Prediction method and device, computer equipment, storage medium and program product

Abstract

The invention relates to a prediction method and device, computer equipment, a storage medium and a program product. The method comprises the following steps: acquiring related data of a to-be-identified object; inputting the related data of the to-be-identified object into a preset prediction network for risk prediction to obtain a prediction result; the preset recognition network comprises a feature extraction sub-network and a prediction sub-network; the feature extraction sub-network is used for performing category feature and numerical feature extraction on the input related data; and the prediction sub-network is used for performing prediction based on the feature information output by the feature extraction sub-network. According to the method, the category features and the numerical features of the related data of the to-be-identified object are extracted through the feature extraction sub-network in the preset prediction network, so that risk prediction is performed based on the category features and the numerical features, and the accuracy of risk prediction is greatly improved.

Description

Technical Field

[0001] The present application relates to the technical field of big data processing, and in particular to a prediction method, apparatus, computer equipment, storage medium and program product. Background Art

[0002] With the rapid development of the combination of social management and scientific and technological applications, the application of intelligent models to comprehensive social governance has become an inevitable trend in the current comprehensive management application. For example, classification models can be used to identify special groups of people, or statistical analysis models can be used to conduct efficient censuses and surveys of the population.

[0003] At present, as an important part of the transportation system, truck drivers have become a high-risk group for traffic accidents due to their high risk, low income, limited social security coverage, and insufficient road rescue service supply. Therefore, in order to reduce the occurrence of traffic accidents, the means of accurately predicting the accident risk of truck drivers has become a research hotspot in the field of traffic management. Summary of the invention

[0004] Based on this, it is necessary to provide a prediction method, device, computer equipment, storage medium and program product that can accurately predict the accident risk of truck drivers in response to the above technical problems.

[0005] In a first aspect, the present application provides a prediction method, comprising:

[0006] Obtain relevant data of the object to be identified;

[0007] The relevant data of the object to be identified is input into the preset prediction network for risk prediction to obtain the prediction result; the preset identification network includes a feature extraction subnetwork and a prediction subnetwork; the feature extraction subnetwork is used to extract category features and numerical features of the input relevant data; the prediction subnetwork is used to make predictions based on the feature information output by the feature extraction subnetwork.

[0008] The above method extracts the category features and numerical features in the relevant data of the object to be identified respectively through a feature extraction subnetwork in a preset prediction network, so as to perform risk prediction based on the category features and numerical features. Compared with the existing method that does not distinguish between category features and numerical features and only performs risk prediction on the relevant data of the object to be identified based on the prediction network, the present application greatly improves the accuracy of risk prediction by distinguishing the relevant data into category features and data features.

[0009] In one embodiment, the feature extraction subnetwork includes a first classification module and a feature extraction module. The relevant data of the object to be identified is input into a preset prediction network for risk prediction to obtain a prediction result, including:

[0010] Inputting relevant data of the object to be identified into the first classification module for classification to obtain numerical feature data and category feature data;

[0011] Inputting the numerical feature data and the category feature data into the feature extraction module for feature extraction to obtain a feature vector; the feature vector includes the category feature and the numerical feature;

[0012] The feature vector is input into the prediction subnetwork for risk prediction to obtain the prediction result.

[0013] The risk prediction method provided in the embodiment of the present application is based on a classification module, a feature extraction module and a prediction network, and considers using different feature extraction methods for different types of data, so that the extracted data features are more accurate, and then risk prediction is performed based on the accurate data features, making the prediction results more accurate.

[0014] In one embodiment, the feature extraction module includes a conversion layer and a connection layer, and the numerical feature data and the category feature data are input into the feature extraction module for feature extraction to obtain a feature vector, including:

[0015] Input the category feature data into the conversion layer for data conversion to obtain dense type feature data;

[0016] The dense type feature data and the numerical feature data are input into the connection layer for concatenation to obtain the feature vector.

[0017] The method for obtaining feature vectors provided in the embodiments of the present application adopts different data processing methods for different types of data, so that the processed data is more accurate, providing a data basis for subsequent risk prediction based on the processed data and a preset prediction network, and making the risk prediction results of the objects to be identified more accurate.

[0018] In one embodiment, the feature extraction module further includes a pooling layer, and the dense type feature data and the numerical feature data are input into the connection layer for splicing to obtain a feature vector, including:

[0019] Input the dense type feature data and the numerical feature data into the connection layer for splicing to obtain the spliced ​​feature data;

[0020] The concatenated feature data and category feature data are input into the pooling layer for cross feature extraction to obtain the feature vector.

[0021] The method for acquiring feature vectors provided in the embodiments of the present application also performs cross-pooling processing on the spliced ​​feature vectors, so that the processed data is more accurate, providing a data basis for subsequent risk prediction based on the processed data and a preset prediction network, and making the risk prediction results of the objects to be identified more accurate.

[0022] In one embodiment, the prediction result includes a predicted risk value, the prediction sub-network includes an estimation module and an adjustment module, and the feature vector is input into the prediction sub-network for risk prediction to obtain a prediction result, including:

[0023] The feature vector is input into the estimation module for risk estimation to obtain an estimated risk value;

[0024] Inputting the numerical feature data and the category feature data into the adjustment module to obtain an adjustment value;

[0025] The estimated risk value is adjusted according to the adjustment value to obtain the predicted risk value.

[0026] The method for obtaining prediction results provided in the embodiments of the present application obtains an estimated risk value based on an estimation module, obtains an adjustment value based on an adjustment module, and adjusts the estimated risk value based on the adjustment value, so that the final predicted risk value is more accurate.

[0027] In one embodiment, the prediction result further includes a classification result, the prediction subnetwork further includes a second classification module, and the method further includes:

[0028] The predicted risk value is input into the second classification module to classify the object to obtain a classification result; the classification result is used to characterize whether the object to be identified is a high-risk type object.

[0029] The classification method provided in the embodiment of the present application performs binary classification based on the predicted risk value to obtain the classification result of the object to be identified, so that the final classification result is more readable, and the user can determine whether the object to be identified belongs to the high-risk type by seeing the evaluation result.

[0030] In a second aspect, the present application also provides a prediction device, comprising:

[0031] An acquisition module, used to acquire relevant data of the object to be identified;

[0032] The prediction module is used to input the relevant data of the object to be identified into the preset prediction network for risk prediction and obtain the prediction result; the preset identification network includes a feature extraction subnetwork and a prediction subnetwork; the feature extraction subnetwork is used to extract category features and numerical features of the input relevant data; the prediction subnetwork is used to make predictions based on the feature information output by the feature extraction subnetwork.

[0033] In a third aspect, the present application further provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the following steps are implemented:

[0034] Obtain relevant data of the object to be identified;

[0035] The relevant data of the object to be identified is input into the preset prediction network for risk prediction to obtain the prediction result; the preset identification network includes a feature extraction subnetwork and a prediction subnetwork; the feature extraction subnetwork is used to extract category features and numerical features of the input relevant data; the prediction subnetwork is used to make predictions based on the feature information output by the feature extraction subnetwork.

[0036] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the following steps are implemented:

[0037] Obtain relevant data of the object to be identified;

[0038] The relevant data of the object to be identified is input into the preset prediction network for risk prediction to obtain the prediction result; the preset identification network includes a feature extraction subnetwork and a prediction subnetwork; the feature extraction subnetwork is used to extract category features and numerical features of the input relevant data; the prediction subnetwork is used to make predictions based on the feature information output by the feature extraction subnetwork.

[0039] In a fifth aspect, the present application further provides a computer program product, including a computer program, which implements the following steps when executed by a processor:

[0040] Obtain relevant data of the object to be identified;

[0041] The relevant data of the object to be identified is input into the preset prediction network for risk prediction to obtain the prediction result; the preset identification network includes a feature extraction subnetwork and a prediction subnetwork; the feature extraction subnetwork is used to extract category features and numerical features of the input relevant data; the prediction subnetwork is used to make predictions based on the feature information output by the feature extraction subnetwork.

[0042] The above-mentioned prediction method, device, computer equipment, storage medium and program product, the above-mentioned method extracts the category characteristics and numerical characteristics of the relevant data of the object to be identified through a feature extraction subnetwork in a preset prediction network, so as to perform risk prediction based on the category characteristics and numerical characteristics. Compared with the existing method that does not distinguish between category characteristics and numerical characteristics and only performs risk prediction on the relevant data of the object to be identified based on the prediction network, the present application greatly improves the accuracy of risk prediction by distinguishing the relevant data into category characteristics and data characteristics. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the drawings required for use in the embodiments of the present application or related technical descriptions will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0044] FIG1 is a diagram of an application environment of a prediction method in one embodiment;

[0045] Figure 2 A schematic diagram of a prediction method in one embodiment;

[0046] Figure 3 is a flow chart of a prediction method in another embodiment;

[0047] Figure 4 is a flow chart of a prediction method in another embodiment;

[0048] Figure 5 is a flow chart of a prediction method in another embodiment;

[0049] Figure 6 is a flow chart of a prediction method in another embodiment;

[0050] Figure 7 is a flow chart of a prediction method in another embodiment;

[0051] Figure 8 is a flow chart of a prediction method in another embodiment;

[0052] FIG. 9 is a structural block diagram of a prediction device in one embodiment. DETAILED DESCRIPTION

[0053] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0054] With the rapid development of the combination of social management and scientific and technological applications, the application of intelligent models to comprehensive social governance has become an inevitable trend in the current comprehensive management application. For example, classification models can be used to identify special groups of people, or statistical analysis models can be used to conduct efficient censuses and surveys of the population.

[0055] At present, as an important part of the transportation system, truck drivers have become a high-risk group for traffic accidents due to their high risk, low income, limited social security coverage, and insufficient road rescue service supply. Therefore, in order to reduce the occurrence of traffic accidents, the means of accurately predicting the accident risk of truck drivers has become a research hotspot in the field of traffic management.

[0056] After the above introduction to the background technology of the prediction method provided by the embodiment of the present application, the implementation environment involved in the prediction method provided by the embodiment of the present application will be briefly described below. The prediction method provided by the embodiment of the present application can be applied to a computer device as shown in Figure 1. The computer device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. Among them, the processor, the memory and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store relevant data of the object to be identified. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a prediction method is implemented.

[0057] Those skilled in the art will understand that the structure shown in FIG. 1 is merely a block diagram of a partial structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. A specific terminal may include more or fewer components than shown in the figure, or combine certain components, or have a different arrangement of components.

[0058] After introducing the application scenarios of the prediction method provided in the embodiments of the present application above, the prediction method described in the present application is mainly introduced below.

[0059] In one embodiment, Figure 2 As shown, a prediction method is provided, which is described by taking the method applied to the computer device in FIG. 1 as an example, and includes the following steps:

[0060] S201, obtaining relevant data of the object to be identified.

[0061] Among them, the objects to be identified refer to all participating objects in the traffic environment, such as truck drivers, bus drivers, taxi drivers, etc.

[0062] Among them, relevant data include offline data and real-time data. Among them, offline data refers to the periodic migration of large quantities of data, which does not require high timeliness and is generally obtained by distributed batch data synchronization. Real-time data is for low-latency data application scenarios and is generally obtained through incremental logs or notification messages. The relevant data in this application refers to the basic information of the participating objects and the relevant information of the participating objects. For example, if the object to be identified is a truck driver, then the relevant data of the object to be identified includes the basic information of the truck driver, physical condition information, economic status information, health information, working environment information, employment relationship information, secondary disaster occurrence information, transportation information, basic information of the vehicle, road rescue service information, rescue supply information, basic accident information, operation and transportation information, annual inspection information, scrapping information, maintenance information, qualification review information, etc.

[0063] In an embodiment of the present application, when it is necessary to evaluate and predict the risk level of various groups in a traffic environment, for example, when it is necessary to predict the risk level of truck drivers in a traffic environment, data warehouse technology (Extract-Load-Transform, ELT) can be used to obtain relevant data of the truck driver group from a database or online announcement.

[0064] Optionally, after obtaining the relevant data of the truck driver group, the relevant data of the truck driver group can also be cleaned and preprocessed, for example, data cleaning processing, data integration processing, data conversion processing and data normalization processing, etc., to obtain the relevant data of the truck drivers after cleaning and preprocessing, so that the subsequent risk status of the object to be identified predicted based on the data after data cleaning and data preprocessing will be more accurate.

[0065] S202: Input relevant data of the object to be identified into a preset prediction network to perform risk prediction and obtain a prediction result.

[0066] The preset prediction network is trained based on the initial prediction model and the sample data of the object to be identified. The preset recognition network includes a feature extraction subnetwork and a prediction subnetwork. The feature extraction subnetwork is used to extract category features and numerical features of the input related data, and the prediction subnetwork is used to make predictions based on the feature information output by the feature extraction subnetwork.

[0067] Among them, categorical features refer to features corresponding to data with high dimensionality and sparseness, such as the basic information features of the driver and the basic information features of the vehicle, etc., and numerical features refer to features corresponding to data with low dimensionality and denseness, such as the economic status information features of the driver and the maintenance information features of the vehicle, etc.

[0068] The prediction result includes the risk assessment value of the object to be identified, or whether the object to be identified has risks.

[0069] In the embodiment of the present application, after the relevant data of the object to be identified is obtained as described above, the relevant data of the object to be identified can be input into a pre-trained preset prediction network to perform risk prediction to obtain a prediction result.

[0070] The prediction method provided in the embodiment of the present application obtains the relevant data of the object to be identified, inputs the relevant data of the object to be identified into a preset prediction network for risk prediction, and obtains a prediction result; the preset identification network includes a feature extraction subnetwork and a prediction subnetwork, the feature extraction subnetwork is used to extract category features and numerical features of the input relevant data, and the prediction subnetwork is used to predict based on the feature information output by the feature extraction subnetwork. The above method extracts the category features and numerical features in the relevant data of the object to be identified respectively through the feature extraction subnetwork in the preset prediction network, so as to predict risks based on the category features and numerical features. Compared with the existing method that does not distinguish between category features and numerical features and only predicts risks on the relevant data of the object to be identified based on the prediction network, the present application greatly improves the accuracy of risk prediction by distinguishing the relevant data into category features and data features.

[0071] In one embodiment, Figure 2 Based on the embodiment shown, the feature extraction subnetwork includes a first classification module and a feature extraction module, which can describe the risk prediction process, such as Figure 3 As shown, the above S202 "inputting relevant data of the object to be identified into a preset prediction network for risk prediction to obtain a prediction result" includes:

[0072] S301, inputting relevant data of the object to be identified into a first classification module for classification to obtain numerical feature data and category feature data.

[0073] Among them, the first classification model is used to classify data based on the sparsity of the data, for example, data with a sparsity greater than a preset sparsity threshold is determined as category feature data, and data with a sparsity not greater than a preset sparsity threshold is determined as numerical feature data.

[0074] Among them, numerical feature data refers to data with low dimension and dense data, such as the driver's economic status information, vehicle maintenance information, etc., and category feature data refers to data with high dimension and sparse data, such as the driver's basic information characteristics, the vehicle's basic information characteristics, etc.

[0075] In an embodiment of the present application, after the relevant data of the object to be identified is obtained as mentioned above, the relevant data of the object to be identified can be input into the first classification model for classification to determine which of the relevant data belong to numerical feature data and which belong to category feature data, thereby obtaining numerical feature data and category feature data.

[0076] S302: Input the numerical feature data and the category feature data into a feature extraction module to extract features and obtain a feature vector.

[0077] Among them, the feature vector includes category features and numerical features.

[0078] Among them, the feature extraction module can be a pre-trained feature extraction network, such as a recurrent neural network (RNN), a convolutional neural network (CNN) and a Transformer network.

[0079] In an embodiment of the present application, after the numerical feature data and the categorical feature data are acquired as described above, the numerical feature data and the categorical feature data can be directly input into a feature extraction module for feature extraction to obtain a feature vector; optionally, the numerical feature data and the categorical feature data can be directly input into a pre-trained feature extraction network for feature extraction to obtain a feature vector.

[0080] S303: Input the feature vector into the prediction subnetwork to perform risk prediction and obtain a prediction result.

[0081] The prediction subnetwork may be a pre-trained prediction network, such as a decision tree network, a support vector machine regression network, etc.

[0082] In the embodiment of the present application, after the feature vector of the relevant data is obtained as described above, the feature vector of the relevant data is input into the prediction subnetwork for risk prediction to obtain the prediction result of the object to be identified.

[0083] The risk prediction method provided in the embodiment of the present application is based on a classification module, a feature extraction module and a prediction network, and considers using different feature extraction methods for different types of data, so that the extracted data features are more accurate, and then risk prediction is performed based on the accurate data features, making the prediction results more accurate.

[0084] In one embodiment, Figure 3 Based on the embodiment shown, the feature extraction module includes a conversion layer and a connection layer, and the feature extraction process can be described as follows: Figure 4As shown, the above S301 "inputting the numerical feature data and the category feature data into the feature extraction module for feature extraction to obtain a feature vector" includes:

[0085] S401, inputting the category feature data into the conversion layer for data conversion to obtain dense type feature data.

[0086] The conversion layer is used to convert sparse data into dense data. For example, the conversion layer may be an Embedding layer. It should be noted that the dimension of the dense data after conversion will be much smaller than that of the sparse data.

[0087] Optionally, the process of the Embedding layer performing dense conversion on sparse data includes: first, defining an Embedding mapping of sparse features, in which the dictionary is {key:value}, key represents each sparse feature, value represents the number of different values ​​of the corresponding column of the data set data_df, as the input dimension of the embedding layer, then, encapsulating the sparse feature information, and constructing the sparse data into a data pipeline, the pipeline includes data information and feature information, and the data information and feature information in the pipeline are used as input parameters of the model, and finally output as the dense conversion result.

[0088] In the embodiment of the present application, after the category feature data in the relevant data of the object to be identified is obtained, the category feature data can be input into the conversion layer for data conversion to obtain dense type feature data.

[0089] S402: Input the dense type feature data and the numerical feature data into the connection layer for concatenation to obtain a feature vector.

[0090] The connection layer is used to concatenate the dense feature data after data conversion and the numerical feature data in the relevant data of the object to be identified to form a new feature vector containing all the features. For example, the connection layer can be a Stacking layer.

[0091] In an embodiment of the present application, after the dense type feature data and the numerical feature data are obtained as described above, the dense type feature data and the numerical feature data can be simultaneously input into the connection layer for splicing processing to obtain the dense type feature data of the spliced ​​feature vector.

[0092] Optionally, if the data is sparse categorical feature data, the categorical feature data needs to be first input into the conversion layer for data conversion to convert the sparse categorical feature data into dense type feature data, and then the dense type feature data is input into the connection layer to be spliced ​​with the dense numerical feature data to obtain the spliced ​​feature data; if the data is dense numerical feature data, the dense numerical feature data can be directly input into the connection layer.

[0093] The method for obtaining feature vectors provided in the embodiments of the present application adopts different data processing methods for different types of data, so that the processed data is more accurate, providing a data basis for subsequent risk prediction based on the processed data and a preset prediction network, and making the risk prediction results of the objects to be identified more accurate.

[0094] In one embodiment, Figure 4 Based on the embodiment shown, the feature extraction module further includes a pooling layer, which can further describe the feature extraction process, such as Figure 5 As shown, the above S402 "inputting dense type feature data and numerical feature data into the connection layer for concatenation to obtain a feature vector" includes:

[0095] S501, input dense type feature data and numerical feature data into the connection layer for splicing to obtain spliced ​​feature data.

[0096] In the embodiment of the present application, after the dense type feature data and the numerical feature data are acquired as described above, the dense type feature data and the numerical feature data can be simultaneously input into the connection layer for splicing processing to obtain the spliced ​​feature data.

[0097] S502: Input the concatenated feature data and category feature data into a pooling layer for cross feature extraction to obtain a feature vector.

[0098] The pooling layer is used to perform cross-pooling on the input data to obtain the feature vector after cross-pooling. The process of cross-pooling in the pooling layer is shown in the following formula (1):

[0099] f BI ( v x )= 1 2 [ ( ∑ i=1 n x i v i ) 2 - ∑ i=1 n (x i v i ) 2 ] （ 1 ）；

[0100] In the formula, refers to the feature vector after pooling, refers to the concatenated feature data. Refers to category feature data.

[0101] In the embodiment of the present application, the concatenated feature data and category feature data are input into the pooling layer for cross feature extraction to obtain a feature vector. Optionally, a CNN model can be embedded in the pooling layer so that only the corresponding positions of the two vectors are multiplied without information concentration, and all vectors are added and pooled and then input into the DNN model. The task of the DNN model is to continue to extract and concentrate high-order information based on the input second-order fusion information vector.

[0102] Furthermore, the DNN model performs DNN linear layer processing. First, based on the sparsity of the vector, the input vector is divided into two parts, namely, a dense numerical vector and a sparse categorical vector. The categorical vector is then converted to long. Different categorical features are respectively embedded[(batch_size, embed_dim)], and then the Embedding layer is used for stacking (discrete feature number, batch_size, embed_dim) / (batch_size, discrete feature number, embed_dim). Here, the shape of the embedding vector sparse_embeds is (batch_size, discrete feature number, embed_dim). Then, the feature cross layer is performed and cross pooling is performed according to the above formula (1). This is the sparse_embeds after Embedding processing. By setting dim=1, dim=1 is compressed (the same position of the row is added and dim=1 is removed), that is, feature crossover is performed; then, the numerical and categorical features are concatenated (batch_size, embed_dim + dense_input dimension), and the concatenated feature vector is input into the Dnn model, and the output result is the feature vector.

[0103] The method for acquiring feature vectors provided in the embodiments of the present application also performs cross-pooling processing on the spliced ​​feature vectors, so that the processed data is more accurate, providing a data basis for subsequent risk prediction based on the processed data and a preset prediction network, and making the risk prediction results of the objects to be identified more accurate.

[0104] In one embodiment, Figure 3 Based on the embodiment shown, the above prediction result includes a predicted risk value, and the above prediction subnetwork includes an estimation module and an adjustment module, which can further describe the process of risk prediction, such as Figure 6As shown, the above S303 "inputting the feature vector into the prediction sub-network for risk prediction to obtain a prediction result" includes:

[0105] S601. Input the feature vector into an estimation module for risk estimation to obtain an estimated risk value.

[0106] In the embodiment of the present application, after the feature vector is obtained, the feature vector can be directly input into a pre-trained prediction model for risk assessment to obtain an estimated risk value output by the prediction model. Optionally, the prediction algorithm of the prediction model can be expressed by the following formula (2):

[0107] y deep =deep(f(x))=deep( f BI ( v x ))=deep( 1 2 [ ( ∑ i=1 n x i v i ) 2 - ∑ i=1 n (x i v i ) 2 ]) (2);

[0109] in, It refers to the estimated risk value after the feature vector is input into the deep prediction model for risk assessment.

[0110] S602: Input the numerical feature data and the category feature data into an adjustment module to obtain an adjustment value.

[0111] In the embodiment of the present application, after obtaining the estimated risk value output by the estimation module, the obtained numerical feature data and category feature data can also be input into the adjustment module to obtain an adjustment value. Optionally, the adjustment algorithm of the adjustment model can be expressed by the following formula (3):

[0112]

[0113] in, It refers to the adjustment value output after the numerical feature data and categorical feature data are input into the adjustment algorithm. is a fixed adjustment parameter value, It refers to numerical feature data or categorical feature data. Refers to the weight value corresponding to numerical feature data or categorical feature data.

[0114] In the embodiment of the present application, after the numerical feature data and the category feature data are acquired as described above, the numerical feature data and the category feature data may be input into an adjustment module for adjustment processing to obtain an adjustment value.

[0115] S603. Adjust the estimated risk value according to the adjustment value to obtain a predicted risk value.

[0116] The predicted risk value may be the degree of possibility that the object to be identified has risks, or may be whether the object to be identified has risks.

[0117] In an embodiment of the present application, after the adjustment value and the estimated risk value are obtained as described above, the estimated risk value can be adjusted according to the adjustment value to obtain a final predicted risk value. Optionally, the sum of the adjustment value and the estimated risk value can be used as the final predicted risk value; optionally, the difference between the adjustment value and the estimated risk value can also be used as the final predicted risk value. It should be noted that the present application does not limit how to adjust the estimated risk value according to the adjustment value.

[0118] The method for obtaining prediction results provided in the embodiments of the present application obtains an estimated risk value based on an estimation module, obtains an adjustment value based on an adjustment module, and adjusts the estimated risk value based on the adjustment value, so that the final predicted risk value is more accurate.

[0119] In one embodiment, Figure 6 Based on the embodiment shown, the above prediction result also includes a classification result, and the above prediction subnetwork also includes a second classification module, such as Figure 7 As shown, the above method also includes:

[0120] S604: Input the predicted risk value into the second classification module to classify the object and obtain a classification result.

[0121] Among them, the second classification module can be a classifier, which is used to determine whether the object to be identified is a high-risk type object based on the predicted risk value; for example, if the above-mentioned predicted risk value is 70% and the preset risk threshold is 80%, the predicted risk value is not greater than the preset risk threshold, indicating that the object to be identified does not belong to the high-risk type; for another example, if the above-mentioned predicted risk value is 90% and the preset risk threshold is 80%, the predicted risk value is greater than the preset risk threshold, indicating that the object to be identified is a high-risk type.

[0122] The classification result is used to indicate whether the object to be identified is a high-risk object.

[0123] In the embodiment of the present application, after the predicted risk value is obtained as described above, the predicted risk value can be input into the second classification model to perform object classification to obtain a classification result of whether the object to be identified belongs to a high-risk type object.

[0124] Optionally, in an embodiment of the present application, the viewing of classification results is divided into two steps. The first step is to set indicators, and the second step is to group the indicators. The purpose of grouping is to construct the obtained classification results on the business side to obtain results that can produce business meanings.

[0125] The classification method provided in the embodiment of the present application performs binary classification based on the predicted risk value to obtain the classification result of the object to be identified, so that the final classification result is more readable, and the user can determine whether the object to be identified belongs to the high-risk type by seeing the evaluation result.

[0126] In one embodiment, Figure 8 As shown, a complete prediction method is also provided, including:

[0127] S10, obtaining relevant data of the object to be identified;

[0128] S11, inputting relevant data of the object to be identified into a first classification module for classification to obtain numerical feature data and category feature data;

[0129] S12, inputting the category feature data into the conversion layer for data conversion to obtain dense type feature data;

[0130] S13, inputting dense type feature data and numerical feature data into the connection layer for splicing to obtain spliced ​​feature data;

[0131] S14, inputting the concatenated feature data and category feature data into a pooling layer for cross feature extraction to obtain a feature vector;

[0132] S15, inputting the feature vector into an estimation module for risk estimation to obtain an estimated risk value;

[0133] S16, inputting the numerical feature data and the category feature data into an adjustment module to obtain an adjustment value;

[0134] S17, adjusting the estimated risk value according to the adjustment value to obtain a predicted risk value;

[0135] S18. Input the predicted risk value into the second classification module to classify the object and obtain the classification result.

[0136] The above method extracts the category features and numerical features of the relevant data of the object to be identified through a feature extraction subnetwork in a preset prediction network, so as to perform risk prediction based on the category features and numerical features. Compared with the existing method that does not distinguish between category features and numerical features and only performs risk prediction on the relevant data of the object to be identified based on the prediction network, the present application greatly improves the accuracy of risk prediction by distinguishing the relevant data into category features and data features.

[0137] It should be understood that, although the various steps in the flowcharts involved in the above-mentioned embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-mentioned embodiments can include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0138] Based on the same inventive concept, the embodiment of the present application also provides a prediction device for implementing the prediction method involved above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above method, so the specific limitations in one or more prediction device embodiments provided below can refer to the limitations of the prediction method above, and will not be repeated here.

[0139] In an exemplary embodiment, as shown in FIG9 , a prediction device is provided, including: an acquisition module 10 and a prediction module 11, wherein:

[0140] The acquisition module 10 is used to acquire relevant data of the object to be identified.

[0141] The prediction module 11 is used to input the relevant data of the object to be identified into the preset prediction network for risk prediction to obtain the prediction result; the preset identification network includes a feature extraction subnetwork and a prediction subnetwork; the feature extraction subnetwork is used to extract category features and numerical features of the input relevant data; the prediction subnetwork is used to make predictions based on the feature information output by the feature extraction subnetwork.

[0142] In an exemplary embodiment, the feature extraction subnetwork includes a first classification module and a feature extraction module, and the prediction module 11 includes: a classification unit, a feature extraction unit and a prediction unit, wherein:

[0143] The classification unit is specifically used to input the relevant data of the object to be identified into the first classification module for classification, so as to obtain numerical feature data and category feature data;

[0144] The feature extraction unit is specifically used to input the numerical feature data and the category feature data into the feature extraction module for feature extraction to obtain a feature vector; the feature vector includes the category feature and the numerical feature;

[0145] The prediction unit is specifically used to input the feature vector into the prediction subnetwork for risk prediction to obtain a prediction result.

[0146] In an exemplary embodiment, the feature extraction module includes a conversion layer and a connection layer, and the feature extraction unit is further used to input the category feature data into the conversion layer for data conversion to obtain dense type feature data; and input the dense type feature data and the numerical feature data into the connection layer for splicing to obtain the feature vector.

[0147] In an exemplary embodiment, the feature extraction module also includes a pooling layer, and the feature extraction unit is specifically used to input dense type feature data and numerical feature data into the connection layer for splicing to obtain spliced ​​feature data; input the spliced ​​feature data and category feature data into the pooling layer for cross feature extraction to obtain a feature vector.

[0148] In an exemplary embodiment, the above-mentioned prediction results include a predicted risk value, the prediction subnetwork includes an estimation module and an adjustment module, and the above-mentioned prediction unit is specifically used to input the feature vector into the estimation module for risk estimation to obtain an estimated risk value; input the numerical feature data and the category feature data into the adjustment module to obtain an adjusted value; adjust the estimated risk value according to the adjustment value to obtain a predicted risk value.

[0149] In an exemplary embodiment, the above-mentioned prediction result also includes a classification result, the above-mentioned prediction subnetwork also includes a second classification module, and the above-mentioned prediction unit is specifically used to input the predicted risk value into the second classification module for object classification to obtain a classification result; the classification result is used to characterize whether the object to be identified is a high-risk type object.

[0150] Each module in the above prediction device can be implemented in whole or in part by software, hardware or a combination thereof. Each module can be embedded in or independent of a processor in a computer device in the form of hardware, or can be stored in a memory in a computer device in the form of software, so that the processor can call and execute the operations corresponding to each module.

[0151] In an exemplary embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be shown in FIG1. ​​The computer device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. The processor, the memory and the input / output interface are connected via a system bus, and the communication interface is connected to the system bus via the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store relevant data of the object to be identified. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal via a network connection. When the computer program is executed by the processor, a prediction method is implemented.

[0152] Those skilled in the art will understand that the structure shown in FIG. 1 is merely a block diagram of a partial structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have a different arrangement of components.

[0153] In an exemplary embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and when the processor executes the computer program, the following steps are implemented:

[0154] Obtain relevant data of the object to be identified;

[0155] The relevant data of the object to be identified is input into the preset prediction network for risk prediction to obtain the prediction result; the preset identification network includes a feature extraction subnetwork and a prediction subnetwork; the feature extraction subnetwork is used to extract category features and numerical features of the input relevant data; the prediction subnetwork is used to make predictions based on the feature information output by the feature extraction subnetwork.

[0156] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0157] Inputting relevant data of the object to be identified into the first classification module for classification to obtain numerical feature data and category feature data;

[0158] Inputting the numerical feature data and the category feature data into the feature extraction module for feature extraction to obtain a feature vector; the feature vector includes the category feature and the numerical feature;

[0159] The feature vector is input into the prediction subnetwork for risk prediction to obtain the prediction result.

[0160] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0161] Input the category feature data into the conversion layer for data conversion to obtain dense type feature data;

[0162] The dense type feature data and the numerical feature data are input into the connection layer for concatenation to obtain the feature vector.

[0163] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0164] Input the dense type feature data and the numerical feature data into the connection layer for splicing to obtain the spliced ​​feature data;

[0165] The concatenated feature data and category feature data are input into the pooling layer for cross feature extraction to obtain the feature vector.

[0166] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0167] The feature vector is input into the estimation module for risk estimation to obtain an estimated risk value;

[0168] Inputting the numerical feature data and the category feature data into the adjustment module to obtain an adjustment value;

[0169] The estimated risk value is adjusted according to the adjustment value to obtain the predicted risk value.

[0170] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0171] The predicted risk value is input into the second classification module to classify the object to obtain a classification result; the classification result is used to characterize whether the object to be identified is a high-risk type object.

[0172] In one embodiment, a computer readable storage medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, the following steps are implemented:

[0173] Obtain relevant data of the object to be identified;

[0174] The relevant data of the object to be identified is input into the preset prediction network for risk prediction to obtain the prediction result; the preset identification network includes a feature extraction subnetwork and a prediction subnetwork; the feature extraction subnetwork is used to extract category features and numerical features of the input relevant data; the prediction subnetwork is used to make predictions based on the feature information output by the feature extraction subnetwork.

[0175] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0176] Inputting relevant data of the object to be identified into the first classification module for classification to obtain numerical feature data and category feature data;

[0177] Inputting the numerical feature data and the category feature data into the feature extraction module for feature extraction to obtain a feature vector; the feature vector includes the category feature and the numerical feature;

[0178] The feature vector is input into the prediction subnetwork for risk prediction to obtain the prediction result.

[0179] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0180] Input the category feature data into the conversion layer for data conversion to obtain dense type feature data;

[0181] The dense type feature data and the numerical feature data are input into the connection layer for concatenation to obtain the feature vector.

[0182] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0183] Input the dense type feature data and the numerical feature data into the connection layer for splicing to obtain the spliced ​​feature data;

[0184] The concatenated feature data and category feature data are input into the pooling layer for cross feature extraction to obtain the feature vector.

[0185] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0186] The feature vector is input into the estimation module for risk estimation to obtain an estimated risk value;

[0187] Inputting the numerical feature data and the category feature data into the adjustment module to obtain an adjustment value;

[0188] The estimated risk value is adjusted according to the adjustment value to obtain the predicted risk value.

[0189] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0190] The predicted risk value is input into the second classification module to classify the object to obtain a classification result; the classification result is used to characterize whether the object to be identified is a high-risk type object.

[0191] In one embodiment, a computer program product is provided, comprising a computer program, which, when executed by a processor, implements the following steps:

[0192] Obtain relevant data of the object to be identified;

[0193] The relevant data of the object to be identified is input into the preset prediction network for risk prediction to obtain the prediction result; the preset identification network includes a feature extraction subnetwork and a prediction subnetwork; the feature extraction subnetwork is used to extract category features and numerical features of the input relevant data; the prediction subnetwork is used to make predictions based on the feature information output by the feature extraction subnetwork.

[0194] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0195] Inputting relevant data of the object to be identified into the first classification module for classification to obtain numerical feature data and category feature data;

[0196] Inputting the numerical feature data and the category feature data into the feature extraction module for feature extraction to obtain a feature vector; the feature vector includes the category feature and the numerical feature;

[0197] The feature vector is input into the prediction subnetwork for risk prediction to obtain the prediction result.

[0198] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0199] Input the category feature data into the conversion layer for data conversion to obtain dense type feature data;

[0200] The dense type feature data and the numerical feature data are input into the connection layer for concatenation to obtain the feature vector.

[0201] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0202] Input the dense type feature data and the numerical feature data into the connection layer for splicing to obtain the spliced ​​feature data;

[0203] The concatenated feature data and category feature data are input into the pooling layer for cross feature extraction to obtain the feature vector.

[0204] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0205] The feature vector is input into the estimation module for risk estimation to obtain an estimated risk value;

[0206] Inputting the numerical feature data and the category feature data into the adjustment module to obtain an adjustment value;

[0207] The estimated risk value is adjusted according to the adjustment value to obtain the predicted risk value.

[0208] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0209] The predicted risk value is input into the second classification module to classify the object to obtain a classification result; the classification result is used to characterize whether the object to be identified is a high-risk type object.

[0210] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0211] A person of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment method can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, an artificial intelligence (AI) processor, etc., but are not limited to this.

[0212] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0213] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A prediction method, characterized in that: The method comprises: Obtain relevant data of the object to be identified; The relevant data of the object to be identified is input into a preset prediction network for risk prediction to obtain a prediction result; the preset identification network includes a feature extraction subnetwork and a prediction subnetwork; the feature extraction subnetwork is used to extract category features and numerical features of the input relevant data; the prediction subnetwork is used to make predictions based on the feature information output by the feature extraction subnetwork.

2. The method according to claim 1, characterized in that The feature extraction subnetwork includes a first classification module and a feature extraction module. The relevant data of the object to be identified is input into a preset prediction network for risk prediction to obtain a prediction result, including: Inputting the relevant data of the object to be identified into the first classification module for classification to obtain numerical feature data and category feature data; Inputting the numerical feature data and the category feature data into the feature extraction module for feature extraction to obtain a feature vector; the feature vector includes the category feature and the numerical feature; The feature vector is input into the prediction subnetwork to perform risk prediction to obtain a prediction result.

3. The method according to claim 2, characterized in that The feature extraction module includes a conversion layer and a connection layer. The numerical feature data and the category feature data are input into the feature extraction module for feature extraction to obtain a feature vector, including: Inputting the category feature data into the conversion layer for data conversion to obtain dense type feature data; The dense type feature data and the numerical feature data are input into the connection layer for concatenation to obtain the feature vector.

4. The method according to claim 3, characterized in that The feature extraction module further includes a pooling layer, which inputs the dense type feature data and the numerical feature data into the connection layer for splicing to obtain the feature vector, including: Inputting the dense type feature data and the numerical feature data into the connection layer for splicing to obtain spliced ​​feature data; The concatenated feature data and the category feature data are input into the pooling layer for cross feature extraction to obtain the feature vector.

5. The method according to claim 2, characterized in that: The prediction result includes a predicted risk value, the prediction sub-network includes an estimation module and an adjustment module, and the feature vector is input into the prediction sub-network for risk prediction to obtain a prediction result, including: Inputting the feature vector into the estimation module to perform risk estimation to obtain an estimated risk value; Inputting the numerical feature data and the category feature data into the adjustment module to obtain an adjustment value; The estimated risk value is adjusted according to the adjustment value to obtain the predicted risk value.

6. The method according to claim 5, characterized in that The prediction result also includes a classification result, the prediction subnetwork also includes a second classification module, and the method further includes: The predicted risk value is input into the second classification module to classify the object to obtain the classification result; the classification result is used to characterize whether the object to be identified is a high-risk type object.

7. A prediction device, characterized in that: The device comprises: An acquisition module, used to acquire relevant data of the object to be identified; The prediction module is used to input the relevant data of the object to be identified into a preset prediction network for risk prediction to obtain a prediction result; the preset identification network includes a feature extraction subnetwork and a prediction subnetwork; the feature extraction subnetwork is used to extract category features and numerical features of the input relevant data; the prediction subnetwork is used to make predictions based on the feature information output by the feature extraction subnetwork.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 6 are implemented.

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

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images