the push rules are pre-set based on historical data and actual experience. Preferably, for long-term preference users, it means that the total purchase frequency is low, and they do not log in frequently, and the average browsing time is short. For such users, push should be mainly carried out during their active time period. Because the browsing time of long-term preference users is short, in order to ensure maximum economic benefits, the proportion of high-end products should be increased in the push ratio. For users who prefer long-term products, 60% of high-end products, 30% of mid-level products, and 10% of low-end products should be pushed during the active time period; for short-term preference users, it means that the total purchase frequency is high. , frequent logins, and long average browsing time. For such users, the proportion of mid-level and low-level products should be increased during active time periods, and high-level products should be mainly placed in stable and slow time periods to increase product exposure and improve overall economic benefits. For users who prefer short-term products, 30% of high-level products, 40% of mid-level products, and 30% of low-level products should be pushed during active time periods. In the above push rule settings, real-time adjustments should be made based on the actual user situation, and user click-through rate, conversion rate, purchasing behavior and other data should be collected in real time. User feedback data should be analyzed every week to adjust the push ratio and content.
[0097] The technical solutions in the above embodiments of the present application have at least the following technical effects or advantages:
[0098] This application obtains the user's push frequency data for analysis to obtain the user's push frequency and push ratio, and formulates an individual time-sharing marketing strategy based on the user's corresponding push frequency and push ratio, thereby improving the user's click-through rate and purchase conversion rate for pushed products, achieving the effect of accurate push and maximizing economic benefits.
[0099] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.
Claims
1. An object-based marketing promotion method, characterized in that: The method comprises: S100: Collecting user data, marketing product data and operation time data, wherein the operation time data refers to the browsing time nodes and purchase time nodes corresponding to the user's click browsing and purchase behaviors of the product and the browsing duration; S200: Preliminarily sorting the user data, the marketing product data and the operation time data, and calculating the time decay factor of each product according to the operation time data ; Calculated according to the time attenuation factor and , and generate user feature vectors and marketing product feature vectors; S300: Calculating the product to be marketed based on the marketing product feature vector The similarity between all products that the user has taken actions on , calculate the user's interest preference for marketing products based on the user feature vector and the similarity , and obtain the corrected user portrait feature vector ; Calculate each user portrait feature vector The similarity between the two and the marketing product feature vector G is used to sort in descending order, and the top 50% of users are obtained to form a preliminary user list; S400: Generate a user timetable for each user according to the preliminary user list, divide commodities into long-term commodities and short-term commodities, wherein the long-term commodities refer to commodities with a long usage period and a low purchase frequency, and the short-term commodities refer to commodities with a short usage period and a high purchase frequency; mark commodities purchased by the user according to the long-term commodities and the short-term commodities to form a commodity time vector; generate a user feature report according to the commodity time vector and the user timetable; generate a primary user list and a secondary user list according to the user feature report; S500: Setting a marketing promotion strategy according to the primary user list and the secondary user list; The user feature vector is: ; The marketing product feature vector is: ; in, Indicates the nth product on which the user has taken action. Indicates the user's interest preference for the nth item after time decay. Represents the nth attribute feature of the product, Indicates the value of the weight of the nth attribute after time decay; use the TF-IDF algorithm to and Calculate the value of It shows the changes in users’ interest in products at different times; It shows the changes of attribute weights of commodities at different times; Said Indicates the proportion of products with the nth attribute feature. The specific calculation formula is as follows: in, Indicates the number of items sold with the nth attribute feature. Indicates the total quantity of goods sold; The primary user list refers to users whose similarity between the product to be marketed and the product previously purchased by the user is not less than the similarity threshold and whose time difference between the purchase time node and the current time node is within the corresponding time difference threshold range; the secondary user list refers to users whose similarity between the product to be marketed and the product previously purchased by the user is less than the similarity threshold and whose time difference between the purchase time node and the current time node is greater than the corresponding time difference high threshold or less than the corresponding time difference low threshold; the similarity threshold is pre-set based on historical data and is used to measure the similarity between the product to be marketed and the product purchased by the user; the time difference threshold range is pre-set based on the user's historical purchase frequency data, and the time difference threshold ranges corresponding to the long-term product and the short-term product are different; The commodity level corresponding to the user is obtained according to the primary user list and the secondary user list, the commodity level including high level, medium level and low level, and the commodity level is divided according to a preset purchase rate threshold range, the high level refers to commodities with a purchase rate not less than a high threshold in the purchase rate threshold range, the medium level refers to commodities with a purchase probability within the purchase rate threshold range, and the low level refers to commodities with a purchase probability less than a low threshold in the purchase rate threshold range; the active time period, the stable time period and the slow time period are obtained according to the user schedule, and marketing promotion is carried out according to the time-sharing marketing strategy.
2. The object-based marketing promotion method according to claim 1, characterized in that: The calculation formula of the time decay factor is: in, represents the time decay factor, Represents a weight parameter, which is used to measure the influence of the user's browsing time node and the current time node in time decay. and They represent the decay rate based on the purchase time node and the decay rate based on the browsing time node respectively. Indicates the current time node. Indicates the purchase time node, Indicates the browsing time node.
3. The object-based marketing promotion method according to claim 2, characterized in that: If the purchase time node is before the browsing time node, the time decay factor is calculated based on the purchase time node. If the browsing time node exists, the weight parameter needs to be set according to the time interval and browsing duration. The specific formula is as follows: in, represents the browsing time duration of the user at the browsing time node; k represents the parameter affected by the browsing time, which is used to adjust the browsing time to the weight parameter The parameters that affect the market are pre-set based on historical data and historical experience.
4. The object-based marketing promotion method according to claim 1, characterized in that: The user portrait feature vector .
5. The object-based marketing promotion method according to claim 1, characterized in that: The user schedule includes the user's shopping time period and shopping preferences. The shopping time period includes an active time period, a stable time period and a slow time period. The shopping preferences refer to whether the user has special preferences for commodity attributes when shopping.
6. The object-based marketing promotion method according to claim 1, characterized in that: The push frequency data is obtained according to the operation time data and the user schedule, the push frequency data is normalized and the user's push frequency is predicted using a linear regression method; the user's preference for long-term and short-term products is obtained according to the product time vector, and a push ratio is generated according to the preference; an individual time-sharing marketing strategy is generated according to the push frequency and the push ratio.
Citation Information
Patent Citations
Big-data-based method and system for establishing and analyzing e-commerce user portrait of mobile terminal
CN108021929A
Portrait generation method / system for personalized shopper, storage medium and terminal
CN108197980A
Commodity recommendation list updating method and device, equipment, medium, and product
CN113627995A
Precision marketing method and system based on user portraits
CN117350770A
Commodity combination sales method and system based on e-commerce platform
CN118469609A
