Big data accurate recommendation method, device and equipment and storage medium

Abstract

The application relates to the technical field of big data advertisement recommendation, and discloses a big data accurate recommendation method, device, equipment and storage medium.The method comprises the following steps: obtaining historical behavior data of a user by responding to a client, performing first data processing on the historical behavior data by using a lookalike audience orientation algorithm, and establishing a similar user set; obtaining a feature vector advertisement click rate, performing correlation degree calculation on the click rate by using an Apriori algorithm, and obtaining a correlation degree data set; performing grade division on the correlation degree data set, and obtaining a high correlation degree data set; obtaining the high correlation degree data set, providing product promotion services and product recommendation services for a target user set in the high correlation degree data set, and realizing accurate user recommendation.The method can effectively improve accurate recommendation effect conversion, improve member new user effect driven by recommendation, improve member activation effect driven by recommendation, and improve the growth effect of single order quantity driven by recommendation.

Description

Technical Field

[0001] The present invention relates to the technical field of big data advertising recommendation, and in particular to a method, device, equipment and storage medium for accurate big data recommendation. Background Art

[0002] Nowadays, existing recommendation activities are carried out indiscriminately for all users. Different recommendation activities are not designed for different user groups to achieve different recommendation purposes and solve different user demands. At present, with the continuous development of e-commerce and the continuous improvement of logistics and transportation systems, more and more users choose to consume online. However, due to the different categories, interests, hobbies and income levels of customers, it is determined that recommendation advertisements are sent on a large scale and in a universal manner. However, the advertisements are not sent accurately and accurate recommendations cannot be achieved, which easily leads to low sales recommendation returns and high costs. With the continuous improvement of big data technology, how to perfectly combine big data technology and recommendations to achieve accurate recommendations to enhance customer purchasing experience and purchase loyalty has become the current focus of client recommendation research. Therefore, it is particularly important to invent a recommendation system based on big data. Summary of the Invention

[0003] In view of this, it is necessary to provide a big data accurate recommendation method, device, equipment and storage medium to address the problems of inaccurate advertising delivery, inability to achieve accurate recommendation, and easy result in low sales recommendation return rate and high cost.

[0004] A method, apparatus, device and storage medium for precise recommendation of big data, comprising the following steps: a method for precise recommendation of big data, applied to a client, the method comprising: in response to the client, obtaining historical behavior data of the user, performing a first data processing on the historical behavior data by a lookalike audience targeting algorithm to establish a similar user set; obtaining the similar user set, performing a second data processing on the similar user set by a RALM algorithm to obtain a recommended user set; obtaining the recommended user set, performing a third data processing on the recommended user set by a Top-N sorting algorithm to obtain a target user set; obtaining the target user set, and using the client to send corresponding feature vector advertisements to the target user set; in response to the client, obtaining the click-through rate of the feature vector advertisement, and performing correlation calculation on the click-through rate by an Apriori algorithm to obtain a correlation data set; and performing hierarchical classification on the correlation data set to obtain a high correlation data set; obtaining the high correlation data set, providing product promotion services and product recommendation services to the target user set in the high correlation data set, and realizing precise user recommendation.

[0005] In one embodiment, the response to the client obtains the user's historical behavior data, performs a first data processing on the historical behavior data through a lookalike audience targeting algorithm, and establishes a similar user set, including: obtaining the historical behavior data, using the historical behavior data to perform data integration, and obtaining a user portrait; expanding the display population based on the user portrait to obtain seed user tags, and the user tags include: geography, interests, behaviors, brand preferences, etc.; using a collaborative filtering recommendation algorithm to score the seed user tags, calculate the similarity of the seed user tags, and find the neighboring set of the seed user tags; weighting the neighboring set to obtain the similar user set.

[0006] In one embodiment, the obtaining of the similar user set and performing a second data processing on the similar user set using the RALM algorithm to obtain the recommended user set includes: obtaining the similar user set and a non-similar user set in the client, using the similar user set as a positive sample and the non-similar user set as a negative sample; inputting the positive samples and the negative samples into a preset RALM algorithm model for training to obtain the recommended user set.

[0007] In one embodiment, the step of obtaining the target user set and using the client to send the corresponding feature vector advertisement to the target user set also includes: obtaining the target user set, obtaining user candidate tags of the target user set, and performing nonlinear mapping on the user candidate tags to obtain multiple user feature tags; classifying the user feature tags to obtain multiple classification results, and normalizing the classification results to obtain feature vector data, and sending advertisements to the feature vector data as the feature vector advertisements.

[0008] In one embodiment, the method of obtaining, in response to the client, the feature vector advertisement click-through rate, and calculating the correlation of the click-through rate using the Apriori algorithm to obtain a correlation data set; and hierarchically dividing the correlation data set to obtain a high correlation data set includes: obtaining, in response to the client, the feature vector advertisement click-through rate and the feature vector, making the feature vector and the feature vector advertisement click-through rate into databases X and Y, respectively, and calculating a minimum support threshold Q; retrieving all frequent item sets in the database through continuous iteration to obtain the correlation data set; hierarchically dividing the correlation data set into three levels: high, medium, and low, and obtaining the high correlation data set.

[0009] In one embodiment, the obtaining of the highly correlated data set, providing product promotion services and product recommendation services to the target user set in the highly correlated data set, and achieving accurate user recommendations include: obtaining the target users in the highly correlated data set, formulating a recommendation strategy for the target users, the recommendation strategy including: contact method, delivery channel, push time, push address, recommendation rights, etc.; and using the recommendation strategy to provide the promotion service and product recommendation service to the target users, thereby achieving accurate user recommendations.

[0010] The second embodiment of the present invention provides a big data precision recommendation device including: a first data processing module, used to respond to the client, obtain the user's historical behavior data, perform a first data processing on the historical behavior data through a lookalike audience targeting algorithm, and establish a similar user set; a second data processing module, used to obtain the similar user set, perform a second data processing on the similar user set using a RALM algorithm, and obtain a recommended user set; a third data processing module, used to obtain the recommended user set, perform a third data processing on the recommended user set through a Top-N sorting algorithm, and obtain a target user set.

[0011] A sending module is used to obtain the target user set and use the client to send the corresponding feature vector advertisement to the target user set; an association module is used to respond to the client, obtain the click-through rate of the feature vector advertisement, and use the Apriori algorithm to calculate the correlation of the click-through rate to obtain a correlation data set; and the correlation data set is graded to obtain a high-correlation data set; a precise recommendation module is used to obtain the high-correlation data set, provide product promotion services and product recommendation services to the target user set in the high-correlation data set, and realize precise user recommendations.

[0012] In a second embodiment, the first data processing module further includes an acquisition submodule configured to acquire the historical behavior data, perform data integration using the historical behavior data, and obtain a user profile; perform display population expansion based on the user profile to obtain seed user tags, wherein the user tags include: geography, interests, behaviors, brand preferences, etc.; a calculation submodule configured to score the seed user tags using a collaborative filtering recommendation algorithm, calculate the similarity of the seed user tags, and identify neighboring sets of the seed user tags; and a weighting submodule configured to weight the neighboring sets to obtain the similar user set.

[0013] In a second embodiment, the second data processing module further includes an acquisition submodule for acquiring the similar user set and the non-similar user set in the client, with the similar user set as a positive sample and the non-similar user set as a negative sample; and a training submodule for inputting the positive samples and negative samples into a preset RALM algorithm model for training to obtain a recommended user set.

[0014] In the second embodiment, the sending module further includes an acquisition submodule, which is used to acquire the target user set, obtain user candidate tags of the target user set, and perform nonlinear mapping on the user candidate tags to obtain multiple user feature tags; a classification submodule, which is used to classify the user feature tags to obtain multiple classification results, and normalize the classification results to obtain feature vector data, and send advertisements to the feature vector data as the feature vector advertisements.

[0015] In a second embodiment, the association module further includes a response submodule for responding to the client, obtaining the feature vector advertising click-through rate and the feature vector, making the feature vector and the feature vector advertising click-through rate into databases X and Y respectively, and calculating the minimum support threshold Q; an iteration submodule for retrieving all frequent item sets in the database through continuous iteration to obtain the association data set; and a division submodule for classifying the association data set into three levels: high, medium, and low, and obtaining the high association data set.

[0016] In a second embodiment, the precise recommendation module further includes a formulation submodule for acquiring target users in the highly correlated data set and formulating a recommendation strategy for the target users, wherein the recommendation strategy includes: contact method, delivery channel, push time, push address, recommendation rights, etc.; and a provision submodule for providing the promotion service and product recommendation service to the target users using the recommendation strategy to achieve precise user recommendations.

[0017] A third embodiment of the present invention provides a big data precision recommendation device, comprising: a memory and at least one processor, wherein instructions are stored in the memory, and the memory and the at least one processor are interconnected via a line.

[0018] The at least one processor calls the instructions in the memory to enable the big data precision recommendation device to perform the steps of the above-mentioned big data precision recommendation method.

[0019] A fourth embodiment of the present invention provides a computer-readable storage medium, which stores instructions. When the computer-readable storage medium is run on a computer, it enables the computer to execute the steps of the above-mentioned big data precision recommendation method.

[0020] The above-mentioned big data precision recommendation method, device, equipment and storage medium obtain the user's historical behavior data in response to the client, perform a first data processing on the historical behavior data using the lookalike audience targeting algorithm to establish a similar user set; obtain the similar user set, perform a second data processing on the similar user set using the RALM algorithm to obtain a recommended user set; obtain the recommended user set, perform a third data processing on the recommended user set using the Top-N sorting algorithm to obtain a target user set; obtain the target user set, and use the client to send corresponding feature vector advertisements to the target user set; respond to the client, obtain the click-through rate of the feature vector advertisement, and calculate the correlation of the click-through rate using the Apriori algorithm to obtain a correlation data set; and then classify the correlation data set to obtain a high-correlation data set; obtain the high-correlation data set, and provide product promotion services and product recommendation services to the target user set in the high-correlation data set to achieve user precision recommendation. It can effectively improve the conversion of precise recommendation effects, improve the member acquisition effect driven by recommendations, improve the member activation effect driven by recommendations, and improve the growth effect of individual orders driven by recommendations. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 This is a schematic diagram of the first embodiment of the big data precise recommendation method of the present invention.

[0022] Figure 2 This is a schematic diagram of the second embodiment of the big data precise recommendation method of the present invention.

[0023] Figure 3 This is a schematic diagram of the third embodiment of the big data precise recommendation method of the present invention.

[0024] Figure 4 This is a schematic diagram of the first embodiment of the big data precision recommendation device of the present invention.

[0025] Figure 5 This is a schematic diagram of an embodiment of the big data precision recommendation device of the present invention. DETAILED DESCRIPTION

[0026] The embodiments of the present invention provide a method, apparatus, device, and storage medium for accurate big data recommendation. The method comprises the following steps: obtaining historical behavior data of a user in response to a client, performing a first data processing on the historical behavior data using a lookalike audience targeting algorithm to establish a similar user set; obtaining the similar user set, performing a second data processing on the similar user set using a RALM algorithm to obtain a recommended user set; obtaining the recommended user set, performing a third data processing on the recommended user set using a Top-N sorting algorithm to obtain a target user set; obtaining the target user set, and sending corresponding feature vector advertisements to the target user set using the client; obtaining the click-through rate of the feature vector advertisement in response to the client, and calculating the relevance of the click-through rate using an Apriori algorithm to obtain a relevance dataset; and classifying the relevance dataset to obtain a high-relevance dataset; obtaining the high-relevance dataset, and providing product promotion and recommendation services to the target user set in the high-relevance dataset, thereby achieving accurate user recommendation. The method can effectively improve the conversion rate of accurate recommendation results, enhance the member acquisition effect driven by recommendation, enhance the member activation effect driven by recommendation, and enhance the growth effect of individual orders driven by recommendation.

[0027] The terms "first," "second," "third," "fourth," and so forth (if any) in the specification and claims of the present invention and the appended claims are used to distinguish similar items and are not necessarily used to describe a particular order or precedence. It should be understood that the terms used in this manner are interchangeable where appropriate, so that the embodiments described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "including" or "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product, or apparatus that includes a series of steps or elements is not necessarily limited to those steps or elements expressly listed, but may include other steps or elements not expressly listed or inherent to such process, method, product, or apparatus.

[0028] For ease of understanding, the specific process of the embodiment of the present invention is described below. Figure 1 The first embodiment of the big data accurate recommendation method in the embodiment of the present invention includes:

[0029] Step 101: In response to the client, obtain the user's historical behavior data, perform the first data processing on the historical behavior data through the lookalike audience targeting algorithm, and establish a similar user set.

[0030] In this embodiment, before the client or system accurately recommends advertisements to customers, it first needs to obtain the user's historical behavior data. The historical behavior data can be filtered through user portrait tags to obtain seed user tags. The filtered tags include but are not limited to: geography, interests, behavior, brand preferences, etc.; the seed user tags are scored using a collaborative filtering recommendation algorithm, the similarity of the seed user tags is calculated, and the neighboring set of seed user tags is found; the seed users based on the nearby set are expanded, and after the expansion, a similar user set can be obtained.

[0031] Specifically, the client described in the present invention is a logistics address input system developed based on Android and iOS programming languages. The client is at least one of a mobile terminal, a WeChat terminal, an Android terminal and an iOS terminal; optionally, the client, or mobile phone user terminal, refers to a program corresponding to the server that provides local services to customers. Except for some applications that only run locally, they are generally installed on ordinary mobile phones and need to cooperate with the server to run. The more commonly used mobile phone user terminals include web browsers used for the World Wide Web, email clients for sending and receiving emails, and instant messaging client software. For this type of application, there needs to be corresponding servers and service programs in the network to provide corresponding services, such as database services, email services, etc., so that a specific communication connection needs to be established between the client and the server to ensure the normal operation of the application.

[0032] Step 102: Obtain a similar user set, and perform a second data processing on the similar user set using the RALM algorithm to obtain a recommended user set.

[0033] In this embodiment, after obtaining a similar user set, the client needs to filter and locate similar users to obtain a more accurate recommended user set. This recommended user set serves as a preselected group for ad recommendations. If the client requires large-scale, indiscriminate advertising, this recommended user set can be selected for ad placement. The recommended user set is a further filter of the similar user set using the RALM algorithm, and the user profile labels obtained are unified.

[0034] Step 103: Obtain a recommended user set, perform a third data processing on the recommended user set using a Top-N sorting algorithm, and obtain a target user set.

[0035] In this embodiment, after the client or system obtains the recommended user set, if more accurate ad recommendations are needed, a third data screening of the recommended user set is performed. This third data screening utilizes a Top-N ranking algorithm to calculate the top-N of the recommended user set. The Top-N analysis method uses the Top-N algorithm to obtain the required N data from the research object and then conducts a focused analysis on these N data.

[0036] Specifically, the recommended user set data is first processed locally. For example, to calculate the Top-N number of word occurrences, the number of words in each region can be locally counted, rewritten to a temporary file, and then the data can be merged later. Another Top-N approach is to independently calculate the recommended user set data by key, and the regions are partitioned by key. For example, the total statistical count can be directly obtained by word, which is similar to the Storm processing method. Top-n can reversely deduce from the necessary and sufficient conditions to obtain some target user restrictions, thereby filtering out data, reducing the data volume, and gradually converging to the target user feature range T. Feature range T is a system-preset range including but not limited to: age, gender, height, weight, hobbies, etc. The target user set obtained by the Top-N sorting algorithm can be used for accurate recommendations.

[0037] Step 104: Acquire a target user set, and use the client to send a corresponding feature vector advertisement to the target user set.

[0038] In this embodiment, after the client obtains the target user set, it can send user-specific feature vector advertisements to the target user set. First, the client obtains multiple portrait label feature vectors corresponding to the user's portrait label features; generates a first similarity coefficient between the portrait label feature vector and all portrait label feature vectors; sums the first similarity coefficient corresponding to each portrait label feature vector to obtain a second similarity coefficient; selects the portrait label feature vector corresponding to the second similarity coefficient that is not greater than the similarity coefficient threshold from the multiple portrait label feature vectors. The system analyzes the portrait label feature vectors and delivers advertisements that have a higher similarity to the portrait label feature vector threshold.

[0039] Step 105: In response to the client, obtain the feature vector advertisement click-through rate, and calculate the relevance of the click-through rate using the Apriori algorithm to obtain a relevance data set; and classify the relevance data set to obtain a high-relevance data set.

[0040] In this embodiment, after the corresponding feature vector advertisement is placed, the system first obtains the user click-through rate of the feature vector advertisement and calculates the correlation of the user click-through rate using the Apriori algorithm. First, all user click-through rates are found, and the frequency of occurrence of these user click-through rates is at least the same as the predefined minimum support. Then, strong association rules are generated based on the user click-through rates. These rules must meet the minimum support and minimum credibility. The user click-through rates found in step 1 are then used to generate the desired rules, generating all rules that only contain items in the set, where each rule has only one item on the right side. The definition of the middle rule is adopted here. Once these rules are generated, only those rules that are greater than the minimum credibility given by the system are retained, thereby obtaining a correlation data set. In order to generate all correlation data sets, the system uses a recursive method.

[0041] The system classifies the correlation data set into three levels: high, medium and low. The high correlation data set is the key data set of the present invention.

[0042] The Apriori algorithm is the first association rule mining algorithm. It uses an iterative, layer-by-layer search method to identify relationships between item sets in a database and form rules. The process consists of concatenation (a matrix-like operation) and pruning (removing unnecessary intermediate results). In this algorithm, the concept of an itemset is a collection of items. A set of K items is called a k-itemset. The frequency of an itemset is the number of transactions containing the itemset, which is called the itemset frequency. If an itemset meets the minimum support, it is called a frequent itemset.

[0043] Step 106: Obtain a highly correlated data set, and provide product promotion services and product recommendation services to the target user set in the highly correlated data set to achieve accurate user recommendations.

[0044] In this embodiment, after the system obtains a highly correlated data set, it formulates a recommendation strategy for the target users in the highly correlated data set. The recommendation strategy includes, but is not limited to, contact methods, delivery channels, push time, push address, and recommendation benefits. It utilizes SMS, WeChat template messages, WeChat service notifications, Alipay notifications, app notifications, and pop-up windows to precisely reach the target audience. Contact points are developed based on user pain points, such as proactively sending group discounts to users who have not placed an order for a long time (churned users) to achieve user conversion; or retaining users who have not placed an order in a timely manner; or recommending contact points for discount-sensitive users. For users who complain, customer service can provide timely compensation with discounts; or shopping guides can issue exclusive discounts to bound existing users, ultimately achieving order conversion. Alternatively, exclusive recommendation QR codes can be posted on express packaging and delivery vehicles to provide users with surprises. All of the above contact scenarios and actions are labeled to facilitate continuous recommendation.

[0045] Recommendation channels: Connect multiple channels to achieve unified coupon channel recommendations. By creating recommendation strategies, binding recommendation channels, and generating channel-specific QR codes or access links, users from different channels can access the links and use OneID to accurately recommend to users who meet the requirements. Recommendation channels and recommendation touchpoints are closely linked; online and offline channels are used. Online channels primarily include proprietary application channels and cross-industry alliance channels. Cross-industry channels require system-level integration (such as coupon issuance and verification integration, and points redemption integration). Offline channels primarily rely on outlets and salesperson channels (such as generating outlet-specific referral codes and salesperson-specific referral codes).

[0046] Recommendation goals and strategies: Determine the recommendation cost and goal, create a recommendation strategy, select the target audience, choose an automated recommendation strategy (e.g., execute recommendation benefit a when target audience A meets condition 1; execute recommendation benefit b when target audience A meets condition 2), select the reach method, and the delivery channel; the target audience for automated recommendation delivery is a core element, such as the definition of discount-sensitive users (who place orders with discounts more frequently than normal orders), and analyzing user behavior by ordering time and date to determine the time period with the highest conversion rate for user acquisition activities, and then accurately push ads to avoid invalid recommendations. Alternatively, analyze user behavior on the activity page to analyze user focus and conduct precise delivery. Using recommendation strategies to deliver ads to target users can increase the conversion rate of ads, thereby achieving accurate user recommendations.

[0047] In the implementation of the present invention, by responding to the client, the user's historical behavior data is obtained, and the historical behavior data is processed for the first time by the lookalike audience targeting algorithm to establish a similar user set; the similar user set is obtained, and the similar user set is processed for the second time by the RALM algorithm to obtain a recommended user set; the recommended user set is obtained, and the recommended user set is processed for the third time by the Top-N sorting algorithm to obtain a target user set; the target user set is obtained, and the client is used to send corresponding feature vector advertisements to the target user set; in response to the client, the click-through rate of the feature vector advertisement is obtained, and the click-through rate is calculated using the Apriori algorithm to obtain a correlation data set; and the correlation data set is graded to obtain a high correlation data set; the high correlation data set is obtained, and product promotion services and product recommendation services are provided to the target user set in the high correlation data set to achieve user accurate recommendation. It can effectively improve the conversion of accurate recommendation effects, improve the member acquisition effect driven by recommendation, improve the member activation effect driven by recommendation, and improve the growth effect of the number of individual orders driven by recommendation.

[0048] See also Figure 2 The second embodiment of the big data accurate recommendation method in the embodiment of the present invention includes:

[0049] Step 201: In response to the client, obtain the user's historical behavior data, perform the first data processing on the historical behavior data through the lookalike audience targeting algorithm, and establish a similar user set.

[0050] Step 202: Obtain a similar user set and a non-similar user set in the client, and use the similar user set as a positive sample and the non-similar user set as a negative sample.

[0051] In this embodiment, after the client or system obtains the similar user set, it also collects the non-similar user set, uses the similar users as positive samples and the non-similar user set as negative samples for calculation sampling.

[0052] Step 203: Input the positive samples and negative samples into a preset RALM algorithm model for training to obtain a recommended user set.

[0053] In this embodiment, the similar user set and the non-similar user set are respectively used as positive samples and negative samples, which are input into the preset RALM algorithm for training to obtain the recommended user set. The RALM algorithm is a nonlinear optimization method between the Newton method and the gradient descent method. It is insensitive to over-parameterization problems, can effectively handle redundant parameter problems, and greatly reduces the chance of the cost function falling into a local minimum. These characteristics make the LM algorithm widely used in fields such as computer vision.

[0054] Step 204: Obtain a recommended user set, perform a third data processing on the recommended user set using a Top-N sorting algorithm, and obtain a target user set.

[0055] Step 205: Acquire a target user set, and use the client to send a corresponding feature vector advertisement to the target user set.

[0056] Step 206 : In response to the client, obtain the feature vector advertisement click-through rate, and calculate the relevance of the click-through rate using the Apriori algorithm to obtain a relevance data set; and classify the relevance data set to obtain a high-relevance data set.

[0057] Step 207: Acquire a highly correlated dataset, and provide product promotion services and product recommendation services to the target user set in the highly correlated dataset to achieve accurate user recommendations.

[0058] Steps 204-207 in this embodiment are similar to steps 103-106 in the first embodiment, and step 201 is similar to step 101, and are not described again here.

[0059] In the implementation of the present invention, by responding to the client, the user's historical behavior data is obtained, and the historical behavior data is processed for the first time by the lookalike audience targeting algorithm to establish a similar user set; the similar user set is obtained, and the similar user set is processed for the second time by the RALM algorithm to obtain a recommended user set; the recommended user set is obtained, and the recommended user set is processed for the third time by the Top-N sorting algorithm to obtain a target user set; the target user set is obtained, and the client is used to send corresponding feature vector advertisements to the target user set; in response to the client, the click-through rate of the feature vector advertisement is obtained, and the click-through rate is calculated using the Apriori algorithm to obtain a correlation data set; and the correlation data set is graded to obtain a high correlation data set; the high correlation data set is obtained, and product promotion services and product recommendation services are provided to the target user set in the high correlation data set to achieve user accurate recommendation. It can effectively improve the conversion of accurate recommendation effects, improve the member acquisition effect driven by recommendation, improve the member activation effect driven by recommendation, and improve the growth effect of the number of individual orders driven by recommendation.

[0060] See also Figure 3 The third embodiment of the big data accurate recommendation method in the embodiment of the present invention includes:

[0061] Step 301: In response to the client, obtain the user's historical behavior data, perform the first data processing on the historical behavior data through the lookalike audience targeting algorithm, and establish a similar user set.

[0062] Step 302: Obtain a similar user set, and perform a second data processing on the similar user set using the RALM algorithm to obtain a recommended user set.

[0063] Step 303: Obtain a recommended user set, perform a third data processing on the recommended user set using a Top-N sorting algorithm, and obtain a target user set.

[0064] Step 304: Acquire a target user set, and use the client to send a corresponding feature vector advertisement to the target user set.

[0065] Step 305: In response to the client, obtain the feature vector advertisement click rate and the feature vector, create the feature vector and the feature vector advertisement click rate into databases X and Y respectively, and calculate the minimum support threshold Q.

[0066] In this embodiment, after the client or system obtains the click-through rate of the feature vector advertisement, the feature vector advertisement and the click-through rate are respectively made into database X and database Y, and the minimum support threshold Q of the feature vector advertisement for the click-through rate is calculated in databases X and Y. The threshold Q is within the range defined by the client or system.

[0067] Step 306: Through continuous iteration, all frequent item sets in the database are retrieved to obtain a correlation dataset.

[0068] In this embodiment, by continuously calculating the support threshold Q of the feature vector advertisement and the click rate, the most frequently clicked frequent item sets in the database Y are retrieved, and the association dataset is obtained.

[0069] Step 307: Classify the correlation dataset into three levels: high, medium, and low, and obtain a high correlation dataset.

[0070] In this embodiment, the client or system collects the correlation data set and divides the correlation data set into three levels: high, medium, and low. The high correlation data set is the important data set in the present invention.

[0071] Step 308: Acquire a highly correlated data set, and provide product promotion services and product recommendation services to the target user set in the highly correlated data set to achieve accurate user recommendations.

[0072] Steps 301-304 in this embodiment are similar to steps 101-104 in the first embodiment, and step 309 is similar to step 106, which will not be repeated here.

[0073] In the implementation of the present invention, by responding to the client, the user's historical behavior data is obtained, and the historical behavior data is processed for the first time by the lookalike audience targeting algorithm to establish a similar user set; the similar user set is obtained, and the similar user set is processed for the second time by the RALM algorithm to obtain a recommended user set; the recommended user set is obtained, and the recommended user set is processed for the third time by the Top-N sorting algorithm to obtain a target user set; the target user set is obtained, and the client is used to send corresponding feature vector advertisements to the target user set; in response to the client, the click-through rate of the feature vector advertisement is obtained, and the click-through rate is calculated using the Apriori algorithm to obtain a correlation data set; and the correlation data set is graded to obtain a high correlation data set; the high correlation data set is obtained, and product promotion services and product recommendation services are provided to the target user set in the high correlation data set to achieve user accurate recommendation. It can effectively improve the conversion of accurate recommendation effects, improve the member acquisition effect driven by recommendation, improve the member activation effect driven by recommendation, and improve the growth effect of the number of individual orders driven by recommendation.

[0074] The above describes the method for accurate recommendation of big data in the embodiment of the present invention. The following describes the device for accurate recommendation of big data in the embodiment of the present invention. Figure 4The first embodiment of the big data accurate recommendation device in the embodiment of the present invention includes:

[0075] The first data processing module 401 is used to respond to the client, obtain the user's historical behavior data, perform a first data processing on the historical behavior data through a lookalike audience targeting algorithm, and establish a similar user set.

[0076] The second data processing module 402 is configured to obtain the similar user set, and perform a second data processing on the similar user set using the RALM algorithm to obtain a recommended user set.

[0077] The third data processing module 403 is used to obtain the recommended user set, perform a third data processing on the recommended user set using a Top-N sorting algorithm, and obtain a target user set.

[0078] The sending module 404 is configured to obtain the target user set and send corresponding feature vector advertisements to the target user set using the client.

[0079] The association module 405 is used to respond to the client, obtain the feature vector advertisement click rate, and calculate the correlation of the click rate using the Apriori algorithm to obtain a correlation data set; and classify the correlation data set to obtain a high correlation data set.

[0080] The precise recommendation module 406 is used to obtain the highly correlated data set, provide product promotion services and product recommendation services to the target user set in the highly correlated data set, and achieve precise user recommendations.

[0081] In an embodiment of the present invention, by responding to a client, obtaining the user's historical behavior data, performing a first data processing on the historical behavior data using a lookalike audience targeting algorithm to establish a similar user set; obtaining the similar user set, performing a second data processing on the similar user set using a RALM algorithm to obtain a recommended user set; obtaining the recommended user set, performing a third data processing on the recommended user set using a Top-N sorting algorithm to obtain a target user set; obtaining the target user set, using the client to send a corresponding feature vector advertisement to the target user set; responding to the client, obtaining the click-through rate of the feature vector advertisement, and performing a correlation calculation on the click-through rate using an Apriori algorithm to obtain a correlation data set; and performing hierarchical classification on the correlation data set to obtain a high correlation data set; obtaining the high correlation data set, providing product promotion services and product recommendation services to the target user set in the high correlation data set, and realizing accurate user recommendations. This can effectively improve the conversion of accurate recommendation effects, improve the member acquisition effect driven by recommendations, improve the member activation effect driven by recommendations, and improve the growth effect of the number of individual orders driven by recommendations.

[0082] above Figure 4 The big data precise recommendation apparatus in the embodiment of the present invention is described in detail from the perspective of modular functional entities. The big data precise recommendation device in the embodiment of the present invention is described in detail from the perspective of hardware processing.

[0083] Figure 5 Figure 8 is a schematic diagram of the structure of a device for precise big data recommendation provided by an embodiment of the present invention. The device 800 may vary significantly depending on configuration or performance. It may include one or more central processing units (CPUs) 810 (e.g., one or more processors), memory 820, and one or more storage media 830 (e.g., one or more mass storage devices) for storing applications 833 or data 832. The memory 820 and storage medium 830 may be either transient or persistent storage. The program stored in the storage medium 830 may include one or more modules (not shown), each of which may include a series of instructions operating on the device 800. Furthermore, the processor 810 may be configured to communicate with the storage medium 830, executing the series of instructions stored in the storage medium 830 on the device 800.

[0084] The big data precision recommendation device 800 may also include one or more power supplies 840, one or more wired or wireless network interfaces 850, one or more input and output interfaces 860, and / or one or more operating clients 831, such as Windows Server, Mac OS X, Unix, Linux, FreeBSD, etc. It will be understood by those skilled in the art that Figure 5 The structure of the big data precision recommendation device shown does not constitute a limitation on the big data precision recommendation device provided in this application, and may include more or fewer components than shown in the figure, or combine certain components, or arrange the components differently.

[0085] A big data precision recommendation device is used to implement the following big data precision recommendation method, specifically including: being applied to a client, the big data precision recommendation method includes: responding to the client, obtaining the user's historical behavior data, performing a first data processing on the historical behavior data using a lookalike audience targeting algorithm to establish a similar user set; obtaining the similar user set, performing a second data processing on the similar user set using a RALM algorithm to obtain a recommended user set; obtaining the recommended user set, performing a third data processing on the recommended user set using a Top-N sorting algorithm to obtain a target user set; obtaining the target user set, and using the client to send corresponding feature vector advertisements to the target user set; responding to the client, obtaining the feature vector advertisement click-through rate, and calculating the correlation of the click-through rate using an Apriori algorithm to obtain a correlation data set; and performing hierarchical classification on the correlation data set to obtain a high-correlation data set; obtaining the high-correlation data set, providing product promotion services and product recommendation services to the target user set in the high-correlation data set, and realizing user precision recommendation.

[0086] In one embodiment, in a first implementation of the first aspect of the present invention, the response to the client obtains the user's historical behavior data, performs a first data processing on the historical behavior data through a lookalike audience targeting algorithm, and establishes a similar user set including: obtaining the historical behavior data, using the historical behavior data to perform data integration, and obtaining a user portrait; expanding the display population based on the user portrait to obtain seed user tags, and the user tags include: geography, interests, behaviors, brand preferences, etc.; using a collaborative filtering recommendation algorithm to score the seed user tags, calculate the similarity of the seed user tags, and find the neighboring set of the seed user tags; weighting the neighboring set to obtain the similar user set.

[0087] In one embodiment, in a second implementation of the first aspect of the present invention, obtaining the similar user set, performing a second data processing on the similar user set using the RALM algorithm, and obtaining the recommended user set includes: obtaining the similar user set and the non-similar user set in the client, using the similar user set as a positive sample and the non-similar user set as a negative sample; inputting the positive samples and the negative samples into a preset RALM algorithm model for training to obtain the recommended user set.

[0088] In one embodiment, in the third implementation method of the first aspect of the present invention, the step of obtaining the target user set and using the client to send the corresponding feature vector advertisement to the target user set also includes: obtaining the target user set, obtaining the user candidate labels of the target user set, and performing nonlinear mapping on the user candidate labels to obtain multiple user feature labels; classifying the user feature labels to obtain multiple classification results, and normalizing the classification results to obtain feature vector data, and sending advertisements to the feature vector data as the feature vector advertisements.

[0089] In one embodiment, in a fourth implementation of the first aspect of the present invention, obtaining the highly correlated dataset, providing product promotion services and product recommendation services to the target user set in the highly correlated dataset, and achieving precise user recommendations includes: obtaining the target users in the highly correlated dataset, formulating a recommendation strategy for the target users, the recommendation strategy including: reach method, delivery channel, push time, push address, recommendation rights, etc.; and utilizing the recommendation strategy to provide the promotion services and product recommendation services to the target users, thereby achieving precise user recommendations. The highly correlated dataset is graded into three levels: high, medium, and low, and the highly correlated dataset is obtained.

[0090] In one embodiment, in the fifth implementation method of the first aspect of the present invention, the obtaining of the highly correlated data set, providing product promotion services and product recommendation services to the target user set in the highly correlated data set, and realizing accurate user recommendations include: obtaining the target users in the highly correlated data set, formulating a recommendation strategy for the target users, the recommendation strategy including: contact method, delivery channel, push time, push address, recommendation rights, etc.; using the recommendation strategy to provide the target users with the promotion service and product recommendation service, and realizing accurate user recommendations.

[0091] The present invention also provides a computer-readable storage medium, which may be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium. The computer-readable storage medium stores instructions, which, when executed on a computer, enable the computer to execute the following steps of the big data precision recommendation method.

[0092] Applied to the client, the big data precise recommendation method includes: Applied to the client, the big data precise recommendation method includes: responding to the client, obtaining the user's historical behavior data, performing a first data processing on the historical behavior data through the lookalike audience targeting algorithm, and establishing a similar user set; obtaining the similar user set, performing a second data processing on the similar user set using the RALM algorithm, and obtaining a recommended user set; obtaining the recommended user set, performing a third data processing on the recommended user set through the Top-N sorting algorithm, and obtaining a target user set; obtaining the target user set, and using the client to send corresponding feature vector advertisements to the target user set; responding to the client, obtaining the feature vector advertisement click-through rate, and calculating the correlation of the click-through rate using the Apriori algorithm to obtain a correlation data set; and performing hierarchical classification on the correlation data set to obtain a high-correlation data set; obtaining the high-correlation data set, and providing product promotion services and product recommendation services to the target user set in the high-correlation data set to achieve user precise recommendation.

[0093] In one embodiment, in a first implementation of the first aspect of the present invention, the response to the client obtains the user's historical behavior data, performs a first data processing on the historical behavior data through a lookalike audience targeting algorithm, and establishes a similar user set including: obtaining the historical behavior data, using the historical behavior data to perform data integration, and obtaining a user portrait; expanding the display population based on the user portrait to obtain seed user tags, and the user tags include: geography, interests, behaviors, brand preferences, etc.; using a collaborative filtering recommendation algorithm to score the seed user tags, calculate the similarity of the seed user tags, and find the neighboring set of the seed user tags; weighting the neighboring set to obtain the similar user set.

[0094] In one embodiment, in the second implementation method of the first aspect of the present invention, the step of obtaining the target user set and using the client to send the corresponding feature vector advertisement to the target user set also includes: obtaining the target user set, obtaining the user candidate labels of the target user set, and performing nonlinear mapping on the user candidate labels to obtain multiple user feature labels; classifying the user feature labels to obtain multiple classification results, and normalizing the classification results to obtain feature vector data, and sending advertisements to the feature vector data as the feature vector advertisements.

[0095] In one embodiment, in a third implementation of the first aspect of the present invention, the response to the client obtains the feature vector advertising click-through rate, and calculates the correlation of the click-through rate using the Apriori algorithm to obtain a correlation data set; and the correlation data set is graded to obtain a high correlation data set, including: responding to the client, obtaining the feature vector advertising click-through rate and feature vector, making the feature vector and feature vector advertising click-through rate into databases X and Y respectively, and calculating the minimum support threshold Q; through continuous iteration, retrieving all frequent item sets in the database to obtain the correlation data set; graded the correlation data set into three levels: high, medium, and low, and obtaining the high correlation data set.

[0096] In one embodiment, in the fourth implementation method of the first aspect of the present invention, the obtaining of the highly correlated data set, providing product promotion services and product recommendation services to the target user set in the highly correlated data set, and realizing accurate user recommendations include: obtaining the target users in the highly correlated data set, formulating a recommendation strategy for the target users, the recommendation strategy including: contact method, delivery channel, push time, push address, recommendation rights, etc.; using the recommendation strategy to provide the promotion service and product recommendation service to the target users, and realizing accurate user recommendations.

[0097] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the client, device and unit described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0098] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the portion that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes various media that can store program code, such as a USB flash drive, a mobile hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0099] As described above, the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that the technical solutions described in the above embodiments can still be modified, or some of the technical features thereof can be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method, device, equipment and storage medium for accurate recommendation of big data, characterized in that: The big data precise recommendation method includes: In response to the client, the user's historical behavior data is obtained, and the historical behavior data is processed for the first time using a lookalike audience targeting algorithm to establish a similar user set; Obtain the similar user set and the non-similar user set in the client, use the similar user set as positive samples and the non-similar user set as negative samples; input the positive and negative samples into a preset RALM algorithm model for training to obtain a recommended user set; the recommended user set is a further screening of the similar user set using the RALM algorithm, and the screened user portrait labels are unified; Obtain the recommended user set, and perform a third data processing on the recommended user set using a Top-N sorting algorithm to obtain a target user set; Acquire the target user set, and use the client to send corresponding feature vector advertisements to the target user set; In response to the client, obtaining the feature vector advertisement click-through rate, and calculating the relevance of the click-through rate using the Apriori algorithm to obtain a relevance data set; and classifying the relevance data set to obtain a high-relevance data set; The highly correlated data set is obtained, and product promotion services and product recommendation services are provided to the target user set in the highly correlated data set to achieve accurate user recommendations.

2. The method, apparatus, device, and storage medium for accurate big data recommendation according to claim 1, wherein: The step of obtaining the user's historical behavior data in response to the client, performing a first data processing on the historical behavior data using a lookalike audience targeting algorithm, and establishing a similar user set includes: Acquire the historical behavior data, use the historical behavior data to perform data integration, and obtain a user profile; expand the display population based on the user profile to obtain seed user tags, wherein the user tags include: geography, interests, behavior, brand preferences, etc.; The seed user tags are scored using a collaborative filtering recommendation algorithm, the similarity of the seed user tags is calculated, and a neighboring set of the seed user tags is found; the neighboring set is weighted to obtain the similar user set.

3. The method, apparatus, device and storage medium for accurate big data recommendation according to claim 1, characterized in that: Before obtaining the target user set and sending corresponding feature vector advertisements to the target user set using the client, the following step further comprises: Acquire the target user set, obtain user candidate labels of the target user set, and perform nonlinear mapping on the user candidate labels to obtain multiple user feature labels; The user feature labels are classified to obtain a plurality of classification results, the classification results are normalized to obtain feature vector data, and feature vector advertisements are sent to the feature vector data.

4. The method, apparatus, device, and storage medium for accurate big data recommendation according to claim 1, wherein: In response to the client, obtaining the feature vector advertisement click-through rate, and calculating the relevance of the click-through rate using an Apriori algorithm to obtain a relevance dataset; The correlation data set is graded to obtain a high correlation data set including: In response to the client, the feature vector advertisement click rate and the feature vector advertisement are obtained, the feature vector and the feature vector advertisement click rate are respectively made into databases X and Y, and a minimum support threshold Q of the user click rate is calculated; Through continuous iteration, all frequent item sets in the database are retrieved to obtain the association degree data set; The correlation data set is graded into three grades: high, medium, and low, and the high correlation data set is obtained.

5. The method, apparatus, device and storage medium for accurate big data recommendation according to claim 1, characterized in that: The obtaining of the highly correlated data set and providing product promotion services and product recommendation services to the target user set in the highly correlated data set to achieve accurate user recommendations includes: Obtain target users from the highly correlated data set and formulate a recommendation strategy for the target users, the recommendation strategy including: contact method, delivery channel, push time, push address, recommendation rights, etc.; The recommendation strategy is used to provide the promotion service and product recommendation service to the target user to achieve accurate user recommendation.

6. A big data precision recommendation device, characterized in that: The big data precise recommendation device includes: A first data processing module is configured to respond to the client, obtain the user's historical behavior data, perform a first data processing on the historical behavior data using a lookalike audience targeting algorithm, and establish a similar user set; The second data processing module is used to obtain the similar user set and the non-similar user set in the client, use the similar user set as positive samples and the non-similar user set as negative samples; input the positive and negative samples into a preset RALM algorithm model for training to obtain a recommended user set; the recommended user set is a further screening of the similar user set using the RALM algorithm, and the screened user portrait labels are unified; A third data processing module is used to obtain the recommended user set, perform a third data processing on the recommended user set using a Top-N sorting algorithm, and obtain a target user set; a sending module, configured to obtain the target user set and send corresponding feature vector advertisements to the target user set using the client; an association module, configured to obtain, in response to the client, a click-through rate of the feature vector advertisement, and calculate a correlation degree of the click-through rate using an Apriori algorithm to obtain a correlation degree data set; and to classify the correlation degree data set to obtain a high correlation degree data set; The precise recommendation module is used to obtain the highly correlated data set, provide product promotion services and product recommendation services to the target user set in the highly correlated data set, and realize precise user recommendation.

7. The big data accurate recommendation device according to claim 6, characterized in that: The second data processing module includes: An acquisition submodule, configured to acquire the similar user set and the non-similar user set in the client, and use the similar user set as a positive sample and the non-similar user set as a negative sample; The training submodule is used to input the positive samples and negative samples into a preset RALM algorithm model for training to obtain a recommended user set.

8. A big data precision recommendation device, characterized in that: The big data precision recommendation device includes: a memory and at least one processor, the memory stores instructions, and the memory and the at least one processor are interconnected through lines; the at least one processor calls the instructions in the memory so that the big data precision recommendation device performs the steps of the big data precision recommendation method as described in any one of claims 1-5.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the big data precision recommendation method according to any one of claims 1 to 6 are implemented.

Images