Construction method of component recommendation model for multivariate application scenarios

Abstract

The invention provides a construction method of a component recommendation model for a multi-element application scene. The method comprises the following steps: acquiring a data set; the data set comprises concerned information data when the user selects the component; labeling information data in the data set to obtain a labeling set; the label set comprises 70% of a training set and 30% of a test set; and training the recommendation model by using the training set data to obtain the component recommendation model oriented to the multi-element application scene. The component recommendation model which is constructed through the scheme and faces multiple application scenes can provide all-around component information for model designers, quality managers, experts and material managers and provide component recommendation services.

Description

Technical Field

[0001] The present invention belongs to the field of information management beta technology, and in particular relates to a method for constructing a component recommendation model for multiple application scenarios. Background Art

[0002] In recent years, the rapid development of the domestic basic electronic industry and the gradual enrichment of various product systems have led to the diversification of requirements for components. The selected components must not only meet the requirements of functional performance, but also meet the requirements of cycle, progress and cost. The constraints on component selection are becoming more complex. Different products or different positions of the same product have different requirements for components.

[0003] Components are the basis for the development of single-machine models. During the development process, it is necessary to obtain basic data, process test data, on-orbit application history and other information of components. Currently, a single electronic system involves thousands or hundreds of components, and the components faced by designers include millions of functional types, quality grades, detailed specifications, application precautions, etc. Although each unit has established its own commonly used component selection catalog, the way designers select components is mainly based on form-based catalog queries. The catalog system has a single function and only has a display function, and the data information displayed is relatively small. It is difficult for designers to obtain detailed component information, especially for newly developed domestic components. There is a lack of channels to obtain device information, and it is difficult to obtain basic information such as test data of components during the development process.

[0004] In the process of selecting components, each unit has accumulated a large amount of component information, including data on identification and inspection, application experience, quality issues, application verification, evaluation tests, and risk hazards. These data are exactly what designers pay most attention to when selecting components. However, there is currently a lack of effective tools to display this information to designers, which leads to users still relying mainly on personal experience when selecting components, resulting in a narrow range of component selection and an inability to obtain more effective information. In addition, this information is stored in the form of raw data and lacks a certain correlation. Such a large amount of messy data will affect the selection decision of designers. Summary of the invention

[0005] The purpose of the embodiments of this specification is to provide a method for constructing a component recommendation model for multiple application scenarios.

[0006] The embodiment of the present application is implemented in the following manner:

[0007] In a first aspect, the present application provides a method for constructing a component recommendation model for multiple application scenarios, the method comprising:

[0008] Acquire a data set; the data set includes information data that users are concerned about when selecting components;

[0009] Label the information data in the data set to obtain a label set; the label set includes 70% training set and 30% test set;

[0010] The recommendation model is trained using the training set data to obtain a component recommendation model for multiple application scenarios.

[0011] In one embodiment, the information data in the data set is annotated to obtain an annotation set, including:

[0012] Perform random sampling of no less than 20% of the data set to form a data set to be labeled;

[0013] Determine the indicators that users are most concerned about in each application scenario in multiple application scenarios;

[0014] Mark the indicators that users care about most as the decisive factors for recommendations;

[0015] If the decisive factors of multiple components are the same, they are annotated according to the user's other concerns to obtain an annotation set.

[0016] In one of the embodiments, multiple application scenarios include: design selection, selection review, and quality issue handling.

[0017] In one embodiment, the recommendation model includes a random forest model;

[0018] The random forest model is trained using the training set data to obtain a component recommendation model for multiple application scenarios, including:

[0019] Several samples are randomly selected from the training set using the Bootstrap sampling method; the selected samples are used to build a decision tree;

[0020] Extract features from each sample through feature engineering;

[0021] Find the best split point based on the features;

[0022] Until all decision trees are constructed, a component recommendation model for multiple application scenarios is obtained.

[0023] In one embodiment, the prediction results of the component recommendation model for multiple application scenarios include:

[0024] For classification problems, the majority voting method is used to determine the final prediction result;

[0025] For regression problems, the average of all decision tree predictions is taken as the prediction result.

[0026] In one embodiment, for the classification problem, let C K is the classification prediction result of the Kth decision tree. The prediction result of the component recommendation model for multiple application scenarios is:

[0027] C(x)=mode{C1(x),C2(x),…,C K (x)}

[0028] Among them, mode represents the majority, that is, the most frequently occurring classification result in K decision trees;

[0029] For regression problems, let R k is the regression prediction result of the kth decision tree. The prediction result of the component recommendation model for multiple application scenarios is:

[0030]

[0031] Here, K is the number of decision trees.

[0032] In one embodiment, the recommendation model further includes a collaborative filtering model; the training set data includes user rating data on components; the users include target users and other users;

[0033] The collaborative filtering model is trained using the training set data to obtain a component recommendation model for multiple application scenarios, including:

[0034] Calculate the similarity of the scores between the target user and other users based on the target user's score data on the components and other users' score data on the components;

[0035] Select a group of users with the highest score similarity to the target user as neighbor users;

[0036] Predict the target user's predicted rating for the unrated components based on the rating data of neighbor users and the similarity of ratings between the target user and neighbor users;

[0037] Generate a recommendation list for the target user based on the predicted score;

[0038] Based on the recommendation list generated for the target user, the parameters of the collaborative filtering model are adjusted to obtain a component recommendation model for multiple application scenarios.

[0039] In one embodiment, the method further comprises:

[0040] Obtain user scores, component click counts, and selection records for the component recommendation model for multiple application scenarios to form feedback information;

[0041] The feedback information is used as an input variable and input into a component recommendation model for multiple application scenarios, and the component recommendation model for multiple application scenarios is further iterated and optimized.

[0042] In a second aspect, the present application provides an electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, the method for constructing a component recommendation model for multiple application scenarios as in the first aspect is implemented.

[0043] In a third aspect, the present application provides a readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method for constructing a component recommendation model for multiple application scenarios as in the first aspect.

[0044] It can be seen from the technical solution provided in the above embodiments of this specification that the component recommendation model for multiple application scenarios constructed by this solution can provide model designers, quality management personnel, experts, and material management personnel with comprehensive component information and component recommendation services, with a recommendation accuracy rate of over 80%. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the embodiments of this specification or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this specification. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0046] Figure 1 A flowchart of a method for constructing a component recommendation model for multiple application scenarios provided in this application;

[0047] Figure 2 A schematic diagram of randomly selecting 5 different test data from the test set to verify the accuracy and recall rate of the model provided by this application;

[0048] Figure 3 A schematic diagram of the structure of the electronic device provided in this application. DETAILED DESCRIPTION

[0049] In order to enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below in conjunction with the drawings in the embodiments of this specification. Obviously, the described embodiments are only part of the embodiments of this specification, not all of the embodiments. Based on the embodiments in this specification, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of this specification.

[0050] In the following description, specific details such as specific system structures, technologies, etc. are provided for the purpose of illustration rather than limitation, so as to provide a thorough understanding of the embodiments of the present application. However, it should be clear to those skilled in the art that the present application may also be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to prevent unnecessary details from obstructing the description of the present application.

[0051] It will be apparent to those skilled in the art that various modifications and variations may be made to the specific embodiments of the present application description without departing from the scope or spirit of the present application. Other embodiments derived from the present application description will be apparent to those skilled in the art. The present application description and examples are exemplary only.

[0052] The words “include,” “including,” “have,” “contain,” etc. used in this document are open-ended terms, meaning including but not limited to.

[0053] The present invention is further described in detail below with reference to the accompanying drawings and embodiments.

[0054] Reference Figure 1 , which shows a flow chart of a method for constructing a component recommendation model for multiple application scenarios provided in an embodiment of the present application.

[0055] like Figure 1 As shown, the method for constructing a component recommendation model for multiple application scenarios may include:

[0056] S110, obtaining a data set; the data set includes information data that the user is concerned about when selecting components.

[0057] Specifically, information data that users focus on when selecting components is collected through multiple channels (such as websites, databases, documents, product manuals of research and development units, etc.), including design selection data, basic material data, quality assurance data, quality problem data, quality inspection data, discontinued production data, evaluation and verification data, procurement information data, application information data, etc., such as component category, functional performance, packaging form, parameter indicators, price, application experience, etc.

[0058] S120, labeling the information data in the data set to obtain a labeling set, which includes 70% of the training set and 30% of the test set.

[0059] Specifically, since expert experience plays a dominant role in the early stage of the component recommendation model, this application will use expert experience as a guide in the process of building the model, manually annotate the data after multiple rounds of communication with experts, and use supervised learning to make the model fit the expert decision-making path and component selection reference standards. In the subsequent process of human-computer interaction between users (such as designers or experts, etc.) and the system, the user's historical operation data will be input into the model as a variable for model iteration.

[0060] In one embodiment, step S120 labels the information data in the data set to obtain a label set, which may include:

[0061] Perform random sampling of no less than 20% of the data set to form a data set to be labeled;

[0062] Determine the indicators that users are most concerned about in each application scenario in multiple application scenarios;

[0063] Mark the indicators that users care about most as the decisive factors for recommendations;

[0064] If the decisive factors of multiple components are the same, they are annotated according to the user's other concerns to obtain an annotation set.

[0065] Specifically, a random sampling of no less than 20% of the data set is performed to form a data set to be labeled. For the diversified application scenarios of the design, first determine the indicators that users are most concerned about in each scenario, and use them as the decisive factors for recommendation. When the decisive factors of multiple components are the same, consider other points of concern of users, and recommend no more than 5 components for each sample. For example, in the application scenario of aerospace satellites, it is necessary to give priority to ensuring the reliability of quality, and use the quality grade and quality problems as the decisive factors for recommendation. When the performance indicators of the components are met, components with higher quality grades are recommended first. Components with quality problems are not recommended or are placed in the last few places of recommendation. The final annotation set is formed through such annotation principles.

[0066] In one embodiment, three application scenarios are designed based on the key nodes of component selection: design selection, selection review, and quality problem handling.

[0067] Specifically, the designer's reasonable selection is an important part of component quality assurance. Different types of equipment tasks and the development of different parts of the product have different comprehensive considerations and requirements for component quality levels and cost control requirements. Faced with the multi-stage and diversity of equipment, the "one-size-fits-all" selection model has the possibility of "over-design" and "over-assurance", so different component selection strategies and requirements need to be adopted.

[0068] The design selection scenario is based on the component database. Starting from the target component model that the designer intends to use, the recommendation algorithm calculates the intelligent recommendation for the component in the knowledge base to assist the designer in comparing and screening components. At the same time, it pushes the historical quality data of components during quality inspection and use, so as to highlight the quality risks in the selection stage and ultimately achieve the purpose of quality risk control. This scenario also supports designers to customize the recommendation goals of the recommendation system, such as the recommendation of components with the best cost, the best quality, the best supply cycle, or the best supply.

[0069] The design selection scenario realizes the intelligent recommendation function of components under different equipment selection requirements. Users can choose the following two input methods: custom recommendation target (optimal cost, optimal quality, optimal supply cycle, optimal supply volume) + component model specifications; custom recommendation target + required component category and key parameters. When the user enters the specific model or related information of the target component in the system, on the selection interface pushed to the designer, no more than 5 components are recommended from high to low according to the degree of fit, to help users compare and screen.

[0070] When designers are designing equipment, they tend to focus on the performance of specific components and whether their indicators meet the use requirements, and they pay attention to the bottom line. For the equipment as a whole, the quality assurance of components greatly affects the reliability of the whole machine. When selecting, it is not only necessary to meet the performance indicators, but also to take into account many factors. Therefore, it is necessary to conduct a selection review of components.

[0071] Component selection review is an activity conducted to determine the suitability, adequacy and effectiveness of component selection to achieve the objectives. During the model selection review, the main concerns are the following: whether the selected components are in the selection catalog; whether the quality grade meets the task requirements; whether they are prohibited or restricted components; whether the newly developed or newly selected devices are reasonable, etc.

[0072] Component quality data refers to various data generated during the quality assurance process of components that describe the reliability level and status of components. These data cover the entire process of component production, selection, procurement, inspection, and use. Component quality problem data is a crucial part of improving the reliability of the model. When selecting components, the model will focus on the incidence and handling of quality problems. At present, quality problem data is relatively scattered, and there is no unified platform to effectively integrate all information, so quality problem data cannot be effectively utilized. Design a quality problem handling scenario in the model, which includes information on component quality problems and related handling situations.

[0073] The quality problem handling scenario can meet the function of learning from similar quality problems and realize the display of quality problems of similar components. There are two input and output methods in this scenario: input model specifications to recommend quality problems of similar products; input process and packaging information to output quality problems of products with the same process and packaging. It provides a reference for component quality assurance and provides professional technical support for component quality problems that occur throughout the life cycle of the model.

[0074] S130: Use the training set data to train the recommendation model to obtain a component recommendation model for multiple application scenarios.

[0075] Among them, the recommended models include random forest models.

[0076] Specifically, the random forest model is selected as the classifier, and the input is the name of the required component and one of the application scenarios, such as optimal cost, optimal quality, etc. The output is five recommended component models. The input and output can be changed flexibly, and retraining can be performed after the change.

[0077] Random forest is an ensemble algorithm consisting of multiple decision trees. Its basic idea is to improve the overall prediction performance by integrating multiple classifiers (i.e., decision trees). Random forest can be used for both classification and regression tasks.

[0078] The advantage of random forest is that it can handle complex nonlinear relationships and is suitable for scenarios with complex features and attribute relationships. In component recommendation, if there is a complex nonlinear relationship between the performance indicators of the components and the application requirements, random forest can effectively capture these relationships and recommend suitable components for different application scenarios. Random forest performs well in handling nonlinear relationships and can effectively capture the complex features and attribute relationships between components. It performs classification and regression by integrating multiple decision trees, with high accuracy and generalization ability. In the scenario of component recommendation, random forest can provide an efficient and accurate recommendation solution. The performance of the random forest model is relatively easy to evaluate and optimize. By adjusting parameters such as the number and depth of trees, the recommendation effect of the model can be effectively improved.

[0079] The rationale for choosing random forest lies in the following points:

[0080] Handling high-dimensional data: Random forests can handle high-dimensional data with a large number of features without the need for feature selection or dimensionality reduction.

[0081] Robustness: It can effectively handle missing values ​​and outliers and has good robustness.

[0082] Reduce overfitting: By randomly selecting features and samples for training, random forests enhance the generalization ability of the model, thereby reducing the overfitting problem.

[0083] Parallel computing: It has a fast training speed and is easy to implement in parallel because the trees are independent of each other.

[0084] Simple implementation: The algorithm implementation is relatively simple and easy to understand and deploy.

[0085] Balanced error: For unbalanced data sets, random forests can balance the error, especially in the case of classification imbalance.

[0086] Therefore, considering the characteristics of the actual application scenario and the performance requirements of the model, this application selects random forest as the algorithm for the intelligent recommendation model of component selection.

[0087] In one embodiment, the training set data is used to train a random forest model to obtain a component recommendation model for multi - application scenarios, including:

[0088] Randomly select a number of samples from the training set using the Bootstrap sampling method; the selected samples are used to construct a decision tree;

[0089] Extract the features in each sample through feature engineering;

[0090] Find the best split point according to the features;

[0091] Until each decision tree is constructed; obtain a component recommendation model for multi - application scenarios.

[0092] In one embodiment, the prediction results of the component recommendation model for multi - application scenarios include:

[0093] For classification problems, use the majority voting method to determine the final prediction result;

[0094] For regression problems, take the average of all decision tree prediction results as the prediction result.

[0095] Specifically, sample selection: Randomly select N samples from the original training set using the Bootstrap sampling method, and these samples are used to construct a decision tree.

[0096] Feature selection: Extract features that are helpful for classification through feature engineering, which includes numerical features, categorical features, and features derived from the original data. If the original data has M features, then at each node of each decision tree, randomly select m features (m << M), and then use these features to find the best split point. All features will be standardized or normalized to ensure that the model performance is not affected by different scales.

[0097] Decision tree construction: For each decision tree, repeat steps 1 and 2 until each tree is built. Usually, decision trees are not pruned during the construction process.

[0098] Result prediction: For classification problems, the majority voting method is used to decide the final prediction; for regression problems, the average of all decision tree predictions is taken.

[0099] For classification problems, let C K is the classification prediction result of the Kth decision tree. The prediction result of the component recommendation model for multiple application scenarios is:

[0100] C(x)=mode{C1(x),C2(x),…,C K (x)}

[0101] Among them, mode represents the majority, that is, the most frequently occurring classification result in K decision trees;

[0102] For regression problems, let R k is the regression prediction result of the kth decision tree. The prediction result of the component recommendation model for multiple application scenarios is:

[0103]

[0104] Here, K is the number of decision trees.

[0105] The advantage of random forest is that it can handle high-dimensional data, is insensitive to small noise in training data, and is not prone to overfitting. In addition, random forest can give the importance score of features, which helps with feature selection.

[0106] Exemplarily, the random forest model is trained using training set data to obtain a component recommendation model for multiple application scenarios, including:

[0107] Collect data about components, including their performance parameters, cost, supply status, etc. These data will be used as input features of the model. At the same time, it is also necessary to define the characteristics of each application scenario, such as optimal cost, optimal supply, etc. Preprocess the collected data, including data cleaning, missing value processing, encoding conversion, etc.

[0108] Use the preprocessed data to build a random forest model. In this process, multiple decision trees are created, each of which is trained based on randomly selected features and data subsets. The model can be optimized by adjusting model parameters such as the number of trees. It has been verified that the model works best when n_estimators = 400 and random seed random_state = 42. Increasing the number of decision trees usually improves the performance of the model, but it also increases the computational cost. Choosing a moderate value can strike a balance between performance and computational efficiency. Setting random_state ensures that the behavior of the random forest is consistent each time you run the code, which helps ensure the reproducibility of the results.

[0109] The data set is divided into a training set and a test set. The training set data is used to train the random forest model, and the test set data is used to verify the model. The performance of the model can be evaluated using indicators such as accuracy, precision, and recall. After the model training is completed, the model can be used to classify and recommend new component requirements. Enter the required component name and its application scenario, and the model will output five recommended component models.

[0110] Among them, the recommendation model also includes a collaborative filtering model.

[0111] In order to break through the barriers between equipment and design units and achieve the goal of knowledge sharing among all users, this application also uses a collaborative filtering model to incorporate the designer's design selection results into the model.

[0112] Collaborative filtering (CF) is a common recommendation algorithm that analyzes user behaviors and preferences, discovers similarities between users, and recommends items that may be of interest to users. Collaborative filtering algorithms are mainly divided into two categories: user-based collaborative filtering and item-based collaborative filtering. This application mainly uses user-based collaborative filtering.

[0113] The collaborative filtering recommendation method mainly consists of three parts, namely user data collection module, user data mining module, and recommendation result generation module. User data collection is mainly responsible for collecting the interactive data of users during the query process, providing data sources for the subsequent system data analysis module. The collected data includes two parts: one is the explicit rating data of the user on the query results, and the other is the implicit behavior data that can reflect the user's preferences, mainly the user's browsing data on a certain component. The user data mining module is responsible for mining the user's interest in the recommended components and their attributes from the user's behavior data, preprocessing the collected data, and generating the "user-component" demand matrix model. Among them, data preprocessing is mainly responsible for cleaning irrelevant data in user behavior, sorting it out, and unifying the data format. Then the "user-component" demand matrix model is constructed and corrected to prepare for recommendation prediction. The recommendation result generation module mainly predicts the user's interest value for unrated components based on the "user-component" demand matrix model, improves the similarity calculation method, and uses the threshold judgment method to improve the algorithm efficiency.

[0114] User explicit feedback is directly reflected in the user's rating of search results. After searching, users can rate the recommended search results. The rating data directly reflects the user's satisfaction. A rating of 1 indicates dissatisfaction and a rating of 5 indicates full satisfaction. Satisfaction increases with the increase of the rating value. This data will be saved in the rating table in the database, so it can be directly extracted from the database. Implicit feedback behavior cannot directly show the user's preferences, but by collecting some user behaviors, these user behaviors can reflect the user's interests and preferences from the side, such as the user's page browsing behavior. User implicit feedback behavior data is obtained by data embedding. The implementation of data embedding is to implant relevant codes where statistical data is needed, and count the corresponding user behaviors in this part of the code, that is, monitor the relevant user operation behaviors, and count the user's browsing time and number of browsing times on components. After user data collection and user data mining to generate the user's demand matrix, the user-based collaborative filtering method is used for recommendation.

[0115] In one embodiment, the training set data includes user rating data on components; wherein the users include target users and other users.

[0116] The collaborative filtering model is trained using the training set data to obtain a component recommendation model for multiple application scenarios, including:

[0117] Calculate the similarity of the scores between the target user and other users based on the target user's score data on the components and other users' score data on the components;

[0118] Select a group of users with the highest score similarity to the target user as neighbor users;

[0119] Predict the target user's predicted rating for the unrated components based on the rating data of neighbor users and the similarity of ratings between the target user and neighbor users;

[0120] Generate a recommendation list for the target user based on the predicted score;

[0121] Based on the recommendation list generated for the target user, the parameters of the collaborative filtering model are adjusted to obtain a component recommendation model for multiple application scenarios.

[0122] Specifically, when obtaining a data set, in addition to obtaining the characteristic data information of components, it is also necessary to collect the user's interaction data on components, such as ratings, click history, etc. Understandably, the interaction data can also be cleaned and converted to ensure the quality of user behavior data, handle missing values ​​and outliers, and convert the data format to adapt to the collaborative filtering algorithm.

[0123] The cosine similarity can be used to calculate the score similarity between the target user and other users. The calculation formula is:

[0124]

[0125] Where: u and v are the rating vectors of user u and user v respectively; · represents the dot product of the vectors; ∥·∥ represents the modulus of the vector.

[0126] The value range of cosine similarity is [-1, 1], where 1 means complete similarity, -1 means complete dissimilarity, and 0 means no correlation.

[0127] The calculation formula for predicting the target user's predicted rating for unrated components is:

[0128]

[0129] in, Give target user u a predicted score for component i; is the average rating of the target user u; N(u) is the set of users similar to the target user u; w uv is the score similarity between target user u and user v; r vi is the actual rating of component i by user v; is the average rating of user v.

[0130] The prediction score formula consists of two parts: the average score of the target user and a correction term. The correction term is the weighted sum of the rating deviations of similar users (i.e., neighboring users) on the component, where the weight is the similarity of the ratings between users.

[0131] According to the predicted scores, a preset number of components with the highest scores are selected to generate a recommendation list, wherein the preset number can be set according to actual needs.

[0132] The parameters of the collaborative filtering model, such as the similarity threshold and the number of neighbors, are adjusted through the feedback from target users to optimize the recommendation effect and obtain a component recommendation model for multiple application scenarios.

[0133] This embodiment uses a collaborative filtering model to provide users with personalized recommendations based on their own behavior and interest patterns; as user behavior data continues to accumulate, the collaborative filtering model can dynamically learn changes in user preferences.

[0134] In one embodiment, the method further includes: obtaining user usage scores, component click times, and selection records for the component recommendation model for multiple application scenarios to form feedback information;

[0135] The feedback information is used as an input variable and input into a component recommendation model for multiple application scenarios, and the component recommendation model for multiple application scenarios is further iterated and optimized.

[0136] Specifically, a user feedback mechanism is established to store user usage scores, component clicks, browsing time, selection records and other information in a database to form feedback information, which is later input into the recommendation model as a variable for further iterative optimization of the model to improve the recommendation accuracy and user-friendliness of the recommendation model and provide users with components that are more in line with their usage habits.

[0137] It is understandable that the test set can be used to evaluate the performance of the component recommendation model for multiple application scenarios (hereinafter referred to as the model) constructed above. The quality of the model can be measured by indicators such as accuracy (the proportion of correct predictions by the model) and recall (the proportion of positive class instances predicted correctly by the model). The accuracy is determined by judging whether the predicted five recommended component models are consistent with the component models marked in the data set.

[0138] like Figure 2 As shown: 5 different test data are randomly selected in the test set to verify the model accuracy (Accuracy Rate) and recall rate (Recall Rate). It can be seen that each accuracy rate and recall rate are above 80%, that is, the component recommendation model for multiple application scenarios constructed by the above embodiment has high performance.

[0139] The component recommendation model for multiple application scenarios constructed through this application can provide model designers, quality management personnel, experts, and material management personnel with comprehensive component information and component recommendation services, with a recommendation accuracy rate of over 80%.

[0140] Figure 3 FIG. 1 is a schematic diagram of the structure of an electronic device provided by an embodiment of the present invention. Figure 3 As shown, a structural schematic diagram of an electronic device 300 suitable for implementing an embodiment of the present application is shown.

[0141] like Figure 3 As shown, the electronic device 300 includes a central processing unit (CPU) 301, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 302 or a program loaded from a storage part 308 into a random access memory (RAM) 303. In the RAM 303, various programs and data required for the operation of the device 300 are also stored. The CPU 301, the ROM 302, and the RAM 303 are connected to each other via a bus 304. An input / output (I / O) interface 305 is also connected to the bus 304.

[0142] The following components are connected to the I / O interface 305: an input section 306 including a keyboard, a mouse, etc.; an output section 307 including a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker, etc.; a storage section 308 including a hard disk, etc.; and a communication section 309 including a network interface card such as a LAN card, a modem, etc. The communication section 309 performs communication processing via a network such as the Internet. A drive 310 is also connected to the I / O interface 305 as needed. A removable medium 311, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 310 as needed, so that a computer program read therefrom is installed into the storage section 308 as needed.

[0143] In particular, according to the embodiments of the present disclosure, the above reference Figure 1 The described process can be implemented as a computer software program. For example, an embodiment of the present disclosure includes a computer program product, which includes a computer program tangibly contained on a machine-readable medium, and the computer program includes program code for executing the above-mentioned method for constructing a component recommendation model for multiple application scenarios. In such an embodiment, the computer program can be downloaded and installed from a network through the communication part 309, and / or installed from a removable medium 311.

[0144] The flowcharts and block diagrams in the accompanying drawings illustrate the possible architecture, functions and operations of the systems, methods and computer program products according to various embodiments of the present invention. In this regard, each box in the flowchart or block diagram may represent a module, a program segment, or a portion of code, and the aforementioned module, program segment, or a portion of code contains one or more executable instructions for implementing the specified logical functions. It should also be noted that in some alternative implementations, the functions marked in the boxes may also be different from those in the examples.

[0145] The units or modules involved in the embodiments described in the present application may be implemented by software or hardware. The units or modules described may also be arranged in a processor. The names of these units or modules do not constitute limitations on the units or modules themselves in certain circumstances.

[0146] The systems, devices, modules or units described in the above embodiments may be implemented by computer chips or entities, or by products with certain functions. A typical implementation device is a computer. Specifically, the computer may be, for example, a personal computer, a laptop, a mobile phone, a smart phone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or a combination of any of these devices.

[0147] As another aspect, the present application further provides a storage medium, which may be the storage medium included in the aforementioned device in the above embodiment; or it may be a storage medium that exists independently and is not assembled into the device. The storage medium stores one or more programs, and the aforementioned programs are used by one or more processors to execute the method for constructing a component recommendation model for multiple application scenarios described in the present application.

[0148] Storage media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer-readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer-readable media does not include temporary computer-readable media (transitory media), such as modulated data signals and carrier waves.

[0149] It should be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of further restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0150] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

Claims

1. A method for constructing a component recommendation model for multiple application scenarios, characterized in that: The method comprises: Acquire a data set; the data set includes information data that users are concerned about when selecting components; Annotating the information data in the data set to obtain an annotation set; the annotation set includes 70% of a training set and 30% of a test set; The recommendation model is trained using the training set data to obtain the component recommendation model for multiple application scenarios.

2. The method according to claim 1, characterized in that The step of labeling the information data in the data set to obtain a labeling set includes: Performing random sampling of no less than 20% of the data set to form a data set to be labeled; Determine the indicators that users are most concerned about in each application scenario in multiple application scenarios; Marking the indicators that the user is most concerned about as decisive factors for recommendation; If the decisive factors of multiple components are the same, they are annotated according to other concerns of the user to obtain the annotation set.

3. The method according to claim 2, characterized in that The multiple application scenarios include: design selection, selection review, and quality issue handling.

4. The method according to claim 1, characterized in that: The recommendation model includes a random forest model; The using the training set data to train the random forest model to obtain the component recommendation model for multiple application scenarios includes: A number of samples are randomly selected from the training set using the Bootstrap sampling method; the selected samples are used to construct a decision tree; Extracting features from each of the samples through feature engineering; Finding the best splitting point according to the features; Until all the decision trees are constructed, the component recommendation model for multiple application scenarios is obtained.

5. The method according to claim 4, characterized in that The prediction results of the component recommendation model for multiple application scenarios include: For classification problems, the majority voting method is used to determine the final prediction result; For regression problems, the average of all the prediction results of the decision trees is taken as the prediction result.

6. The method according to claim 5, characterized in that For classification problems, let C K is the classification prediction result of the Kth decision tree, and the prediction result of the component recommendation model for multiple application scenarios is: C(x)=mode{C1(x),C2(x),…,C K (x)} Among them, mode represents the majority, that is, the most frequently occurring classification result in K decision trees; For regression problems, let R k is the regression prediction result of the kth decision tree, and the prediction result of the component recommendation model for multiple application scenarios is: Here, K is the number of decision trees.

7. The method according to claim 4, characterized in that The recommendation model also includes a collaborative filtering model; the training set data includes user rating data on components; the users include target users and other users; The using the training set data to train the collaborative filtering model to obtain the component recommendation model for multiple application scenarios includes: Calculating the similarity of the scores between the target user and other users based on the target user's score data on the component and the other users' score data on the component; Select a group of users with the highest score similarity to the target user as neighbor users; Predicting the target user's predicted rating for the unrated component based on the rating data of the neighbor user and the rating similarity between the target user and the neighbor user; Generating a recommendation list for the target user according to the predicted score; Based on the recommendation list generated for the target user, the parameters of the collaborative filtering model are adjusted to obtain the component recommendation model for multiple application scenarios.

8. The method according to claim 1, characterized in that The method further comprises: Obtaining user usage scores, component click times, and selection records for the component recommendation model for multiple application scenarios to form feedback information; The feedback information is used as an input variable and input into the component recommendation model for multiple application scenarios, and the component recommendation model for multiple application scenarios is further iteratively optimized.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, it implements the method for constructing a component recommendation model for multiple application scenarios as described in any one of claims 1-8.

10. A readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, it implements the method for constructing a component recommendation model for multiple application scenarios as described in any one of claims 1-8.

Images