Financial data prediction method and device, computer equipment and storage medium

Abstract

The invention discloses a financial data prediction method and device, computer equipment and a storage medium, and the method can obtain different types of to-be-processed data obtained by a plurality of different obtaining modes, and calls a prediction model according to the data types contained in the to-be-processed data. Wherein the prediction model is a combined model of a data analysis algorithm model and a machine learning model. And inputting the to-be-processed data into the prediction model to obtain prediction result data output by the prediction model. And generating a result preview interface according to the prediction result data. According to the method, comprehensive analysis and prediction can be carried out by using a data analysis algorithm and a machine learning algorithm based on the multi-modal to-be-processed data, so that more accurate prediction result data is obtained, and the accuracy and reliability of the prediction process are improved.

Description

Technical Field

[0001] The present application relates to the field of financial technology, and in particular to a financial data forecasting method, apparatus, computer equipment and storage medium. Background Art

[0002] Financial data refers to data generated by an enterprise during its economic activities, which is used to reflect the enterprise's financial status, operating results and cash flow. These data can be used to prepare financial statements and assist enterprises in making decisions. Financial data may include digital data such as funds, debts, equity, and expenses.

[0003] Taking funds as an example, financial management requires statistics and forecasts on fund flows, and the statistical and forecast results are displayed through terminal devices. Based on the displayed statistical and forecast results, the impact of fund flow fluctuations on the operation and management process can be dealt with in advance. That is, decisions can be made in advance based on the statistical and forecast results displayed by the terminal devices, reducing the risks of fund shortages and inability to pay debts on time that may be caused by unstable fund flows, as well as reducing excessive fund reserves and improving the efficiency of the company's fund use.

[0004] Financial data forecasting can analyze and judge the fluctuation trend of financial data within a certain period based on historical capital data, and display the forecast results through terminal devices. However, due to the single data source, forecasting algorithm and result evaluation index of historical capital data, the accuracy and reliability of financial data forecasting based on historical capital data are insufficient, resulting in a large deviation between the financial forecast results displayed by the terminal device and the actual financial data, reducing the accuracy of the content display process. Summary of the invention

[0005] In view of this, embodiments of the present application provide a financial data forecasting method, apparatus, computer equipment, and storage medium to solve the problem of low accuracy in displaying financial data forecasting results.

[0006] According to one aspect of the present application, a financial data forecasting method is provided, the method comprising:

[0007] Acquire data to be processed, wherein the data to be processed includes one or more combinations of first data, second data, and third data, wherein the first data is financial data obtained by querying a preset financial database; the second data is business data obtained through interface monitoring; and the third data is influencing factor data obtained through data capture;

[0008] Read the data type contained in the data to be processed, and call the prediction model according to the data type, wherein the prediction model includes at least one first prediction sub-model and a second prediction sub-model; the type and quantity of the first prediction sub-model and the second prediction sub-model contained in the prediction model are associated with the data type; the first prediction sub-model has a built-in data analysis algorithm function; and the second prediction sub-model is a machine learning model obtained by training based on sample data;

[0009] Inputting the data to be processed into the prediction model to obtain prediction result data output by the prediction model;

[0010] A result preview interface is generated according to the prediction result data, and the result preview interface includes a view component generated based on the user-defined warning parameters and the prediction result data.

[0011] According to another aspect of the present application, a financial data forecasting device is provided, the device comprising:

[0012] A data acquisition module is used to acquire data to be processed, wherein the data to be processed includes one or more combinations of first data, second data and third data, wherein the first data is financial data obtained by querying a preset financial database; the second data is business data obtained through interface monitoring; and the third data is influencing factor data obtained through data capture;

[0013] A model calling module, used to read the data type contained in the data to be processed, and to call the prediction model according to the data type, wherein the prediction model includes at least one first prediction sub-model and a second prediction sub-model; the types and quantities of the first prediction sub-model and the second prediction sub-model contained in the prediction model are associated with the data type; the first prediction sub-model has a built-in data analysis algorithm function; and the second prediction sub-model is a machine learning model obtained by training based on sample data;

[0014] A prediction module, used for inputting the data to be processed into the prediction model to obtain prediction result data output by the prediction model;

[0015] A display module is used to generate a result preview interface according to the prediction result data, wherein the result preview interface includes a view component generated based on user-defined warning parameters and the prediction result data.

[0016] According to another aspect of the present application, a computer device is provided, including a storage medium, a processor, and a computer program stored on the storage medium and executable on the processor, wherein the processor implements the above-mentioned financial data forecasting method when executing the program.

[0017] According to another aspect of the present application, a storage medium is provided, on which a computer program is stored, and when the program is executed by a processor, the above-mentioned financial data forecasting method is implemented.

[0018] By means of the above technical scheme, a financial data forecasting method, device, computer equipment and storage medium provided in the embodiment of the present application can obtain different types of data to be processed obtained by a plurality of different acquisition methods, and call a forecasting model according to the data type contained in the data to be processed. Among them, the forecasting model includes at least one first forecasting sub-model and a second forecasting sub-model, and the type and quantity of the first forecasting sub-model and the second forecasting sub-model contained in the forecasting model are associated with the data type contained in the data to be processed. Then the data to be processed is input into the forecasting model to obtain the forecasting result data output by the forecasting model. Thus, a result preview interface is generated according to the forecasting result data. The method can use data analysis algorithms and machine learning algorithms to perform comprehensive analysis and forecasting based on multi-modal data to be processed to obtain more accurate forecasting result data, improve the accuracy and reliability of the forecasting process, and improve the problem of low accuracy in the presentation process of financial data forecasting results.

[0019] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0021] Figure 1 An application scenario diagram of a financial data prediction method provided by an embodiment of the present application is shown;

[0022] Figure 2 A schematic diagram of a process flow of a financial data forecasting method provided by an embodiment of the present application is shown;

[0023] Figure 3 A schematic diagram of a process for obtaining data to be processed provided in an embodiment of the present application is shown;

[0024] Figure 4 A schematic diagram of a calling model process provided by an embodiment of the present application is shown;

[0025] Figure 5 A schematic diagram of the output prediction result data flow provided by an embodiment of the present application is shown;

[0026] Figure 6A schematic diagram of a data preprocessing process provided by an embodiment of the present application is shown;

[0027] Figure 7 A schematic diagram of the process of training the second prediction sub-model provided in an embodiment of the present application is shown;

[0028] Figure 8 A schematic diagram of the verification prediction result process provided by an embodiment of the present application is shown;

[0029] Fig. 9 A schematic diagram of the structure of a financial data forecasting device provided in an embodiment of the present application is shown;

[0030] Fig.10 A schematic diagram of the structure of a computer device provided in an embodiment of the present application is shown. DETAILED DESCRIPTION

[0031] The present application will be described in detail below with reference to the accompanying drawings and in combination with embodiments. It should be noted that the embodiments and features in the embodiments of the present application can be combined with each other without conflict.

[0032] In the embodiments of the present application, the financial data generally refers to data generated during economic activities that can be identified, processed, transmitted and stored by computers and other data processing equipment. The financial data includes, but is not limited to, cash flow data, goods flow data, debt flow data, etc. In the embodiments of the present application, the financial data forecasting method is explained using cash flow forecasting as an example. It should be understood that the method can also be applied to other types of financial data forecasting. Furthermore, based on the financial data forecasting method described in the embodiments of the present application, data forecasting methods in other fields that can be associated with by those skilled in the art without paying creative labor also belong to the scope of protection of the present application. Specific examples in other fields will not be listed one by one in this application.

[0033] The financial data forecast refers to the process of estimating and predicting the financial status, operating results and cash flow in a certain period of time in the future by using existing data, statistical methods, measurement models and empirical analysis. Therefore, financial data forecasts may include but are not limited to: revenue forecasts, cost forecasts, profit forecasts, cash flow forecasts, asset-liability forecasts, financial ratio analysis, etc. The forecast results of financial data forecasts can be used to assist decision-making, optimize resource allocation, etc., to improve financial management, reduce operating risks, and provide decision-making basis for investors, creditors and other stakeholders.

[0034] A financial data forecasting method provided in the embodiment of the present application can be applied to a financial management system. The financial management system is a software and hardware platform for displaying financial data and financial data forecasting results. The financial management system can rely on computers, mobile terminals, servers, industrial hosts, smart wearable devices, etc. to provide users with a human-computer interaction interface, and respond to users performing operations such as collection, processing, transmission, and storage of financial data based on the provided human-computer interaction interface to meet specific user needs.

[0035] In a feasible implementation, the financial management system can be run on a single device, that is, the user can realize all the functions of the financial management system through a single client device. For example, the financial management system is an application installed on a personal computer. The financial management system application has a built-in financial data prediction algorithm program. When performing financial data prediction, the computer can obtain financial data input by the user or stored locally, and perform prediction on the financial data to obtain corresponding prediction result data. The prediction result data is then displayed based on the display program provided by the application.

[0036] In another feasible implementation, the financial management system can be operated on multiple devices, that is, the coordination between multiple devices is required to realize the full functions of the financial management system. Figure 1 As shown, the financial management system includes a server and multiple clients connected to the server. The multiple clients can be operated by different users respectively. The financial data generated by the operation can be sent to the server through the network after preliminary processing locally on the client. The financial data is then further processed, distributed and stored by the server. Different clients can call different financial data from the server based on different user permissions. And the server can send specific data to the client according to specific financial data management needs.

[0037] It should be noted that in the embodiment of the present application, the financial management system is used as the execution subject of the financial data prediction method. Unless otherwise specified, the financial management system can be implemented by running relevant applications on the client, or by running relevant applications on the server, or by cooperating with the client to run relevant applications. Therefore, from the perspective of hardware devices, the steps corresponding to the financial data prediction method can be executed by the client, or by the server, or some steps can be executed by the client and some steps can be executed by the server.

[0038] By running the financial data forecasting method based on the financial management system, it is possible to obtain the financial data forecast results within a certain future period by collecting, organizing, analyzing and calculating the existing data. Among them, the existing data includes historical data and real-time data of a specific type of financial data. Taking the cash flow data forecasting process as an example, the financial management system can collect historical cash flow data within the past 12 months, and combine it with data analysis and forecasting algorithms such as time series analysis and regression analysis to predict the cash flow of the next month, that is, to obtain the cash flow forecast results for the next month.

[0039] When financial data is predicted based solely on historical data, the data prediction result is a pure mathematical analysis result, and the revenue and expenditure rules of financial data do not completely conform to mathematical relationships such as time series relationships or regression function relationships, resulting in large errors in the prediction activities of some data. Therefore, in order to improve the accuracy of data prediction results, the financial management system can also collect other types of existing data.

[0040] In a feasible implementation, the existing data may also include business data. Business data refers to data generated in business activities involving finance. Business data may not be directly reflected in digital form related to financial data, but may be multimodal data of various types. For example, business data may include data generated in the current business operations such as the production quantity, orders, letters of intent, personnel structure, and business development plans of the enterprise. These data can be collected and organized by the financial management system to assist in analysis and prediction, and obtain more accurate financial data prediction results.

[0041] In addition to the impact that the internal business activities of the enterprise may have on future financial data, the external environment of the enterprise may also have an impact on future financial data. Therefore, in order to obtain more accurate financial data forecast results, in a feasible implementation, the existing data may also include influencing factor data. Among them, influencing factor data generally refers to the data of external environmental factors of the enterprise that can affect the financial data forecast process. For example, influencing factor data may include financial market fluctuation data, seasonal data, security events, etc. These data can also be collected by the financial management system to assist in analysis and forecasting.

[0042] To this end, a financial data forecasting method is provided in this embodiment. Figure 2 As shown, the method includes:

[0043] S101. Obtain data to be processed.

[0044] The financial management system can obtain the data to be processed in response to the data prediction instructions input by the user. Among them, the data prediction instructions generally refer to the interactive instructions used to control the financial management system to perform the financial data prediction function. The user can input the data prediction instructions according to the interactive mode supported by the client or server. For example, the client can present a user interface related to financial management by running the financial management system application. The user interface may include at least one prediction function control, and the user can click the prediction function control on the user interface provided by the client to input the data prediction instruction. For another example, for a client and a financial management system that support voice interaction functions, the user can input the data prediction instruction through voice data with the input content being "Please help me predict the cash flow situation for the next quarter".

[0045] In some embodiments, the data forecasting instruction may also include a data type specified by the user, and the forecasting method used. For example, in the user interface of the financial management system, a drop-down list control for specifying a data type and a check box control for specifying a forecasting method may be included. The user can specify that the data type to be forecasted is cash flow by clicking the drop-down list control and selecting the type item named "cash flow" in the list. And by selecting the check box controls with the contents "timing analysis" and "AI analysis", specify that the data forecasting method is data timing analysis and machine learning algorithm analysis. Then enter the data forecasting instruction by clicking the forecasting function control.

[0046] Data to be processed refers to the existing data collected by the financial management system in response to the data forecasting instructions input by the user and used to participate in the financial data forecasting activities. Figure 3 As shown, the data to be processed includes one or more combinations of first data, second data and third data. The first data is financial data obtained by querying a preset financial database. The second data is business data obtained through interface monitoring. The third data is influencing factor data obtained through data capture.

[0047] The financial management system can collect and organize financial data in daily business activities, and then store the collected financial data through the financial database. For example, sales revenue data and operating cash flow data in the past period of time. Depending on the specific forecasting period, sales revenue data can be monthly or quarterly sales amounts. Operating cash flow data is used to record the cash inflows and outflows generated by the daily business activities of the enterprise. After these financial data are generated, they can be recorded in the financial database. When forecasting financial data, the financial management system can generate a data query request according to the needs of the forecasting task, and then send the data query request to the financial database so that the financial database can feedback the required financial data according to the data query request. For the convenience of description, these historical financial data obtained by querying the financial database are called first data.

[0048] In addition to obtaining financial data through queries, the financial management system can also connect to other business systems through an application programming interface (API), and collect business data corresponding to the API through data acquisition methods such as data requests and data capture. For ease of description, these business data obtained through interface monitoring are called second data. For example, by accessing the API interface of the market management system, you can obtain real-time and historical market-related data of the enterprise, that is, the second data. The second data may include data such as external market data and economic indicators. It can be seen that the second data obtained by monitoring the API can greatly enrich the data source and enhance the accuracy and timeliness of financial data forecasts.

[0049] In a feasible implementation, the financial management system can also capture data on influencing factors that can affect changes in financial data from network data of public networks or other systems by means of data crawling. These influencing factor data are referred to as third data. For example, the financial management system can have a built-in web crawler tool that can collect public market information and related corporate data. The influencing factor-related data obtained from the Internet by the web crawler tool can be used to analyze market dynamics and related corporate behaviors, and then assist in analyzing changes in financial data.

[0050] In order to facilitate the acquisition of data to be processed, the financial management system can also use big data technology to write the captured data to the big data platform through a stream processing framework (such as flink, etc.), and monitor the written data and data updates in real time. The collected data is monitored and updated in real time to obtain the dynamic changes in the status of financial related activities.

[0051] It should be noted that the historical financial data (first data), business data (second data) and influencing factor data (third data) contained in the data to be processed may be affected by prerequisite factors such as the data prediction instruction input by the user, the user identity, and the user's use time. That is, under different prerequisite conditions, the data content contained in the data to be processed is different. For example, the data to be processed may include only the first data. The data to be processed may also include the first data and the second data. The data to be processed may also include the first data and the third data. The data to be processed may also include the first data, the second data and the third data at the same time. In addition, in addition to the first data, the second data and the third data mentioned above, the data to be processed may also include other types of data, such as non-public data, legal data, business operation data, etc. that can affect the fluctuation of financial data. Other types of data that can be associated with the above-mentioned first data, second data and third data all belong to the scope of protection of this application.

[0052] S102: Read the data type contained in the data to be processed, and call the prediction model according to the data type.

[0053] After obtaining the data to be processed, the financial management system can read the data type contained in the data to be processed. The data type refers to the specific form of the data to be processed, including but not limited to text type, number type, picture type, audio and video type, etc. The financial management system can determine the data type by reading the tag used to characterize the data type in the data to be processed. For example, after obtaining the file corresponding to the data to be processed, the financial management system can read the extension of the data file to be processed to determine the data type.

[0054] Obviously, the data types contained in the data to be processed are not limited to one form. When the data to be processed is multimodal data, the financial management system can read multiple data types. For example, if the data to be processed includes historical cash flow, contracts, news videos, etc., then by reading the data type, it can be determined that the data types contained in the current data to be processed are digital type, text type, and video type.

[0055] like Figure 4As shown, according to the type of data read, the financial management system can call the corresponding prediction model. The financial management system can build in multiple prediction models that process different types of data and have different uses according to different data types and different prediction results to form a model library. Since the data to be processed can be multimodal data, the prediction model includes at least one first prediction sub-model and a second prediction sub-model. Among them, the first prediction sub-model has a built-in data analysis algorithm function. The second prediction sub-model is a machine learning model obtained by training based on sample data. In order to obtain more accurate financial data prediction results, the type and quantity of the first prediction sub-model and the second prediction sub-model contained in the prediction model are associated with the data type.

[0056] For example, for the data to be processed of pure digital type, the digital type data can be predicted through data analysis algorithm functions such as time series analysis and regression analysis. Therefore, when the data to be processed contains digital data, the first prediction sub-model suitable for digital analysis can be called. Among them, the first prediction sub-model has a built-in regression analysis function. For the data to be processed of text type, since data analysis algorithm functions such as time series analysis and regression analysis cannot predict text data, when the data to be processed contains text data, the second prediction sub-model suitable for text analysis can be called. Among them, the second prediction sub-model has a pre-trained machine learning neural network model built in.

[0057] In addition, the financial management system can also call the corresponding prediction model based on the prediction method specified by the user in the data prediction instruction. For example, when the user enters the data prediction instruction by selecting the two options of "regression analysis" and "AI analysis" in the user interface, the financial management system can call the first prediction sub-model and the second prediction sub-model in the model library to form a prediction model combining the first prediction sub-model and the second prediction sub-model.

[0058] S103: Input the data to be processed into the prediction model to obtain prediction result data output by the prediction model.

[0059] After calling the prediction model, the financial management system can input the data to be processed into the prediction model, so as to perform prediction analysis based on the data to be processed through the prediction model and obtain prediction result data output by the prediction model.

[0060] like Figure 5As shown, after the data to be processed is input into the prediction model, the prediction model can classify the data according to the data type in the data to be processed to obtain data sets suitable for different sub-models. For the first prediction sub-model based on the data analysis algorithm function, the front-end processing module of the prediction model can read the digital type data in the data to be processed to obtain a numerical data set, and then input the numerical data set into the first prediction sub-model, so that the first prediction sub-model performs data analysis such as time series analysis and regression analysis on the numerical data, thereby obtaining analysis results of the numerical type.

[0061] For example, the financial management system can obtain independent variable data for data analysis in the data to be processed, including financial data such as sales revenue, cost expenditure, accounts receivable turnover days, accounts payable turnover days, and obtain macroeconomic indicators and industry indicators. And extract historical cash flow data from the data to be processed as dependent variable data. Then, based on the extracted independent variable data and dependent variable data combined with the regression analysis algorithm, regression analysis is performed to calculate the regression function relationship between the dependent variable and the independent variable. Therefore, according to the regression function relationship, by defining the independent variable, the dependent variable data in the future cycle is calculated.

[0062] In a feasible implementation, when the financial management system performs time series analysis, historical data can be extracted from the data to be processed. The historical data is numerical data with time series recorded at preset time intervals. The historical data is then sorted in chronological order to obtain a time series data set. The time series data set is then smoothed to obtain a smoothed data sequence. The first prediction sub-model based on the time series analysis algorithm is called according to the smoothed data sequence, so that the first prediction sub-model generates prediction result data according to the smoothed data sequence.

[0063] For example, according to the time series analysis algorithm, after collecting the historical cash inflow and outflow data of the enterprise, the financial management system can arrange the collected cash flow data in chronological order. For the historical cash flow data of the enterprise, these data are arranged in chronological order and may also include seasonal factor data, such as the sales peak in a specific season affecting cash inflow, and trend factor data, such as the impact of the long-term development trend of the enterprise on cash flow.

[0064] For cash flow data arranged in chronological order, the financial management system can use unit root test algorithms such as the Augmented Dickey-Fuller (ADF) test to test for stationarity. After the test, if the arranged data is not stable, differential processing is required to make the sorted cash flow data stable. Among them, when performing differential processing, the moving average method, exponential smoothing method, etc. can be used to smooth the time series data to eliminate short-term fluctuations and better observe long-term trends.

[0065] Then select the appropriate time series model according to the characteristics of the data, such as the Autoregressive Moving Average (ARMA) model, the Autoregressive Integrated Moving Average (ARIMA) model, etc.

[0066] Based on cash flow data, the model parameters called by the determination of historical data can be determined through analysis algorithms such as grid search, such as the number of autoregressive terms p, the number of differences d, and the number of moving average terms q in ARIMA (p, d, q). Then use diagnostic algorithms such as Ljung-Box test to diagnose the established model and check whether the residual is white noise. In addition, the reserved historical data can be used for verification.

[0067] For example, the first prediction sub-model can obtain the average cash flow of the past few months as a preliminary prediction of the cash flow for the next month based on the moving average prediction algorithm. If the cash inflows in the past three months are 1 million yuan, 1.2 million yuan, and 1.1 million yuan respectively, then the cash inflow for the next month predicted by the simple moving average method is (100+120+110) / 3=1.1 million yuan.

[0068] For the second prediction sub-model based on the machine learning algorithm, the front-end processing module of the prediction model can read the text type data in the data to be processed to obtain a text data set. Then the text data set is input into the second prediction sub-model, so that the second prediction sub-model performs classification probability calculation based on the input text data set, thereby obtaining a classification result for a specific classification label.

[0069] Before using machine learning technology to predict financial data, the financial management system can call a pre-trained neural network model. For example, the second prediction sub-model is a neural network model obtained by training historical financial data and business data as input and cash flow as output. The neural network model can automatically learn complex patterns and nonlinear relationships in the data to improve the accuracy of the prediction.

[0070] After calling the pre-trained second prediction sub-model, the financial management system can input the independent variable data used for prediction into the second prediction sub-model, and output the predicted cash flow value after calculation by the neural network. For example, the second prediction sub-model is a feedforward neural network model with a three-layer structure. The input layer of the model has 5 nodes, corresponding to the input sales revenue, cost expenditure, accounts receivable turnover days, accounts payable turnover days and a macroeconomic indicator, the hidden layer of the model has 10 nodes, and the output layer of the model has 1 node, which corresponds to the prediction result of cash flow. When performing cash flow prediction, the corresponding independent variable data can be input into the neural network, and the predicted cash flow can be obtained after the forward propagation calculation of the neural network.

[0071] After the data to be processed is input into the prediction model, the prediction model can perform data prediction based on multiple sub-models such as the first prediction sub-model and the second prediction sub-model, and each sub-model can output prediction result data accordingly. Among them, the prediction results output by different sub-models can be of different types. For example, the first prediction sub-model can output digital prediction values ​​such as cash flow prediction values, while the second prediction sub-model outputs prediction values ​​of types such as influencing factors and calculation weights. The financial management system can combine the output results of different types of models or calculate them in mutual influence, so as to obtain richer prediction result data.

[0072] The prediction results output by different sub-models can also be of the same type. For example, the first prediction sub-model and the second prediction sub-model both output predicted values ​​for cash flow. For the same type of prediction results output by different sub-models, the financial management system can verify each other based on the output prediction results, and perform iterative calculations based on the verification results to narrow the differences between different prediction results. When the differences between the results output by different sub-models meet the preset difference range, the prediction result data is output.

[0073] It can be seen that in the embodiment of the present application, the financial management system can combine multiple different prediction models through model fusion technology to improve the stability and reliability of the prediction.

[0074] S104: Generate a result preview interface according to the prediction result data.

[0075] After obtaining the forecast results output by the forecast model, the financial management system can generate a result preview interface based on the forecast result data to display the forecast result data to the user. In the result preview interface, visual forecast result display effects such as charts and reports can be provided to facilitate users to intuitively understand the trend of cash flow changes.

[0076] The generated result preview interface can also support user-defined warning thresholds. When the forecast result exceeds the warning range, the financial management system will display the data items that exceed the warning range through specific highlighting effects such as highlighting and marking in red. And it can also automatically issue an alarm to remind the user to take corresponding measures. Therefore, the result preview interface can include a view component generated based on the user-defined warning parameters and the forecast result data.

[0077] In order to generate a result preview interface, in a feasible implementation, the financial management system can obtain the component type and customized warning parameters specified by the user. The component type and automatic warning parameters can be specified based on the user interface provided by the financial management system. That is, the user can input data, select the prediction model and parameters, and select the way to view the prediction results in the user interface.

[0078] Then, the visualization tool corresponding to the component type is called, and the visualization tool is used to generate a view component according to the forecast result data. For example, the user can select the "trend chart" mode for viewing the forecast result in the user interface. After obtaining the component type specified by the user as a trend chart, the financial management system can call the trend chart coordinate drawing tool, and use time as the horizontal axis of the coordinate and cash flow as the vertical axis of the coordinate. Then, historical cash flow and forecast cash flow are added to the coordinate drawing tool in chronological order, thereby generating a view component including a trend chart effect.

[0079] After generating the view component, the financial management system can set the display effect of the view component according to the user-defined warning parameters. That is, after obtaining the customized warning parameters and the forecast result data, the forecast result data and the customized warning parameters can be compared to obtain the parameter comparison results. Then, according to the parameter comparison results, the display parameters of the view component are set. For example, the user-defined warning parameter cash flow is a maximum of 1.5 million, and the predicted cash flow output by the forecast model is 1.8 million. By comparing the predicted cash flow and the warning parameters, it can be determined that the predicted cash flow exceeds the warning parameters. Therefore, the display background color of the view component can be set to red to highlight the view component.

[0080] It can be seen from the above technical solutions that the financial data forecasting method provided in the above embodiments can obtain different types of data to be processed obtained by a variety of different acquisition methods, and call the forecasting model according to the data type contained in the data to be processed. The data to be processed is then input into the forecasting model to obtain the forecast result data output by the forecasting model. Thereby, a result preview interface is generated according to the forecast result data. The financial data forecasting method improves the accuracy of the financial data forecasting results by adopting a variety of data collection methods, combining data analysis algorithms and machine learning techniques, and establishing a result evaluation indicator system that supports customization. The obtained results can assist enterprises in better financial management and help enterprises better plan funds.

[0081] Furthermore, as a refinement and expansion of the specific implementation of the above embodiment, in order to fully illustrate the specific implementation process of this embodiment, some embodiments of the present application also provide another financial data forecasting method. The difference between this financial data forecasting method and the above embodiment is that after obtaining the data to be processed, the financial management system can also perform data preprocessing on the obtained data to be processed, that is, Figure 6 As shown, after the step of obtaining the data to be processed, the financial data forecasting method further includes:

[0082] S201. Perform data cleaning on the data to be processed based on preset data cleaning items.

[0083] Wherein, the preset data cleaning items include at least one of removing duplicate data, supplementing missing values, and correcting erroneous data. Exemplarily, when the financial management system removes duplicate data, it can first perform data evaluation, that is, evaluate the scale and content of the data to be processed, and understand the scale and type of duplicate data in the data set. Then, define the execution method of the deduplication process by defining duplicate data, calling algorithms or tools, and determining deduplication strategies. Then perform data comparison, that is, determine duplicate data and non-duplicate data by comparison. And after deleting redundant duplicate data, update the data set to update the processed data set to the data to be processed, replacing the original data set.

[0084] The financial management system can also use missing value filling methods such as mean filling, median filling or interpolation to fill in missing values ​​in the data to be processed. For example, for missing values ​​in sales revenue data, the average value of sales revenue in the months before and after the time period can be used to fill in.

[0085] When the financial management system corrects erroneous data, the financial management system can correct the data that is obviously not within the value range based on the value range of each data. The financial management system can also compare the same project data in multiple business documents and correct the erroneous data based on the comparison results.

[0086] S202, obtaining data input format information of the prediction model;

[0087] After data cleaning of the data to be processed, the financial management system can obtain the data input format information of the called prediction model. Since different prediction models support different data formats, after calling the prediction model, the data input format information can be determined based on the identification information of the sub-models contained in the prediction model. For example, after calling the first prediction sub-model for regression analysis, the financial management system can read the model ID of the first prediction sub-model, such as MD10001. Then, based on the model ID, query the model database for the data input format information supported by the first prediction sub-model, such as .txt, .xls, etc.

[0088] S203: Perform format conversion on the data to be processed after data cleaning according to the data input format information.

[0089] After obtaining the data input format information, the financial management system can also call format conversion tools such as astype to perform format conversion on the data to be processed, that is, the data to be processed in other formats can be converted into the format corresponding to the data output format information.

[0090] For the second prediction sub-model based on machine learning, the financial management system can also perform format conversion processes such as encoding categorical variables, screening features, aggregating data, and generating new features on the data to be processed after data cleaning. For example, use pd.get_dummies for one-hot encoding, or use groupby and transform methods for feature aggregation.

[0091] Through data conversion, the original format of the processed data can be converted into a format acceptable to the model, so that the financial management system can provide a unified data model. After the data is sent or captured, the data form required to meet the financial data forecasting process can be obtained, so that more data can be used to predict financial data, which is conducive to improving the accuracy and reliability of financial data forecasts.

[0092] Furthermore, as a refinement and expansion of the specific implementation of the above embodiment, in order to fully illustrate the specific implementation process of this embodiment, some embodiments of the present application also provide a financial data prediction method, which is different from the above embodiment in that it also includes a model training process, that is, Figure 7 As shown, the method also includes:

[0093] S301, obtaining sample data.

[0094] Corresponding to the data to be processed, the sample data includes first sample data, second sample data and third sample data. The first sample data is financial data with a prediction classification label, the second sample data is business data with a prediction classification label, and the third sample data is influencing factor data with a prediction classification label.

[0095] For example, a financial management system can construct a trained model including a multi-layer neural network, with historical financial data and business data as input and cash flow as output for training. The neural network of the trained model can be a model structure with an input layer, two hidden layers and an output layer. Among them, the input layer includes input nodes corresponding to characteristic variables such as sales revenue, cost expenditure, accounts receivable balance, accounts payable balance, etc. The output layer includes output nodes corresponding to cash flow. The financial management system can then obtain sample data of sales revenue, cost expenditure, accounts receivable balance, accounts payable balance, etc. from public data sets, custom data sets, online platforms, or using databases such as openml.org, wherein the sample data is marked with classification labels for different numerical cash flows.

[0096] S302: Input the sample data into the trained model to obtain training result data output by the trained model.

[0097] After obtaining the sample data, the characteristic variable content in the sample data can be input into the trained model to obtain the training result data of the trained model. For example, after obtaining the sample data, the sales revenue, cost expenditure, accounts receivable balance, and accounts payable balance in the sample data can be input into the trained model to obtain the predicted cash flow output by the trained model for the sales revenue, cost expenditure, accounts receivable balance, and accounts payable balance.

[0098] In a feasible implementation, in the step of inputting sample data into the trained model, the normalized interval of the second prediction sub-model can be obtained first. Then the sample data is mapped to the normalized interval to obtain normalized data. Then, according to a preset ratio, the normalized data is divided into a training set, a validation set, and a test set. Among them, the training set is used to train the trained model; the validation set is used to adjust the model parameters of the trained model; and the test set is used to evaluate the model performance of the trained model. Therefore, the training set can be input into the trained model.

[0099] For example, after obtaining sample data, the sample data can be preprocessed, that is, the data can be normalized and mapped to the interval [0, 1] or [-1, 1] to speed up the training of the neural network and improve the stability of the model. The ratio of the training set, validation set, and test set can be 8:1:1, that is, most of the data is used as the training set to train the neural network. A part of the data is used as the validation set to adjust the hyperparameters of the model. A part of the data is used as the test set to evaluate the performance of the model.

[0100] S303: Calculate training loss based on the predicted classification label and the training result data.

[0101] After obtaining the training result data, the training loss can be calculated by comparing the predicted classification label and the training result data. Among them, the training loss is used to characterize the difference between the predicted value and the actual value used to evaluate the model in machine learning. The specific calculation method of the training loss can be determined according to factors such as the type of training sample data and the back propagation method of the neural network model. Therefore, the loss function can be a combination of one or more of Mean Squared Error (MSE), Mean Absolute Error (MAE), Cross-Entropy Loss, Binary Cross-Entropy Loss, and Categorical Cross-Entropy Loss.

[0102] Exemplarily, when the loss function is a mean square error loss function, the mean square error can be calculated based on the following formula as the training loss:

[0103]

[0104] Among them, y i is the classification label value, is the training result value, and n is the number of samples.

[0105] S304: If the training loss is greater than a loss threshold, modify the model parameters of the trained model according to the training loss.

[0106] After obtaining the training loss based on the loss function calculation, the training loss can be compared with the preset loss threshold. If the training loss is greater than the loss threshold, it means that the current training process has not converged, so the training loss can be used for back propagation to modify the model parameters of the trained model. The modified model parameters can include the weights and biases between neurons in the neural network model. Then, by inputting the training data, the training loss is obtained and compared with the preset loss threshold, and multiple iterations of training are performed.

[0107] S305: If the training loss is less than or equal to the loss threshold, output the model parameters of the trained model to obtain the second prediction sub-model.

[0108] Through multiple iterations of training, that is, training the neural network with a large amount of historical data, and constantly adjusting the weights and biases between neurons, when the training loss is less than or equal to the loss threshold, it means that the current trained model can accurately predict future financial data, thereby obtaining model parameters that can make the model converge. At this time, the model parameters are output and applied to the second prediction sub-model of the same structure, and a second prediction sub-model with accurate output results can be obtained.

[0109] It should be noted that the above-mentioned model training process can be run independently, that is, model training and data prediction are independent of each other. When performing financial data prediction, the financial management system can obtain model parameters from the server or the client for training, and build a second prediction sub-model based on the obtained model parameters, that is, call the prediction model, and perform data prediction through the prediction model. The model training process can also be carried out continuously with the financial data prediction process, that is, after obtaining the second prediction sub-model, new training data can be built based on real-time data and user feedback, and training can be performed on the client or server running the financial management system, so that the prediction model obtained by training can better meet the personalized needs of users.

[0110] It can be seen that in the above embodiment, the financial management system can perform a comprehensive analysis of multimodal data to be processed through a neural network model based on machine learning, so as to obtain more accurate prediction results based on fully considering the impact of multiple modal data on the prediction results.

[0111] In some feasible implementations, the financial management system can also establish a result evaluation indicator system, such as mean absolute error, root mean square error, etc. The result evaluation indicator system can evaluate and verify the forecast results. And according to the evaluation results, the forecast model can be optimized and adjusted to continuously improve the accuracy of the forecast. Figure 8 As shown, the financial data forecasting method can also obtain real result data and forecast result data within a preset verification period. The real result data is of the same type as the preset result data. For example, when the forecast result data is cash flow data, the real result data is also cash flow data.

[0112] Then calculate the prediction error in the verification cycle, that is, calculate the absolute value of the difference between the real result data and the predicted result data. And calculate the mean error (MAE) and root mean square error (RMSE) based on the prediction errors in multiple verification cycles. The mean absolute error and root mean square error are used to represent the difference data between the predicted value and the actual value.

[0113] For example, the average error can be calculated as follows:

[0114]

[0115] Among them, y i is the true value of the i-th sample, is the predicted value of the i-th sample, and n is the number of samples.

[0116] Similarly, the root mean square error can be calculated as follows:

[0117]

[0118] Among them, y i is the true value of the i-th sample, is the predicted value of the i-th sample, and n is the number of samples.

[0119] For example, in a cash flow forecasting model, mean absolute error and root mean square error are used to characterize the difference between the predicted cash flow value and the actual cash flow value. For a company's cash flow forecast for a specific period of time, the forecast value may be 1 million yuan of cash inflow for the next month calculated based on the model, while the actual value is 1.1 million yuan of cash inflow after the end of the next month. The difference between these two values ​​is the error data. MAE and RMSE can be calculated based on the predicted value and the actual value to determine the difference between the predicted value and the actual value.

[0120] If the average error and the root mean square error are both within the allowable error range, the forecast result data is output. For example, suppose an enterprise uses a cash flow forecasting model to forecast cash flow for the next three months. In the first month, the model predicts a cash inflow of 800,000 yuan, and the actual cash inflow is 780,000 yuan; in the second month, the model predicts a cash inflow of 900,000 yuan, and the actual cash inflow is 920,000 yuan; in the third month, the model predicts a cash inflow of 1 million yuan, and the actual cash inflow is 1.05 million yuan.

[0121] The mean absolute error (MAE) can be calculated based on the three-month forecast and true values. The absolute error for each month can be calculated first, which is |80-78|=20,000 yuan, |90-92|=20,000 yuan, and |100-105|=50,000 yuan. Then the mean absolute error can be calculated, MAE=(2+2+5) / 3=30,000 yuan.

[0122] At the same time, the root mean square error (RMSE) is calculated based on the predicted values ​​and true values ​​for three months. The square of the error for each month can be calculated first, which is (80-78) 2 =4, (90-92) 2 =4, (100-105) 2 =25. Then calculate the average square error, (4+4+25) / 3=11. Finally calculate the root mean square error, Ten thousand yuan.

[0123] By calculating the MAE and RMSE values, we can see that the average difference between the predicted value and the actual value of the model is about 30,000 yuan, and the root mean square difference is about 33,200 yuan. If this error is within the acceptable range for the enterprise, then the model can be considered to have a certain degree of accuracy and practicality.

[0124] Similarly, if the average error or the root mean square error is not within the allowable error range, the input data or model needs to be adjusted, that is, the model parameters of the prediction model are modified, the data type in the data to be processed is modified, and at least one of the prediction models is re-called. Therefore, by comparing the calculated MAE value and RMSE value with the preset allowable error range, if it is determined that the error is large, it is possible to consider adjusting the model parameters, improving the data input, or trying other model structures to improve the accuracy of the prediction.

[0125] Furthermore, as a specific implementation of the above-mentioned financial data prediction method, some embodiments of the present application also provide a financial data prediction device, such as Fig. 9 As shown, the device comprises:

[0126] A data acquisition module is used to acquire data to be processed, wherein the data to be processed includes one or more combinations of first data, second data and third data, wherein the first data is financial data obtained by querying a preset financial database; the second data is business data obtained through interface monitoring; and the third data is influencing factor data obtained through data capture;

[0127] A model calling module, used to read the data type contained in the data to be processed, and to call the prediction model according to the data type, wherein the prediction model includes at least one first prediction sub-model and a second prediction sub-model; the types and quantities of the first prediction sub-model and the second prediction sub-model contained in the prediction model are associated with the data type; the first prediction sub-model has a built-in data analysis algorithm function; and the second prediction sub-model is a machine learning model obtained by training based on sample data;

[0128] A prediction module, used for inputting the data to be processed into the prediction model to obtain prediction result data output by the prediction model;

[0129] A display module is used to generate a result preview interface according to the prediction result data, wherein the result preview interface includes a view component generated based on user-defined warning parameters and the prediction result data.

[0130] It should be noted that for other corresponding descriptions of the functional units involved in a financial data prediction device provided in an embodiment of the present application, reference can be made to the corresponding descriptions in the financial data prediction method described in the above embodiment, and will not be repeated here.

[0131] In the above embodiment, the financial data prediction device may include a data acquisition module, a model calling module, a prediction module and a display module. Among them, the data acquisition module can be used to obtain multimodal data to be processed, and the model calling module can be used to call an applicable prediction model according to the type of data contained in the data to be processed. The prediction module can be used to perform comprehensive analysis and prediction based on the multimodal data to be processed in the data analysis algorithm and the machine learning algorithm to obtain more accurate prediction result data, improve accuracy and reliability. The display module can be used to provide a concise and clear user interface, so that users can easily input data, select prediction models and parameters, and view prediction results, further improving the problem of low accuracy in the display process of financial data prediction results.

[0132] like Fig.10As shown, the embodiment of the present application also provides a computer device, which can be specifically a personal computer, a server, a network device, etc. The computer device includes a bus, a processor, a memory and a communication interface, and can also include an input and output interface and a display device. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store location information. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, the steps in each method embodiment are implemented.

[0133] Those skilled in the art will appreciate that the structure of the above-mentioned computer device is only a partial structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components, or combine certain components, or have a different arrangement of components.

[0134] In one embodiment, a computer-readable storage medium is provided. The computer-readable storage medium may be non-volatile or volatile, and stores a computer program thereon. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments are implemented.

[0135] In one embodiment, a computer program product is provided, including a computer program, which implements the steps in the above method embodiments when executed by a processor.

[0136] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties.

[0137] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but are not limited to this.

[0138] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0139] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A financial data forecasting method, characterized in that: The method comprises: Acquire data to be processed, wherein the data to be processed includes one or more combinations of first data, second data, and third data, wherein the first data is financial data obtained by querying a preset financial database; the second data is business data obtained through interface monitoring; and the third data is influencing factor data obtained through data capture; Read the data type contained in the data to be processed, and call the prediction model according to the data type, wherein the prediction model includes at least one first prediction sub-model and a second prediction sub-model; the type and quantity of the first prediction sub-model and the second prediction sub-model contained in the prediction model are associated with the data type; the first prediction sub-model has a built-in data analysis algorithm function; and the second prediction sub-model is a machine learning model obtained by training based on sample data; Inputting the data to be processed into the prediction model to obtain prediction result data output by the prediction model; A result preview interface is generated according to the prediction result data, and the result preview interface includes a view component generated based on the user-defined warning parameters and the prediction result data.

2. The method according to claim 1, characterized in that The method further comprises: Performing data cleaning on the data to be processed based on preset data cleaning items, wherein the preset data cleaning items include at least one of removing duplicate data, supplementing missing values, and correcting erroneous data; Obtaining data input format information of the prediction model; According to the data input format information, format conversion is performed on the data to be processed after data cleaning.

3. The method according to claim 1, characterized in that: The method further comprises: Acquire sample data, the sample data including first sample data, second sample data and third sample data, the first sample data is financial data with a prediction classification label; the second sample data is business data with a prediction classification label; the third sample data is influencing factor data with a prediction classification label; Inputting the sample data into a trained model to obtain training result data output by the trained model; Calculate training loss based on the predicted classification label and the training result data; If the training loss is greater than a loss threshold, modifying model parameters of the trained model according to the training loss; If the training loss is less than or equal to the loss threshold, the model parameters of the trained model are output to obtain the second prediction sub-model.

4. The method according to claim 3, characterized in that Inputting the sample data into the trained model includes: Obtaining a normalized interval of the second prediction sub-model; Mapping the sample data to the normalization interval to obtain normalized data; According to a preset ratio, the normalized data is divided into a training set, a validation set and a test set, wherein the training set is used to train the trained model; the validation set is used to adjust the model parameters of the trained model; and the test set is used to evaluate the model performance of the trained model; The training set is input into the trained model.

5. The method according to claim 1, characterized in that Calling a prediction model according to the data type includes: Extracting historical data from the data to be processed, wherein the historical data is numerical data with time series recorded at preset time intervals; Sorting the historical data in chronological order to obtain a time series data set; Performing smoothing processing on the time series data set to obtain a smoothed data sequence; A first prediction sub-model based on a time series analysis algorithm is called according to the smoothed data sequence, and the first prediction sub-model is used to generate prediction result data according to the smoothed data sequence.

6. The method according to claim 1, characterized in that Inputting the data to be processed into a prediction model to obtain prediction result data output by the prediction model includes: Acquire real result data and predicted result data within a preset verification period, wherein the real result data is of the same type as the preset result data; Calculating the prediction error within the verification period, the prediction error being the absolute value of the difference between the actual result data and the predicted result data; Calculating an average error and a root mean square error based on the prediction errors over a plurality of validation cycles; If the average error and the root mean square error are both within the allowable error range, output the prediction result data; If the mean error or the root mean square error is not within the allowable error interval, the model parameters of the prediction model are modified, and / or the data type in the data to be processed is modified, and / or the prediction model is re-called.

7. The method according to claim 1, characterized in that Generating a result preview interface according to the prediction result data includes: Get the component type and custom warning parameters specified by the user; Calling a visualization tool corresponding to the component type, and using the visualization tool to generate a view component according to the prediction result data; Comparing the prediction result data with the customized warning parameters to obtain a parameter comparison result; According to the parameter comparison result, the display parameters of the view component are set.

8. A financial data forecasting device, characterized in that: The device comprises: A data acquisition module is used to acquire data to be processed, wherein the data to be processed includes one or more combinations of first data, second data and third data, wherein the first data is financial data obtained by querying a preset financial database; the second data is business data obtained through interface monitoring; and the third data is influencing factor data obtained through data capture; A model calling module, used to read the data type contained in the data to be processed, and to call the prediction model according to the data type, wherein the prediction model includes at least one first prediction sub-model and a second prediction sub-model; the types and quantities of the first prediction sub-model and the second prediction sub-model contained in the prediction model are associated with the data type; the first prediction sub-model has a built-in data analysis algorithm function; and the second prediction sub-model is a machine learning model obtained by training based on sample data; A prediction module, used for inputting the data to be processed into the prediction model to obtain prediction result data output by the prediction model; A display module is used to generate a result preview interface according to the prediction result data, wherein the result preview interface includes a view component generated based on user-defined warning parameters and the prediction result data.

9. A computer device comprising a storage medium, a processor, and a computer program stored in the storage medium and executable on the processor, characterized in that: When the processor executes the computer program, the method according to any one of claims 1 to 7 is implemented.

10. A storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images