CN112102095A - Fund product recommendation method, device and equipment - Google Patents
Fund product recommendation method, device and equipment Download PDFInfo
- Publication number
- CN112102095A CN112102095A CN202010980314.8A CN202010980314A CN112102095A CN 112102095 A CN112102095 A CN 112102095A CN 202010980314 A CN202010980314 A CN 202010980314A CN 112102095 A CN112102095 A CN 112102095A
- Authority
- CN
- China
- Prior art keywords
- fund
- user
- training
- characteristic information
- data set
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 61
- 238000012549 training Methods 0.000 claims abstract description 131
- 230000006399 behavior Effects 0.000 claims description 125
- 238000013136 deep learning model Methods 0.000 claims description 26
- 230000015654 memory Effects 0.000 claims description 22
- 238000012545 processing Methods 0.000 claims description 17
- 238000003860 storage Methods 0.000 claims description 12
- 238000011478 gradient descent method Methods 0.000 claims description 6
- 238000011068 loading method Methods 0.000 claims description 5
- 238000010586 diagram Methods 0.000 description 11
- 230000006870 function Effects 0.000 description 9
- 230000005540 biological transmission Effects 0.000 description 7
- 238000004590 computer program Methods 0.000 description 4
- 238000007405 data analysis Methods 0.000 description 4
- 238000007726 management method Methods 0.000 description 4
- 230000007246 mechanism Effects 0.000 description 4
- 230000000694 effects Effects 0.000 description 3
- 230000004048 modification Effects 0.000 description 3
- 238000012986 modification Methods 0.000 description 3
- 230000008569 process Effects 0.000 description 3
- 230000009471 action Effects 0.000 description 2
- 238000004422 calculation algorithm Methods 0.000 description 2
- 238000004891 communication Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000001914 filtration Methods 0.000 description 2
- 238000005111 flow chemistry technique Methods 0.000 description 2
- 239000012634 fragment Substances 0.000 description 2
- OKTJSMMVPCPJKN-UHFFFAOYSA-N Carbon Chemical compound [C] OKTJSMMVPCPJKN-UHFFFAOYSA-N 0.000 description 1
- 244000035744 Hura crepitans Species 0.000 description 1
- 108010001267 Protein Subunits Proteins 0.000 description 1
- 230000002159 abnormal effect Effects 0.000 description 1
- 230000004075 alteration Effects 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000013527 convolutional neural network Methods 0.000 description 1
- 125000004122 cyclic group Chemical group 0.000 description 1
- 238000013075 data extraction Methods 0.000 description 1
- 238000013500 data storage Methods 0.000 description 1
- 238000013501 data transformation Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 229910021389 graphene Inorganic materials 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000007787 long-term memory Effects 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000010295 mobile communication Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
- 230000006403 short-term memory Effects 0.000 description 1
- 230000001360 synchronised effect Effects 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q40/00—Finance; Insurance; Tax strategies; Processing of corporate or income taxes
- G06Q40/06—Asset management; Financial planning or analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/24—Querying
- G06F16/245—Query processing
- G06F16/2457—Query processing with adaptation to user needs
Abstract
The specification provides a fund product recommendation method, device and equipment. The method comprises the steps of obtaining a behavior record of a specified user, wherein the behavior record comprises user characteristic information and fund characteristic information; inputting the behavior record into an access frequency prediction model to obtain the access frequency of the designated user to the fund product corresponding to the fund characteristic information; the access times prediction model is obtained by training according to data in a training data set; the training data set comprises behavior records of different users, each behavior record comprises user characteristic information and fund characteristic information, each behavior record corresponds to a category label, the category labels are determined according to the access times of the users to fund products, and the access times corresponding to different category labels are different; and returning a fund product recommendation result according to the access times of the specified user to the fund product corresponding to the fund characteristic information. The accuracy of fund product recommendation can be improved by utilizing the embodiment of the specification.
Description
Technical Field
The present application relates to the field of data processing technologies, and in particular, to a method, an apparatus, and a device for recommending fund products.
Background
With the development of economy and the advancement of science and technology, the financial consciousness of people is gradually enhanced, and more people invest idle funds into purchasing fund products. Therefore, it is becoming more and more important to recommend appropriate funds for users from a vast amount of financial products.
In the prior art, the fund recommendation mainly comprises a hot recommendation and a collaborative filtering recommendation. The popular recommendation is mainly based on the popular fund products counted in a certain area in the recent period of time. Collaborative filtering recommendations are specifically categorized into two categories: one is to recommend funds purchased by other customers similar to the customer; another is to recommend a fund similar to the fund purchased by the customer. However, both of these approaches have difficulty covering long tails and combining deep-level user fund features, making fund product recommendation less accurate.
Therefore, there is a need in the art for a solution to the above problems.
Disclosure of Invention
The embodiment of the specification provides a fund product recommendation method, device and equipment, and the fund product recommendation accuracy can be improved.
The fund product recommendation method, device and equipment provided by the specification are realized in the following modes.
A fund product recommendation method, comprising: acquiring a behavior record of a designated user, wherein the behavior record comprises user characteristic information and fund characteristic information; inputting the behavior record into an access frequency prediction model to obtain the access frequency of the designated user to the fund product corresponding to the fund characteristic information; the access times prediction model is obtained by training according to data in a training data set; the training data set comprises behavior records of different users, each behavior record comprises user characteristic information and fund characteristic information, each behavior record corresponds to a category label, the category labels are determined according to the access times of the users to fund products, and the access times corresponding to different category labels are different; and returning a fund product recommendation result according to the access times of the specified user to the fund product corresponding to the fund characteristic information.
A training method of an access time prediction model comprises the following steps: acquiring a user behavior data set; the user behavior data set comprises behavior records of different users, and each behavior record comprises user characteristic information and fund characteristic information; distributing a category label to each behavior record in the user behavior data set according to the access times of the user to the fund product corresponding to the fund feature information to obtain a training data set; the access times corresponding to different types of tags are different; and training a preset deep learning model by using the training data set to obtain an access time prediction model.
A fund product recommendation device, comprising: the behavior record acquisition module is used for acquiring the behavior record of the specified user, wherein the behavior record comprises user characteristic information and fund characteristic information; the access times obtaining module is used for inputting the behavior record into an access times prediction model and obtaining the access times of the designated user to the fund product corresponding to the fund characteristic information; the access times prediction model is obtained by training according to data in a training data set; the training data set comprises behavior records of different users, each behavior record comprises user characteristic information and fund characteristic information, each behavior record corresponds to a category label, the category labels are determined according to the access times of the users to fund products, and the access times corresponding to different category labels are different; and the recommending module is used for returning the fund product recommending result according to the access times of the specified user to the fund product corresponding to the fund characteristic information.
A training apparatus for a visit number prediction model, comprising: the behavior data set acquisition module is used for acquiring a user behavior data set; the user behavior data set comprises behavior records of different users, and each behavior record comprises user characteristic information and fund characteristic information; a training data set obtaining module, configured to assign a category label to each behavior record in the user behavior data set according to the number of times that a user accesses a fund product corresponding to the fund feature information, so as to obtain a training data set; the access times corresponding to different types of tags are different; and the prediction model obtaining module is used for training a preset deep learning model by using the training data set to obtain an access time prediction model.
A fund product recommendation device comprising a processor and a memory storing processor-executable instructions that, when executed, perform the steps of any one of the method embodiments of the present specification.
A training apparatus for an access times prediction model, comprising a processor and a memory for storing processor-executable instructions, which when executed by the processor, perform the steps of any one of the method embodiments of the present specification.
The specification provides a fund product recommendation method, device and equipment. In some embodiments, the data of the training data set is divided into a plurality of training subsets, and each training subset is allocated to different nodes for training, so that the network load during model training can be reduced, and the training efficiency can be improved. By setting the training period, multi-dimensional and deep feature combination and feature intersection can be effectively excavated, and long-tailed fund products can be better covered, so that the precision of the training model is improved, and the guarantee is improved for accurately recommending fund products subsequently. By inputting the fund characteristic information corresponding to a plurality of fund products to the access times prediction model at one time, the access times of a specified user to the plurality of fund products can be quickly and conveniently obtained, and therefore the subsequent fund recommendation efficiency and accuracy are improved. By inputting the acquired behavior record into the pre-trained visit number prediction model, a proper fund product can be recommended to the user quickly, so that the recommendation efficiency is improved. By adopting the embodiment provided by the specification, the accuracy of recommendation of fund products can be improved.
Drawings
The accompanying drawings, which are included to provide a further understanding of the specification, are incorporated in and constitute a part of this specification, and are not intended to limit the specification. In the drawings:
FIG. 1 is a schematic flow chart diagram illustrating one embodiment of a method for recommending fund products provided herein;
FIG. 2 is a block diagram of an overall framework for a fund product recommendation method provided herein;
FIG. 3 is a block diagram illustrating the structure of one embodiment of a fund product recommendation device provided herein;
FIG. 4 is a block diagram illustrating an embodiment of a training apparatus for a visit number prediction model provided in the present specification;
FIG. 5 is a block diagram of the hardware architecture of one embodiment of a fund product recommendation server as provided herein.
Detailed Description
In order to make those skilled in the art better understand the technical solutions in the present specification, the technical solutions in the embodiments of the present specification will be clearly and completely described below with reference to the drawings in the embodiments of the present specification, and it is obvious that the described embodiments are only a part of the embodiments in the present specification, and not all of the embodiments. All other embodiments that can be obtained by a person skilled in the art on the basis of one or more embodiments of the present description without inventive step shall fall within the scope of protection of the embodiments of the present description.
The following describes an embodiment of the present disclosure with a specific application scenario as an example. Specifically, fig. 1 is a schematic flow chart of an embodiment of a fund product recommendation method provided in the present specification. Although the present specification provides the method steps or apparatus structures as shown in the following examples or figures, more or less steps or modules may be included in the method or apparatus structures based on conventional or non-inventive efforts.
One embodiment provided by the present specification can be applied to a client, a server, and the like. The client may include a terminal device, such as a smart phone, a tablet computer, and the like. The server may include a single computer device, or may include a server cluster formed by a plurality of servers, or a server structure of a distributed system, and the like.
It should be noted that the following description of the embodiments does not limit the technical solutions in other extensible application scenarios based on the present specification. Detailed description of the preferred embodimentsa specific embodiment of a method for recommending fund products as provided herein, as shown in fig. 1, may comprise the following steps.
S0: and acquiring a behavior record of the appointed user, wherein the behavior record comprises user characteristic information and fund characteristic information.
In the embodiment of the present specification, the specified user may indicate a user who needs to recommend a fund product. Behavior records may be used to dynamically record data corresponding to some behavior of a user. The behavior record may include user characteristic information and fund characteristic information. The user characteristic information may include age, gender, region, user category, and the like. For example, the user characteristic information may include the age, gender, whether it is a credit card customer, whether it is a long payment customer, whether it is a fast credit customer, whether it is a money customer, whether it is a third party inventory customer, whether it is a forwarder, etc. The fund characteristic information may include a fund product identification. The fund product identifier may be used to identify the fund product. The fund product identifier may be comprised of one or more of numbers, alphabetic characters, and the like.
In some implementations, a flow processing engine may be used to read a record of behavior of a given user from the Kafka message queue in real-time. The Kafka message queue can include log data dynamically collected by the front end, and the log data includes behavior records of the user. Kafka is an open source streaming platform that can handle all the action stream data of a consumer in a web site. Kafka unifies online and offline message processing through a parallel loading mechanism of Hadoop, and real-time messages can be provided through clustering.
In some implementation scenarios, log data can be dynamically acquired through a front end or through a server log acquisition tool, then the acquired log data is transmitted to an acquisition management platform through nft (network file transmission), the acquisition management platform is transmitted to an electronic bank data analysis application area through nft, and finally the electronic bank data analysis application area preprocesses the log data through an ETL tool and puts the preprocessed data into a Kafka message queue. Among them, ETL (Extract-Transform-Load) can be used to describe the process of extracting (Extract), converting (Transform), and loading (Load) data from the source end to the destination end. The log data is preprocessed through the ETL tool, so that some redundant information in the log data can be removed, and the subsequent data processing efficiency is improved.
In some embodiments of this specification, before acquiring the behavior record of the specified user, the method may include: starting a Docker container included by each node; the Docker container comprises TensorFlow serving; and loading the visit number prediction model on the TensorFlow service. The Docker container is an open-source application container engine, and developers can package applications and dependency packages of the applications into a lightweight and portable container and then distribute the container to any popular Linux machine, and virtualization can also be achieved. The container is completely using sandbox mechanism, there is no interface between them, and the container performance overhead is very low. The TensorFlow is an open source software library that employs data flow graphs (dataflow graphs) for numerical calculations. Nodes (Nodes) represent mathematical operations in the graph, and lines (edges) in the graph represent the multidimensional data array, i.e., tensor, that is interconnected between the Nodes. The TensorFlow flexible architecture can deploy computing on a variety of platforms, e.g., one or more CPUs (or GPUs) in a desktop computer, a server, a mobile device, and so forth. TensorFlow serving can bring the trained model directly online and provide Restful services. And the visit number prediction model is obtained by pre-training according to data in the training data set. The training data set can comprise behavior records of different users, each behavior record comprises user characteristic information and fund characteristic information, each behavior record corresponds to a category label, the category labels are determined according to the access times of the users to fund products, and the access times corresponding to different category labels are different.
In some implementation scenarios, after each node starts a Docker container, the Docker container maps a port to a physical machine, and maps a disk to a network file system (nfs) shared storage. In some implementation scenarios, active/standby service and load balancing may be implemented by Keepalived and HAProxy. Here, Keepalived is software similar to layer3, 4&5 switching mechanism, i.e. layer3, layer 4 and layer 5 switching in normal times. The Keepalived is used for detecting the state of the web server, if one web server crashes or works out of order, the Keepalived detects the web server and eliminates the web server with the fault from the system, and when the web server works normally, the Keepalived automatically adds the web server into a server group, all the works are automatically completed without manual intervention, and only the web server with the fault is repaired manually. HAProxy is a free and open source software written in C language that provides high availability, load balancing, and TCP (Transmission Control Protocol) and HTTP based application proxies.
In some implementation scenarios, the behavior record of the specified user may include one user characteristic information and fund characteristic information corresponding to a plurality of fund products. For example, a record of behavior for a given user may include one user profile and 50 fund profiles. Therefore, the appropriate fund products can be recommended for the user subsequently based on the fund characteristic information corresponding to the plurality of fund products at one time, and the recommendation efficiency is improved. The amount of the fund feature information can be set according to an actual scene.
S2: inputting the behavior record into an access frequency prediction model to obtain the access frequency of the designated user to the fund product corresponding to the fund characteristic information; the access times prediction model is obtained by training according to data in a training data set; the training data set comprises behavior records of different users, each behavior record comprises user characteristic information and fund characteristic information, each behavior record corresponds to a category label, the category labels are determined according to the access times of the users to fund products, and the access times corresponding to different category labels are different.
In the embodiment of the present specification, after the behavior record of the specified user is obtained, the behavior record may be input into the access frequency prediction model, and the access frequency of the specified user to the fund product corresponding to the fund feature information is obtained, so that a guarantee is provided for subsequently recommending a proper fund product.
In some embodiments of the present description, the visit number prediction model may need to be trained in advance before entering the behavior record into the visit number prediction model.
In some implementation scenarios, the access times prediction model may be obtained by: acquiring a user behavior data set; the user behavior data set comprises behavior records of different users, and each behavior record comprises user characteristic information and fund characteristic information; distributing a category label to each behavior record in the user behavior data set according to the access times of the user to the fund product corresponding to the fund feature information to obtain a training data set; the access times corresponding to different types of tags are different; and training a preset deep learning model by using the training data set to obtain an access time prediction model.
In some implementation scenarios, the flow processing engine may be used to read data in the Kafka message queue in real time to obtain the user behavior data set. The Kafka message queue may include log data dynamically collected by the front end, where the log data may include behavior records of different users. In some implementation scenarios, data in the Kafka message queue for a preset time period may be read as the user behavior data set. For example, one month data may be read as a user behavior data set.
In some implementation scenarios, log data can be dynamically acquired through a front end or through a server log acquisition tool, then the acquired log data is transmitted to an acquisition management platform through nft (network file transmission), the acquisition management platform is transmitted to an electronic bank data analysis application area through nft, and finally the electronic bank data analysis application area preprocesses the log data through an ETL tool and puts the preprocessed data into a Kafka message queue.
In some implementation scenarios, after the user behavior data set is obtained, a category label may be assigned to each behavior record in the user behavior data set according to the number of times of access by the user to the fund product corresponding to the fund feature information, so as to obtain a training data set. Each piece of data included in the training data set corresponds to one category label. Different data may correspond to the same tag or to different tags. The training data set may be used to train the model.
In some implementation scenarios, after the user behavior data set and the training data set are obtained, nfs shared storage to the training cluster may be associated and exported by Hive.
In some implementation scenarios, the training a preset deep learning model by using the training data set to obtain a visit number prediction model may include: dividing the training data set into a plurality of training subsets; assigning each training subset to a different node; each node is stored with the preset deep learning model in advance, and different nodes are used for determining different parameters of the preset deep learning model; calculating the gradient of data in the training subset in each node by using the initial parameters of the preset deep learning model, the training subset in each node and the preset deep learning model; and updating the parameters of the preset deep learning model by using a gradient descent method based on the gradient of the data in the training subset in all the nodes to obtain an access frequency prediction model.
In some implementation scenarios, after obtaining the training data set, in order to reduce network load during training the model and improve training efficiency, the data of the training data set may be divided into a plurality of training subsets, and then each training subset may be assigned to a different node for training. For example, in some implementation scenarios, the GPU cluster may be used to read training data after stream processing, then scatter the training data (introduce data randomness), and finally segment the training data to each Worker node for local storage, so as to effectively reduce the network load during model training. The GPU cluster is a computer cluster which comprises a plurality of Worker nodes, and each Worker node is provided with a Graphic Processing Unit (GPU). The computational power of modern GPUs is exploited by general purpose computations on the graphics processing unit, which enables the GPU cluster to perform very fast computations.
In some implementations, the number of training cycles may be set for each node. This may improve the accuracy of the training model.
In some implementation scenarios, after each training subset is allocated to different nodes, each Worker node can be scheduled to calculate the gradient of local fragment data in parallel, the gradient is uniformly submitted to a parameter server, the parameter server asynchronously summarizes the received gradient, and then the preset deep learning model parameters are updated through a gradient descent method, so that an access frequency prediction model is obtained. The preset deep learning model may include a convolutional neural network, a cyclic neural network, a long-term and short-term memory network, and the like. Initial parameters of the preset deep learning model can be stored in a parameter server in advance. Different nodes may share parameters in the parameter server. In some implementation scenarios, different parameters may be stored in different parameter servers. After updating the preset deep learning model parameters, the updated parameters may be stored in the parameter server. The Gradient Descent method may include a Batch Gradient Descent method (BGD), a random Gradient Descent method (SGD), and the like. It should be noted that, a common deep learning model is data parallelization, that is, the tensrflow task is trained on different small-batch data sets by using the same training model, and then shared parameters of the model are updated on a parameter server. The TensorFlow supports two model training modes of synchronous training and asynchronous training.
In some implementations, multiple models may be trained at one time, and the performance of each model may be compared to select the best performing model as the final visit prediction model. In some implementation scenarios, weights may be assigned to the obtained multiple models, and then the multiple models may be used as the final access prediction model. Wherein, the access times forecasting model can be used for forecasting the browsing times of a user on a certain fund product.
In some implementation scenarios, after obtaining the access time prediction model, parameter information of the access time prediction model may be derived, and then a version number is assigned to the parameter information to generate a reasonable model file corresponding to the access time prediction model. The reasonable model file can be used for recording data such as input data, output data and the like when the access number prediction model is trained. In some implementation scenarios, the reasonable model file may be shared and stored through nfs after being converted into Serving format.
In some implementations, after obtaining the access times prediction model and the reasonable model file, the access times prediction model and the reasonable model file may be stored in a Hadoop distributed file system.
In some implementation scenarios, after the access frequency prediction model is trained, the acquired behavior record of the specified user can be input into the model, and the access frequency of the specified user to the fund product corresponding to the fund feature information is acquired, so that guarantee is provided for subsequently recommending a proper fund product.
In some implementation scenarios, the behavior record of the obtained specified user may include one user characteristic information and fund characteristic information corresponding to a plurality of fund products; accordingly, the access times of the designated user to the plurality of fund products can be obtained by inputting the behavior records into an access time prediction model.
In the embodiment of the specification, the data of the training data set is divided into a plurality of training subsets, and each training subset is distributed to different nodes for training, so that the network load during model training can be reduced, and the training efficiency can be improved. By setting the training period, multi-dimensional and deep feature combination and feature intersection can be effectively excavated, and long-tailed fund products can be better covered, so that the precision of the training model can be improved. By inputting the fund characteristic information corresponding to a plurality of fund products to the access times prediction model at one time, the access times of a specified user to the plurality of fund products can be quickly and conveniently obtained, and therefore the subsequent fund recommendation efficiency and accuracy are improved.
S4: and returning a fund product recommendation result according to the access times of the specified user to the fund product corresponding to the fund characteristic information.
In the embodiment of the present specification, after the number of times of access to the fund product corresponding to the fund feature information by the specified user is obtained, a fund product recommendation result may be returned to the user according to the number of times of access to the fund product corresponding to the fund feature information by the specified user.
In some implementation scenarios, when the number of times of access to the plurality of fund products by the specified user is obtained, the returning a fund product recommendation result according to the number of times of access to the fund product corresponding to the fund feature information by the specified user may include: judging whether a fund product meeting preset conditions exists in the plurality of fund products or not; the preset condition is used for limiting the access times of the fund product; and returning the identifier of the fund product meeting the preset condition when the existence is confirmed.
In some implementation scenarios, the determining whether there is a target fund product that meets a preset condition in the plurality of fund products may further include: and when the fund product does not exist, randomly returning the identifiers of the fund products with the preset quantity.
For example, in some implementation scenarios, after the access times of the designated user to the fund product corresponding to the fund feature information are obtained, the relationship between the obtained access times and the preset access times may be determined, and when the obtained access times are greater than or equal to the preset access times, the fund product with the access times greater than or equal to the preset access times may be returned to the front end for display. It is understood that when the obtained number of accesses is less than the preset number of accesses, the preset information may be returned to the front-end display. The preset information may be an identifier of the random fund product, a prompt message that no relevant fund product is found, and the like.
In some implementation scenarios, when there are a plurality of fund products with access times greater than or equal to the preset access times, the fund products with the access times higher than the preset access times may be selected for feedback. The preset number can be set according to an actual scene, and the specification does not limit the preset number.
In some implementations, if no fund recommendation is returned within a preset time, an exception log may be displayed. The preset time may be set according to an actual scene, for example, 500 milliseconds. The condition that no fund recommendation result is returned within the preset time can include service timeout of 500 milliseconds, message analysis error or abnormal content value, etc.
In some implementations, the fund online transaction may be conducted. For example, the latest popular product identifier and the feature information corresponding to a certain client identifier may be read from the database, and then the popular product identifier and the feature information of the client are packaged, and then the Serving interface is invoked to predict the browsing times of the client on the popular product, and finally the recommendation result is returned according to the browsing times. The fund online transaction can be configured with a keepalive entry address, and the address can be set according to an actual scene. In some implementation scenarios, a new and old version switcher may also be added for fund online transactions.
It is to be understood that the foregoing is only exemplary, and the embodiments of the present disclosure are not limited to the above examples, and other modifications may be made by those skilled in the art within the spirit of the present disclosure, and the scope of the present disclosure is intended to be covered by the claims as long as the functions and effects achieved by the embodiments are the same as or similar to the present disclosure.
As shown in fig. 2, fig. 2 is an overall frame diagram of the fund product recommendation method provided in the present specification. The front-end channel may include a plurality of websites (e.g., website 1, website 2, website 3, etc.), and dynamic data of the user at the websites may be collected by the log collection tool and then transmitted to the transaction service cluster. The transaction service cluster may perform load balancing. And the trading platform in the trading service cluster can provide trading service according to the generated model. The trading platform may also store trading service data in HBase and Redis in big data and algorithm clusters. In the big data and algorithm cluster, training data can be obtained from the HDFS, then the training data is segmented into each Pod for model training, and after a final model is obtained, the model can be exported to the NAS. When transaction services are required using the final model, the model may be loaded to a TensorFlow serving provisioning service. Wherein each Pod, TensorFlow serving, NAS is located in the microservice distributed and GPU virtualized area. A TensorFlow Training model is stored in advance in each Pod. In the process of obtaining training data from the HDFS, data can be obtained from Kafka through Flink, then the data are stored in Redis and HBase, then the data are transmitted to the HDFS through Hive, and finally the HDFS fragments the data to each Pod for model training. Among them, HBase is a distributed, column-oriented open-source database. Redis is a Key-Value storage system. HDFS is a Hadoop distributed file system. Pod is the fundamental unit of the kubernets system, is the smallest component created or deployed by a user, and is also the resource object on which containerized applications run on the kubernets system. Kubernets is an open source platform for automated container operations including deployment, scheduling, and inter-node cluster extension. Kubernets is a container control platform that can abstract all underlying infrastructure. NAS (Network Attached Storage) is a special data Storage server, which takes data as a center, completely separates Storage devices from the server, and centrally manages the data, thereby releasing bandwidth, improving performance, reducing cost, and protecting investment. Hive is a data warehouse tool based on Hadoop, can be used for data extraction, transformation and loading, and is a mechanism capable of storing, querying and analyzing large-scale data stored in Hadoop. Flink is an open source stream processing framework that executes arbitrary stream data programs in a data parallel and pipelined manner, and Flink's pipelined runtime system can execute batch and stream processing programs. Hadoop is a distributed system infrastructure. The TensorFlow Training model may be a pre-set deep learning model.
From the above description, it can be seen that the embodiments of the present application can achieve the following technical effects: the data of the training data set are divided into a plurality of training subsets, and each training subset is distributed to different nodes for training, so that the network load during model training can be reduced, and the training efficiency is improved. By setting the training period, multi-dimensional and deep feature combination and feature intersection can be effectively excavated, and long-tailed fund products can be better covered, so that the precision of the training model can be improved. By inputting the fund characteristic information corresponding to a plurality of fund products to the access times prediction model at one time, the access times of a specified user to the plurality of fund products can be quickly and conveniently obtained, and therefore the subsequent fund recommendation efficiency and accuracy are improved. By inputting the acquired behavior record into the pre-trained visit number prediction model, a proper fund product can be recommended to the user quickly, so that the recommendation efficiency is improved.
Based on the fund product recommendation method, one or more embodiments of the specification further provide a fund product recommendation device. The apparatus may include systems (including distributed systems), software (applications), modules, components, servers, clients, etc. that use the methods described in the embodiments of the present specification in conjunction with any necessary apparatus to implement the hardware. Based on the same innovative conception, embodiments of the present specification provide an apparatus as described in the following embodiments. Since the implementation scheme of the apparatus for solving the problem is similar to that of the method, the specific implementation of the apparatus in the embodiment of the present specification may refer to the implementation of the foregoing method, and repeated details are not repeated. As used hereinafter, the term "unit" or "module" may be a combination of software and/or hardware that implements a predetermined function. Although the means described in the embodiments below are preferably implemented in software, an implementation in hardware, or a combination of software and hardware is also possible and contemplated.
Specifically, fig. 3 is a schematic block diagram of an embodiment of a fund product recommendation device provided in this specification, and as shown in fig. 3, the fund product recommendation device provided in this specification may include: a behavior record obtaining module 120, an access times obtaining module 122 and a recommending module 124.
A behavior record obtaining module 120, configured to obtain a behavior record of a specified user, where the behavior record includes user characteristic information and fund characteristic information;
the access frequency obtaining module 122 may be configured to input the behavior record into an access frequency prediction model, and obtain the access frequency of the designated user to the fund product corresponding to the fund feature information; the access times prediction model is obtained by training according to data in a training data set; the training data set comprises behavior records of different users, each behavior record comprises user characteristic information and fund characteristic information, each behavior record corresponds to a category label, the category labels are determined according to the access times of the users to fund products, and the access times corresponding to different category labels are different;
the recommending module 124 may be configured to return a fund product recommending result according to the number of times that the specified user accesses the fund product corresponding to the fund feature information.
It should be noted that the above-mentioned description of the apparatus according to the method embodiment may also include other embodiments, and specific implementation manners may refer to the description of the related method embodiment, which is not described herein again.
Fig. 4 is a schematic block diagram of an embodiment of a training apparatus for a visit number prediction model provided in this specification, as shown in fig. 4. The training device for the visit number prediction model provided by the specification can comprise: a behavior data set obtaining module 210, a training data set obtaining module 212, and a prediction model obtaining module 214.
A behavior data set obtaining module 210, which may be used to obtain a user behavior data set; the user behavior data set comprises behavior records of different users, and each behavior record comprises user characteristic information and fund characteristic information;
a training data set obtaining module 212, configured to assign a category label to each behavior record in the user behavior data set according to the number of times that the user accesses a fund product corresponding to the fund feature information, so as to obtain a training data set; the access times corresponding to different types of tags are different;
the prediction model obtaining module 214 may be configured to train a preset deep learning model by using the training data set, so as to obtain a visit number prediction model.
It should be noted that the above-mentioned description of the apparatus according to the method embodiment may also include other embodiments, and specific implementation manners may refer to the description of the related method embodiment, which is not described herein again.
The present specification also provides an embodiment of a fund product recommendation apparatus, comprising a processor and a memory for storing processor-executable instructions, wherein the processor, when executing the instructions, may implement any one of the method embodiments described above. For example, the instructions when executed by the processor implement steps comprising: acquiring a behavior record of a designated user, wherein the behavior record comprises user characteristic information and fund characteristic information; inputting the behavior record into an access frequency prediction model to obtain the access frequency of the designated user to the fund product corresponding to the fund characteristic information; the access times prediction model is obtained by training according to data in a training data set; the training data set comprises behavior records of different users, each behavior record comprises user characteristic information and fund characteristic information, each behavior record corresponds to a category label, the category labels are determined according to the access times of the users to fund products, and the access times corresponding to different category labels are different; and returning a fund product recommendation result according to the access times of the specified user to the fund product corresponding to the fund characteristic information.
It should be noted that the above-mentioned apparatuses may also include other embodiments according to the description of the method or apparatus embodiments. The specific implementation manner may refer to the description of the related method embodiment, and is not described in detail herein.
The present specification also provides an embodiment of a training apparatus for an access number prediction model, including a processor and a memory for storing processor-executable instructions, where the processor executes the instructions to implement any one of the above method embodiments. For example, the instructions when executed by the processor implement steps comprising: acquiring a user behavior data set; the user behavior data set comprises behavior records of different users, and each behavior record comprises user characteristic information and fund characteristic information; distributing a category label to each behavior record in the user behavior data set according to the access times of the user to the fund product corresponding to the fund feature information to obtain a training data set; the access times corresponding to different types of tags are different; and training a preset deep learning model by using the training data set to obtain an access time prediction model.
It should be noted that the above-mentioned apparatuses may also include other embodiments according to the description of the method or apparatus embodiments. The specific implementation manner may refer to the description of the related method embodiment, and is not described in detail herein.
The method embodiments provided in the present specification may be executed in a mobile terminal, a computer terminal, a server or a similar computing device. Taking an example of the system running on a server, fig. 5 is a block diagram of a hardware structure of an embodiment of a fund product recommendation server provided in the present specification, where the server may be a fund product recommendation device or a fund product recommendation device in the above embodiments. As shown in fig. 5, the server 10 may include one or more (only one shown) processors 100 (the processors 100 may include, but are not limited to, a processing device such as a microprocessor MCU or a programmable logic device FPGA, etc.), a memory 200 for storing data, and a transmission module 300 for communication functions. It will be understood by those skilled in the art that the structure shown in fig. 5 is only an illustration and is not intended to limit the structure of the electronic device. For example, the server 10 may also include more or fewer components than shown in FIG. 5, and may also include other processing hardware, such as a database or multi-level cache, a GPU, or have a different configuration than shown in FIG. 5, for example.
The memory 200 may be used to store software programs and modules of application software, such as program instructions/modules corresponding to the fund product recommendation method in the embodiments of the present specification, and the processor 100 executes various functional applications and data processing by executing the software programs and modules stored in the memory 200. Memory 200 may include high speed random access memory and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some examples, memory 200 may further include memory located remotely from processor 100, which may be connected to a computer terminal through a network. Examples of such networks include, but are not limited to, the internet, intranets, local area networks, mobile communication networks, and combinations thereof.
The transmission module 300 is used for receiving or transmitting data via a network. Specific examples of the network described above may include a wireless network provided by a communication provider of the computer terminal. In one example, the transmission module 300 includes a Network adapter (NIC) that can be connected to other Network devices through a base station so as to communicate with the internet. In one example, the transmission module 300 may be a Radio Frequency (RF) module, which is used for communicating with the internet in a wireless manner.
The foregoing description has been directed to specific embodiments of this disclosure. Other embodiments are within the scope of the following claims. In some cases, the actions or steps recited in the claims may be performed in a different order than in the embodiments and still achieve desirable results. In addition, the processes depicted in the accompanying figures do not necessarily require the particular order shown, or sequential order, to achieve desirable results. In some embodiments, multitasking and parallel processing may also be possible or may be advantageous.
The method or apparatus provided by the present specification and described in the foregoing embodiments may implement service logic through a computer program and record the service logic on a storage medium, where the storage medium may be read and executed by a computer, so as to implement the effect of the solution described in the embodiments of the present specification. The storage medium may include a physical device for storing information, and typically, the information is digitized and then stored using an electrical, magnetic, or optical media. The storage medium may include: devices that store information using electrical energy, such as various types of memory, e.g., RAM, ROM, etc.; devices that store information using magnetic energy, such as hard disks, floppy disks, tapes, core memories, bubble memories, and usb disks; devices that store information optically, such as CDs or DVDs. Of course, there are other ways of storing media that can be read, such as quantum memory, graphene memory, and so forth.
The foregoing method or apparatus for recommending a fund product provided in this specification may be implemented by a processor executing corresponding program instructions in a computer, for example, implemented by using a c + + language of a windows operating system on a PC side, a linux system, or implemented by using android and iOS system programming languages on an intelligent terminal, or implemented by using processing logic based on a quantum computer.
It should be noted that descriptions of the apparatuses and devices described above according to the related method embodiments in the specification may also include other embodiments, and specific implementation manners may refer to descriptions of corresponding method embodiments, which are not described in detail herein.
The embodiments in the present application are described in a progressive manner, and the same and similar parts among the embodiments can be referred to each other, and each embodiment focuses on the differences from the other embodiments. In particular, for the hardware + program class embodiment, since it is substantially similar to the method embodiment, the description is simple, and the relevant points can be referred to the partial description of the method embodiment.
For convenience of description, the above devices are described as being divided into various modules by functions, and are described separately. Of course, when implementing one or more of the present description, the functions of some modules may be implemented in one or more software and/or hardware, or the modules implementing the same functions may be implemented by a plurality of sub-modules or sub-units, etc.
The present invention has been described with reference to flowchart illustrations and/or block diagrams of methods, apparatus, devices according to embodiments of the invention. It will be understood that the implementation can be by computer program instructions which can be provided to a processor of a general purpose computer, special purpose computer, embedded processor or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified. These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
As will be appreciated by one skilled in the art, one or more embodiments of the present description may be provided as a method, system, or computer program product. Accordingly, one or more embodiments of the present description may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects.
The above description is merely exemplary of one or more embodiments of the present disclosure and is not intended to limit the scope of one or more embodiments of the present disclosure. Various modifications and alterations to one or more embodiments described herein will be apparent to those skilled in the art. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the scope of the claims.
Claims (15)
1. A method for recommending fund products, comprising:
acquiring a behavior record of a designated user, wherein the behavior record comprises user characteristic information and fund characteristic information;
inputting the behavior record into an access frequency prediction model to obtain the access frequency of the designated user to the fund product corresponding to the fund characteristic information; the access times prediction model is obtained by training according to data in a training data set; the training data set comprises behavior records of different users, each behavior record comprises user characteristic information and fund characteristic information, each behavior record corresponds to a category label, the category labels are determined according to the access times of the users to fund products, and the access times corresponding to different category labels are different;
and returning a fund product recommendation result according to the access times of the specified user to the fund product corresponding to the fund characteristic information.
2. The method of claim 1, wherein prior to obtaining the record of the behavior of the specified user, the method comprises:
starting a Docker container included by each node; the Docker container comprises TensorFlow serving;
and loading the visit number prediction model on the TensorFlow service.
3. The method of claim 2, wherein the Docker container maps a port to a physical machine; the Docker container maps disks to network file system shared storage.
4. The method according to claim 3, wherein the behavior record of the obtained specified user includes user characteristic information and fund characteristic information corresponding to a plurality of fund products;
correspondingly, the behavior record is input into an access frequency prediction model, and the access frequency of the appointed user to the plurality of fund products is obtained.
5. The method according to claim 4, wherein when the access times of the specified user to the plurality of fund products are obtained, the returning a fund product recommendation result according to the access times of the specified user to the fund products corresponding to the fund feature information comprises:
judging whether a fund product meeting preset conditions exists in the plurality of fund products or not; the preset condition is used for limiting the access times of the fund product;
and returning the identifier of the fund product meeting the preset condition when the existence is confirmed.
6. The method of claim 5, wherein said determining whether there is a target fund product from the plurality of fund products that meets a preset condition further comprises:
and when the fund product does not exist, randomly returning the identifiers of the fund products with the preset quantity.
7. A training method of an access time prediction model is characterized by comprising the following steps:
acquiring a user behavior data set; the user behavior data set comprises behavior records of different users, and each behavior record comprises user characteristic information and fund characteristic information;
distributing a category label to each behavior record in the user behavior data set according to the access times of the user to the fund product corresponding to the fund feature information to obtain a training data set; the access times corresponding to different types of tags are different;
and training a preset deep learning model by using the training data set to obtain an access time prediction model.
8. The method of claim 7, wherein the obtaining the user behavior data set comprises:
reading data in the Kafka message queue in real time by using a stream processing engine to obtain a user behavior data set; the Kafka message queue comprises log data dynamically acquired by a front end, and the log data comprises behavior records of users.
9. The method according to claim 7, wherein the training a preset deep learning model by using the training data set to obtain a visit number prediction model comprises:
dividing the training data set into a plurality of training subsets;
assigning each training subset to a different node; each node is stored with the preset deep learning model in advance, and different nodes are used for determining different parameters of the preset deep learning model;
calculating the gradient of data in the training subset in each node by using the initial parameters of the preset deep learning model, the training subset in each node and the preset deep learning model;
and updating the parameters of the preset deep learning model by using a gradient descent method based on the gradient of the data in the training subset in all the nodes to obtain an access frequency prediction model.
10. The method of claim 9, wherein obtaining the access times prediction model comprises:
deriving parameter information of the access times prediction model;
and distributing the version number to the parameter information, and generating an inference model file corresponding to the access time prediction model.
11. The method according to claim 10, wherein initial parameters of the preset deep learning model are stored in a parameter server in advance; different nodes share the parameters in the parameter server; and the access times prediction model and the reasonable model file are stored in a Hadoop distributed file system.
12. A fund product recommendation device, comprising:
the behavior record acquisition module is used for acquiring the behavior record of the specified user, wherein the behavior record comprises user characteristic information and fund characteristic information;
the access times obtaining module is used for inputting the behavior record into an access times prediction model and obtaining the access times of the designated user to the fund product corresponding to the fund characteristic information; the access times prediction model is obtained by training according to data in a training data set; the training data set comprises behavior records of different users, each behavior record comprises user characteristic information and fund characteristic information, each behavior record corresponds to a category label, the category labels are determined according to the access times of the users to fund products, and the access times corresponding to different category labels are different;
and the recommending module is used for returning the fund product recommending result according to the access times of the specified user to the fund product corresponding to the fund characteristic information.
13. An apparatus for training a visit number prediction model, comprising:
the behavior data set acquisition module is used for acquiring a user behavior data set; the user behavior data set comprises behavior records of different users, and each behavior record comprises user characteristic information and fund characteristic information;
a training data set obtaining module, configured to assign a category label to each behavior record in the user behavior data set according to the number of times that a user accesses a fund product corresponding to the fund feature information, so as to obtain a training data set; the access times corresponding to different types of tags are different;
and the prediction model obtaining module is used for training a preset deep learning model by using the training data set to obtain an access time prediction model.
14. A fund product recommendation device comprising a processor and a memory for storing processor executable instructions that when executed by the processor perform the steps of the method of any one of claims 1 to 6.
15. Training device for an access times prediction model, comprising a processor and a memory for storing processor-executable instructions, which when executed by the processor implement the steps of the method according to any of claims 7 to 11.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010980314.8A CN112102095A (en) | 2020-09-17 | 2020-09-17 | Fund product recommendation method, device and equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010980314.8A CN112102095A (en) | 2020-09-17 | 2020-09-17 | Fund product recommendation method, device and equipment |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112102095A true CN112102095A (en) | 2020-12-18 |
Family
ID=73760228
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010980314.8A Pending CN112102095A (en) | 2020-09-17 | 2020-09-17 | Fund product recommendation method, device and equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112102095A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112966189A (en) * | 2021-04-14 | 2021-06-15 | 刘蒙 | Fund product recommendation system |
| CN113656702B (en) * | 2021-08-27 | 2023-07-14 | 建信基金管理有限责任公司 | User behavior prediction method and device |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108197633A (en) * | 2017-11-24 | 2018-06-22 | 百年金海科技有限公司 | Deep learning image classification based on TensorFlow is with applying dispositions method |
| CN109670104A (en) * | 2018-11-12 | 2019-04-23 | 深圳壹账通智能科技有限公司 | Information-pushing method, unit and storage medium based on machine learning |
| WO2019153518A1 (en) * | 2018-02-08 | 2019-08-15 | 平安科技（深圳）有限公司 | Information pushing method and device, computer device and storage medium |
| CN110851706A (en) * | 2019-10-10 | 2020-02-28 | 百度在线网络技术（北京）有限公司 | Training method and device for user click model, electronic equipment and storage medium |
| CN111538910A (en) * | 2020-06-23 | 2020-08-14 | 上海摩莱信息科技有限公司 | Intelligent recommendation method and device and computer storage medium |
-
2020
- 2020-09-17 CN CN202010980314.8A patent/CN112102095A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108197633A (en) * | 2017-11-24 | 2018-06-22 | 百年金海科技有限公司 | Deep learning image classification based on TensorFlow is with applying dispositions method |
| WO2019153518A1 (en) * | 2018-02-08 | 2019-08-15 | 平安科技（深圳）有限公司 | Information pushing method and device, computer device and storage medium |
| CN109670104A (en) * | 2018-11-12 | 2019-04-23 | 深圳壹账通智能科技有限公司 | Information-pushing method, unit and storage medium based on machine learning |
| CN110851706A (en) * | 2019-10-10 | 2020-02-28 | 百度在线网络技术（北京）有限公司 | Training method and device for user click model, electronic equipment and storage medium |
| CN111538910A (en) * | 2020-06-23 | 2020-08-14 | 上海摩莱信息科技有限公司 | Intelligent recommendation method and device and computer storage medium |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112966189A (en) * | 2021-04-14 | 2021-06-15 | 刘蒙 | Fund product recommendation system |
| CN112966189B (en) * | 2021-04-14 | 2024-01-26 | 北京基智科技有限公司 | Fund product recommendation system |
| CN113656702B (en) * | 2021-08-27 | 2023-07-14 | 建信基金管理有限责任公司 | User behavior prediction method and device |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US11386128B2 (en) | Automatic feature learning from a relational database for predictive modelling | |
| CN106371975B (en) | A kind of O&M automation method for early warning and system | |
| US10929899B2 (en) | Dynamic pricing of application programming interface services | |
| US20120116743A1 (en) | Optimizing storage cloud environments through adaptive statistical modeling | |
| US11403131B2 (en) | Data analysis for predictive scaling of container(s) based on prior user transaction(s) | |
| US11295247B2 (en) | Discovery and generation of organizational key performance indicators utilizing glossary repositories | |
| US20230289828A1 (en) | Data processing method, computer device, and readable storage medium | |
| US20200034750A1 (en) | Generating artificial training data for machine-learning | |
| US11308077B2 (en) | Identifying source datasets that fit a transfer learning process for a target domain | |
| EP4198771A1 (en) | Data processing method and apparatus, computer readable medium, and electronic device | |
| CN107273979B (en) | Method and system for performing machine learning prediction based on service level | |
| US20210191759A1 (en) | Elastic Execution of Machine Learning Workloads Using Application Based Profiling | |
| CN112379869A (en) | Standardized development training platform | |
| US10282411B2 (en) | System, method, and recording medium for natural language learning | |
| CN112102095A (en) | Fund product recommendation method, device and equipment | |
| JP2024500730A (en) | Training interpretable deep learning models using disentangled learning | |
| US20180349476A1 (en) | Evaluating theses using tree structures | |
| CN113721898A (en) | Machine learning model deployment method, system, computer device and storage medium | |
| US11195620B2 (en) | Progress evaluation of a diagnosis process | |
| Han | Investigations into elasticity in cloud computing | |
| Daradkeh et al. | Dynamic k-means clustering of workload and cloud resource configuration for cloud elastic model | |
| US20230009237A1 (en) | Multi-dimensional data labeling | |
| US20220318671A1 (en) | Microservice compositions | |
| US20210357781A1 (en) | Efficient techniques for determining the best data imputation algorithms | |
| US11455467B2 (en) | Relation extraction using full dependency forests |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |