Search

CN-122022066-A - Index early warning information generation method and device based on time sequence prediction model

CN122022066ACN 122022066 ACN122022066 ACN 122022066ACN-122022066-A

Abstract

The application relates to the technical field of data analysis and discloses an index early warning information generation method and device based on a time sequence prediction model, wherein the method comprises the steps of extracting trend characteristics, period characteristics and event characteristics of multi-source service data; generating a predicted business index value through a time sequence prediction model, a trend feature, a period feature and an event feature, and generating target index early warning information according to the predicted business index value and an actual business index value. By means of the method, the multi-source business data are fused, and the core features are extracted to provide information input for the time sequence prediction model. The predicted value is compared with the actual value, and a potential risk trend is identified before the index actually goes wrong, so that the defect of lag reaction or high false alarm rate in the traditional means is overcome. The application can be applied to the business fields of finance science and technology, medical health care and the like, and improves the index early warning accuracy of the data early warning system aiming at multi-source heterogeneous data.

Inventors

  • ZHANG SHENGRONG

Assignees

  • 中国平安财产保险股份有限公司

Dates

Publication Date
20260512
Application Date
20260306

Claims (10)

  1. 1. The index early warning information generation method based on the time sequence prediction model is characterized by comprising the following steps of: collecting multi-source business data, and extracting trend characteristics, period characteristics and event characteristics of the multi-source business data; Generating a predicted business index value of a future preset time period through a preset time sequence prediction model, the trend characteristic, the period characteristic and the event characteristic; and acquiring an actual business index value, and generating target index early warning information according to the predicted business index value and the actual business index value.
  2. 2. The method for generating indicator warning information based on a time series prediction model according to claim 1, wherein the generating a predicted traffic indicator value for a future preset time period by presetting the time series prediction model, the trend feature, the period feature, and the event feature includes: Generating an initial predicted component according to the preset time sequence prediction model, the trend characteristic, the periodic characteristic and the event characteristic; And carrying out weighted fusion on each initial prediction component according to preset weights to generate the prediction business index value.
  3. 3. The method for generating indicator warning information based on a time series prediction model according to claim 2, wherein the preset time series prediction model includes a trend feature channel, a period feature channel, and an event feature channel, the initial prediction component includes a trend prediction component, a period prediction component, and an event influence prediction component, and the generating an initial prediction component according to the preset time series prediction model, according to the trend feature, the period feature, and the event feature includes: fitting the trend characteristics in the trend characteristic channel through a linear regression algorithm to generate the trend prediction component; reconstructing the periodic characteristics in the periodic characteristic channel through a harmonic synthesis algorithm to generate the periodic prediction component; And in the event feature channel, calculating an influence value of the event feature on the preset time period in the future through a time-varying impact response function, and generating the event influence component.
  4. 4. The method for generating indicator early-warning information based on a time-series prediction model according to claim 3, wherein before the step of weighting and fusing each initial prediction component according to a preset weight to generate the predicted traffic indicator value, the method comprises the steps of: calculating a weighted percentage error of each initial predicted component in a history period according to a preset sliding time window, and normalizing the reciprocal of the weighted percentage error to generate a basic weight coefficient; Determining a service scene weight adjustment factor according to the service type of the multi-source service data; and adjusting the basic weight coefficient according to the service scene weight adjustment factor to generate the preset weight.
  5. 5. The method for generating the indicator early-warning information based on the time-series prediction model according to claim 4, wherein the step of weighting and fusing the initial prediction components according to preset weights to generate the predicted traffic indicator value comprises the following steps: Calculating the product between each initial prediction component and the preset weight to be used as a prediction index value corresponding to each initial prediction classification; And summing the prediction index values to generate the prediction business index value.
  6. 6. The method for generating the target indicator early-warning information based on the time sequence prediction model according to claim 1, wherein the target indicator early-warning information includes primary indicator early-warning information, secondary indicator early-warning information and tertiary indicator early-warning information, the step of acquiring an actual business indicator value and generating the target indicator early-warning information according to the predicted business indicator value and the actual business indicator value includes: Subscribing a system message queue of the multi-source service data through a data stream processing engine to acquire the actual service index value; Calculating absolute deviation values and relative deviation rates of the actual service index values and the predicted service index values; Generating the first-level index early warning information under the condition that the absolute deviation value is larger than or equal to a preset deviation value threshold value and the relative deviation rate is larger than or equal to a preset deviation rate threshold value; Generating the secondary index early warning information under the condition that the absolute deviation value is greater than or equal to the preset deviation value threshold value or the relative deviation rate is greater than or equal to the preset deviation rate threshold value; and generating the three-level index early warning information under the condition that the absolute deviation value is smaller than the preset deviation value threshold and the relative deviation rate is smaller than the preset deviation rate threshold.
  7. 7. The method for generating indicator warning information based on a time series prediction model according to any one of claims 1 to 6, wherein the collecting multi-source service data and extracting trend features, period features and event features of the multi-source service data includes: window segmentation is carried out on the multi-source business data, and local trend slopes are extracted in each window through linear regression fitting to serve as trend features; Identifying periodic frequency components of the multi-source service data through a wavelet transformation algorithm, and extracting seasonal fluctuation features, month period features and holiday effect features of the multi-source service data as the periodic features; And constructing an event feature template through a preset service knowledge base, extracting a special event from the multi-source service data, and generating the event feature according to the event feature template and the special event.
  8. 8. An index early warning information generating device based on a time sequence prediction model is characterized by comprising: The characteristic extraction module is used for monitoring the acquisition module, collecting multi-source service data and extracting trend characteristics, period characteristics and event characteristics of the multi-source service data; The business index value prediction module is used for generating a predicted business index value of a future preset time period through a preset time sequence prediction model, the trend characteristic, the period characteristic and the event characteristic; the target index early warning information generation module is used for acquiring the actual business index value and generating target index early warning information according to the predicted business index value and the actual business index value.
  9. 9. A computer device, the computer device comprising a memory and a processor; the memory is used for storing a computer program; The processor is configured to execute the computer program and implement the index early-warning information generation method based on a time-series prediction model according to any one of claims 1 to 7 when the computer program is executed.
  10. 10. A computer-readable storage medium, wherein the computer-readable storage medium stores a computer program which, when executed by a processor, causes the processor to implement the index early-warning information generation method based on a time-series prediction model as claimed in any one of claims 1 to 7.

Description

Index early warning information generation method and device based on time sequence prediction model Technical Field The application relates to the technical field of data analysis, in particular to an index early warning information generation method, device, equipment and medium based on a time sequence prediction model. Background With the deep evolution of digital economy and aging society, the fields of finance science and technology and medical health care are facing data-driven intelligent wind control and accurate service transformation pressure. The traditional technical architecture has obvious common defects in the aspects of cross-field risk early warning and resource coordination, and serious data island phenomenon exists in the fields of financial science and technology and medical health care business. In the financial and scientific business, the data standards of the internal underwriting, claim settlement and financial system of enterprises are different, the fraud cases of the external industries and the macroscopic economic indexes are difficult to be effectively integrated, and the medium and small companies especially lack a unified data management system. In the innovation of finance and technology, customer behavior data, credit investigation data and transaction flow among banks, securities and payment institutions are split due to the fact that compliance barriers and interface standards are missing, and the input dimension of an air control model is limited. The medical health care business field is more complicated, the electronic health files are distributed in hospitals, communities and care institutions, and physiological time sequence data (heart rate, blood sugar and activity) collected by the wearable equipment are in heterogeneous conflict with clinical diagnosis and treatment records, medical insurance settlement data and care logs in structure, frequency and semantics, so that the data integrity, consistency and credibility assessment mechanism is deficient. The traditional technical scheme can only solve the problem that the surface layer formats are aligned, and causal association and time sequence dislocation between cross-source data cannot be identified, so that systematic deviation exists in an early warning model input layer. Therefore, in the business fields of finance science and technology, medical health care and the like, how to improve the index early warning accuracy of the data early warning system aiming at the multi-source heterogeneous data becomes the technical problem to be solved. Disclosure of Invention The application provides an index early warning information generation method, device, equipment and medium based on a time sequence prediction model, so as to improve the index early warning accuracy of a data early warning system aiming at multi-source heterogeneous data. In a first aspect, the present application provides a method for generating indicator early warning information based on a time sequence prediction model, where the method includes: collecting multi-source business data, and extracting trend characteristics, period characteristics and event characteristics of the multi-source business data; Generating a predicted business index value of a future preset time period through a preset time sequence prediction model, the trend characteristic, the period characteristic and the event characteristic; and acquiring an actual business index value, and generating target index early warning information according to the predicted business index value and the actual business index value. In a second aspect, the present application further provides an indicator early-warning information generating device based on a time sequence prediction model, where the device includes: The characteristic extraction module is used for monitoring the acquisition module, collecting multi-source service data and extracting trend characteristics, period characteristics and event characteristics of the multi-source service data; The business index value prediction module is used for generating a predicted business index value of a future preset time period through a preset time sequence prediction model, the trend characteristic, the period characteristic and the event characteristic; the target index early warning information generation module is used for acquiring the actual business index value and generating target index early warning information according to the predicted business index value and the actual business index value. In a third aspect, the application further provides a computer device, which comprises a memory and a processor, wherein the memory is used for storing a computer program, and the processor is used for executing the computer program and realizing the index early warning information generation method based on the time sequence prediction model when executing the computer program. In a fourth aspect, the present application also provide