Enterprise expert talent prediction method and device based on big data

Abstract

The invention discloses an enterprise expert talent prediction method and device based on big data, and the method comprises the steps: obtaining enterprise employee data, obtaining a target data set, carrying out the prediction of the target data set according to an enterprise expert talent target model, and obtaining the enterprise expert talent prediction based on big data; the enterprise expert talent target model is pre-constructed, and the method comprises the following steps: obtaining a sample data set; dividing the sample data set into a training sample set and a test sample set in proportion, and screening out a reliable negative sample set from the unmarked sample sets; performing iterative training on the positive sample set and the reliable negative sample set; and performing model evaluation and optimization on the enterprise expert talent training model according to the test sample set. According to the method, enterprise expert talents can be predicted based on big data, so that an enterprise human resource department is helped to quickly lock a talent resource pool, and a reference basis is provided for enterprise talent selection.

Description

Technical Field

[0001] The present invention relates to a method and device for predicting enterprise expert talents based on big data. Background Art

[0002] As the company accumulates more and more personnel information, more and more data becomes available for reference in internal talent selection. The existing talent screening rules within the company can no longer meet the diverse needs of existing talents.

[0003] The existing internal talent selection in enterprises are all based on selection rules set by the human resources department according to past experience. The company conducts internal talent selection through established rules and on-site employee defense. This selection system has a high human cost for selecting talents and a small talent resource pool. It is difficult to discover expert talents other than those specified in the above selection rules, and it is easy to miss corporate talents and the selection cost is high. Summary of the invention

[0004] The technical problem to be solved by the present invention is to propose a method and device for predicting enterprise expert talents based on big data in response to the above-mentioned deficiencies in the prior art. The method can predict enterprise expert talents based on big data, thereby helping the enterprise human resources department to quickly lock in the talent resource pool and provide a reference basis for enterprise talent selection.

[0005] In a first aspect, the present invention provides a method for enterprise expert talent prediction based on big data, the method comprising the steps of:

[0006] Step S1: Acquire enterprise employee data to obtain a target data set;

[0007] Step S2: predicting the target data set according to the enterprise expert talent target model to achieve enterprise expert talent prediction based on big data.

[0008] Furthermore, before step S1, the method further includes:

[0009] Step S0: Constructing an enterprise expert talent target model, which includes:

[0010] Step S01: Obtain a sample data set;

[0011] Step S02: Divide the sample data set into a training sample set and a test sample set in proportion, wherein the training sample set includes a positive sample set and an unlabeled sample set;

[0012] Step S03: Filter out a reliable negative sample set from the unlabeled sample set according to the features contained in the positive sample set;

[0013] Wherein: the positive samples are enterprise experts and talents, the reliable negative samples are ordinary employees, the positive sample set is a set including enterprise experts and talents, the unlabeled sample set is a set not including enterprise experts and talents, and the reliable negative sample set is a set including ordinary employees;

[0014] Step S04: iteratively train the positive sample set and the reliable negative sample set to obtain an enterprise expert talent training model;

[0015] Step S05: Perform model evaluation and optimization on the enterprise expert talent training model according to the test sample set to obtain the enterprise expert talent target model.

[0016] Furthermore, in step S02, dividing the sample data set into the training sample set and the test sample set in proportion is dividing the sample data set into the training sample set and the test sample set in a ratio of 7:3, which specifically includes the following steps:

[0017] Set the size of the sample data set to N, where N is a natural number greater than 1;

[0018] Randomly shuffle the order of the sample data set;

[0019] The first 0.7N samples in the randomly shuffled sample data set are used as the training sample set, and the last 0.3N samples in the randomly shuffled sample data set are used as the test sample set.

[0020] Furthermore, in step S03, based on the features included in the positive sample set, a reliable negative sample set is screened out from the unlabeled sample set, specifically including:

[0021] According to the features contained in the positive sample set, the unlabeled samples in the unlabeled sample set are classified based on the Bayesian classifier, so as to determine the positive sample probability corresponding to the unlabeled samples in the unlabeled sample set, and the unlabeled samples with positive sample probability lower than the threshold are determined as reliable negative samples, and all reliable negative samples form a reliable negative sample set;

[0022] Among them: the features contained in the positive sample set specifically include enterprise expert talent achievement information, enterprise expert talent award information, enterprise expert talent training information, enterprise expert talent performance appraisal information, and enterprise expert talent qualification certification information; the positive sample probability is the probability of becoming an enterprise expert talent.

[0023] Furthermore, in step S04, iterative training is performed on the positive sample set and the reliable negative sample set to obtain an enterprise expert talent training model, which includes the following specific steps:

[0024] Step S41: extracting all sample features, the sample features including positive sample features and reliable negative sample features; the positive sample features are sample features of expert talents, and the reliable negative sample features are sample features of ordinary employees;

[0025] Step S42: Mark each positive sample as 1 and each reliable negative sample as -1 to construct a training data set;

[0026] Step S43: iteratively train the preset training model using the training data set to obtain the enterprise expert talent training model;

[0027] The preset training models include SVM model, Random Forest model, K-nearest neighbor model, Logistic Regression model, and A type of Bayesian model.

[0028] Furthermore, in step S05, the enterprise expert talent training model is evaluated and optimized according to the test sample set to obtain the enterprise expert talent target model, which includes the following specific steps:

[0029] S51: Performing model evaluation on the enterprise expert talent training model according to the test sample set to obtain an evaluation result of the enterprise expert talent training model;

[0030] S52: According to the evaluation results of the enterprise expert talent training model, the parameters of the enterprise expert talent training model are processed in a serial pipeline manner through the Pipeline script to find the best combination of multiple parameter indicators of the enterprise expert talent training model, and the GridSearch function is used to search for the best combination of multiple parameter indicators of the enterprise expert talent training model, so as to automatically screen the optimal parameters and obtain the enterprise expert talent target model.

[0031] Furthermore, in step S1, obtaining enterprise employee data to obtain a target data set includes the following steps:

[0032] Collecting enterprise employee data, including employee performance data, employee training data, employee achievement data, employee award data, employee certification data, and employee professional skills data;

[0033] Conduct data traversal and data processing on the collected enterprise employee data;

[0034] in,

[0035] The data processing includes:

[0036] Fill or delete the company's employee data;

[0037] or,

[0038] One-hot encoding is used to convert character variables in enterprise employee data into numerical variables;

[0039] in,

[0040] The filling of the enterprise employee data is to fill the enterprise employee data with the mean, median or mode;

[0041] The deleting of the enterprise employee data includes: deleting fields in the enterprise employee data with missing values ​​exceeding 50%, or screening and deleting variables in the enterprise employee data with a correlation coefficient between variables exceeding 0.8.

[0042] In a second aspect, the present invention provides a big data-based enterprise expert talent prediction device, the device comprising:

[0043] An acquisition unit, used to acquire enterprise employee data to obtain a target data set;

[0044] The prediction unit is connected to the acquisition unit and is used to predict the target data set according to the enterprise expert talent target model to achieve enterprise expert talent prediction based on big data.

[0045] Furthermore, the device further comprises a construction unit, which is connected to the prediction unit and is used to construct an enterprise expert talent target model;

[0046] The building block comprises:

[0047] An acquisition module, used to acquire a sample data set;

[0048] A proportion module, connected to the acquisition module, for dividing the sample data set into a training sample set and a test sample set in proportion, wherein the training sample set includes a positive sample set and an unlabeled sample set;

[0049] A screening module, connected to the ratio module, for screening out a reliable negative sample set from the unlabeled sample set according to the features contained in the positive sample set;

[0050] Wherein: the positive samples are enterprise experts and talents, the reliable negative samples are ordinary employees, the positive sample set is a set including enterprise experts and talents, the unlabeled sample set is a set not including enterprise experts and talents, and the reliable negative sample set is a set including ordinary employees;

[0051] An iteration module, connected to the screening module, for iteratively training the positive sample set and the reliable negative sample set to obtain an enterprise expert talent training model;

[0052] The optimization module is connected to the acquisition module and the iteration module respectively, and is used to evaluate and optimize the enterprise expert talent training model according to the test sample set to obtain the enterprise expert talent target model.

[0053] Furthermore, the iteration module includes:

[0054] The extraction submodule extracts all sample features, including positive sample features and reliable negative sample features; the positive sample features are sample features of expert talents, and the reliable negative sample features are sample features of ordinary employees;

[0055] A marking submodule, connected to the extraction submodule, is used to mark each positive sample as 1 and each reliable negative sample as -1, so as to construct a training data set;

[0056] An iterative submodule, connected to the marking submodule, for iteratively training a preset training model using a training data set to obtain a first model of enterprise expert talents;

[0057] The preset training models include SVM model, Random Forest model, K-nearest neighbor model, Logistic Regression model, and A type of Bayesian model.

[0058] Beneficial effects of the present invention:

[0059] 1. The present invention can predict enterprise expert talents based on big data, thereby helping the enterprise human resources department to quickly lock in the talent resource pool and provide a reference for enterprise talent selection;

[0060] 2. The characteristics of enterprise experts learned by the model can more accurately form an efficient team, give full play to the advantages of enterprise patent talents, and thus improve overall performance;

[0061] 3. The present invention can accurately screen out expert talents, thereby helping enterprises understand the flow trend of expert talents and make talent reserve and introduction plans in advance;

[0062] 4. Enterprises can provide personalized training and development opportunities for employees based on the prediction results of the present invention, helping them grow into expert talents and improving the overall capabilities of the organization. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] Figure 1 A schematic diagram of the enterprise expert talent prediction process in an embodiment of the present invention;

[0064] Figure 2A schematic diagram of a method for predicting enterprise expert talents in an embodiment of the present invention;

[0065] Figure 3 Schematic diagram of a device for predicting enterprise expert talents in an embodiment of the present invention;

[0066] Figure 4 FIG. 4 is another schematic diagram of the enterprise expert talent prediction device according to an embodiment of the present invention.

[0067] Among them, the figure numerals are: 10, acquisition unit, 20, prediction unit, 30, construction unit. DETAILED DESCRIPTION

[0068] In order to enable those skilled in the art to better understand the technical solution of the present invention, the embodiments of the present invention will be further described in detail below with reference to the accompanying drawings.

[0069] It should be understood that the specific embodiments and drawings described herein are only used to explain the present invention rather than to limit the present invention.

[0070] It can be understood that, in the absence of conflict, the various embodiments of the present invention and the various features in the embodiments can be combined with each other.

[0071] It can be understood that, for the convenience of description, the drawings of the present invention only show the parts related to the present invention, while the parts irrelevant to the present invention are not shown in the drawings.

[0072] It can be understood that each unit and module involved in the embodiments of the present invention may correspond to only one physical structure, or may be composed of multiple physical structures, or multiple units and modules may be integrated into one physical structure.

[0073] It can be understood that, without conflict, the functions and steps marked in the flowcharts and block diagrams of the present invention may occur in an order different from that marked in the drawings.

[0074] It is understood that the flowcharts and block diagrams of the present invention illustrate the possible architectures, functions, and operations of the systems, devices, equipment, and methods according to the various embodiments of the present invention. Each box in the flowchart or block diagram may represent a unit, module, program segment, or code, which contains executable instructions for implementing the specified functions. Moreover, each box or combination of boxes in the block diagram and flowchart may be implemented by a hardware-based system that implements the specified functions, or may be implemented by a combination of hardware and computer instructions.

[0075] It can be understood that the units and modules involved in the embodiments of the present invention can be implemented by software or hardware. For example, the units and modules can be located in a processor.

[0076] Embodiment 1:

[0077] like Figure 1 and Figure 2 As shown, this embodiment provides a method for enterprise expert talent prediction based on big data, which includes steps S1 to S2:

[0078] Step S1: Acquire enterprise employee data to obtain a target data set;

[0079] Specifically, obtaining the enterprise employee data to obtain the target data set includes the following steps:

[0080] Collect enterprise employee data, including employee performance data, employee training data, employee achievement data, employee award data, employee certification data, and employee professional skills data;

[0081] Conduct data traversal and data processing on the collected enterprise employee data;

[0082] in,

[0083] Data processing includes:

[0084] Fill or delete the company's employee data;

[0085] or,

[0086] One-hot encoding is used to convert character variables in enterprise employee data into numerical variables;

[0087] in,

[0088] Filling the data of enterprise employees means filling the data of enterprise employees with mean, median or mode;

[0089] Deleting the enterprise employee data includes: deleting fields with more than 50% missing values ​​in the enterprise employee data, or screening and deleting variables with a correlation coefficient between variables exceeding 0.8 in the enterprise employee data.

[0090] Step S2: predicting the target data set according to the enterprise expert talent target model to achieve enterprise expert talent prediction based on big data.

[0091] Specifically, before step S1, step S0 of constructing an enterprise expert talent target model is also included. Step S0 specifically includes steps S01 to S05:

[0092] S01: Obtain a sample data set.

[0093] S02: Divide the sample data set into a training sample set and a test sample set in proportion, where the training sample set includes a positive sample set and an unlabeled sample set.

[0094] Specifically, in step S02, the sample data set is divided into a training sample set and a test sample set in proportion, that is, the sample data set is divided into a training sample set and a test sample set in a ratio of 7:3, which specifically includes the following steps:

[0095] Set the size of the sample data set to N, where N is a natural number greater than 1;

[0096] Randomly shuffle the order of the sample data set;

[0097] The first 0.7N samples in the randomly shuffled sample data set are used as the training sample set, and the last 0.3N samples in the randomly shuffled sample data set are used as the test sample set.

[0098] S03: According to the features contained in the positive sample set, a reliable negative sample set is screened out from the unlabeled sample set.

[0099] Among them: positive samples are enterprise experts and talents, reliable negative samples are ordinary employees, the positive sample set is a set including enterprise experts and talents, the unlabeled sample set is a set not including enterprise experts and talents, and the reliable negative sample set is a set including ordinary employees;

[0100] Specifically, in step S03, based on the features included in the positive sample set, a reliable negative sample set is screened out from the unlabeled sample set, which specifically includes:

[0101] According to the features contained in the positive sample set, the unlabeled samples in the unlabeled sample set are classified based on the Bayesian classifier, so as to determine the positive sample probability corresponding to the unlabeled samples in the unlabeled sample set, and the unlabeled samples with positive sample probability lower than the threshold are determined as reliable negative samples, and all reliable negative samples form a reliable negative sample set;

[0102] Among them: the features contained in the positive sample set include enterprise expert talent achievement information, enterprise expert talent award information, enterprise expert talent training information, enterprise expert talent performance appraisal information, and enterprise expert talent qualification certification information; the positive sample probability is the probability of becoming an enterprise expert talent.

[0103] S04: Iteratively train the positive sample set and the reliable negative sample set to obtain the enterprise expert talent training model.

[0104] Specifically, in step S04, iterative training is performed on the positive sample set and the reliable negative sample set to obtain the enterprise expert talent training model, including the following specific steps:

[0105] Step S41: extracting all sample features, including positive sample features and reliable negative sample features; positive sample features are sample features of expert talents, and reliable negative sample features are sample features of ordinary employees;

[0106] Step S42: Mark each positive sample as 1 and each reliable negative sample as -1 to construct a training data set;

[0107] Step S43: iteratively train the preset training model using the training data set to obtain the enterprise expert talent training model;

[0108] The preset training models include SVM model, Random Forest model, K-nearest neighbor model, Logistic Regression model, and A type of Bayesian model.

[0109] The training models include SVM model, Random Forest model, K-nearest neighbor model, Logistic Regression model, and One of the models is the Bayesian model.

[0110] Support Vector Machine (SVM) is a commonly used supervised learning algorithm, which is widely used in classification and regression problems; Random Forest model is an integrated learning algorithm, which builds multiple decision trees and makes predictions by voting or averaging; K-nearest neighbor model (KN model) is an instance-based learning algorithm, which stores the feature vectors of all samples during the training phase, and then measures the distance between the feature vectors of new samples and the samples in the training set, and selects the K samples with the closest distance for prediction; Logistic Regression model is a commonly used linear classification model, which is suitable for binary classification problems and can predict the probability that a sample belongs to a certain category; The Bayesian model (Naive Bayesian model) is a probability model based on Bayes' theorem.

[0111] 1. When using scikit-learn (sklearn) for classification tasks and training with a support vector machine (SVM) model, the steps are as follows:

[0112] 1.1 Import the required libraries and modules:

[0113] Python

[0114] from sklearn import svm

[0115] from sklearn.model_selection import train_test_split

[0116] from sklearn.metrics import accuracy_score

[0117] 1.2 Prepare the dataset:

[0118] Prepare a labeled training dataset that includes input features and corresponding class labels. Typically, the input features are represented as a two-dimensional array (or matrix), while the labels are represented as a one-dimensional array.

[0119] 1.3 Data preprocessing:

[0120] If there are missing values, outliers, or feature scaling is required in the dataset, you can use the data preprocessing tools in sklearn to perform the corresponding processing.

[0121] 1.4 Split the dataset:

[0122] Divide the data set into training set and test set. You can use the train_test_split function to divide the data set into two parts according to a certain ratio.

[0123] 1.5 Create SVM model object:

[0124] Python

[0125] model = svm.SVC()

[0126] The parameters in the SVC() function, such as the kernel function (kernel), regularization parameter (C), etc., can be adjusted as needed.

[0127] 1.6 Model training:

[0128] Python

[0129] model.fit(X_train,y_train)

[0130] Among them, X_train is the input feature of the training set, and y_train is the category label of the training set.

[0131] 1.7 Model prediction:

[0132] Python

[0133] y_pred=model.predict(X_test)

[0134] Among them, X_test is the input feature of the test set.

[0135] 1.8 Model Evaluation:

[0136] Python

[0137] accuracy=accuracy_score(y_test,y_pred)

[0138] Using appropriate indicators to evaluate the model can also further optimize the model parameters based on actual conditions.

[0139] 2. When using scikit-learn (sklearn) for classification tasks and using the RandomForest model for training, the steps are as follows:

[0140] 2.1 Import the required libraries and modules:

[0141] Python

[0142] from sklearn.ensemble import RandomForestClassifier

[0143] from sklearn.model_selection import train_test_split

[0144] from sklearn.metrics import accuracy_score

[0145] 2.2 Prepare the dataset:

[0146] Prepare a labeled training dataset that includes input features and corresponding class labels. Typically, the input features are represented as a two-dimensional array (or matrix), while the labels are represented as a one-dimensional array.

[0147] 2.3 Data preprocessing:

[0148] If there are missing values, outliers, or feature scaling is required in the dataset, you can use the data preprocessing tools in sklearn to perform the corresponding processing.

[0149] 2.4 Split the dataset:

[0150] Divide the data set into training set and test set. You can use the train_test_split function to divide the data set into two parts according to a certain ratio.

[0151] 2.5 Create a random forest model object:

[0152] Python

[0153] model=RandomForestClassifier()

[0154] You can adjust the parameters in the RandomForestClassifier() function as needed, such as the number of decision trees (n_estimators), the maximum number of features (max_features), etc.

[0155] 2.6 Model training:

[0156] Python

[0157] model.fit(X_train,y_train)

[0158] Among them, X_train is the input feature of the training set, and y_train is the category label of the training set.

[0159] 2.7 Model prediction:

[0160] Python

[0161] y_pred=model.predict(X_test)

[0162] Among them, X_test is the input feature of the test set.

[0163] 2.8 Model Evaluation:

[0164] Python

[0165] accuracy=accuracy_score(y_test,y_pred)

[0166] Using appropriate indicators to evaluate the model can also further optimize the model parameters based on actual conditions.

[0167] 3. When using scikit-learn (sklearn) for classification tasks, the steps for training with the K-nearest neighbors (KNN) model are as follows:

[0168] 3.1 Import the required libraries and modules:

[0169] Python

[0170] from sklearn.neighbors import KNeighborsClassifier

[0171] from sklearn.model_selection import train_test_split

[0172] from sklearn.metrics import accuracy_score

[0173] 3.2 Prepare the dataset:

[0174] Prepare a labeled training dataset that includes input features and corresponding class labels. Typically, the input features are represented as a two-dimensional array (or matrix), while the labels are represented as a one-dimensional array.

[0175] 3.3 Data preprocessing:

[0176] If there are missing values, outliers, or feature scaling is required in the dataset, you can use the data preprocessing tools in sklearn to perform the corresponding processing.

[0177] 3.4 Split the dataset:

[0178] Divide the data set into training set and test set. You can use the train_test_split function to divide the data set into two parts according to a certain ratio.

[0179] 3.5 Create KNN model object:

[0180] Python

[0181] model=KNeighborsClassifier(n_neighbors=k)

[0182] Among them, k is a hyperparameter in KNN, which represents the number of selected nearest neighbors.

[0183] 3.6 Model training:

[0184] Python

[0185] model.fit(X_train,y_train)

[0186] Among them, X_train is the input feature of the training set, and y_train is the category label of the training set.

[0187] 3.7 Model prediction:

[0188] Python

[0189] y_pred=model.predict(X_test)

[0190] Among them, X_test is the input feature of the test set.

[0191] 3.8 Model Evaluation:

[0192] Python

[0193] accuracy=accuracy_score(y_test,y_pred)

[0194] By using appropriate indicators to evaluate the model, the model parameters can be further optimized according to the actual situation, such as adjusting the K value, using different distance measurement methods, etc.

[0195] 4. When using scikit-learn (sklearn) for classification tasks, the steps for training with the LogisticRegression model are as follows:

[0196] 4.1 Import the required libraries and modules:

[0197] Python

[0198] from sklearn.linear_model import LogisticRegression

[0199] from sklearn.model_selection import train_test_split

[0200] from sklearn.metrics import accuracy_score

[0201] 4.2 Prepare the dataset:

[0202] Prepare a labeled training dataset that includes input features and corresponding class labels. Typically, the input features are represented as a two-dimensional array (or matrix), while the labels are represented as a one-dimensional array.

[0203] 4.3 Data Preprocessing:

[0204] If there are missing values, outliers, or feature scaling is required in the dataset, you can use the data preprocessing tools in sklearn to perform the corresponding processing.

[0205] 4.4 Split the dataset:

[0206] Divide the data set into training set and test set. You can use the train_test_split function to divide the data set into two parts according to a certain ratio.

[0207] 4.5 Create a logistic regression model object:

[0208] Python

[0209] model = LogisticRegression()

[0210] 4.6 Model training:

[0211] Python

[0212] model.fit(X_train,y_train)

[0213] Among them, X_train is the input feature of the training set, and y_train is the category label of the training set.

[0214] 4.7 Model prediction:

[0215] Python

[0216] y_pred=model.predict(X_test)

[0217] Among them, X_test is the input feature of the test set.

[0218] 4.8 Model Evaluation:

[0219] Python

[0220] accuracy=accuracy_score(y_test,y_pred)

[0221] Use appropriate indicators to evaluate the model and further optimize the model parameters according to the actual situation, such as adjusting regularization parameters and dealing with sample imbalance problems.

[0222] 5. When using scikit-learn (sklearn) for classification tasks, use Naive Bayes ( When training a Bayesian model, the steps are as follows:

[0223] 5.1 Import the required libraries and modules:

[0224] Python

[0225] from sklearn.naive_bayes import GaussianNB

[0226] from sklearn.model_selection import train_test_split

[0227] from sklearn.metrics import accuracy_score

[0228] 5.2 Prepare the dataset:

[0229] Prepare a labeled training dataset that includes input features and corresponding class labels. Typically, the input features are represented as a two-dimensional array (or matrix), while the labels are represented as a one-dimensional array.

[0230] 5.3 Data Preprocessing:

[0231] If there are missing values, outliers, or feature scaling is required in the dataset, you can use the data preprocessing tools in sklearn to perform the corresponding processing.

[0232] 5.4 Split the dataset:

[0233] Divide the data set into training set and test set. You can use the train_test_split function to divide the data set into two parts according to a certain ratio.

[0234] 5.5 Create a Naive Bayes model object:

[0235] Python

[0236] model=GaussianNB()

[0237] 5.6 Model training:

[0238] Python

[0239] model.fit(X_train,y_train)

[0240] Among them, X_train is the input feature of the training set, and y_train is the category label of the training set.

[0241] 5.7 Model prediction:

[0242] Python

[0243] y_pred=model.predict(X_test)

[0244] Among them, X_test is the input feature of the test set.

[0245] 5.8 Model Evaluation:

[0246] Python

[0247] accuracy=accuracy_score(y_test,y_pred)

[0248] Use appropriate indicators to evaluate the model, and further optimize the model parameters according to the actual situation, such as Gaussian naive Bayes (GaussianNB), multinomial naive Bayes (MultinomialNB) or Bernoulli naive Bayes (BernoulliNB), and make corresponding parameter adjustments.

[0249] S05: Evaluate and optimize the enterprise expert talent training model based on the test sample set to obtain the enterprise expert talent target model.

[0250] Specifically, in step S05, the enterprise expert talent training model is evaluated and optimized according to the test sample set to obtain the enterprise expert talent target model, including the following specific steps:

[0251] S51: Performing model evaluation on the enterprise expert talent training model according to the test sample set to obtain an evaluation result of the enterprise expert talent training model;

[0252] S52: According to the evaluation results of the enterprise expert talent training model, the parameters of the enterprise expert talent training model are processed in a serial pipeline manner through the Pipeline script to find the best combination of multiple parameter indicators of the enterprise expert talent training model, and the GridSearch function is used to search for the best combination of multiple parameter indicators of the enterprise expert talent training model, so as to automatically screen the optimal parameters and obtain the enterprise expert talent target model.

[0253] Specifically, the data is divided into two parts (based on the previous screening), one part is the clear expert talent data, and the other part is the common employee data with high reliability obtained by the first step model. These two parts of data are input into the classification model for training. The classification model used and the prediction results (only considering expert talents) are as follows:

[0254]

[0255] Then tune the model:

[0256] Because the senior talents are known and there is no definite inference about the unselected talents, the results of the model output cannot be judged according to the general precision, recall, accuracy, and F1 score. According to the characteristics of the known data, the performance of the model is evaluated using a validation set without negative samples. The judgment criteria are:

[0257] ( Where Pr[Y=1] is the actual probability of a positive sample), which is equivalent to

[0258] ( in Pr[f(x)＝1] is the probability that a sample is classified as a positive sample). Among them, r represents the recall rate and p represents the precision rate. r can be calculated using the positive samples of the validation set, and pr[f(x)=1] can be calculated using the entire validation set. This criterion can be used to judge the performance of the classifier like the F value. The probability of ordinary employees is uncertain, but the probability of expert talents is certain, so the accuracy of the model is the inferred probability of expert talents divided by the actual probability of expert talents. The way to tune is to adjust the parameters of the model to improve this accuracy.

[0259] Use the Pipeline script to implement serial pipeline processing of parameter optimization, and GridSearch to realize the combination of multiple parameter indicators in the parameter optimization process, and automatically screen the optimal parameters. Pipeline and GridSearch are both built-in functions of sklearn: GridSearch is a grid search, which is an exhaustive search method for specifying parameter values. The optimal learning algorithm is obtained by optimizing the parameters of the estimation function through cross-validation. That is, the possible values ​​of each parameter are arranged and combined, and all possible combination results are listed to generate a "grid". Then each combination is used for SVM training, and the performance is evaluated using cross-validation. After the fitting function tries all parameter combinations, it returns a suitable classifier and automatically adjusts to the best parameter combination. Please refer to the literature for details; the Pipeline script is a wrapper function that encapsulates the data preprocessing and modeling processes and executes them together.

[0260] It is a common way to use Pipeline scripts to perform serial pipeline processing and optimization on the parameters of the enterprise expert talent training model. The following is one of the specific implementation methods: Data preprocessing: First, preprocess the original data, such as data cleaning, feature extraction and conversion, etc. You can use a series of data preprocessing techniques, such as missing value processing, normalization, standardization, etc.; Feature selection: According to the relevance and importance of the features, select the features that have the greatest impact on model training. Commonly used feature selection methods include chi-square test, mutual information, model-based feature selection, etc.; Parameter optimization: Using pipeline technology, you can connect multiple steps in series and tune the parameters, so that Use optimization algorithms such as grid search and random search to search for the optimal parameter combination in a given parameter space; Model training: Based on the results of parameter optimization, use the training data to train the model. Various machine learning algorithms can be used, such as support vector machines, decision trees, neural networks, etc.; Model evaluation: Use evaluation indicators to evaluate the performance of the trained model. Common evaluation indicators include accuracy, recall rate, F1-score, etc.; Model optimization: Based on the evaluation results, further optimize the model, such as adjusting the model structure, adding regularization, etc.; Through such serial pipeline processing, in the process of parameter optimization, the optimal model parameters can be found more efficiently to improve the effect and performance of the model.

[0261] Using GridSearch can realize the free combination of multiple parameter indicators in the parameter optimization process. GridSearch is a function in the scikit-learn library that can be used to systematically traverse multiple parameter combinations and determine the best parameter settings through cross-validation.

[0262] As a specific implementation method of using GridSearch to find the best parameter combination:

[0263]

[0264]

[0265] In the above code, an SVC model object is defined and the parameter grid to be searched is specified. Then multiple metrics are defined, including accuracy, precision, and recall. Finally, a GridSearch object is created and the parameter search is performed. After performing the parameter search, the best parameter combination can be obtained using the best_params_ attribute and the corresponding metric result can be obtained using the best_score_ attribute.

[0266] Embodiment 2:

[0267] like Figure 3As shown, this embodiment provides an enterprise expert talent prediction device based on big data, the device comprising:

[0268] An acquisition unit 10 is used to acquire enterprise employee data to obtain a target data set;

[0269] The prediction unit 20 is connected to the acquisition unit 10 and is used to predict the target data set according to the enterprise expert talent target model to achieve enterprise expert talent prediction based on big data.

[0270] like Figure 4 As shown, specifically, the device further includes a construction unit 30 connected to the prediction unit, and the construction unit 30 is used to construct an enterprise expert talent target model;

[0271] The building unit 30 comprises:

[0272] An acquisition module, used to acquire a sample data set;

[0273] A proportion module, connected to the acquisition module, for dividing the sample data set into a training sample set and a test sample set in proportion, wherein the training sample set includes a positive sample set and an unlabeled sample set;

[0274] A screening module, connected to the ratio module, is used to screen out a reliable negative sample set from the unlabeled sample set according to the features contained in the positive sample set;

[0275] Among them: positive samples are enterprise experts and talents, reliable negative samples are ordinary employees, the positive sample set is a set including enterprise experts and talents, the unlabeled sample set is a set not including enterprise experts and talents, and the reliable negative sample set is a set including ordinary employees;

[0276] The iteration module is connected with the screening module and is used to iteratively train the positive sample set and the reliable negative sample set to obtain the enterprise expert talent training model;

[0277] Specifically, the iteration module includes:

[0278] The extraction submodule extracts all sample features, including positive sample features and reliable negative sample features; positive sample features are sample features of expert talents, and reliable negative sample features are sample features of ordinary employees;

[0279] The marking submodule is connected to the extraction submodule and is used to mark each positive sample as 1 and each reliable negative sample as -1 to construct a training data set;

[0280] The iteration submodule is connected to the marking submodule and is used to iteratively train the preset training model using the training data set to obtain the first model of enterprise expert talents;

[0281] The preset training models include SVM model, Random Forest model, K-nearest neighbor model, Logistic Regression model, and A type of Bayesian model.

[0282] The optimization module is connected to the acquisition module and the iteration module respectively, and is used to evaluate and optimize the enterprise expert talent training model according to the test sample set to obtain the enterprise expert talent target model.

[0283] It is to be understood that the above embodiments are merely exemplary embodiments used to illustrate the principles of the present invention, but the present invention is not limited thereto. For those of ordinary skill in the art, various modifications and improvements can be made without departing from the spirit and essence of the present invention, and these modifications and improvements are also considered to be within the scope of protection of the present invention.

Claims

1. A method for predicting enterprise expert talents based on big data, characterized in that: The method comprises the steps of: Step S1: Acquire enterprise employee data to obtain a target data set; Step S2: predicting the target data set according to the enterprise expert talent target model to achieve enterprise expert talent prediction based on big data.

2. The enterprise expert talent prediction method based on big data according to claim 1 is characterized in that: Before step S1, the method further includes: Step S0: Constructing an enterprise expert talent target model, which includes: Step S01: Obtain a sample data set; Step S02: Divide the sample data set into a training sample set and a test sample set in proportion, wherein the training sample set includes a positive sample set and an unlabeled sample set; Step S03: Filter out a reliable negative sample set from the unlabeled sample set according to the features contained in the positive sample set; Wherein: the positive samples are enterprise experts and talents, the reliable negative samples are ordinary employees, the positive sample set is a set including enterprise experts and talents, the unlabeled sample set is a set not including enterprise experts and talents, and the reliable negative sample set is a set including ordinary employees; Step S04: iteratively train the positive sample set and the reliable negative sample set to obtain an enterprise expert talent training model; Step S05: Perform model evaluation and optimization on the enterprise expert talent training model according to the test sample set to obtain the enterprise expert talent target model.

3. The enterprise expert talent prediction method based on big data according to claim 2 is characterized in that: In step S02, dividing the sample data set into a training sample set and a test sample set in proportion is dividing the sample data set into a training sample set and a test sample set in a ratio of 7:3, which specifically includes the following steps: Set the size of the sample data set to N, where N is a natural number greater than 1; Randomly shuffle the order of the sample data set; The first 0.7N samples in the randomly shuffled sample data set are used as the training sample set, and the last 0.3N samples in the randomly shuffled sample data set are used as the test sample set.

4. The enterprise expert talent prediction method based on big data according to claim 2 is characterized in that: In step S03, based on the features contained in the positive sample set, a reliable negative sample set is screened out from the unlabeled sample set, specifically including: According to the features contained in the positive sample set, the unlabeled samples in the unlabeled sample set are classified based on the Bayesian classifier, so as to determine the positive sample probability corresponding to the unlabeled samples in the unlabeled sample set, and the unlabeled samples with positive sample probability lower than the threshold are determined as reliable negative samples, and all reliable negative samples form a reliable negative sample set; Among them: the features contained in the positive sample set specifically include enterprise expert talent achievement information, enterprise expert talent award information, enterprise expert talent training information, enterprise expert talent performance appraisal information, and enterprise expert talent qualification certification information; the positive sample probability is the probability of becoming an enterprise expert talent.

5. The enterprise expert talent prediction method based on big data according to claim 2 is characterized in that: In step S04, iterative training is performed on the positive sample set and the reliable negative sample set to obtain an enterprise expert talent training model, including the following specific steps: Step S41: extracting all sample features, the sample features including positive sample features and reliable negative sample features; the positive sample features are sample features of expert talents, and the reliable negative sample features are sample features of ordinary employees; Step S42: Mark each positive sample as 1 and each reliable negative sample as -1 to construct a training data set; Step S43: iteratively train the preset training model using the training data set to obtain the enterprise expert talent training model; The preset training models include SVM model, Random Forest model, K-nearest neighbor model, Logistic Regression model, and A type of Bayesian model.

6. The enterprise expert talent prediction method based on big data according to claim 2 is characterized in that: In step S05, the enterprise expert talent training model is evaluated and optimized according to the test sample set to obtain the enterprise expert talent target model, including the following specific steps: S51: Performing model evaluation on the enterprise expert talent training model according to the test sample set to obtain an evaluation result of the enterprise expert talent training model; S52: According to the evaluation results of the enterprise expert talent training model, the parameters of the enterprise expert talent training model are processed in a serial pipeline manner through the Pipeline script to find the best combination of multiple parameter indicators of the enterprise expert talent training model, and the GridSearch function is used to search for the best combination of multiple parameter indicators of the enterprise expert talent training model, so as to automatically screen the optimal parameters and obtain the enterprise expert talent target model.

7. The enterprise expert talent prediction method based on big data according to any one of claims 1 to 6, characterized in that: In step S1, obtaining enterprise employee data to obtain a target data set includes the following steps: Collecting enterprise employee data, including employee performance data, employee training data, employee achievement data, employee award data, employee certification data, and employee professional skills data; Conduct data traversal and data processing on the collected enterprise employee data; in, The data processing includes: Fill or delete the company's employee data; or, One-hot encoding is used to convert character variables in enterprise employee data into numerical variables; in, The filling of the enterprise employee data is to fill the enterprise employee data with the mean, median or mode; The deleting of the enterprise employee data includes: deleting fields in the enterprise employee data with missing values ​​exceeding 50%, or screening and deleting variables in the enterprise employee data with a correlation coefficient between variables exceeding 0.

8.

8. A big data-based enterprise expert talent prediction device, characterized in that: The device comprises: An acquisition unit, used to acquire enterprise employee data to obtain a target data set; The prediction unit is connected to the acquisition unit and is used to predict the target data set according to the enterprise expert talent target model to achieve enterprise expert talent prediction based on big data.

9. The enterprise expert talent prediction device based on big data according to claim 8, characterized in that: The device also includes a construction unit, which is connected to the prediction unit and is used to construct an enterprise expert talent target model; The building block comprises: An acquisition module, used to acquire a sample data set; A proportion module, connected to the acquisition module, for dividing the sample data set into a training sample set and a test sample set in proportion, wherein the training sample set includes a positive sample set and an unlabeled sample set; A screening module, connected to the ratio module, for screening out a reliable negative sample set from the unlabeled sample set according to the features contained in the positive sample set; Wherein: the positive samples are enterprise experts and talents, the reliable negative samples are ordinary employees, the positive sample set is a set including enterprise experts and talents, the unlabeled sample set is a set not including enterprise experts and talents, and the reliable negative sample set is a set including ordinary employees; An iteration module, connected to the screening module, for iteratively training the positive sample set and the reliable negative sample set to obtain an enterprise expert talent training model; The optimization module is connected to the acquisition module and the iteration module respectively, and is used to evaluate and optimize the enterprise expert talent training model according to the test sample set to obtain the enterprise expert talent target model.

10. The enterprise expert talent prediction device based on big data according to claim 9, characterized in that: The iteration module includes: The extraction submodule extracts all sample features, including positive sample features and reliable negative sample features; the positive sample features are sample features of expert talents, and the reliable negative sample features are sample features of ordinary employees; A marking submodule, connected to the extraction submodule, is used to mark each positive sample as 1 and each reliable negative sample as -1, so as to construct a training data set; An iterative submodule, connected to the marking submodule, for iteratively training a preset training model using a training data set to obtain a first model of enterprise expert talents; The preset training models include SVM model, Random Forest model, K-nearest neighbor model, Logistic Regression model, and A type of Bayesian model.

Images