Service feature based communication service forecasting method

Abstract

The invention relates to the technical field of communication and discloses a service feature based communication service forecasting method. The method includes following steps: S1, preprocessing service volume data to be forecast; S2, subjecting the service volume data after being preprocessed to sample construction, sample generalization and sample screening; S3, clustering the service volume data after being processed in the S2 according to service feature of samples to obtain a forecasting model of the service volume data with multiple area types; S4, performing service forecasting to the service volume data of each area type, and estimating network areas where the service volume data are positioned; S5, estimating network resources needed in the network areas; and S6, comparing the estimated network resources with existing network resources in the network areas, and performing network plan optimization according to comparison results. By the service feature based communication service forecasting method, sudden factors influencing the service volume data are smoothed, and influences, of smoothing, on forecasting results are eliminated after forecasting, so that forecasting accuracy is improved.

Description

technical field [0001] The invention relates to the technical field of communication, in particular to a communication service prediction method based on service characteristics. Background technique [0002] At present, the wireless communication service prediction technology adopted by most operators is limited to the fitting prediction of simple functions, and the input parameters and prediction models are relatively simple. In fact, it is a rough estimate of a trend. It plays a certain positive role in the rapid construction of the network. Today, with changes in user behavior and network scale and structure, the original simple prediction technology is no longer suitable. [0003] To sum up, the current main wireless communication service forecasting methods include inertial forecasting, Kalman filtering, and traffic OLAP (On-Line Analytical Processing, Online Analytical Processing) analysis. Among them, inertial prediction and Kalman filtering are relatively simple, ...

AI Technical Summary

Problems solved by technology

[0010] 2. Current service forecasting attempts to predict the traffic volume of a single site; however, due to the mobility characteristics of users in cellular systems, it is unreasonable to perform service forecasting for a single site, and it is difficult to achieve a stable result

[0011] 3. Existing business forecasts do not use coverage blind spots as potential traffic generation spaces to jointly generate service forecast results, making it difficult to effectively use service forecast results to guide network evaluation planning and optimization

[0012] 4. Existing forecasting methods hope to be able to use historical data to predict business volume for a long time in the future. Therefore, if the impact of external sudden factors and changes in business volume trends during this period are not captured, it is difficult to accurately Prediction

[0013] 5. The existing forecasting scheme expects to continuously adjust the parameters of the forecasting model according to the input data, thus introducing a large amount of forecasting model training overhead and delay

However, the research on business characteristics and the construction of business samples are rarely mentioned or not deep enough

However, the sample characteristics of the business are the key factors affecting the performance of the prediction algorithm. It can be said that without good samples, there will be no good prediction samples, and it is difficult to design a good prediction algorithm.

Images