Model training method, acceptance rate estimation method and storage medium
By using semi-supervised learning method in model training, the unlabeled samples were scored and added, which solved the problem of insufficient participation of unexposed samples in navigation route recommendation model training, and improved the accuracy of the model and the accuracy of the recommendation results.
Patent Information
- Application Number
- CN202311444927.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2023-11-01
- Publication Date
- 2025-05-13
AI Technical Summary
In the prior art, during the training process, the classification model cannot participate in training because the unexposed navigation route samples cannot participate in the training, resulting in the model's deviation in the navigation route scoring, and the accuracy of the recommended results is reduced.
The semi-supervised learning method is used to score the label-free samples, and the label-free samples are added to the sample set based on the confidence probability and the preset threshold, and the accuracy of the model is improved through iterative training.
By incorporating label-free samples into the training process, the exposure deviation problem of model training samples is solved, and the accuracy of model training and acceptance rate estimates are improved.
Smart Images
Figure CN119990484A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of machine learning technology, and in particular to a model training method, an acceptance rate estimation method, and a storage medium. Background Art
[0002] Since the traffic conditions of the road network will change over time, existing application software with map navigation function will dynamically determine whether there is a better navigation route than the navigation route currently used by the user (for example, a navigation route that saves more time than the current navigation route) while the user is using the navigation function to travel. If a better navigation route exists, the application software may inform the user of the new navigation route through voice or graphic prompts. The user may choose to use the new navigation route or continue to use the current navigation route and reject the new navigation route recommendation.
[0003] The prior art usually uses a pre-trained classification model to score new navigation routes recommended to users, and then selects the route with the highest score to recommend to users based on the scoring results. However, the inventors found that the samples used in the prior art to train the classification model are samples that have been recommended to users and accepted or rejected by the users, but those navigation routes that have not been recommended to users, that is, unexposed navigation routes, are never able to participate in the training of the classification model. That is, there are missing training samples for the classification model (exposure bias), which will cause the classification model to fail to be fully trained, resulting in deviations in the classification model's scoring of navigation routes and reduced accuracy of the model recommendation results. Summary of the invention
[0004] The embodiments of the present application provide a model training method, an acceptance rate estimation method and a storage medium to solve the problem of exposure bias and improve the accuracy of model training and the accuracy of acceptance rate estimation.
[0005] In a first aspect, an embodiment of the present application provides a model training method, comprising:
[0006] Constructing an unlabeled sample set for the unlabeled samples, scoring the unlabeled samples in the unlabeled sample set using the trained model to obtain the confidence probability of the unlabeled samples;
[0007] Comparing the confidence probability of the unlabeled sample with a preset confidence probability threshold;
[0008] According to the result of comparing the confidence probability with the preset confidence probability threshold, adding the unlabeled sample to a sample set, wherein the sample set includes samples that have been used to train the model;
[0009] The model is iteratively trained using samples in the sample set.
[0010] In a second aspect, an embodiment of the present application provides a method for estimating acceptance rate, comprising:
[0011] For a candidate navigation route, using the specified features of the candidate navigation route as input of a pre-trained acceptance rate prediction model to obtain an acceptance rate score given by the acceptance rate prediction model to the candidate navigation route;
[0012] Recommend the candidate navigation routes with the highest acceptance rate score to the user;
[0013] Wherein, the acceptance rate prediction model is trained using the model training method described above.
[0014] In a third aspect, an embodiment of the present application provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, any of the methods described above is implemented.
[0015] Compared with the prior art, this application has the following advantages:
[0016] An unlabeled sample set is formed for the unlabeled samples, and the unlabeled samples in the unlabeled sample set are scored using the trained model to obtain the confidence probability of the unlabeled samples; the confidence probability is compared with a preset confidence probability threshold; based on the comparison result, the unlabeled sample is added to the sample set, and the model is iteratively trained using the samples in the sample set. This training method based on semi-supervised learning also adds the unlabeled samples as learning samples into the training process, which solves the problem of exposure bias of model training samples and can improve the accuracy of model training.
[0017] Using the acceptance rate prediction model trained by the method provided by this application to estimate the candidate navigation routes to obtain acceptance rate scores can improve the accuracy of the acceptance rate scores, so that the navigation routes recommended to users meet user needs and improve user experience.
[0018] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below. BRIEF DESCRIPTION OF THE DRAWINGS
[0019] In the accompanying drawings, unless otherwise specified, the same reference numerals throughout the multiple drawings represent the same or similar parts or elements. These drawings are not necessarily drawn to scale. It should be understood that these drawings only depict some embodiments according to the present application and should not be regarded as limiting the scope of the present application.
[0020] Figure 1 A flowchart of a model training method according to an embodiment of the present application;
[0021] Figure 2 A flowchart of a model training method according to another embodiment of the present application;
[0022] Figure 3 is a schematic diagram of a model training process in another embodiment of the present application;
[0023] Figure 4 is a flow chart of an acceptance rate estimation method according to another embodiment of the present application;
[0024] Figure 5 A block diagram of an electronic device used to implement an embodiment of the present application. DETAILED DESCRIPTION
[0025] In the following, only some exemplary embodiments are briefly described. As those skilled in the art will appreciate, the described embodiments may be modified in various ways without departing from the concept or scope of the present application. Therefore, the drawings and descriptions are considered to be exemplary in nature and not restrictive.
[0026] To facilitate understanding of the technical solutions of the embodiments of the present application, the following describes the related technologies of the embodiments of the present application. The following related technologies can be combined with the technical solutions of the embodiments of the present application as optional solutions, and they all belong to the protection scope of the embodiments of the present application.
[0027] First, the terms involved in this application are explained.
[0028] In-trip route recommendation: refers to when an application with map navigation function is in the navigation state, a new navigation route is calculated in real time for the user in the navigation state according to the real-time updated road conditions. If the new navigation route is better than the navigation route currently used by the user, the new navigation route will be recommended to the user. The recommendation method can be to inform the user in the form of a bubble box or voice broadcast on the navigation interface of the application. The user can trigger the switch to the new navigation route by clicking the switch route control or voice interaction, or refuse the recommendation by ignoring the bubble box prompt or voice interaction.
[0029] Semi-supervised learning: A machine learning algorithm between supervised learning and unsupervised learning. Semi-supervised learning attempts to train the model using both unlabeled sample data and labeled sample data.
[0030] Exposure bias: usually refers to a model-based recommendation system in which the target object recommended to the user is always in an exposed state, and the possibility of the user selecting the target object will increase, while the target object not recommended to the user is always in an unexposed state, and the user has no possibility of selecting the target object. Since these unselected target objects have never been selected, the probability of them being recommended to the user will become lower and lower. This phenomenon is called exposure bias.
[0031] In order to solve the problem of exposure bias, the embodiment of the present application provides a model training method, which can be applied to a variety of business scenarios, including but not limited to navigation route recommendation, product recommendation or video recommendation, etc. The method provided by the present application is described in detail below in conjunction with the accompanying drawings.
[0032] The present application embodiment provides a model training method, such as Figure 1 The flowchart of the model training method according to one embodiment of the present application is shown, and the method may include the following steps S101 to S104.
[0033] In step S101, an unlabeled sample set is formed for unlabeled samples, and the unlabeled samples in the unlabeled sample set are scored using a trained model to obtain a confidence probability of the unlabeled samples.
[0034] In this embodiment, the unlabeled sample set may include one or more unlabeled samples, where an unlabeled sample refers to a sample without any label added, that is, it is neither a positive sample nor a negative sample.
[0035] The trained model involved in this embodiment can be any type of classification model, and the sample can be scored to obtain the confidence probability, and the embodiment of the present application does not specifically limit this. Exemplarily, the model can be a LightGBM (Light Gradient Boosting Machine) model or a deep neural network classification model.
[0036] In step S102, the confidence probability of the unlabeled sample is compared with a preset confidence probability threshold.
[0037] In step S103, based on the result of comparing the confidence probability with a preset confidence probability threshold, the unlabeled sample is added to a sample set, where the sample set includes the labeled samples that have been used to train the above model.
[0038] In this embodiment, the preset threshold can be set as needed, specifically one threshold, or multiple thresholds, which is not specifically limited in this embodiment of the present application. The value of the preset threshold is also not limited, such as 70% or 80%.
[0039] In step S104, the above model is iteratively trained using samples in the sample set.
[0040] In one implementation, the above step S103 may include:
[0041] According to the result of comparing the confidence probability with the preset confidence probability threshold, the unlabeled sample is added to the positive sample set or the negative sample set.
[0042] In this embodiment, the positive sample set may include one or more positive samples, wherein the positive sample refers to a sample confirmed to belong to the classification result, and can be identified by adding a positive label to the sample. The negative sample set may include one or more negative samples, wherein the negative sample refers to a sample confirmed not to belong to the classification result, and can be identified by adding a negative label to the sample. Iterative training of the model based on the positive sample set and the negative sample set can improve the model's ability to recognize various types of samples, thereby improving the accuracy of classification.
[0043] Exemplarily, an unlabeled navigation route (a navigation route that has never been recommended to the user before) is used as an unlabeled sample, and the model gives a confidence probability for the navigation route. Based on the result of comparing the confidence probability with a preset threshold, if the model determines that the navigation route can be recommended to the user, a positive label is added to the navigation route and it is added to the positive sample set as a positive sample. If the model recognizes that the navigation route cannot be recommended to the user, a negative label is added to the navigation route and it is added to the negative sample set as a negative sample, thus completing one training step.
[0044] The above method provided in this embodiment forms an unlabeled sample set for unlabeled samples, scores the unlabeled samples in the unlabeled sample set with a trained model, obtains the confidence probability of the unlabeled samples, compares the confidence probability with a preset confidence probability threshold, adds the unlabeled samples to the sample set according to the comparison result, and iteratively trains the model with the samples in the sample set. This training method based on semi-supervised learning adds unlabeled samples as learning samples into the training process, solves the problem of exposure bias of model training samples, and can improve the accuracy of model training.
[0045] The present application also provides a model training method, such as Figure 2 The figure shows a flow chart of a model training method according to another embodiment of the present application. The method may include the following steps S201 to S206.
[0046] In step S201, a trained model is used to score an unlabeled sample in the unlabeled sample set to obtain a confidence probability of the unlabeled sample.
[0047] The model in this embodiment can be any classification model as long as it can score the samples to obtain the confidence probability. There is no specific limitation on the specific structure of the model.
[0048] In step S202, if the confidence probability is higher than a preset upper threshold of the confidence probability, the unlabeled sample is added to the positive sample set.
[0049] In this embodiment, the upper threshold value can be set as needed, and is not specifically limited, such as 80% or 90%, etc. Before adding the unlabeled sample to the positive sample set, a positive label can be added to it to identify the sample as a positive sample.
[0050] In step S203, if the confidence probability is lower than a preset lower limit threshold of the confidence probability, the unlabeled sample is added to the negative sample set.
[0051] In this embodiment, the lower threshold can be set to a value as needed, and is not specifically limited, such as 40% or 50%, etc. Before adding the unlabeled sample to the negative sample set, a negative label can be added to it to identify the sample as a negative sample.
[0052] In step S204, if the confidence probability is between the upper threshold and the lower threshold, the unlabeled sample is retained in the unlabeled sample set.
[0053] In this case, the model cannot identify whether the unlabeled sample is a positive sample or a negative sample. Therefore, no label is added to the unlabeled sample, and it is still retained as an unlabeled sample in the unlabeled sample set, which can be used as a sample for the next training. For example, the upper threshold is set to 80% and the lower threshold is set to 50%. If the confidence probability is 65%, it is confirmed that it is between the upper threshold and the lower threshold, and the unlabeled sample is retained in the unlabeled sample set.
[0054] In step S205, the model is iteratively trained using the positive sample set and the negative sample set.
[0055] In this embodiment, iterative training of the model based on the positive sample set and the negative sample set can improve the model's ability to recognize various types of samples, thereby improving the accuracy of classification.
[0056] In step S206, the above training is iteratively performed using the remaining unlabeled samples in the unlabeled sample set until any specified condition is met.
[0057] In this embodiment, the above-mentioned specified conditions can be pre-set, including but not limited to: the number of iterations reaches a specified number, or the number of unlabeled samples in the unlabeled sample set is zero, or the accuracy of the model reaches a specified value, etc.
[0058] Exemplarily, the specified number of times is pre-set to 100 times, then the training can be iteratively performed 100 times using the unlabeled samples in the unlabeled sample set, and then the training is stopped. At this time, it is considered that the training of the model has met the requirements and the accuracy is in line with expectations. Alternatively, the training is iteratively performed using multiple unlabeled samples in the unlabeled sample set until all unlabeled samples in the unlabeled sample set are labeled and added to the corresponding sample set. At this time, there are no remaining unlabeled samples in the unlabeled sample set, that is, the number of unlabeled samples is zero, then it is considered that the training of the model has met the requirements and the accuracy is in line with expectations, and the training can be stopped. Alternatively, the specified value is pre-set to 85%. In the process of iteratively performing the training using multiple unlabeled samples in the unlabeled sample set, if the accuracy of the model reaches 85%, then it is considered that the training of the model has met the requirements and the accuracy is in line with expectations, and the training can be stopped.
[0059] Figure 3 Schematic diagram of the model training process of one embodiment of the present application. Figure 3 As shown, the model in the dotted box is the initial model trained based on the known positive sample set and negative sample set. Use this model to score the unlabeled samples in the unexposed unlabeled sample set to obtain the confidence probability. If the confidence probability is higher than the upper threshold, the unlabeled sample is added to the positive sample set as a positive sample. If the confidence probability is lower than the lower threshold, the unlabeled sample is added to the negative sample set as a negative sample, thereby forming a new positive sample set and negative sample set. Then use the new positive sample set and negative sample set to iteratively train the model, and use the remaining unlabeled samples in the unlabeled sample set again to continue iterative training until any specified condition is met, then stop training, and obtain a model that meets the expectations.
[0060] The present application provides an acceptance rate prediction method, which may include: for a candidate navigation route, taking a specified feature of the candidate navigation route as input, and using a pre-trained acceptance rate prediction model to give an acceptance rate score to the candidate navigation route; wherein the acceptance rate prediction model is trained using the model training method provided by any of the above embodiments. Figure 4 The flowchart of the acceptance rate estimation method according to another embodiment of the present application is shown. The method may include the following steps S401 to S402.
[0061] In step S401, for a candidate navigation route, a designated feature of the candidate navigation route is used as an input of a pre-trained acceptance rate prediction model to obtain an acceptance rate score given by the acceptance rate prediction model to the candidate navigation route.
[0062] In this embodiment, the above-mentioned model is trained using the model training method provided in any of the above-mentioned embodiments, including but not limited to the Light GBM model or the deep neural network classification model.
[0063] In one embodiment, the above-mentioned specified features include but are not limited to at least one of the following: ETA (Estimated Time of Arrival), the distance from the starting point to the end point of the candidate navigation route, the navigation scene or route preference features.
[0064] Among them, there can be multiple scenarios for the above navigation, including but not limited to: route planning scenarios or in-trip route recommendation scenarios. Route planning scenarios, that is, the user is not in the navigation state, and the navigation route from the starting point to the end point can be obtained through the route planning function. In this scenario, the starting point and the end point of the candidate navigation route are generally selected by the user, and the navigation service plans the candidate navigation route based on the starting point and the end point selected by the user. In-trip route recommendation scenarios, that is, the user is in the process of traveling along the navigation route, and the navigation service recommends the candidate navigation route to the user when it believes that there is a better navigation route. In this scenario, the end point of the candidate navigation route is still the location selected by the user, but the starting point is a certain position of the user in the process of traveling along the navigation route, and there is a better navigation route for the user to choose at this position.
[0065] The above-mentioned route preference characteristics refer to the personal tendency of users to choose recommended navigation routes. For example, users tend to choose navigation routes that avoid congestion during rush hours in the morning and evening, tend to choose navigation routes with the shortest time when going to the airport or train station, and tend to choose familiar navigation routes when going home.
[0066] In step S402, the candidate navigation route with the highest acceptance rate score is recommended to the user.
[0067] In one implementation, the acceptance rate scores of each generated candidate navigation route can be ranked, and one or more optimal navigation routes can be selected and recommended to the user, and the advantages of the recommended navigation routes can be identified and displayed to the user. For example, the acceptance rate scores are ranked from high to low, and the top three ranked navigation routes are recommended to the user, and information such as the shortest distance, shortest time or lowest cost is identified to the user, so that the user can make corresponding selections according to needs.
[0068] In addition, the navigation route recommended to the user can be displayed accordingly on the current map interface, including but not limited to displaying the navigation route on the map interface in conjunction with text or voice display, so that the user can view the recommended information in a timely and convenient manner.
[0069] In this embodiment, if the user selects one of the navigation routes recommended to the user, the navigation route selected by the user will be marked with a positive label, and other navigation routes not selected by the user will be marked with negative labels. Both the navigation routes marked with positive labels and negative labels will be used as samples for iterative training, which can improve the accuracy of model training.
[0070] The above method provided in this embodiment uses the acceptance rate estimation model pre-trained by the model training method provided in any of the above embodiments of this application to estimate the acceptance rate score for the candidate navigation routes, which can improve the accuracy of the acceptance rate score, so that the navigation route recommended to the user meets the user's needs and optimizes the user's travel experience. When using multiple specified features, the above acceptance rate estimation model realizes the acceptance rate scoring of the candidate navigation routes based on multiple dimensions, avoids too single reference dimension, better balances the importance relationship between each feature, and achieves the effect of comprehensive consideration based on multiple factors.
[0071] The present application also provides a model training device, including:
[0072] A scoring module is used to form an unlabeled sample set for unlabeled samples, and use the trained model to score the unlabeled samples in the unlabeled sample set to obtain the confidence probability of the unlabeled samples;
[0073] A comparison module, used to compare the confidence probability of the unlabeled sample with a preset confidence probability threshold;
[0074] An adding module, used for adding the unlabeled sample to a sample set including the labeled samples used for training the model according to the result of comparing the confidence probability with a preset confidence probability threshold;
[0075] The training module is used to iteratively train the model using samples in the sample set.
[0076] Among them, the above model can be a Light GBM or a deep neural network classification model.
[0077] In one embodiment, adding a module includes:
[0078] The adding unit is used to add the unlabeled sample to the positive sample set or the negative sample set according to the result of comparing the confidence probability with the preset confidence probability threshold.
[0079] In one embodiment, the adding unit is specifically used to: if the confidence probability is higher than a preset upper limit threshold of the confidence probability, add the unlabeled sample to the positive sample set; if the confidence probability is lower than a preset lower limit threshold of the confidence probability, add the unlabeled sample to the negative sample set.
[0080] In one implementation, the adding unit is further configured to: if the confidence probability is between an upper threshold and a lower threshold, retain the unlabeled sample in the unlabeled sample set.
[0081] In one embodiment, the training module is further used to iteratively perform training using the remaining unlabeled samples in the unlabeled sample set until any of the following conditions is met: the number of iterations reaches a specified number; or the number of unlabeled samples in the unlabeled sample set is zero; or the accuracy of the model reaches a specified value.
[0082] The model training device provided in this embodiment forms an unlabeled sample set for unlabeled samples, scores the unlabeled samples in the unlabeled sample set with a trained model, obtains the confidence probability of the unlabeled samples, compares the confidence probability with a preset confidence probability threshold, adds the unlabeled samples to the sample set according to the comparison result, and iteratively trains the model with the samples in the sample set. This training method based on semi-supervised learning adds the unlabeled samples as learning samples into the training process, solves the problem of exposure bias of model training samples, and can improve the accuracy of model training.
[0083] The present application also provides an acceptance rate estimation device, including:
[0084] A scoring module is used for taking the designated features of the candidate navigation routes as inputs of a pre-trained acceptance rate prediction model to obtain an acceptance rate score given by the acceptance rate prediction model to the candidate navigation routes;
[0085] A recommendation module, used to recommend the candidate navigation routes with the highest acceptance rate score to the user;
[0086] Among them, the above-mentioned acceptance rate prediction model is trained using the model training method provided by any of the above-mentioned embodiments.
[0087] The specified features may include at least one of the following: an estimated arrival time ETA, a distance from the starting point to the end point of the candidate navigation route, a navigation scene, or a route preference feature. The acceptance rate estimation model is a LightGBM or a deep neural network classification model.
[0088] The acceptance rate estimation device provided in this embodiment uses the acceptance rate estimation model pre-trained by the model training method provided in any of the above embodiments of the present application to estimate the acceptance rate score for the candidate navigation route, which can improve the accuracy of the acceptance rate score, so that the navigation route recommended to the user meets the user's needs and optimizes the user's travel experience. When using multiple specified features, the acceptance rate estimation model can achieve acceptance rate scoring for the candidate navigation route based on multiple dimensions, avoid too single reference dimension, better balance the importance relationship between each feature, and achieve the effect of comprehensive consideration based on multiple factors.
[0089] Figure 5 FIG. 1 is a block diagram of an electronic device used to implement an embodiment of the present application. Figure 5 As shown, the electronic device includes: a memory 510 and a processor 520. The memory 510 stores a computer program that can be run on the processor 520. When the processor 520 executes the computer program, the method in the above embodiment is implemented. The number of the memory 510 and the processor 520 can be one or more.
[0090] The electronic device also includes: a communication interface 530, which is used to communicate with external devices and perform data exchange transmission.
[0091] If the memory 510, the processor 520 and the communication interface 530 are implemented independently, the memory 510, the processor 520 and the communication interface 530 can be connected to each other through a bus and communicate with each other. The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, only one thick line is used in the figure, but it does not mean that there is only one bus or one type of bus.
[0092] Optionally, in a specific implementation, if the memory 510, the processor 520 and the communication interface 530 are integrated on a chip, the memory 510, the processor 520 and the communication interface 530 can communicate with each other through an internal interface.
[0093] An embodiment of the present application provides a computer-readable storage medium storing a computer program, which implements the method provided in the embodiment of the present application when the program is executed by a processor.
[0094] An embodiment of the present application also provides a chip, which includes a processor for calling and executing instructions stored in the memory from the memory, so that a communication device equipped with the chip executes the method provided by the embodiment of the present application.
[0095] An embodiment of the present application also provides a chip, including: an input interface, an output interface, a processor and a memory, wherein the input interface, the output interface, the processor and the memory are connected via an internal connection path, and the processor is used to execute the code in the memory. When the code is executed, the processor is used to execute the method provided in the embodiment of the application.
[0096] It should be understood that the processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc. It is worth noting that the processor may be a processor supporting the Advanced RISC Machines (ARM) architecture.
[0097] Further, optionally, the above-mentioned memory may include a read-only memory and a random access memory. The memory may be a volatile memory or a non-volatile memory, or may include both volatile and non-volatile memory. Among them, the non-volatile memory may include a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), or a flash memory. The volatile memory may include a random access memory (RAM), which is used as an external cache. By way of exemplary but not limiting description, many forms of RAM are available. For example, static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDR SDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous link dynamic random access memory (SLDRAM) and direct memory bus random access memory (DR RAM).
[0098] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function according to the present application is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium.
[0099] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine different embodiments or examples described in this specification and the features of different embodiments or examples, unless they are contradictory.
[0100] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include at least one of the features. In the description of this application, the meaning of "plurality" is two or more, unless otherwise clearly and specifically defined.
[0101] Any process or method described in the flow chart or otherwise described herein can be understood as a module, fragment or portion of a code representing one or more executable instructions for implementing the steps of a specific logical function or process. And the scope of the preferred embodiment of the present application includes other implementations, in which the functions may not be performed in the order shown or discussed, including in a substantially simultaneous manner or in a reverse order according to the functions involved.
[0102] The logic and / or steps described in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, which can be specifically implemented in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or used in combination with these instruction execution systems, devices or apparatuses.
[0103] It should be understood that the various parts of the present application can be implemented with hardware, software, firmware or a combination thereof. In the above embodiments, multiple steps or methods can be implemented with software or firmware stored in a memory and executed by a suitable instruction execution system. All or part of the steps of the above embodiment method can be completed by instructing the relevant hardware through a program, which can be stored in a computer-readable storage medium, and when the program is executed, it includes one of the steps of the method embodiment or a combination thereof.
[0104] In addition, each functional unit in each embodiment of the present application can be integrated into a processing module, or each unit can exist physically separately, or two or more units can be integrated into one module. The above-mentioned integrated module can be implemented in the form of hardware or in the form of a software functional module. If the above-mentioned integrated module is implemented in the form of a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium. The storage medium can be a read-only memory, a disk or an optical disk, etc.
[0105] It should be noted that the embodiments of the present application may involve the use of user data. In actual applications, user-specific personal data can be used in the scheme described herein within the scope permitted by applicable laws and regulations, subject to the requirements of applicable laws and regulations of the country where the user is located (for example, with the user's explicit consent, effective notification to the user, etc.).
[0106] The above is only an exemplary embodiment of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of various changes or substitutions within the technical scope recorded in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application shall be based on the protection scope of the claims.
Claims
1. A model training method, characterized in that: The method comprises: Constructing an unlabeled sample set for the unlabeled samples, scoring the unlabeled samples in the unlabeled sample set using the trained model to obtain the confidence probability of the unlabeled samples; Comparing the confidence probability of the unlabeled sample with a preset confidence probability threshold; According to the result of comparing the confidence probability with the preset confidence probability threshold, adding the unlabeled sample to a sample set, wherein the sample set includes labeled samples that have been used to train the model; The model is iteratively trained using samples in the sample set.
2. The method according to claim 1, characterized in that: The model is a Light GBM model or a deep neural network classification model.
3. The method according to claim 1, characterized in that: The adding the unlabeled sample to the sample set according to the result of comparing the confidence probability with the preset confidence probability threshold comprises: According to the result of comparing the confidence probability with the preset confidence probability threshold, the unlabeled sample is added to the positive sample set or the negative sample set.
4. The method according to claim 3, characterized in that The adding the unlabeled sample to the positive sample set or the negative sample set according to the result of comparing the confidence probability with the preset confidence probability threshold comprises: If the confidence probability is higher than a preset upper threshold of the confidence probability, the unlabeled sample is added to the positive sample set; If the confidence probability is lower than a preset lower limit threshold of the confidence probability, the unlabeled sample is added to the negative sample set.
5. The method according to claim 4, characterized in that The method comprises: If the confidence probability is between the upper threshold and the lower threshold, the unlabeled sample is retained in the unlabeled sample set.
6. The method according to any one of claims 1 to 5, characterized in that: The method further comprises: The training is iteratively performed using the remaining unlabeled samples in the unlabeled sample set until any of the following conditions is met: The number of iterations reaches the specified number; The number of unlabeled samples in the unlabeled sample set is zero; The accuracy of the model reaches the specified value.
7. A method for estimating acceptance rate, characterized in that: The method comprises: For a candidate navigation route, using the specified features of the candidate navigation route as input of a pre-trained acceptance rate prediction model to obtain an acceptance rate score given by the acceptance rate prediction model to the candidate navigation route; Recommend the candidate navigation routes with the highest acceptance rate score to the user; The acceptance rate prediction model is trained using the model training method described in any one of claims 1 to 6.
8. The method according to claim 7, characterized in that The specified features include at least one of the following: an estimated time of arrival (ETA), a distance from a starting point to an end point of a candidate navigation route, a navigation scene, or a route preference feature.
9. The method according to claim 7, characterized in that: The acceptance rate prediction model is a Light GBM model or a deep neural network classification model.
10. A computer-readable storage medium, wherein a computer program is stored in the computer-readable storage medium, and when the computer program is executed by a processor, the method according to any one of claims 1 to 9 is implemented.