Method and device for recommending hotspot data

Abstract

The present invention discloses a method and device for recommending hot data, and relates to the field of computer technology. A specific implementation of the method includes: calculating the heat score of data to generate a heat label of the data; using the heat label to mark the feature vector of the data to generate training data; training the training data set to obtain a hot data recommendation model; using the hot data recommendation model to determine the hot data and recommend it. This implementation uses cutting-edge deep learning technology to construct a training data set with significant potential growth through a heuristic method, predicts hot data that may be popular and have a large growth rate in the short term in the future, and displays it to users. While attracting user attention and improving user experience, it also makes the recommendation of hot data more real-time and diverse.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to a method and device for recommending hotspot data. Background Art

[0002] With the development of big data and the Internet, more and more users like to obtain hot data from the Internet. For example, they can obtain real-time hot news from websites, popular products from online shopping platforms, etc. Accordingly, in order to attract users, various websites or platforms will also recommend hot data to users.

[0003] Taking online shopping as an example, generally speaking, different users have different levels of interest in different categories of content, especially inactive users or newly registered users. It is meaningful to quickly attract their attention, grasp their shopping needs, and provide them with refreshing and timely enjoyment of products.

[0004] When making data recommendations for users, existing technologies mainly consider the popularity of the data or the mid- to long-term popularity of all products. When the user's historical behavior cannot be captured, the general hot data is directly displayed.

[0005] In the process of implementing the present invention, the inventors found that there are at least the following problems in the prior art:

[0006] Directly displaying general hot data may result in missing data that is not currently popular but has a significant upward trend, so the recommendation results are not scientific and reasonable. Summary of the invention

[0007] In view of this, an embodiment of the present invention provides a method and device for recommending hot data, which can apply cutting-edge deep learning technology, construct a training data set with significant potential growth through heuristic methods, predict hot data that may be popular and have a large growth rate in a short period of time in the future, and display it to users. While attracting user attention and improving user experience, it also makes the recommendation of hot data more real-time and diverse.

[0008] To achieve the above objective, according to one aspect of an embodiment of the present invention, a method for recommending hotspot data is provided.

[0009] A method for recommending hot data includes: calculating the heat score of data to generate a heat label of the data; using the heat label to mark the feature vector of the data to generate training data; training the training data set to obtain a hot data recommendation model; and using the hot data recommendation model to determine the hot data and make recommendations.

[0010] Optionally, calculating the heat score of data to generate a heat tag for the data includes: calculating an access increase rate of the data in a recent period; calculating the heat score of the data based on the access increase rate; and generating a heat tag for the data by comparing the heat score with a preset threshold.

[0011] Optionally, calculating the increase rate of access to data in a recent period includes: calculating an average first number of accesses per day to the data in the recent period, and an average second number of accesses per day to the data in a specified period, wherein the specified period is longer than the recent period and includes the recent period; and taking the ratio of the first number of accesses to the second number of accesses as the increase rate of access to the data in the recent period.

[0012] Optionally, generating a heat label for the data by comparing the heat score with a preset threshold includes: comparing the heat score with a preset threshold, and if the heat score is greater than the preset threshold, setting the heat label for the data to 1; otherwise, setting the heat label for the data to 0.

[0013] Optionally, training the training data set to obtain a hot data recommendation model includes: dividing the training data in the training data set into a discrete data set and a continuous data set; using different networks to perform feature processing on the discrete data set and the continuous data set, respectively, and splicing the processed features; inputting the spliced ​​features into the reverse joint training layer to obtain the hot data recommendation model.

[0014] Optionally, for the discrete data set, feature selection is first performed using leaf nodes of a gradient descent tree, and then a low-dimensional dense feature vector is generated for the selected features through a factorization machine for feature processing.

[0015] Optionally, for the continuous data set, a deep learning neural network is used to perform feature processing.

[0016] According to another aspect of an embodiment of the present invention, a device for recommending hotspot data is provided.

[0017] A hot data recommendation device includes: a label generation module, used to calculate the heat score of data to generate the heat label of the data; a data labeling module, used to label the feature vector of the data using the heat label to generate training data; a model training module, used to obtain a hot data recommendation model by training the training data set; a data recommendation module, used to determine the hot data and recommend it using the hot data recommendation model.

[0018] According to yet another aspect of an embodiment of the present invention, an electronic device for recommending hotspot data is provided.

[0019] An electronic device for recommending hotspot data includes: one or more processors; a storage device for storing one or more programs, when the one or more programs are executed by the one or more processors, the one or more processors implement the hotspot data recommendation method provided by an embodiment of the present invention.

[0020] According to still another aspect of the embodiments of the present invention, a computer readable medium is provided.

[0021] A computer readable medium stores a computer program, which, when executed by a processor, implements a method for recommending hotspot data provided by an embodiment of the present invention.

[0022] An embodiment of the above invention has the following advantages or beneficial effects: generating a heat label of the data by calculating the heat score of the data; marking the feature vector of the data with the heat label to generate training data; obtaining a hot data recommendation model by training the training data set; determining the hot data and recommending it using the hot data recommendation model. Cutting-edge deep learning technology can be applied to construct a training data set with significant potential growth through heuristic methods, predicting the hot data that may be popular and have a large growth rate in the short term in the future, and displaying it to users. While attracting user attention and improving user experience, it also makes the recommendation of hot data more real-time and diverse.

[0023] The further effects of the above-mentioned non-conventional optional manner will be described below in conjunction with the specific implementation manner. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] The accompanying drawings are used to better understand the present invention and do not constitute an improper limitation of the present invention.

[0025] Figure 1 is a schematic diagram of main steps of a method for recommending hotspot data according to an embodiment of the present invention;

[0026] Figure 2 It is a schematic diagram of the implementation principle flow of an embodiment of the present invention;

[0027] Figure 3 It is a schematic diagram of the functional implementation principle of an embodiment of the present invention;

[0028] Figure 4 is a schematic diagram of the model training process of an embodiment of the present invention;

[0029] Figure 5 is a schematic diagram of main modules of a device for recommending hotspot data according to an embodiment of the present invention;

[0030] Figure 6 is an exemplary system architecture diagram to which embodiments of the present invention may be applied;

[0031] Figure 7 It is a schematic diagram of the structure of a computer system of a terminal device or a server suitable for implementing an embodiment of the present invention. DETAILED DESCRIPTION

[0032] The following is a description of exemplary embodiments of the present invention in conjunction with the accompanying drawings, including various details of the embodiments of the present invention to facilitate understanding, which should be considered as merely exemplary. Therefore, it should be recognized by those of ordinary skill in the art that various changes and modifications may be made to the embodiments described herein without departing from the scope and spirit of the present invention. Similarly, for clarity and conciseness, the description of well-known functions and structures is omitted in the following description.

[0033] Figure 1 FIG. 1 is a schematic diagram of the main steps of the method for recommending hotspot data according to an embodiment of the present invention. Figure 1 As shown, the hot spot data recommendation method of the embodiment of the present invention mainly includes the following steps S101 to S104.

[0034] Step S101: Calculate the heat score of the data to generate a heat label of the data;

[0035] Step S102: labeling the feature vector of the data using heat labels to generate training data;

[0036] Step S103: training the training data set to obtain a hot data recommendation model;

[0037] Step S104: Use the hot data recommendation model to determine the hot data and make recommendations.

[0038] At present, when determining hot data, the method of statistical analysis and calculation is generally used to obtain the data with the most visits as hot data. However, the embodiment of the present invention uses a heuristic method to calculate the heat score of the data to mark the data, and then trains the marked data to obtain a hot data recommendation model, and uses the hot data recommendation model to determine the hot data for recommendation. It can predict the data with a large increase in heat value in a short period of time in the future and recommend it to users, while attracting user attention and improving user experience, making the data recommendation results more real-time and diverse.

[0039] According to an embodiment of the present invention, when calculating the heat score of data to generate the heat label of the data, the specific implementation includes, for example:

[0040] Calculate the access increase rate of data in the recent period;

[0041] Calculate the popularity score of the data based on the visit increase rate;

[0042] Generate a heat label for the data by comparing the heat score with a preset threshold.

[0043] When calculating the access increase rate of data in a recent period, for example, the following steps are included:

[0044] Calculate the average number of first visits per day for the data in the most recent period, and the average number of second visits per day for the data in a specified period, where the specified period is longer than and includes the most recent period;

[0045] The ratio of the first access amount to the second access amount is taken as the access increase rate of the data in the recent period.

[0046] For example: Assuming the most recent period is the last 7 days, and the specified period is the last month (including the last 7 days), by calculating the average first visits per day in the last 7 days and the average second visits per day in the last month, the visit increase rate of the data in the last 7 days can be obtained. In this way, data with a small number of visits but a high visit increase rate can be obtained, that is, data with an obvious upward trend in visits. For example: for some new data, the total amount of data visits may not be large, but the upward trend is obvious, so it can also be determined as hot data.

[0047] After that, the popularity score of the data can be calculated according to the access increase rate of the data. The access operation of the data may include, for example, click, browse, collect, edit, and other operations on the data. Therefore, when calculating the popularity score of the data, in addition to considering the access increase rate of the data, the number of clicks, browses, collections, edits, and other operations of the data may also be considered. Therefore, by setting different weights for the access increase rate, the number of clicks, the number of browses, the number of collections, the number of edits, and the like, and then performing weighted summation, the popularity score of the data can be obtained. Among them, in the embodiment of the present invention, since it is important to consider the rising trend of the popularity of the data, the weight of the access increase rate is set to be the largest, for example, it may be 80%.

[0048] After that, you can generate the heat label of the data by comparing the heat score with the preset threshold. Specifically, you can follow the steps below:

[0049] The heat score is compared with the preset threshold. If the heat score is greater than the preset threshold, the heat label of the data is set to 1; otherwise, the heat label of the data is set to 0.

[0050] The preset threshold is determined, for example, based on the heat scores of all data, after sorting them from high to low, and then based on the set proportion or number of hot spot data. It can also be a fixed value set based on experience, and so on.

[0051] According to another embodiment of the present invention, training the training data set to obtain a hot data recommendation model may specifically include:

[0052] Divide the training data in the training data set into discrete data sets and continuous data sets;

[0053] Use different networks to process features of discrete data sets and continuous data sets respectively, and concatenate the processed features;

[0054] The concatenated features are input into the reverse joint training layer to obtain the hot data recommendation model.

[0055] Specifically, for discrete data sets, feature selection is first performed using the leaf nodes of the gradient descent tree, and then the selected features are used to generate low-dimensional dense feature vectors through a factor decomposition machine for feature processing; for continuous data sets, deep learning neural networks are used for feature processing.

[0056] The implementation process of the present invention is described below in conjunction with specific embodiments. In the introduction of the embodiments of the present invention, the recommendation of popular products on an e-commerce platform is taken as an example.

[0057] Figure 2 Schematic diagram of the implementation principle of the embodiment of the present invention. Figure 2 As shown in the figure, the left side is the hot product mining client, which mainly includes functional modules such as hot product display, short-term hot product recommendation, real-time user click feedback, and product portrait mining; the right side of the figure is the hot product recommendation server, which mainly includes functional modules such as calculating the short-term product sales increase, constructing training data sets in an heuristic way, building deep learning models, pre-release testing of algorithms, and joint debugging and online release.

[0058] Figure 3 Schematic diagram of the functional implementation principle of the embodiment of the present invention. Figure 3 As shown, in an embodiment of the present invention, the recommendation system for popular products mainly includes the following six modules, namely: model background module, feature engineering module, seasonal popular product module, training set construction module, deep learning model improvement module and business application module.

[0059] 1. Model background module: mainly includes general data, supplementary popular recall, rising product display and recommendation engine recall.

[0060] When recommending popular products to users, especially to new users or inactive users, most of the time, general data (for example, the most viewed products in the last three months) is obtained to make recommendations directly.

[0061] However, this method may miss popular products that are not popular at present but have an obvious upward trend. In order to make up for this deficiency, the present invention can screen out popular products with a large upward trend by calculating the recent orders and browsing performance of the products. This recall data can mine seasonal popular products and supplement the popular recall of general data.

[0062] The present invention makes statistics on commodities with obvious rising trends in a short period of time (such as the past 7 days), and makes predictions on commodities that may have obvious rising trends in the short period of time in the future, and displays them to users after comprehensive calculation.

[0063] The seasonal hot-selling products counted and predicted by the present invention can be used as the general recall layer at the bottom of the recommendation engine architecture to supplement the timeliness of general data and make product recommendations more seasonal and real-time.

[0064] 2. Feature engineering module: The main features collected include the features corresponding to the number of product views at different times, the number of product attentions at different times, the number of product add-ons at different times, and the number of product orders at different times.

[0065] For the number of product views at different times, the present invention collects statistics on the total site views of the product in the past 30 days, the mobile terminal views of the product in the past 30 days, the PC terminal views of the product in the past 30 days, the total site views of the product in the past 7 days, the mobile terminal views of the product in the past 7 days, the PC terminal views of the product in the past 7 days, etc. as the underlying data of feature engineering.

[0066] For the amount of attention paid to a product at different times, the present invention can count the amount of attention paid to the product in the past 30 days and the amount of attention paid to the product in the past 7 days as the underlying data of feature engineering of the product attention dimension.

[0067] For the amount of added products (added to shopping cart) at different times, the present invention counts the amount of added products on mobile terminals in the past 30 days and 7 days, the proportion of added products after browsing on mobile terminals in the past 30 days and 7 days, the amount of added products on PC terminals in the past 30 days and 7 days, the proportion of added products after browsing on PC terminals in the past 30 days and 7 days, the number of times the product was deleted from the shopping cart in the past 30 days / 7 days, etc., as the underlying data of the shopping cart dimension of feature engineering.

[0068] For the order volumes of goods at different times, the present invention collects statistics on the incoming order volumes of goods in the past 30 days and 7 days, the completed order volumes of goods in the past 30 days and 7 days, etc. as the underlying data of the feature engineering order dimension.

[0069] 3. Seasonal hot product module: mainly includes extracting hot-selling products, discovering trends, mining hot-selling products and multi-dimensional integration.

[0070] Among them, extracting hot-selling products, for example, can be extracting the 50 products with the highest sales in each third-level category in the commodity big data table in the past 7 days. Discovering trends, for example, can be extracting the top 10 products in each third-level category in the commodity big data table whose 7-day average sales are increased relative to the 30-day average sales. The present invention defines these products as products with a significant upward trend. Mining hot products, for example, can be extracting the top 50 products ranked by the average daily sales in the past 7 days in the secondary category dimension. The present invention defines these products as hot-selling products. Multi-dimensional fusion, for example, can be taking hot-selling products under all categories, using the increase value of the average daily browsing in the past 7 days relative to the average daily browsing in the past 30 days as an indicator, and reversing and filtering with the same SPU. Among them, SPU (Standard Product Unit): Standardized product unit, is the smallest unit of commodity information aggregation, and is a set of reusable and easily searchable standardized information, which describes the characteristics of a product. In layman's terms, products with the same attribute values ​​and characteristics can be called an SPU.

[0071] 4. Training set construction module: mainly includes heat threshold calculation, product screening, indicator calculation and label feature association.

[0072] When calculating the popularity threshold, the sales threshold, order threshold, pageview threshold, etc. corresponding to the products in each category can be counted according to the 95th percentile. Then, the product screening is performed. The present invention uses a heuristic method to perform product screening, which is mainly divided into the following three steps:

[0073] Step 1: Associate the 7-day feature table of products with the product table to obtain a set of products with sales;

[0074] Step 2: Associate the result of the first step with the product wide table to obtain the secondary and tertiary categories corresponding to the product. The product wide table stores various attributes of the product, such as product price, product color, product size, product brand, etc.

[0075] Step 3: Associate the previously calculated heat threshold, and retain the product set that is higher than the threshold based on sales, orders, and browsing thresholds.

[0076] Afterwards, with the product number as the primary key, associate the 7-day feature table and the 30-day feature table of the product; calculate the average daily page views for 7 days and 30 days and the improvement of the indicator in 7 days compared to 30 days; calculate the average daily sales volume for 7 days and 30 days and the improvement of the indicator in 7 days compared to 30 days; calculate the sales proportion of the third-level categories in 7 days and 30 days and the improvement of the indicator in 7 days compared to 30 days, and so on, so as to calculate the popularity score of the product.

[0077] Finally, based on the above index calculations, the present invention calculates a comprehensive popularity score for the product, and generates a product label by comparing the popularity score with a set threshold, and then associates the product label with multi-dimensional underlying product features to construct a training data set. The set threshold can be

[0078] 5. Deep learning model improvement module, mainly including embedding layer, feature abstraction layer, forward propagation layer and reverse joint training layer.

[0079] The present invention uses the popularity label of the product and the underlying features of the associated product as the input of the deep learning model to predict the probability of the product becoming a hot item in a short period of time in the future. The solution designed by the present invention is to combine the neural network and automatic feature selection (inputting data into GBDT and FM for feature selection), and then use the result of the feature selection together with the output of the deep neural network as input, and calculate the probability that the product is a short-term hot item through the activation function sigmoid function of the reverse joint training layer. Among them, GBDT stands for gradient descent tree, which is one of the best algorithms for fitting the real distribution in traditional machine learning algorithms; FM (Factor Machine) algorithm is a machine learning algorithm based on matrix decomposition, in order to solve the problem of feature combination in large-scale sparse data.

[0080] Figure 4 Schematic diagram of the model training process of an embodiment of the present invention. Figure 4 As shown, the model training process of the embodiment of the present invention mainly includes several functional layers such as an embedding layer, a feature abstraction layer, a forward propagation layer and a reverse joint training layer.

[0081] First, the original features of the training data set are divided into two parts: continuous features (also called numerical features) and discrete features (also called categorical features). For discrete features, the left branch in the input graph is processed, and for continuous features, the right branch in the input graph is processed. Then, the processed features are spliced ​​and input into the reverse joint training layer to obtain the hot data recommendation model.

[0082] In the present invention, low-order discrete features are processed by GBDT+FM to convert discrete features into low-dimensional dense embedding vectors for splicing. The purpose of the original features entering GBDT is to use leaf nodes for feature selection, and then construct hidden vectors through FM to better learn the relationship between features, and can solve the problems of large data volume and data sparsity. High-order continuous abstract features are extracted through DNN. The present invention designs three hidden layers (fully connected layers), uses the Relu activation function, and then splices the output of the third fully connected layer and the output of FM together, and passes them into the fully connected layer of joint training.

[0083] After the original features in the training data set pass through GBDT+FM and DNN, the final output is obtained and enters the activation function to predict the probability of the product being a short-term hit. The activation function is a nonlinear transformation of the neural network after matrix calculation, which outputs the probability of the product being a hit.

[0084] Finally, the concatenated features are input into the reverse joint training layer. The reverse joint training of the present invention selects the optimizer Adam, and continuously iterates and updates the model parameters through the back propagation algorithm, so that the model is optimized towards the correct learning goal round by round.

[0085] In this way, you can Figure 4 The hot data recommendation model is obtained through the model training process.

[0086] 6. Business application module, which introduces the application effects of the hot product recommendation module of the embodiment of the present invention, mainly including: increasing the timeliness of products, enriching the recommended categories, improving user experience and optimizing recommendation results.

[0087] The present invention extracts short-term seasonal hot-selling trend products, which can increase the timeliness of product recommendations for users. Users will feel that the platform's recommendations are very suitable for the current fashion style and will be more willing to browse and buy. Compared with the prior art that recommends mid- and long-term hot products to users as general data, the present invention mines out products with significant short-term rising trends, and the categories will be richer. The present invention can enhance the user experience by extracting hot sales, discovering trends, and mining hot products. Users can feel the flexibility, diversity, and real-time interaction of the recommended products. The present invention mines products with obvious short-term rising trends as a supplement to the general hot data of the product recommendation business, enriching the product categories, increasing the opportunities for new product display, and optimizing the model effect.

[0088] Figure 5 FIG. 1 is a schematic diagram of the main modules of a device for recommending hotspot data according to an embodiment of the present invention. Figure 5 As shown, the hot data recommendation device 500 of the embodiment of the present invention mainly includes a label generation module 501, a data marking module 502, a model training module 503 and a data recommendation module 504.

[0089] The label generation module 501 is used to calculate the heat score of the data to generate the heat label of the data;

[0090] A data labeling module 502 is used to label the feature vector of the data using a heat label to generate training data;

[0091] The model training module 503 is used to obtain a hot data recommendation model by training the training data set;

[0092] The data recommendation module 504 is used to determine hot data and make recommendations using a hot data recommendation model.

[0093] According to an embodiment of the present invention, the label generation module 501 may also be used for:

[0094] Calculate the access increase rate of data in the recent period;

[0095] Calculating the popularity score of the data according to the access increase rate;

[0096] The heat label of the data is generated by comparing the heat score with a preset threshold.

[0097] According to another embodiment of the present invention, when calculating the access increase rate of data in the recent period, the tag generation module 501 may also be used to:

[0098] Calculate the average daily first visits of the data in a recent period, and the average daily second visits of the data in a specified period, wherein the specified period is longer than the recent period and includes the recent period;

[0099] The ratio of the first access amount to the second access amount is taken as the access increase rate of the data in a recent period.

[0100] According to another embodiment of the present invention, when the label generation module 501 generates the heat label of the data by comparing the heat score with a preset threshold, it can also be used to:

[0101] The heat score is compared with a preset threshold value. If the heat score is greater than the preset threshold value, the heat label of the data is set to 1; otherwise, the heat label of the data is set to 0.

[0102] According to yet another embodiment of the present invention, the model training module 503 may also be used to:

[0103] Dividing the training data in the training data set into a discrete data set and a continuous data set;

[0104] Using different networks to process features of the discrete data set and the continuous data set respectively, and concatenating the processed features;

[0105] The concatenated features are input into the reverse joint training layer to obtain the hot data recommendation model.

[0106] According to another embodiment of the present invention, for the discrete data set, feature selection is first performed using leaf nodes of a gradient descent tree, and then a low-dimensional dense feature vector is generated for the selected features through a factorization machine for feature processing.

[0107] According to yet another embodiment of the present invention, a deep learning neural network is used to perform feature processing on the continuous data set.

[0108] According to the technical solution of the embodiment of the present invention, the heat score of the data is calculated to generate the heat label of the data; the feature vector of the data is marked with the heat label to generate training data; the training data set is trained to obtain a hot data recommendation model; the hot data recommendation model is used to determine the hot data and recommend it. Cutting-edge deep learning technology can be applied to construct a training data set with potential significant increase through heuristic methods, and the hot data that may be popular and have a large growth rate in the short term in the future is predicted and displayed to users. While attracting user attention and improving user experience, the recommendation of hot data is also made more real-time and diverse.

[0109] Figure 6 An exemplary system architecture 600 is shown to which the method for recommending hotspot data or the apparatus for recommending hotspot data according to the embodiment of the present invention can be applied.

[0110] like Figure 6 As shown, system architecture 600 may include terminal devices 601, 602, 603, network 604 and server 605. Network 604 is used to provide a medium for communication links between terminal devices 601, 602, 603 and server 605. Network 604 may include various connection types, such as wired, wireless communication links or optical fiber cables, etc.

[0111] Users can use terminal devices 601, 602, and 603 to interact with server 605 through network 604 to receive or send messages, etc. Various communication client applications can be installed on terminal devices 601, 602, and 603, such as shopping applications, web browser applications, search applications, instant messaging tools, email clients, social platform software, etc. (only examples).

[0112] The terminal devices 601 , 602 , and 603 may be various electronic devices having a display screen and supporting web browsing, including but not limited to smart phones, tablet computers, laptop computers, and desktop computers, etc.

[0113] The server 605 may be a server that provides various services, such as a backend management server (only an example) that provides support for shopping websites browsed by users using the terminal devices 601, 602, and 603. The backend management server may analyze and process the received data such as product information query requests, and feed back the processing results (such as target push information, product information - only an example) to the terminal device.

[0114] It should be noted that the hotspot data recommendation method provided in the embodiment of the present invention is generally executed by the server 605 , and accordingly, the hotspot data recommendation device is generally set in the server 605 .

[0115] It should be understood that Figure 6 The number of terminal devices, networks and servers in the embodiment is only for illustration. Any number of terminal devices, networks and servers may be provided according to implementation requirements.

[0116] Reference below Figure 7 , which shows a schematic diagram of the structure of a computer system 700 of a terminal device or a server suitable for implementing an embodiment of the present invention. Figure 7 The terminal device or server shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present invention.

[0117] like Figure 7 As shown, the computer system 700 includes a central processing unit (CPU) 701, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 702 or a program loaded from a storage part 708 into a random access memory (RAM) 703. In the RAM 703, various programs and data required for the operation of the system 700 are also stored. The CPU 701, the ROM 702, and the RAM 703 are connected to each other via a bus 704. An input / output (I / O) interface 705 is also connected to the bus 704.

[0118] The following components are connected to the I / O interface 705: an input section 706 including a keyboard, a mouse, etc.; an output section 707 including a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker, etc.; a storage section 708 including a hard disk, etc.; and a communication section 709 including a network interface card such as a LAN card, a modem, etc. The communication section 709 performs communication processing via a network such as the Internet. A drive 710 is also connected to the I / O interface 705 as needed. A removable medium 711, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 710 as needed, so that a computer program read therefrom is installed into the storage section 708 as needed.

[0119] In particular, according to the embodiments disclosed in the present invention, the process described above with reference to the flowchart can be implemented as a computer software program. For example, the embodiments disclosed in the present invention include a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes a program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from the network through the communication part 709, and / or installed from the removable medium 711. When the computer program is executed by the central processing unit (CPU) 701, the above-mentioned functions defined in the system of the present invention are executed.

[0120] It should be noted that the computer-readable medium shown in the present invention may be a computer-readable signal medium or a computer-readable storage medium or any combination of the above two. The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present invention, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, device or device. In the present invention, a computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, which carries a computer-readable program code. This propagated data signal may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. The computer-readable signal medium may also be any computer-readable medium other than a computer-readable storage medium, which may send, propagate or transmit a program for use by or in conjunction with an instruction execution system, apparatus or device. The program code contained on the computer-readable medium may be transmitted using any appropriate medium, including but not limited to: wireless, wire, optical cable, RF, etc., or any suitable combination of the above.

[0121] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present invention. In this regard, each box in the flow chart or block diagram can represent a module, a program segment, or a part of a code, and the above-mentioned module, program segment, or a part of a code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram or flow chart, and the combination of the boxes in the block diagram or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0122] The units or modules involved in the embodiments of the present invention may be implemented in software or hardware. The units or modules described may also be provided in a processor. For example, they may be described as follows: a processor includes a label generation module, a data labeling module, a model training module, and a data recommendation module. Among them, the names of these units or modules do not constitute limitations on the units or modules themselves in certain circumstances. For example, the label generation module may also be described as a "module for calculating the heat score of data to generate a heat label for the data."

[0123] As another aspect, the present invention also provides a computer-readable medium, which may be included in the device described in the above embodiment; or it may exist independently without being assembled into the device. The above computer-readable medium carries one or more programs, and when the above one or more programs are executed by a device, the device includes: calculating the heat score of the data to generate the heat label of the data; using the heat label to label the feature vector of the data to generate training data; training the training data set to obtain a hot data recommendation model; using the hot data recommendation model to determine the hot data and recommend it.

[0124] According to the technical solution of the embodiment of the present invention, the heat score of the data is calculated to generate the heat label of the data; the feature vector of the data is marked with the heat label to generate training data; the training data set is trained to obtain a hot data recommendation model; the hot data recommendation model is used to determine the hot data and recommend it. Cutting-edge deep learning technology can be applied to construct a training data set with potential significant increase through heuristic methods, and the hot data that may be popular and have a large growth rate in the short term in the future is predicted and displayed to users. While attracting user attention and improving user experience, the recommendation of hot data is also made more real-time and diverse.

[0125] The above specific implementations do not constitute a limitation on the protection scope of the present invention. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions may occur depending on design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for recommending hotspot data, characterized in that: include: Calculating the heat score of the data to generate the heat label of the data, including: calculating the access increase rate of the data in a recent period; calculating the heat score of the data according to the access increase rate; and generating the heat label of the data by comparing the heat score with a preset threshold; Using the heat label to label the feature vector of the data to generate training data; The hot data recommendation model is obtained by training the training data set, including: dividing the training data in the training data set into a discrete data set and a continuous data set; using different networks to perform feature processing on the discrete data set and the continuous data set respectively, and splicing the processed features; inputting the spliced ​​features into the reverse joint training layer to obtain the hot data recommendation model; The hot data recommendation model is used to determine hot data and make recommendations.

2. The method according to claim 1, characterized in that: Calculating the increase rate of data access in the recent period includes: Calculate the average daily first visits of the data in a recent period, and the average daily second visits of the data in a specified period, wherein the specified period is longer than the recent period and includes the recent period; The ratio of the first access amount to the second access amount is taken as the access increase rate of the data in a recent period.

3. The method according to claim 1, characterized in that Generating a heat label of the data by comparing the heat score with a preset threshold comprises: The heat score is compared with a preset threshold value. If the heat score is greater than the preset threshold value, the heat label of the data is set to 1; otherwise, the heat label of the data is set to 0.

4. The method according to claim 1, characterized in that: For the discrete data set, feature selection is first performed using the leaf nodes of the gradient descent tree, and then a low-dimensional dense feature vector is generated for the selected features through a factorization machine for feature processing.

5. The method according to claim 1, characterized in that For the continuous data set, a deep learning neural network is used for feature processing.

6. A device for recommending hotspot data, characterized in that: include: The tag generation module is used to calculate the heat score of the data to generate the heat tag of the data, including: calculating the access increase rate of the data in the recent period; calculating the heat score of the data according to the access increase rate; and generating the heat tag of the data by comparing the heat score with a preset threshold; A data labeling module, used to label the feature vector of the data using the heat label to generate training data; The model training module is used to obtain a hot data recommendation model by training a training data set, including: dividing the training data in the training data set into a discrete data set and a continuous data set; using different networks to perform feature processing on the discrete data set and the continuous data set respectively, and splicing the processed features; inputting the spliced ​​features into a reverse joint training layer to obtain a hot data recommendation model; The data recommendation module is used to determine the hot data and make recommendations using the hot data recommendation model.

7. An electronic device for recommending hotspot data, characterized in that: include: one or more processors; a storage device for storing one or more programs, When the one or more programs are executed by the one or more processors, the one or more processors implement the method according to any one of claims 1 to 5.

8. A computer readable medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the method according to any one of claims 1 to 5 is implemented.

Images