Method and system for accelerating feature engineering derivative index construction

Through the system of automatic feature derivation, feature selection and model evaluation modules, the problems of inefficient feature engineering and insufficient automation are solved, efficient and accurate feature generation and model performance improvement are achieved, and different data types and business scenarios are adapted to different data types and business scenarios.

CN118939987APending Publication Date: 2024-11-12SHANGHAI XURONG NETWORK TECH CO LTD
View PDF 4 Cites 0 Cited by

Patent Information

Application Number
CN202410953486.4
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-16
Publication Date
2024-11-12

AI Technical Summary

Technical Problem

The existing feature engineering technology is inefficient, insufficient automation, error-prone and lacks adaptability, making it difficult to meet the needs of complex data processing, affecting model performance and stability.

Method used

Provide a system to accelerate the construction of derivative indicators of feature engineering, including automatic feature derivative module, feature selection module and model evaluation module. By identifying data type allocation operation rules, feature data is automatically generated, and indicators that improve model performance are selected through algorithms. Training and evaluation of machine learning models.

Benefits of technology

显著提升了衍生指标的生成速度、准确性和模型性能,减少了人工错误,提高了特征工程的自动化水平和适应性,满足复杂场景的需求。

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118939987A_ABST
    Figure CN118939987A_ABST
Patent Text Reader

Abstract

The invention relates to a method and a system for accelerating the construction of a feature engineering derivative index, and the system comprises an automatic feature derivation module which is used for distributing a corresponding operation rule through the recognition of the data type of received data, and carrying out the calculation of the data according to the operation rule, and obtaining feature data; the feature selection module is used for selecting a derivative index with the maximum improvement effect on the model performance from the feature data through an algorithm; and the model evaluation module is used for training a machine learning model according to the derivative indexes and evaluating the training of the machine learning model to obtain an evaluation result. The method has the advantages that the generation speed, the accuracy and the model performance of the derivative indexes are improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of computer technology applications, and in particular to a method and system for accelerating the construction of feature engineering derived indicators. Background Art

[0002] In the field of data science and machine learning, feature engineering is a crucial step in the model building process. Currently, feature engineering mainly relies on the following methods:

[0003] (1) Data scientists manually design and construct derived indicators based on their understanding of the data and business. This method is highly flexible but inefficient, especially when dealing with large-scale data sets.

[0004] (2) Although some derived indicators are automatically generated by writing scripts (such as Python, R, etc.), although this improves efficiency, it still requires a lot of manual adjustment and verification.

[0005] (3) Currently, some data processing tools (such as SQL, Excel, Pandas, etc.) provide partial automation support, but these tools have limited functions and cannot meet the needs of complex scenarios.

[0006] As can be seen from the above, the existing technology is inefficient in manually constructing derived indicators when performing feature engineering, which is very time-consuming, especially on large-scale data sets. The degree of automation is low, and the existing tools have single functions and are difficult to cope with complex data processing requirements. Human errors are easily introduced during the manual construction process, affecting model performance and stability. The lack of adaptability to different data types and business scenarios limits the scope of application.

[0007] At present, no effective solution has been proposed to address the problems of low efficiency, insufficient automation, proneness to errors and lack of adaptability in existing feature engineering technologies. Summary of the invention

[0008] The purpose of the present invention is to provide a method and system for accelerating the construction of feature engineering derived indicators in view of the deficiencies in the prior art, so as to solve the problems of low efficiency, insufficient automation, easy errors and lack of adaptability in the existing feature engineering technology.

[0009] To achieve the above object, the technical solution adopted by the present invention is:

[0010] In a first aspect, the present invention provides a system for accelerating the construction of feature engineering derived indicators, comprising: an automatic feature derivation module, a feature selection module and a model evaluation module, wherein the automatic feature derivation module is used to assign corresponding operation rules by identifying the data type of the received data, and calculate the data according to the operation rules to obtain feature data; the feature selection module is used to select the derivative indicators that have the greatest effect on improving model performance from the feature data through an algorithm; the model evaluation module is used to train a machine learning model based on the derived indicators, and evaluate the training of the machine learning model to obtain an evaluation result.

[0011] In some embodiments, the automatic feature derivation module includes: a data identification unit, which is used to identify the data type based on at least one received data to obtain the data type of each data; a rule allocation unit, which is used to allocate corresponding operation rules according to the identified data type; an operation unit, which is used to operate on the data of the data type according to the operation rules to obtain derived indicators; a feature combination unit, which is used to perform feature combination on the derived indicators to generate feature data.

[0012] In some of the embodiments, the data types include: numerical type, time type and categorical type.

[0013] In some of the embodiments, the feature selection module includes: a screening unit, used to calculate the correlation between feature data and the target variable, and screen feature data with high correlation; a calculation unit, used to calculate the predictive ability of feature data with high correlation for the target variable based on the information gain method, and evaluate the importance of the feature; a feature selection unit, used to screen from the feature data based on importance and correlation, obtain the optimal feature subset, and determine the optimal feature subset as the derivative indicator that has the greatest effect on improving model performance.

[0014] In some of the embodiments, optionally, the model evaluation module includes: an evaluation unit, used to train the machine learning model based on the derived indicators and evaluate the performance parameters of the machine learning model; and a verification unit, used to evaluate the stability and generalization ability of the model through a cross-validation method to obtain an evaluation result.

[0015] In some of the embodiments, the system also includes: a result display module and a user interface module, wherein the result display module is used to display the derived indicators and the performance impact of the derived indicators on the model; the user interface module is used to provide a data upload interface, a rule customization interface and a rule visualization interface.

[0016] In a second aspect, the present invention provides a method for accelerating the construction of feature engineering derived indicators, which is applied to a system for accelerating the construction of feature engineering derived indicators, including: assigning corresponding operation rules by identifying the data type of the received data, and calculating the data according to the operation rules to obtain feature data; selecting the derivative indicator that has the greatest effect on improving model performance from the feature data through an algorithm; training a machine learning model based on the derived indicators, and evaluating the training of the machine learning model to obtain an evaluation result.

[0017] In some of the embodiments, the data type of the received data is identified and assigned a corresponding operation rule, and the data is calculated according to the operation rule to obtain characteristic data, including: identifying the data type according to at least one received data to obtain the data type of each data; assigning the corresponding operation rule according to the identified data type; calculating the data of the data type according to the operation rule to obtain derived indicators; and combining the derived indicators to generate characteristic data.

[0018] In some of the embodiments, the data types include: numerical type, time type and categorical type.

[0019] In some of the embodiments, selecting a derivative indicator that has the greatest effect on improving model performance from feature data through an algorithm includes: calculating the correlation between feature data and a target variable, and screening feature data with high correlation; calculating the predictive ability of feature data with high correlation for the target variable based on an information gain method, and evaluating the importance of the feature; screening from feature data based on importance and correlation to obtain an optimal feature subset, and determining the optimal feature subset as the derivative indicator that has the greatest effect on improving model performance.

[0020] The present invention adopts the above technical scheme, through an automatic feature derivation module, which is used to assign corresponding operation rules by identifying the data type of the received data, and calculate the data according to the operation rules to obtain feature data; a feature selection module is used to select derivative indicators that have the greatest effect on improving model performance from the feature data through an algorithm; a model evaluation module is used to train the machine learning model according to the derived indicators, and evaluate the training of the machine learning model to obtain an evaluation result. Compared with the prior art, it has the following technical effects: the generation speed, accuracy and model performance of the derivative indicators are improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 is a schematic diagram of a system for accelerating the construction of feature engineering derived indicators according to an embodiment of the present invention;

[0022] Figure 2 is a schematic diagram of another system for accelerating the construction of feature engineering derived indicators according to an embodiment of the present invention;

[0023] Figure 3 It is a flowchart of a method for accelerating the construction of feature engineering derived indicators according to an embodiment of the present invention. DETAILED DESCRIPTION

[0024] In order to make the purpose, technical solutions and advantages of the present application clearer, the present application is described and illustrated 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 intended to limit the present application. Based on the embodiments provided in the present application, all other embodiments obtained by ordinary technicians in the field without making creative work are within the scope of protection of the present application.

[0025] Obviously, the drawings described below are only some examples or embodiments of the present application. For ordinary technicians in this field, the present application can also be applied to other similar scenarios based on these drawings without creative work. In addition, it can also be understood that although the efforts made in this development process may be complicated and lengthy, for ordinary technicians in this field related to the content disclosed in this application, some changes in design, manufacturing or production based on the technical content disclosed in this application are just conventional technical means, and should not be understood as insufficient content disclosed in this application.

[0026] Reference to "embodiments" in this application means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those of ordinary skill in the art that the embodiments described in this application may be combined with other embodiments without conflict.

[0027] Unless otherwise defined, the technical terms or scientific terms involved in this application should be understood by people with ordinary skills in the technical field to which this application belongs. The words "one", "a", "a", "the" and the like involved in this application do not indicate a quantitative limitation, and may represent the singular or plural. The terms "include", "comprise", "have" and any of their variations involved in this application are intended to cover non-exclusive inclusions; for example, a process, method, system, product or device that includes a series of steps or units (units) is not limited to the listed steps or units, but may also include steps or units that are not listed, or may also include other steps or units inherent to these processes, methods, products or devices. The words "connect", "connected", "coupled" and the like involved in this application are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. The "multiple" involved in this application refers to two or more. "And / or" describes the association relationship of associated objects, indicating that there can be three relationships, for example, "A and / or B" can mean: A exists alone, A and B exist at the same time, and B exists alone. The character " / " generally indicates that the objects before and after are in an "or" relationship. The terms "first", "second", "third", etc. involved in this application are only used to distinguish similar objects and do not represent a specific ordering of the objects.

[0028] Example 1

[0029] An exemplary embodiment of the present invention is as follows Figure 1 As shown, Figure 1 is a schematic diagram of a system for accelerating the construction of feature engineering derived indicators according to an embodiment of the present invention. The system for accelerating the construction of feature engineering derived indicators provided by an embodiment of the present application includes:

[0030] An automatic feature derivation module 12, a feature selection module 14 and a model evaluation module 16, wherein the automatic feature derivation module 12 is used to assign corresponding operation rules by identifying the data type of the received data, and calculate the data according to the operation rules to obtain feature data; the feature selection module 14 is used to select the derivative indicators that have the greatest effect on improving the model performance from the feature data through an algorithm; the model evaluation module 16 is used to train the machine learning model according to the derived indicators, and evaluate the training of the machine learning model to obtain an evaluation result.

[0031] In some embodiments, the automatic feature derivation module 12 includes: a data identification unit, which is used to identify the data type according to at least one received data to obtain the data type of each data; a rule allocation unit, which is used to allocate corresponding operation rules according to the identified data type; an operation unit, which is used to operate on the data of the data type according to the operation rules to obtain derived indicators; a feature combination unit, which is used to perform feature combination on the derived indicators to generate feature data.

[0032] In some of the embodiments, the data types include: numerical type, time type and categorical type.

[0033] In some of the embodiments, the feature selection module 14 includes: a screening unit, used to calculate the correlation between feature data and the target variable, and screen feature data with high correlation; a calculation unit, used to calculate the predictive ability of feature data with high correlation for the target variable based on the information gain method, and evaluate the importance of the feature; a feature selection unit, used to screen from the feature data based on importance and correlation, obtain the optimal feature subset, and determine the optimal feature subset as the derivative indicator that has the greatest effect on improving model performance.

[0034] In some of the embodiments, optionally, the model evaluation module 16 includes: an evaluation unit, used to train the machine learning model based on the derived indicators and evaluate the performance parameters of the machine learning model; and a verification unit, used to evaluate the stability and generalization ability of the model through a cross-validation method to obtain an evaluation result.

[0035] In some of the embodiments, a system for accelerating the construction of feature engineering derived indicators provided by the embodiments of the present application also includes: a result display module, wherein the result display module and a user interface module, wherein the result display module is used to display the derived indicators and the performance impact of the derived indicators on the model; and a user interface module is used to provide a data upload interface, a rule customization interface, and a rule visualization interface.

[0036] In summary, if Figure 2 As shown, Figure 2 is a schematic diagram of another system for accelerating the construction of feature engineering derived indicators according to an embodiment of the present invention, wherein:

[0037] Data upload: Upload the original dataset and select the option to automatically generate derived metrics.

[0038] Automatic feature derivation module 12: automatically generates a series of derived indicators through predefined mathematical operations, logical operations and statistical operations. The specific steps include:

[0039] Step 1: Automatically identify the data type of the input original variable (such as numerical type, categorical type, time type, etc.).

[0040] Step 2: Generate derived indicators by applying appropriate operation rules according to the data type of the original input variable. For example, generate derived indicators such as square, square root, logarithm, etc. from numerical features; generate derived indicators such as year, month, day, week, etc. from time features.

[0041] Step 3: Combine the original input variables in pairs to derive bivariate features, such as feature crossover, addition, subtraction, multiplication, and division.

[0042] Among them, data recognition: the system automatically recognizes the numerical and time features in the data set.

[0043] Operation rules: For numerical features, the system uses square, square root, logarithm and other operations to generate.

[0044] Derived indicators: For time-type features, the system generates derivative indicators such as year, month, day, and week.

[0045] Feature combination: The system automatically generates feature cross combinations, such as "Feature A*Feature B".

[0046] Feature selection module 14: Automatically select the derivative indicators that have the greatest effect on improving model performance through algorithms. The specific steps include:

[0047] Step 1: Calculate the correlation between the derived new indicator and the target variable, and select the derived variables with higher correlation.

[0048] Step 2: Use information gain methods (such as random forests) to evaluate the predictive ability of the derived new indicators on the target variable, that is, to evaluate the importance of the features.

[0049] Step 3: Combine the feature importance and feature relevance calculated in the above two steps to select the optimal feature subset and the features that contribute most to the model performance.

[0050] Among them, correlation analysis: calculate the correlation between each derived indicator and the target variable, and screen features with higher correlation.

[0051] Information gain calculation: Evaluate the information gain of each feature and further screen out high-value features.

[0052] Feature Importance Evaluation: Use the random forest model to evaluate the importance of features and select the features that contribute most to the model performance.

[0053] Model Evaluation Module 16:

[0054] Model training: Use the filtered features to train the machine learning model and evaluate the model’s performance indicators such as accuracy and F1 score.

[0055] Cross-validation: The stability and generalization ability of the model are evaluated through the cross-validation method to ensure the reliability of the model.

[0056] Results display: The final selected derivative indicators and their impact on model performance. Users can view the importance score of each feature and the model performance improvement.

[0057] Among them, the user interface module provides a friendly user interface, allowing users to easily operate and monitor the feature generation process.

[0058] Specific features include:

[0059] Dataset upload: Users upload the original dataset through the interface.

[0060] Rule customization: Allows users to customize feature generation rules and constraints, such as specifying the operation method of specific features.

[0061] Process visualization: Real-time display of feature generation and selection process to help users understand and adjust generation strategies.

[0062] Result display: The final selected derivative indicators and their impact on model performance are intuitively displayed to facilitate user decision-making.

[0063] (i.e., the functions of the result display module and the user interface module in the embodiment of the present application)

[0064] Specifically, the system for accelerating the construction of feature engineering derived indicators in the embodiment of the present application can be as follows in specific applications:

[0065] Automatic Feature Derivation Module 12:

[0066] 1) The user uploads the original data set, and the system automatically identifies the data type of each feature in the data set (such as numerical type, categorical type, time type, etc.) and classifies it.

[0067] #Identify data types

[0068] numeric_features=data.select_dtypes(include=[np.number]).columns

[0069] time_features=data.select_dtypes(include=['datetime']).columns

[0070] categorical_features=data.select_dtypes(include=['object','category']).columns

[0071] The select_dtypes method of the library is used to identify the numerical, temporal, and categorical features in the dataset and obtain the column names of these features.

[0072] 2) Perform feature derivative processing based on the data type of each feature.

[0073] (1) For numerical features, mathematical operations are applied to generate derivative indicators, such as square (x^2), square root (sqrt(x)), logarithm (log(x)), etc. Statistical features can also be generated, such as mean, variance, standard deviation, etc.

[0074] for feature in numeric_features:

[0075] data[f'{feature}_square']=data[feature]**2

[0076] data[f'{feature}_sqrt']=np.sqrt(data[feature])

[0077] data[f'{feature}_log']=np.log1p(data[feature])

[0078] For numerical features, the code generates derived indicators by cyclically applying mathematical operations such as square, square root, and logarithm.

[0079] (2) For time-related features, extract time-related features, such as year, month, day, week, quarter, etc. Generate time difference features, such as the difference between two time features.

[0080] for feature in time_features:

[0081] data[feature]=pd.to_datetime(data[feature])

[0082] data[f'{feature}_year']=data[feature].dt.year

[0083] data[f'{feature}_month']=data[feature].dt.month

[0084] data[f'{feature}_day']=data[feature].dt.day

[0085] data[f'{feature}_weekday']=data[feature].dt.weekday

[0086] For time-type features, the code first converts the features into datetime type, and then extracts time-related derivative indicators such as year, month, day, and week.

[0087] (3) For categorical features, frequency encoding, target encoding and other methods are used to generate derived indicators. Generate One-Hot Encoding features.

[0088] data=pd.get_dummies(data,columns=categorical_features)

[0089] Among them, the get_dummies method is used to apply one-hot encoding to the categorical features to generate One-Hot encoding features.

[0090] 3) Feature combination generation

[0091] (1) Feature intersection: Automatically generate cross combinations of features, such as “Feature A*Feature B”, “Feature C+Feature D”, etc.

[0092] (2) Complex operations: Generate more complex feature combinations based on user-defined rules.

[0093] #Feature combination

[0094] from itertools import combinations

[0095] for feature1,feature2 in combinations(numeric_features,2):

[0096] data[f'{feature1}_{feature2}_product']=data[feature1]*data[feature2]

[0097] data[f'{feature1}_{feature2}_sum']=data[feature1]+data[feature2]

[0098] The code uses the combinations method of itertools to generate cross combinations of features, such as feature products and feature sums.

[0099] Feature selection module 14:

[0100] 1) Correlation analysis: The system calculates the correlation between each derived indicator and the target variable (such as Pearson correlation coefficient, Spearman rank correlation coefficient, etc.).

[0101] #Assume the target variable is 'target'

[0102] correlation=data.corr()['target'].sort_values(ascending=False)

[0103] Among them, the correlation between the derived indicators and the target variable is calculated, and the features are sorted by the size of the correlation coefficient. 2) Calculate information gain: The system evaluates the information gain of each feature and calculates its contribution to the target variable.

[0104] from sklearn.ensemble import RandomForestRegressor

[0105] #Separate features and target variables

[0106] X=data.drop(columns=['target'])

[0107] y = data['target']

[0108] #Training random forest model

[0109] model=RandomForestRegressor(n_estimators=100,random_state=42)

[0110] model.fit(X,y)

[0111] #Get feature importance

[0112] feature_importances=pd.Series(model.feature_importances_,index=X.columns)

[0113] Among them, the random forest model is used to evaluate the information gain of the feature, that is, the contribution of the feature to the prediction ability of the model.

[0114] 3) Select important features: Based on the feature importance score, select the features that contribute most to the model performance.

[0115] threshold=0.01

[0116] important_features=feature_importances[feature_importances>threshold].index

[0117] #Combining the results of correlation analysis

[0118] correlated_features = correlation[correlation.abs()>0.1].index #Assume the threshold is 0.1

[0119] selected_features=important_features.intersection(correlated_features)

[0120] Among them, the final feature subset is selected based on feature importance and correlation with the target variable.

[0121] Model Evaluation Module 16:

[0122] The model performance is evaluated by calculating the accuracy and F1 score of the model.

[0123] def evaluate_model(X,y):

[0124] X_train,X_test,y_train,y_test=train_test_split(X,y,test_size=0.2,random_state=42)

[0125] model=RandomForestRegressor(n_estimators=100,random_state=42)

[0126] model.fit(X_train,y_train)

[0127] y_pred=model.predict(X_test)

[0128] #Model performance evaluation

[0129] accuracy=accuracy_score(y_test,y_pred.round())

[0130] f1=f1_score(y_test,y_pred.round(),average='macro')

[0131] print(f'Accuracy:{accuracy}')

[0132] print(f'F1 Score:{f1}')

[0133] Among them, a function evaluate_model is defined to evaluate the accuracy and F1 score of the model, which can provide performance indicators of the model on the test set.

[0134] These code snippets show the entire process of automated feature engineering, from data preprocessing to feature generation, feature selection, and model evaluation. In this way, the efficiency of feature engineering and the performance of the model can be significantly improved.

[0135] The system provided by the embodiment of the present application accelerates the construction of derived indicators of feature engineering and automates the generation and selection of features, which greatly reduces the manual operation time. Derived indicators are automatically generated by predefined operation rules and user-defined rules, which greatly improves the efficiency of feature generation. The full process automation support from feature generation, selection to evaluation reduces the steps of manual intervention and improves the automation level of feature engineering. Automated and intelligent processes reduce the error rate of manual operations and ensure that the generated derived indicators are more accurate and reliable. Through systematic feature selection and evaluation methods, the errors caused by human factors are reduced. Through flexible feature generation and selection mechanisms, it adapts to different data types and business scenarios, and improves the universality and practicality of feature engineering. Users can customize feature generation rules according to specific needs to meet diverse business needs. Through automated feature generation and selection, the present invention can generate high-quality derived indicators and significantly improve the performance of the model. The model evaluation module can automatically select the optimal feature combination to ensure that the performance of the model on new data is stable and excellent.

[0136] The present invention adopts the above technical scheme, through an automatic feature derivation module, which is used to assign corresponding operation rules by identifying the data type of the received data, and calculate the data according to the operation rules to obtain feature data; a feature selection module is used to select derivative indicators that have the greatest effect on improving model performance from the feature data through an algorithm; a model evaluation module is used to train the machine learning model according to the derived indicators, and evaluate the training of the machine learning model to obtain an evaluation result. Compared with the prior art, it has the following technical effects: the generation speed, accuracy and model performance of the derivative indicators are improved.

[0137] Example 2

[0138] An exemplary embodiment of the present invention is as follows Figure 3 As shown, Figure 31 is a flow chart of a method for accelerating the construction of a feature engineering derived indicator according to an embodiment of the present invention, which is applied to a system for accelerating the construction of a feature engineering derived indicator. The method for accelerating the construction of a feature engineering derived indicator provided by an embodiment of the present application includes:

[0139] Step S302, assigning corresponding operation rules by identifying the data type of the received data, and calculating the data according to the operation rules to obtain characteristic data;

[0140] In some of the embodiments, the data type of the received data is identified and assigned a corresponding operation rule, and the data is calculated according to the operation rule to obtain characteristic data, including: identifying the data type according to at least one received data to obtain the data type of each data; assigning the corresponding operation rule according to the identified data type; calculating the data of the data type according to the operation rule to obtain derived indicators; and combining the derived indicators to generate characteristic data.

[0141] In some of the embodiments, the data types include: numerical type, time type and categorical type.

[0142] Step S304, selecting the derivative indicator that has the greatest effect on improving model performance from the feature data through an algorithm;

[0143] In some of the embodiments, selecting a derivative indicator that has the greatest effect on improving model performance from feature data through an algorithm includes: calculating the correlation between feature data and a target variable, and screening feature data with high correlation; calculating the predictive ability of feature data with high correlation for the target variable based on an information gain method, and evaluating the importance of the feature; screening from feature data based on importance and correlation to obtain an optimal feature subset, and determining the optimal feature subset as the derivative indicator that has the greatest effect on improving model performance.

[0144] Step S306: train the machine learning model based on the derived indicators, and evaluate the training of the machine learning model to obtain an evaluation result.

[0145] In some of the embodiments, optionally, the machine learning model is trained based on the derived indicators to evaluate the performance parameters of the machine learning model; the verification unit is used to evaluate the stability and generalization ability of the model through a cross-validation method to obtain an evaluation result.

[0146] In some of the embodiments, a method for accelerating the construction of feature engineering derived indicators provided in the embodiments of the present application can also display the derived indicators and the performance impact of the derived indicators on the model; provide a data upload interface, a rule customization interface and a rule visualization interface.

[0147] In combination with step S302 to step S306, a method for accelerating the construction of feature engineering derived indicators provided in an embodiment of the present application is applied to Figure 1 and Figure 2 The system shown here accelerates the construction of feature engineering derived indicators.

[0148] The present invention adopts the above technical scheme, assigns corresponding operation rules by identifying the data type of the received data, and calculates the data according to the operation rules to obtain feature data; selects the derivative indicators that have the greatest effect on improving the model performance from the feature data through the algorithm; trains the machine learning model according to the derivative indicators, and evaluates the training of the machine learning model to obtain evaluation results. Compared with the existing technology, it has the following technical effects: improving the generation speed, accuracy and model performance of derivative indicators.

[0149] The above description is only a preferred embodiment of the present invention, and does not limit the implementation mode and protection scope of the present invention. For those skilled in the art, it should be aware that all solutions obtained by equivalent substitutions and obvious changes made using the description and illustrations of the present invention should be included in the protection scope of the present invention.

Claims

1. A system to accelerate the construction of feature engineering derived indicators, It is characterized in that include: Automatic feature derivation module, feature selection module and model evaluation module, among which, The automatic feature derivation module is used to assign corresponding operation rules by identifying the data type of the received data, and calculate the data according to the operation rules to obtain feature data; The feature selection module is used to select the derivative indicator that has the greatest effect on improving model performance from the feature data through an algorithm; The model evaluation module is used to train the machine learning model according to the derived indicators, and evaluate the training of the machine learning model to obtain an evaluation result.

2. The system for accelerating the construction of feature engineering derived indicators according to claim 1, It is characterized in that The automatic feature derivation module comprises: A data identification unit, used to identify a data type according to at least one received data to obtain a data type of each data; A rule allocation unit, used to allocate corresponding operation rules according to the identified data type; A calculation unit, used for calculating the data of the data type according to the calculation rule to obtain a derived index; The feature combination unit is used to perform feature combination on the derived indicators to generate feature data.

3. The system for accelerating the construction of feature engineering derived indicators according to claim 2, It is characterized in that The data types include: numerical type, time type and subtype.

4. The system for accelerating the construction of feature engineering derived indicators according to claim 1 or 2, It is characterized in that The feature selection module comprises: A screening unit, used to calculate the correlation between the feature data and the target variable, and screen the feature data with high correlation; A calculation unit, used to calculate the predictive ability of highly correlated feature data on the target variable according to the information gain method, and obtain the importance of the evaluation feature; A feature selection unit is used to screen the feature data according to the importance and the relevance to obtain an optimal feature subset, and determine the optimal feature subset as a derivative indicator that has the greatest effect on improving the model performance.

5. The system for accelerating the construction of feature engineering derived indicators according to claim 4, It is characterized in that The model evaluation module includes: An evaluation unit, configured to train a machine learning model based on the derived indicators and evaluate performance parameters of the machine learning model; The validation unit is used to evaluate the stability and generalization ability of the model through a cross-validation method to obtain the evaluation result.

6. The system for accelerating the construction of feature engineering derived indicators according to claim 1, It is characterized in that The system also includes: a result display module and a user interface module, wherein: The result display module is used to display the derived indicators and the impact of the derived indicators on the performance of the model; The user interface module is used to provide a data upload interface, a rule customization interface and a rule visualization interface.

7. A method to speed up the construction of feature engineering derived indicators, It is characterized in that Systems used to accelerate the construction of feature engineering derived metrics, including: By identifying the data type of the received data, assigning a corresponding operation rule, and calculating the data according to the operation rule, the characteristic data is obtained; Selecting the derivative indicator that has the greatest effect on improving model performance from the feature data through an algorithm; The machine learning model is trained according to the derived indicators, and the training of the machine learning model is evaluated to obtain an evaluation result.

8. The method for accelerating the construction of feature engineering derived indicators according to claim 7, It is characterized in that The step of assigning a corresponding operation rule to the data type of the received data and calculating the data according to the operation rule to obtain the characteristic data includes: Identify the data type according to at least one received data to obtain the data type of each data; Assign corresponding operation rules according to the identified data type; Performing operations on the data of the data type according to the operation rules to obtain a derived indicator; The derived indicators are subjected to feature combination to generate feature data.

9. The method for accelerating the construction of feature engineering derived indicators according to claim 8, It is characterized in that The data types include: numerical type, time type and subtype.

10. The method for accelerating the construction of feature engineering derived indicators according to claim 7 or 8, It is characterized in that The algorithm selects the derivative index that has the greatest effect on improving the model performance from the feature data, including: Calculate the correlation between the feature data and the target variable, and select feature data with high correlation; The information gain method is used to calculate the predictive power of highly relevant feature data on the target variable, and the importance of the evaluation feature is obtained; The feature data is screened according to the importance and the relevance to obtain an optimal feature subset, and the optimal feature subset is determined as a derivative indicator that has the greatest effect on improving the performance of the model.

Citation Information

Patent Citations

  • Method and device for determining characteristic engineering variable

    CN108875815A

  • Enterprise credit assessment automatic rule mining method and device based on improved decision tree

    CN115860918A

  • Credit risk management field feature cross derivation method

    CN116861207A

  • Risk control characteristic automatic screening method for multi-channel fund integrated credit business

    CN116862643A