CN-122022005-A - Complaint quantity prediction method, complaint quantity prediction device, complaint quantity prediction equipment, storage medium and product

CN122022005ACN 122022005 ACN122022005 ACN 122022005ACN-122022005-A

Abstract

The application discloses a complaint quantity prediction method, a device, equipment, a storage medium and a product, which belong to the field of data processing and are used for improving the accuracy of predicted complaint quantity. The method comprises the steps of obtaining historical complaint time sequence data, wherein the historical complaint time sequence data comprises historical complaint data of a plurality of continuous time points, the historical complaint data comprises complaint time and complaint quantity, extracting features of the historical complaint time sequence data to obtain time sequence feature vectors, wherein the time sequence feature vectors are feature vectors of the historical complaint time sequence data, performing frequency domain conversion operation on the time sequence feature vectors, determining time sequence periods and periodic feature vectors based on frequency domain conversion results, wherein the time sequence periods are periods included in the time sequence feature vectors, the periodic feature vectors are feature vectors of the time sequence feature vectors in the time sequence periods, and inputting the periodic feature vectors into a pre-trained complaint quantity prediction model to obtain complaint quantity prediction results.

Inventors

YOU WENTING
ZHU XINYANG
WANG CHUANG
YANG CHAO
WANG YUTAO
BAO YUANYUAN
WANG YONGFANG
WANG LEI
LI DAWEI
DONG XIAOLI
ZHOU YAKUN
WU ZHIJIAN

Assignees

中移在线服务有限公司
中国移动通信集团有限公司

Dates

Publication Date: 20260512
Application Date: 20251225

Claims (10)

1. A method of complaint volume prediction, the method comprising: acquiring historical complaint time sequence data, wherein the historical complaint time sequence data comprises historical complaint data of a plurality of continuous time points, and the historical complaint data comprises complaint time and complaint; extracting features of the historical complaint time sequence data to obtain time sequence feature vectors, wherein the time sequence feature vectors are feature vectors of the historical complaint time sequence data; performing frequency domain conversion operation on the time sequence feature vector, and determining a time sequence period and a period feature vector based on a frequency domain conversion result, wherein the time sequence period is a period included by the time sequence feature vector, and the period feature vector is a feature vector of the time sequence feature vector in the time sequence period; and inputting the periodic feature vector into a pre-trained complaint quantity prediction model to obtain a complaint quantity prediction result.
2. The method of claim 1, wherein the feature extracting the historical complaint volume time series data to obtain a time series feature vector comprises: Respectively carrying out normalization processing on the complaint quantity and the complaint time in the historical complaint quantity data to obtain a first time sequence feature and a second time sequence feature; Performing position coding processing on the historical complaint volume data to obtain a third time sequence characteristic; the timing feature vector is determined based on the first timing feature, the second timing feature, and the third timing feature.
3. The method of claim 2, wherein the determining the timing feature vector based on the first timing feature, the second timing feature, and the third timing feature comprises: adding the first time sequence feature, the second time sequence feature and the third time sequence feature to obtain a fourth time sequence feature, wherein the fourth time sequence feature is used for representing the original data feature of the historical complaint volume data; inputting the fourth time sequence characteristic into a preset self-attention network to obtain a fifth time sequence characteristic, wherein the fifth time sequence characteristic is used for representing long-distance data characteristics of the historical complaint volume data; The timing feature vector is determined based on the fourth timing feature and the fifth timing feature.
4. The method of claim 1, wherein performing a frequency domain conversion operation on the timing feature vector and determining a timing period and a period feature vector based on a frequency domain conversion result comprises: performing frequency domain conversion operation on the time sequence feature vector to obtain time sequence frequency features, wherein the time sequence frequency features are frequency features of the time sequence feature vector; Performing cycle identification operation on the time sequence frequency characteristics to determine a plurality of time sequence cycles; Extracting features of the historical complaint volume data in each time sequence period to obtain complaint volume data features; The periodic feature vector is determined based on the complaint volume data feature and the timing feature vector.
5. The method of claim 4, wherein the determining the periodic feature vector based on the complaint volume data feature and the timing feature vector comprises: scoring each time sequence period based on a preset period scoring strategy to obtain a period score; Weighting the complaint volume data characteristics in each time sequence period based on the period scores to obtain complaint volume weighting characteristics; the periodic feature vector is determined based on the complaint volume weighted feature and the timing feature vector.
6. The method of claim 1, wherein prior to said inputting the periodic feature vector into a pre-trained complaint volume prediction model to obtain a complaint volume prediction result, the method further comprises: Determining a plurality of pieces of historical complaint amount data at preset positions in the historical complaint amount time sequence data as sample complaint amount time sequence data; extracting features of the sample complaint amount time sequence data to obtain sample feature vectors, wherein the sample feature vectors are feature vectors of the sample complaint amount time sequence data; inputting the periodic feature vector and the sample feature vector into the complaint volume prediction model to be trained to obtain a first prediction result, wherein the complaint volume prediction model to be trained consists of a self-attention network; Determining a loss result based on the first prediction result and a first real result through a preset loss function, wherein the first real result is real complaint volume data corresponding to the first prediction result; And iteratively updating the complaint quantity prediction model to be trained based on the loss result to obtain the pre-trained complaint quantity prediction model.
7. A complaint volume predicting apparatus, the apparatus comprising: The first acquisition module is used for acquiring historical complaint time sequence data, wherein the historical complaint time sequence data comprises historical complaint data of a plurality of continuous time points, and the historical complaint data comprises complaint time and complaint quantity; the first extraction module is used for carrying out feature extraction on the historical complaint time sequence data to obtain time sequence feature vectors, wherein the time sequence feature vectors are feature vectors of the historical complaint time sequence data; The first determining module is used for performing frequency domain conversion operation on the time sequence feature vector, determining a time sequence period and a period feature vector based on a frequency domain conversion result, wherein the time sequence period is a period included by the time sequence feature vector, and the period feature vector is a feature vector of the time sequence feature vector in the time sequence period; The first prediction module is used for inputting the periodic feature vector into a pre-trained complaint quantity prediction model to obtain a complaint quantity prediction result.
8. An electronic device, the device comprising: Processor, and A memory arranged to store computer executable instructions configured for execution by the processor, the executable instructions comprising steps for performing the complaint volume prediction method of any one of claims 1 to 6.
9. A storage medium storing computer-executable instructions for causing a computer to perform the complaint volume prediction method of any one of claims 1 to 6.
10. A computer program product comprising a computer program which, when executed by a processor, implements the complaint volume prediction method of any one of claims 1 to 6.

Description

Complaint quantity prediction method, complaint quantity prediction device, complaint quantity prediction equipment, storage medium and product Technical Field The application belongs to the field of data processing, and particularly relates to a complaint quantity prediction method, a complaint quantity prediction device, complaint quantity prediction equipment, a storage medium and a complaint quantity prediction product. Background In recent years, along with the development of artificial intelligence technology, complaint volume prediction is widely applied in the fields of customer service optimization, resource allocation and the like. The complaint quantity prediction is mainly to predict the complaint quantity of a period of time in the future in advance by analyzing the historical data, so as to provide basis for enterprise decision. At present, most complaint volume predictions adopt traditional statistical analysis methods, such as a time series algorithm like a moving average method and an ARIMA model, and regression analysis, and when complaint volume data is processed, the periodicity and the trend of the data can be captured, and the influence of a single factor is quantized, but when the complaint volume data is subjected to nonlinear and complex complaint data, the accuracy of complaint volume prediction is lower. Therefore, a method capable of improving the accuracy of predicting the amount of complaints is required. Disclosure of Invention The embodiment of the application provides a complaint quantity prediction method which can improve the accuracy of the predicted complaint quantity. In a first aspect, an embodiment of the present application provides a complaint volume prediction method, which includes obtaining historical complaint volume time sequence data, where the historical complaint volume time sequence data includes historical complaint volume data of a plurality of continuous time points, the historical complaint volume data includes complaint time and complaint volume, performing feature extraction on the historical complaint volume time sequence data to obtain a time sequence feature vector, where the time sequence feature vector is a feature vector of the historical complaint volume time sequence data, performing frequency domain conversion operation on the time sequence feature vector, and determining a time sequence period and a period feature vector based on a frequency domain conversion result, where the time sequence period is a period included in the time sequence feature vector, and the period feature vector is a feature vector of the time sequence feature vector in the time sequence period, and inputting the period feature vector to a pre-trained complaint volume prediction model to obtain a complaint volume prediction result. In a second aspect, the embodiment of the application provides a complaint amount prediction device, which comprises a first acquisition module, a first extraction module and a first prediction module, wherein the first acquisition module is used for acquiring historical complaint amount time sequence data, the historical complaint amount time sequence data comprises historical complaint amount data of a plurality of continuous time points, the historical complaint amount data comprises complaint time and complaint amount, the first extraction module is used for carrying out feature extraction on the historical complaint amount time sequence data to obtain a time sequence feature vector, the time sequence feature vector is a feature vector of the historical complaint amount time sequence data, the first determination module is used for carrying out frequency domain conversion operation on the time sequence feature vector and determining a time sequence period and a period feature vector based on a frequency domain conversion result, the time sequence period is a period included by the time sequence feature vector, and the period feature vector is a feature vector of the time sequence feature vector in the time sequence period, and the first prediction module is used for inputting the period feature vector into a pre-trained complaint amount prediction model to obtain a complaint amount prediction result. In a third aspect, an embodiment of the present application provides an electronic device, including a processor, a memory, and a program or instruction stored on the memory and executable on the processor, the program or instruction implementing the steps of the method according to the first aspect when executed by the processor. In a fourth aspect, embodiments of the present application provide a readable storage medium having stored thereon a program or instructions which when executed by a processor perform the steps of the method according to the first aspect. In a fifth aspect, embodiments of the present application provide a computer program product which, when executed by a processor, implements the steps of the method according to th