Enterprise data prediction method, device and system

Abstract

The invention discloses an enterprise data prediction method and device, equipment and a storage medium, and relates to the field of big data, and the method comprises the steps: obtaining a historical revenue data set of a to-be-predicted enterprise; performing data derivation processing based on the historical revenue data to obtain expanded feature data; and inputting the expanded feature data into a revenue prediction model to obtain a revenue prediction result, the revenue prediction model being an algorithm model based on a gradient boosting decision tree. According to the scheme provided by the invention, the accuracy of enterprise data prediction is improved.

Description

Technical Field

[0001] The present invention relates to the field of big data technology, and in particular to an enterprise data prediction method, device, equipment and storage medium. Background Art

[0002] In order to provide data support for the development of enterprises, enterprise revenue forecast is a commonly used data analysis method. Enterprise revenue forecast is a data analysis method that uses internal statistical data, financial data and relevant market research information of the enterprise to predict and analyze the changing trends of the assets and profitability of the enterprise.

[0003] The current revenue forecasting methods for listed companies include historical data analysis, industry comparison, regression analysis, expert judgment, machine learning, etc. However, all of the above methods have certain limitations, making it difficult to accurately forecast corporate revenue and unable to provide accurate data support for corporate development. Summary of the invention

[0004] The present invention provides an enterprise data prediction method, device, equipment and storage medium, which improves the accuracy of enterprise data prediction.

[0005] According to one aspect of the present invention, there is provided a method for predicting enterprise data, comprising:

[0006] Obtain the historical revenue data set of the enterprise to be predicted;

[0007] Perform data derivation processing based on historical revenue data to obtain expanded feature data;

[0008] The expanded feature data is input into the revenue prediction model to obtain the revenue prediction result. The revenue prediction model is an algorithm model based on the gradient boosting decision tree.

[0009] In one embodiment, the method further comprises:

[0010] The revenue prediction model is trained based on the training data set, wherein the training data set is segmented based on K-means clustering and then a five-fold cross-validation training is performed.

[0011] In one embodiment, training a revenue prediction model based on a training data set includes:

[0012] Perform data derivation processing on the training data set to obtain expanded training feature data;

[0013] Perform K-means clustering on the expanded training feature data;

[0014] Each clustered data set is divided into five subsets, and the revenue prediction model is trained using the divided subsets.

[0015] In one embodiment, the training data set includes a historical revenue data set of the enterprise to be predicted and other enterprises in the same industry as the enterprise to be predicted.

[0016] In one embodiment, before performing data derivation processing based on historical revenue data to obtain expanded feature data, the method further includes:

[0017] Perform data preprocessing on historical revenue datasets;

[0018] Based on historical revenue data, data is derived and processed to obtain expanded feature data, including:

[0019] Based on the historical revenue data after data preprocessing, data derivation processing is performed to obtain expanded feature data.

[0020] In one embodiment, data preprocessing is performed on the historical revenue data set, including:

[0021] Perform default value processing, outlier processing and data normalization on historical revenue data.

[0022] In one embodiment, data derivation processing is performed based on historical revenue data to obtain expanded feature data, including:

[0023] Perform feature screening on historical revenue data;

[0024] Construct derived features based on expert experience;

[0025] Construct new time series features based on the features filtered out.

[0026] According to another aspect of the present invention, there is provided an enterprise data prediction device, comprising:

[0027] The acquisition module is used to obtain the historical revenue data set of the enterprise to be predicted;

[0028] A processing module is used to perform data derivation processing based on historical revenue data to obtain expanded feature data;

[0029] The prediction module is used to input the expanded feature data into the revenue prediction model to obtain the revenue prediction result. The revenue prediction model is an algorithm model based on the gradient boosting decision tree.

[0030] According to another aspect of the present invention, there is provided an electronic device, comprising: a processor, and a memory communicatively connected to the processor;

[0031] Memory stores computer-executable instructions;

[0032] The processor executes the computer execution instructions stored in the memory to implement the above-mentioned enterprise data prediction method.

[0033] According to another aspect of the present invention, a computer-readable storage medium is provided, in which computer-executable instructions are stored. When the computer-executable instructions are executed by a processor, they are used to implement the above-mentioned enterprise data prediction method.

[0034] According to another aspect of the present invention, a computer program product is provided, including a computer program stored in a computer-readable storage medium, at least one processor can read the computer program from the computer-readable storage medium, and at least one processor implements the above-mentioned enterprise data prediction method when executing the computer program.

[0035] The technical solution of the embodiment of the present invention obtains the historical revenue data set of the enterprise to be predicted, and performs data derivation processing based on the historical revenue data to obtain expanded feature data, and then inputs the expanded feature data into the revenue prediction model to obtain the revenue prediction result. The revenue prediction model is an algorithm model based on the gradient boosting decision tree, which provides a more accurate enterprise data prediction method and improves the accuracy of enterprise data prediction.

[0036] It should be understood that the contents described in this section are not intended to identify the key or important features of the embodiments of the present invention, nor are they intended to limit the scope of the present invention. Other features of the present invention will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0038] Figure 1 A flowchart of an enterprise data prediction method provided by an embodiment of the present invention;

[0039] Figure 2 A schematic diagram of the structure of an enterprise data prediction device provided by an embodiment of the present invention;

[0040] Figure 3 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0041] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiment of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiment of the present invention. Obviously, the described embodiment is only an embodiment of a part of the present invention, not all of the embodiments. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work should fall within the scope of protection of the present invention. The acquisition, storage, use, and processing of data in the technical scheme of this application comply with the relevant provisions of national laws and regulations.

[0042] Figure 1 A flow chart of an enterprise data prediction method provided by an embodiment of the present invention is as follows: Figure 1 As shown, the enterprise data prediction method provided by this embodiment includes:

[0043] Step S110, obtaining a historical revenue dataset of the enterprise to be predicted.

[0044] The enterprise data prediction method provided in the embodiment of the present application is applied to any computer or server with processing capabilities, and the computer or server is used to analyze and predict enterprise data. Analyzing and predicting enterprise data can provide data support for the business development of the enterprise and improve resource utilization efficiency.

[0045] The enterprise data prediction method provided in the embodiment of the present application can predict various business development data of the enterprise. In the embodiment of the present application, the enterprise revenue prediction is used as an example for explanation. Traditional enterprise revenue prediction often has forecast cycles such as quarterly revenue prediction, half-year revenue prediction, and full-year revenue prediction. Taking the quarterly revenue prediction of the above-mentioned company as an example, the traditional revenue prediction methods of the above-mentioned company include historical data analysis method, industry comparison method, regression analysis method, expert judgment method, machine learning method, etc. Among them, the historical data analysis method is based on the historical financial data of the listed company, using the average growth rate or time series analysis to predict the company's quarterly revenue; there is an industry comparison method, which predicts the company's quarterly revenue by referring to the financial performance of other companies in the same industry, combined with market development trends and industry data; there is a regression analysis method, which uses a statistical regression model to analyze factors affecting revenue, such as company assets and liabilities, monthly statistics of the company's main business, and the company's cash flow situation; expert judgment, combined with industry experts and internal experience of the company, predicts the company's revenue for the next quarter; machine learning, using machine learning models, digs out effective information from complex data sets, thereby predicting the company's revenue.

[0046] However, the historical data analysis method lacks foresight, relies too much on historical data, and does not analyze the revenue fluctuations of other companies in the same industry, and cannot reflect future market changes. In contrast to the historical data analysis method, the industry comparison method ignores the company's own historical data and only looks forward to the overall development trend of the industry. The prediction results obtained are often underfitting. In the regression analysis method, if important influencing factors are not included in the model, the model's prediction accuracy will be significantly reduced, and the model may ignore the nonlinear relationship between the influencing factors and the revenue value. Expert judgment is affected by the expert's personal subjective bias, and the judgments of different experts may be contradictory, and the prediction results between experts lack uniformity. Machine learning has relatively high requirements for data sets, and the model may be at risk of overfitting, that is, it performs well on the training set and poorly on the test set.

[0047] The enterprise data prediction method provided in this embodiment first obtains the historical revenue data set of the enterprise to be predicted. The historical revenue data set of the enterprise to be predicted is a data set composed of various historical data related to the business development of the enterprise. The historical revenue data set includes revenue-related data of the enterprise to be predicted within a certain time range. The longer the data time range in the historical revenue data set, the higher the accuracy of the final prediction result. The feature data included in the historical revenue data set can be multiple, and the more types of data, the higher the prediction accuracy. The historical revenue data set, for example, includes 12 feature data in total, including operating income, operating costs, investment income, non-operating income, interest income, total assets, quarterly stock prices, trading volume, transaction amount, company liabilities and loans in each quarter, and long-term deferred expenses of the enterprise to be predicted in units of quarters. Of course, the historical revenue data set can also include more or less feature data.

[0048] Step S120, performing data derivation processing based on historical revenue data to obtain expanded feature data.

[0049] It is difficult to reflect the revenue development trend of an enterprise by only using the acquired historical revenue data. Therefore, in this embodiment, data derivation processing is also performed based on the historical revenue data of the enterprise to be predicted, and the historical post-production data is expanded. The expansion of historical revenue data is based on the actual historical revenue data, and the revenue data with correlation is processed and obtained, which reflects the coupling characteristics of feature parts, or the historical revenue data at different times is processed and obtained, which reflects the time series characteristics of the feature itself. After the data derivation processing, the amount of data of the expanded feature data obtained is larger, which can more accurately reflect the revenue status of the enterprise.

[0050] Before performing data derivation processing based on historical revenue data to obtain expanded feature data, the historical revenue data set can be preprocessed first, and then data derivation processing is performed based on the preprocessed historical revenue data to obtain expanded feature data. Data preprocessing of historical revenue data sets is a means to reduce the impact of invalid data, and data preprocessing includes various means. For example, default value processing, outlier processing, and data normalization processing are performed on historical revenue data.

[0051] In detail, there are two strategies for default value processing of historical revenue data. One is to directly discard fields with default values ​​exceeding 80% of the total sample. The other is to fill the remaining fields with default values ​​with the mode in the field. For example, the quartile method can be used to identify outliers in historical revenue data, and the mode filling is adopted to count the proportion of abnormal fields in all fields. If the proportion exceeds 80%, the sample is directly discarded. The steps of the quartile method are as follows: 1. Sort the data from small to large. 2. Calculate the quartiles, the first quartile Q1: 25% position value, the second quartile Q2: 50% position value, and the third quartile Q3: 75% position value. 3. Calculate the interquartile range and determine the outlier limit. Interquartile range: IQR = Q3-Q1, lower limit: Q1-1.5×IQR, upper limit: Q3+1.5×IQR. For example, a specific method can be used to normalize the historical revenue data after default value processing and outlier processing. For example, the formula Normalization is performed, where μ is the mean, σ is the standard deviation, and X' is the normalized data.

[0052] Furthermore, the steps of performing data derivation processing based on the historical revenue data to obtain expanded feature data include the following steps: first, performing feature screening on the historical revenue data; then constructing derived features based on expert experience; and finally constructing new time series features based on the features screened.

[0053] Specifically, feature screening of historical revenue data can be feature screening based on the Pearson correlation coefficient, that is, calculating the Pearson correlation coefficient between the feature field and the revenue value in the historical revenue data set, where a coefficient value between -0.3 and 0.3 indicates a weak correlation, and a coefficient absolute value greater than 0.7 indicates a strong correlation. The Pearson correlation coefficient calculation formula is: Where cov(X,Y) is the covariance, σ X , σ Y Screening for standard deviation characteristics revealed that the company's quarterly operating income coefficient value was the highest, with a coefficient value of approximately 0.4.

[0054] For example, the following three aspects of expert experience features can be transformed by constructing derived features based on expert experience: 1. Features that reflect the relationship between operating income and operating costs: operating income / operating costs, operating income-operating costs. 2. Features that reflect the relationship between liabilities and assets: total liabilities / total assets, total liabilities-total assets. 3. Features that reflect the development trend of assets and liabilities: asset growth rate, liability growth rate.

[0055] For example, the time series feature construction can select the top 10 features with the highest correlation with revenue (such as operating income, operating costs, total liabilities, total assets, quarterly liabilities, quarterly loans, quarterly stock prices, quarterly stock trading volume, quarterly stock trading amount, and company interest income) selected from the feature screening, perform quarterly time series difference processing on these 10 features, calculate the growth rate of operating income for each quarter, and thus obtain 11 new time series features.

[0056] After the historical revenue data set includes the above 12 features, after feature derivation processing, for example, a total of 29 features are obtained, namely, quarterly operating income, operating costs, investment income, non-operating income, interest income, total assets, quarterly stock prices, trading volume, trading amount, the company's quarterly liabilities, loans, long-term deferred expenses, quarterly differences in operating income, quarterly differences in operating costs, quarterly differences in total liabilities, quarterly differences in total assets, quarterly differences in quarterly liabilities, quarterly differences in quarterly loans, quarterly differences in quarterly stock prices, quarterly differences in quarterly stock trading volume, quarterly differences in quarterly stock trading amount, quarterly differences in the company's interest income, quarterly growth rate of operating income, operating income / operating costs, operating income-operating costs, total liabilities / total assets, total liabilities-total assets, asset growth rate, and liability growth rate.

[0057] Step S130, input the expanded feature data into the revenue prediction model to obtain a revenue prediction result. The revenue prediction model is an algorithm model based on a gradient boosting decision tree.

[0058] Finally, the expanded feature data is input into the pre-trained revenue forecasting model to obtain the revenue forecast results. Among them, the revenue forecasting model is an algorithm model based on the gradient boosting decision tree, that is, the revenue forecasting model is trained based on the gradient boosting decision tree mechanism, and has the characteristics of high efficiency, accuracy and high scalability. Since the model based on the gradient boosting decision tree algorithm analyzes the nonlinear relationship between influencing factors and indicator values, this method can also solve the defects of the regression analysis method. By inputting the expanded feature data into the trained revenue forecasting model, the revenue forecast results of the company to be forecasted can be achieved. According to the design of the model, more accurate revenue forecast results for the company's quarterly, annual and other time periods can be output, thereby assisting the company to adjust its business strategy and change the direction of resource investment, thereby improving resource utilization efficiency.

[0059] In order to achieve accurate prediction of enterprise data, training of revenue prediction model is the key to ensure prediction accuracy. Before using revenue prediction model to predict enterprise data, it is necessary to first train the revenue prediction model based on the training data set. The training data set is segmented based on K-means clustering, and then five-fold cross-validation training is performed.

[0060] The training data set is a collection of data used to train the revenue forecasting model, which includes a variety of data. In order to ensure the accuracy of the training of the revenue forecasting model, the training data set can first be segmented based on K-means clustering. Data segmentation is to divide the data in the training data set into different data subsets according to different characteristics and then train the revenue forecasting model. The trained model after data segmentation has a higher degree of fit. Then, a five-fold cross-validation training is performed during the training process. The five-fold cross-validation is to randomly divide the data set into five equal-sized subsets, select one subset as the test set in turn, and the rest as the training set, repeat five times, and use different performance metrics to evaluate the model each time, and use the average of the five results as the final evaluation of the model performance. The five-fold cross-validation can effectively reduce the risk of overfitting of the model. Therefore, the revenue forecasting model is trained in combination with K-means clustering and five-fold cross-validation, and the resulting revenue forecasting model can more accurately forecast corporate revenue.

[0061] The specific steps for training the revenue prediction model may include: first, performing data derivation processing on the training data set to obtain expanded training feature data. The method for performing derivation processing on the training data set is the same as step S102. Then, K-means clustering is performed on the expanded training feature data. Finally, each clustered data set is divided into five subsets, and the revenue prediction model is trained using the divided subsets. Among them, the training data set includes the historical revenue data set of the enterprise to be predicted and other enterprises in the same industry as the enterprise to be predicted. In addition to using the historical revenue data of the enterprise to be predicted in the training data set, the historical revenue data of other enterprises in the same industry as the enterprise to be predicted is also used to train the revenue prediction model, which can improve the accuracy of revenue data prediction. For example, the data set can be divided from the two dimensions of target value turnover and industry through K-means clustering to generate a data set of companies in the same industry with similar turnover, thereby enhancing the similarity between the model training set and the test set.

[0062] The enterprise data prediction method provided in this embodiment obtains the historical revenue data set of the enterprise to be predicted, and performs data derivation processing based on the historical revenue data to obtain expanded feature data, and then inputs the expanded feature data into the revenue prediction model to obtain the revenue prediction result. The revenue prediction model is an algorithm model based on the gradient boosting decision tree, which provides a more accurate enterprise data prediction method and improves the accuracy of enterprise data prediction.

[0063] The following is a complete example to illustrate the model training process in the enterprise data prediction method provided in the embodiment of the present application.

[0064] First, the training data set is preprocessed. Taking the securities, insurance, and banking industries as examples, 50 companies are selected from each industry to obtain revenue-related data. Then, default value processing, outlier processing, and data normalization are performed.

[0065] Two strategies are used for default values:

[0066] (1) Fields with default values ​​exceeding 80% of the total sample are directly discarded.

[0067] (2) For the remaining fields with default values, fill them with the majority value in the field.

[0068] For outlier processing, the quartile method is used to identify outliers, and the mode filling is adopted to count the proportion of sample abnormal fields in all fields. If the proportion exceeds 80%, the sample is directly discarded. The steps of the quartile method are as follows:

[0069] (1) Data sorting: sort the data from small to large.

[0070] (2) Calculate the quartiles: the first quartile Q1: 25% position value, the second quartile Q2: 50% position value, and the third quartile Q3: 75% position value.

[0071] (3) Calculate the interquartile range and determine the outlier limits. Interquartile range: IQR = Q3-Q1, lower limit: Q1-1.5×IQR, upper limit: Q3+1.5×IQR.

[0072] For data normalization, the relevant algorithm is used to normalize the data set after default value processing and outlier processing. The formula is: Where μ is the mean, σ is the standard deviation, and X' is the normalized data.

[0073] Then, based on the Pearson correlation coefficient, the data set after data preprocessing is subjected to feature screening. First, the data set after data preprocessing is obtained, and the Pearson correlation coefficient between the feature field and the revenue value in the data set is calculated. The coefficient value between -0.3 and 0.3 is weakly correlated, and the absolute value of the coefficient greater than 0.7 is strongly correlated. The calculation formula of the Pearson correlation coefficient is: Where cov(X,Y) is the covariance, σ X , σ Y is the standard deviation.

[0074] Feature screening found that the company's quarterly operating income coefficient value was the highest, with a coefficient value of approximately 0.4.

[0075] At the same time, derived feature construction based on expert experience

[0076] The constructed expert experience features are as follows:

[0077] (1) Characteristics that reflect the relationship between operating income and operating costs: operating income / operating costs, operating income-operating costs.

[0078] (2) Characteristics that reflect the relationship between liabilities and assets: total liabilities / total assets, total liabilities-total assets.

[0079] (3) Characteristics that reflect the development trends of assets and liabilities: asset growth rate, liability growth rate.

[0080] Time series feature construction

[0081] After feature screening, the top 10 features with the highest correlation with operating income (operating income, operating cost, total liabilities, total assets, quarterly liabilities, quarterly loans, quarterly stock prices, quarterly stock trading volume, quarterly stock trading amount, and company interest income) are obtained, and these 10 features are processed by quarterly time series difference, and the growth rate of operating income for each quarter is calculated, thereby obtaining 11 new time series features.

[0082] When building the model, we first segmented the data set based on K-means clustering. After feature derivation processing, a total of 29 features were obtained, namely, quarterly operating income, operating costs, investment income, non-operating income, interest income, total assets, quarterly stock prices, trading volume, trading amount, the company's quarterly liabilities, loans, long-term deferred expenses, quarterly differences in operating income, quarterly differences in operating costs, quarterly differences in total liabilities, quarterly differences in total assets, quarterly differences in quarterly liabilities, quarterly differences in quarterly loans, quarterly differences in quarterly stock prices, quarterly differences in quarterly stock trading volume, quarterly differences in quarterly stock trading amount, quarterly differences in the company's interest income, quarterly growth rate of operating income, operating income / operating costs, operating income-operating costs, total liabilities / total assets, total liabilities-total assets, asset growth rate, and liability growth rate.

[0083] The feature-derived data set is segmented according to industry and operating income to ensure that similar companies can be clustered into one category, thereby strengthening the similarity between the model training set and the test set and improving the prediction accuracy of the model on the test set.

[0084] The steps of K-means clustering are as follows:

[0085] (1) Select the number of clusters k and initialize the center point.

[0086] (2) For each sample point, assign it to the nearest center point to obtain a cluster.

[0087] (3) Calculate the mean of all sample points in each cluster and re-update the center point.

[0088] (4) Repeat steps (2) and (3) until the center point no longer changes.

[0089] Finally, a revenue prediction model based on gradient boosting decision tree-five-fold crossover is established

[0090] After segmenting the data set based on K-means clustering (k is a multiple of 6 for the number of industries), 6 data sets were obtained. One of the data sets was randomly divided into 5 subsets of the same size, F1, F2, F3, F4, and F5. The model parameters based on the gradient boosting decision tree were initialized and five rounds of training were performed on it as the benchmark model.

[0091] Initialize the parameter grid of the gradient boosting decision tree: num_leaves, the maximum number of leaf nodes in the tree. If it is too large, it is easy to overfit. max_depth, the maximum depth of the tree, is used to limit the complexity of the tree. min_data_in_leaf, the minimum number of samples per leaf node, prevent overfitting. Learning parameters: learning_rate, learning rate, n_estimators, the number of training trees.

[0092] The model performance indicator is the mean square error (MSE), calculated as: Where n is the number of samples, y i is the true value, y i ' is the predicted value.

[0093] The model training and validation process is as follows: five rounds of training and validation are performed. In the i-th round, F i As the validation set, the rest of the data set is used as the training set. Record the performance indicators of the model. Performance evaluation: After completing five rounds, calculate the average of the five rounds of validation results to obtain the overall performance evaluation of the model. Adopt the grid search method, continuously conduct model training and evaluation, iteratively optimize the model parameters to the optimal, and finally obtain the revenue prediction model with the highest prediction accuracy.

[0094] Finally, Sun Lianhou’s revenue forecasting model is used for model forecasting to obtain the historical revenue data set of the company to be predicted. After data preprocessing and feature derivation, the data set is input into the revenue forecasting model based on gradient boosting decision tree-five-fold crossover in 3.2 to obtain the final quarterly revenue forecast results.

[0095] The enterprise data prediction method provided by the embodiment of the present invention provides an enterprise data prediction method based on a gradient boosting decision tree. Accurate revenue prediction is particularly important for the financial industry, especially the banking industry, such as pre-loan risk assessment of the credit system, risk decision-making during the loan, and post-loan collection. Enterprise revenue prediction is also used in financial businesses such as credit, anti-fraud, and investment and financing decisions. Accurate prediction results help improve the bank's financial decision-making, monitoring and early warning capabilities.

[0096] The revenue forecasting method based on the gradient boosting decision tree provided in the embodiment of the present invention utilizes public data sources, and the data is preprocessed, and the number of features is expanded through the method of expert experience + time series feature construction, and the coupling characteristics between features and the time series characteristics of the features themselves are mined. According to the integrated features, the data set is divided from the two dimensions of target value turnover and industry through K-means clustering to generate a data set of companies in the same industry with similar turnover, thereby strengthening the similarity between the model training set and the test set, and improving the prediction accuracy of the gradient boosting decision tree model. Finally, the model parameters are adjusted through the five-fold crossover + grid search method, which can effectively reduce the risk of model overfitting.

[0097] Figure 2 A schematic diagram of the structure of an enterprise data prediction device provided by an embodiment of the present invention is shown in FIG. Figure 2 As shown, the enterprise data prediction device provided by this embodiment includes:

[0098] The acquisition module 21 is used to obtain the historical revenue data set of the enterprise to be predicted; the processing module 22 is used to perform data derivation processing based on the historical revenue data to obtain expanded feature data; the prediction module 23 is used to input the expanded feature data into the revenue prediction model to obtain the revenue prediction result. The revenue prediction model is an algorithm model based on the gradient boosting decision tree.

[0099] The enterprise data prediction device provided in this embodiment is used to implement Figure 1 The enterprise data prediction method provided by the illustrated embodiment has similar implementation principles and technical effects, which will not be described in detail here.

[0100] In one embodiment, the enterprise data prediction device also includes a training module for training a revenue prediction model based on a training data set, wherein the training data set is segmented based on K-means clustering and then subjected to five-fold cross-validation training.

[0101] In one embodiment, the training module is specifically used to perform data derivation processing on the training data set to obtain expanded training feature data; perform K-means clustering on the expanded training feature data; divide each clustered data set into five subsets, and use the divided subsets to train the revenue prediction model.

[0102] In one embodiment, the training data set includes a historical revenue data set of the enterprise to be predicted and other enterprises in the same industry as the enterprise to be predicted.

[0103] In one embodiment, the processing module 22 is further used to perform data preprocessing on the historical revenue data set; and perform data derivation processing based on the historical revenue data after data preprocessing to obtain expanded feature data.

[0104] In one embodiment, the processing module 22 is specifically used to perform default value processing, abnormal value processing and data normalization processing on the historical revenue data.

[0105] In one embodiment, the processing module 22 is specifically used to perform feature screening on historical revenue data; construct derived features based on expert experience; and construct new time series features based on the features screened.

[0106] Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present application. Figure 3 As shown, the electronic device may include: a transceiver 31 , a processor 32 , and a memory 33 .

[0107] The processor 32 executes the computer execution instructions stored in the memory, so that the processor 32 executes the scheme in the above embodiment. The processor 32 can be a general-purpose processor, including a central processing unit CPU, a network processor (NP), etc.; it can also be a digital signal processor DSP, an application-specific integrated circuit ASIC, a field programmable gate array FPGA or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components.

[0108] The memory 33 is connected to the processor 32 via a system bus and completes communication between them. The memory 33 is used to store computer program instructions.

[0109] The transceiver 31 may be used to obtain tasks to be executed and configuration information of the tasks to be executed.

[0110] The system bus can be a peripheral component interconnect (PCI) bus or an extended industry standard architecture (EISA) bus, etc. The system bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, only one thick line is used in the figure, but it does not mean that there is only one bus or one type of bus. The transceiver is used to realize the communication between the database access device and other computers (such as clients, read-write libraries, and read-only libraries). The memory may include random access memory (RAM) and may also include non-volatile memory.

[0111] The electronic device provided in the embodiment of the present application may be a computer or server deployed by the enterprise data prediction device in the above-mentioned embodiment.

[0112] An embodiment of the present application also provides a computer-readable storage medium, in which computer instructions are stored. When the computer instructions are executed on a computer, the computer executes the technical solution of the enterprise data prediction method of the above embodiment.

[0113] An embodiment of the present application also provides a computer program product, which includes a computer program stored in a computer-readable storage medium. At least one processor can read the computer program from the computer-readable storage medium. When at least one processor executes the computer program, it can implement the technical solution of the enterprise data prediction method in the above embodiment.

[0114] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps described in the present invention can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solution of the present invention can be achieved, and this document does not limit this.

[0115] The above specific implementations do not constitute a limitation on the protection scope of the present invention. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for predicting enterprise data, characterized in that: include: Obtain the historical revenue data set of the enterprise to be predicted; Performing data derivation processing based on the historical revenue data to obtain expanded feature data; The expanded feature data is input into a revenue prediction model to obtain a revenue prediction result, wherein the revenue prediction model is an algorithm model based on a gradient boosting decision tree.

2. The method according to claim 1, characterized in that The method further comprises: The revenue prediction model is trained based on a training data set, wherein the training data set is segmented based on K-means clustering and then subjected to five-fold cross-validation training.

3. The method according to claim 2, characterized in that The training of the revenue prediction model based on the training data set includes: Performing data derivation processing on the training data set to obtain expanded training feature data; Performing K-means clustering on the expanded training feature data; Each clustered data set is divided into five subsets, and the revenue prediction model is trained using the divided subsets.

4. The method according to claim 3, characterized in that The training data set includes a historical revenue data set of the enterprise to be predicted and other enterprises in the same industry as the enterprise to be predicted.

5. The method according to any one of claims 1 to 4, before performing data derivation processing based on the historical revenue data to obtain expanded feature data, further comprising: Performing data preprocessing on the historical revenue data set; The data derivation processing based on the historical revenue data to obtain expanded feature data includes: Data derivation processing is performed based on the historical revenue data after the data preprocessing to obtain expanded feature data.

6. The method according to claim 5, characterized in that The data preprocessing of the historical revenue data set includes: The historical revenue data is subjected to default value processing, outlier processing and data normalization processing.

7. The method according to any one of claims 1 to 4, wherein the step of performing data derivation processing based on the historical revenue data to obtain expanded feature data comprises: Performing feature screening on the historical revenue data; Construct derived features based on expert experience; Construct new time series features based on the features filtered out.

8. An enterprise data prediction device, characterized in that: include: The acquisition module is used to obtain the historical revenue data set of the enterprise to be predicted; A processing module, used for performing data derivation processing based on the historical revenue data to obtain expanded feature data; The prediction module is used to input the expanded feature data into the revenue prediction model to obtain the revenue prediction result. The revenue prediction model is an algorithm model based on the gradient boosting decision tree.

9. An electronic device, characterized in that: include: A processor, and a memory communicatively connected to the processor; The memory stores computer-executable instructions; The processor executes the computer-executable instructions stored in the memory to implement the enterprise data prediction method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-executable instructions, which, when executed by a processor, are used to implement the enterprise data prediction method according to any one of claims 1 to 7.

11. A computer program product, characterized in that It includes a computer program stored in a computer-readable storage medium. At least one processor can read the computer program from the computer-readable storage medium. When at least one processor executes the computer program, it can implement the enterprise data prediction method as described in any one of claims 1 to 7.

Images