CN-121981338-A - Salary prediction method and device and readable storage medium

CN121981338ACN 121981338 ACN121981338 ACN 121981338ACN-121981338-A

Abstract

The application provides a salary prediction method, a device and a readable storage medium, wherein the method comprises the steps of obtaining user input; the method comprises the steps of carrying out intention analysis based on user input to obtain user intention, carrying out pay prediction through a pre-trained long-short-term memory network LSTM model according to the analyzed user intention to obtain a pay prediction result, wherein the LSTM model comprises a number pay prediction sub-model corresponding to the number pay query and a product pay prediction sub-model corresponding to the product pay query, and outputting the pay prediction result. The application can realize accurate prediction of employee compensation data.

Inventors

LIU YANG
ZHANG KAICHENG
GONG KUN
Ran Mengxuan
ZHANG ZHIJIE
WANG YUANJIE

Assignees

中国联合网络通信集团有限公司

Dates

Publication Date: 20260505
Application Date: 20260320

Claims (10)

1. A method of compensation prediction, the method comprising: S1, acquiring user input; s2, carrying out intention analysis based on user input to obtain user intention, wherein the categories of the user intention comprise number salary inquiry and product salary inquiry; s3, carrying out salary prediction through a pre-trained long-short-term memory network LSTM model according to the user intention obtained through analysis to obtain a salary prediction result, wherein the LSTM model comprises a number salary prediction sub-model corresponding to the number salary query and a product salary prediction sub-model corresponding to the product salary query; and S4, outputting a compensation prediction result.
2. The compensation prediction method according to claim 1, wherein S1 includes: if the user input is voice input, converting the voice into text, and analyzing the text by using a Natural Language Processing (NLP) technology to obtain a language structure and grammar; if the user input is text input, the text is analyzed by using a natural language processing NLP technology, and a language structure and grammar are obtained.
3. The compensation prediction method according to claim 1, wherein S2 includes: based on user input, performing intention analysis through a pre-trained BERT model to obtain user intention; the BERT model is obtained by training based on user history input and corresponding intention classification annotation data.
4. The salary prediction method according to claim 1, wherein in S3, the process of obtaining the number salary predictor model includes: s311, data preprocessing, namely acquiring number related data and preprocessing the data; s312, feature engineering, namely carrying out feature extraction on the preprocessed number related data to obtain number related features; S313, constructing an LSTM model, namely creating the LSTM layer, adding a full connection layer, and defining a loss function and an optimizer to obtain a number compensation prediction initial model; And S314, model training and evaluation, namely carrying out model training and model evaluation optimization on the initial model of the number salary prediction through the number related features to obtain a number salary prediction sub-model.
5. The compensation prediction method of claim 1, wherein in S3, the obtaining process of the product compensation prediction sub-model includes: S321, data preprocessing, namely acquiring product related data and performing data preprocessing; S322, feature engineering, namely carrying out feature extraction on the preprocessed product related data to obtain product related features; S323, constructing an LSTM model, namely creating an LSTM layer, adding a full connection layer, and defining a loss function and an optimizer to obtain a product salary prediction initial model; S324, model training and evaluation, namely carrying out model training and model evaluation optimization on the initial model of the product salary prediction through the relevant characteristics of the product to obtain a product salary prediction sub-model.
6. The compensation prediction method according to claim 4 or 5, wherein the data preprocessing includes at least one of: the missing value processing comprises missing value identification, missing value deletion and model prediction filling; Outlier processing; Data type conversion.
7. The compensation prediction method according to claim 4 or 5, wherein in LSTM model construction, a mean square error MSE is selected as a loss function, and Adam optimizer is selected to dynamically adjust a learning rate of each parameter by calculating a first moment estimate and a second moment estimate of the gradient.
8. A compensation prediction apparatus, the apparatus comprising: an input acquisition module configured to acquire user input; An intention analysis module configured to perform intention analysis based on user input to obtain user intention, wherein the category of the user intention includes a number salary query and a product salary query; The system comprises a salary prediction module, a salary prediction module and a storage module, wherein the salary prediction module is used for carrying out salary prediction through a pre-trained long-short-term memory network LSTM model according to the analyzed user intention to obtain a salary prediction result, and the LSTM model comprises a number salary prediction sub-model corresponding to number salary inquiry and a product salary prediction sub-model corresponding to product salary inquiry; And a result output module configured to output a compensation prediction result.
9. A compensation prediction device comprising a memory and a processor, the memory having a computer program stored therein, the processor being arranged to run the computer program to implement a compensation prediction method as claimed in any one of claims 1 to 7.
10. A computer readable storage medium, wherein a computer program is stored on the computer readable storage medium, which when executed by a processor, implements the salary prediction method according to any one of claims 1-7.

Description

Salary prediction method and device and readable storage medium Technical Field The present application relates to the field of data processing technologies, and in particular, to a salary prediction method, apparatus, and readable storage medium. Background Salary prediction is important for company and personal development, and not only is the operation efficiency and staff management of the enterprise concerned, but also the personal development and working power of staff are directly influenced. Timely and accurate salary prediction can bring continuous working power to staff, greatly improve the work enthusiasm and efficiency of the staff and is beneficial to improving the work efficiency of the team. However, the salary prediction result is commonly affected by various conditions, such as employee business development, enterprise background, etc., and the complex interaction between these factors makes it difficult to accurately predict the salary, since the salary is affected by many variables, and there is a complex nonlinear relationship between these variables, it is often difficult for a simple linear model to capture all the information therein. Therefore, how to more accurately conduct compensation prediction becomes a problem to be solved. Disclosure of Invention The present application is directed to solving the above-mentioned problems of the prior art, and provides a salary prediction method, apparatus and readable storage medium, which are used for solving the problems of the prior art. In a first aspect, the present application provides a method of compensation prediction, the method comprising: S1, acquiring user input; s2, carrying out intention analysis based on user input to obtain user intention, wherein the categories of the user intention comprise number salary inquiry and product salary inquiry; s3, carrying out salary prediction through a pre-trained long-short-term memory network LSTM model according to the user intention obtained through analysis to obtain a salary prediction result, wherein the LSTM model comprises a number salary prediction sub-model corresponding to the number salary query and a product salary prediction sub-model corresponding to the product salary query; and S4, outputting a compensation prediction result. In some embodiments, S1 comprises: if the user input is voice input, converting the voice into text, and analyzing the text by using a Natural Language Processing (NLP) technology to obtain a language structure and grammar; if the user input is text input, the text is analyzed by using a natural language processing NLP technology, and a language structure and grammar are obtained. In some embodiments, S2 comprises: based on user input, performing intention analysis through a pre-trained BERT model to obtain user intention; the BERT model is obtained by training based on user history input and corresponding intention classification annotation data. In some embodiments, in S3, the obtaining of the number compensation predictor model includes: s311, data preprocessing, namely acquiring number related data and preprocessing the data; s312, feature engineering, namely carrying out feature extraction on the preprocessed number related data to obtain number related features; S313, constructing an LSTM model, namely creating the LSTM layer, adding a full connection layer, and defining a loss function and an optimizer to obtain a number compensation prediction initial model; And S314, model training and evaluation, namely carrying out model training and model evaluation optimization on the initial model of the number salary prediction through the number related features to obtain a number salary prediction sub-model. In some embodiments, in S3, the obtaining of the product salary predictor model includes: S321, data preprocessing, namely acquiring product related data and performing data preprocessing; S322, feature engineering, namely carrying out feature extraction on the preprocessed product related data to obtain product related features; S323, constructing an LSTM model, namely creating an LSTM layer, adding a full connection layer, and defining a loss function and an optimizer to obtain a product salary prediction initial model; S324, model training and evaluation, namely carrying out model training and model evaluation optimization on the initial model of the product salary prediction through the relevant characteristics of the product to obtain a product salary prediction sub-model. In some embodiments, the data preprocessing includes at least one of: the missing value processing comprises missing value identification, missing value deletion and model prediction filling; Outlier processing; Data type conversion. In some embodiments, in LSTM model construction, the mean square error MSE is selected as the loss function, and Adam optimizers are selected to dynamically adjust the learning rate of each parameter by computing the first and second moment estima