The Telecommunications Operating Income An Alysis And Forecast Based On Data Mining

In r ecent years, the reform of the telec ommunicat ions industry is eve r-changing,incr easingly fierce competition betwee n the various oper ators. Fa cing the customers, how to provide bet ter qual ity services, inc rease t he number of use rs,ac hieve revenue growth, is pl aced in front of opera tors.The income forecast for operators can help them predict fut ure income.Tel ecommunica tions enterpri ses operating income predicted results is to calculate the invest ment be nefit, the pr emise of enterpri se profit, but also for the enterpr ise invest ment decision, t he network construction planning to provide the reference.Pr ediction method of forecast is directly related to the realizat ion of the goal and the ac curacy of the pre dict ion results. Operators by predicted results can be set i n advance the future marke ting s trategy for a period of time, therefore, for the forecast of telec om bus iness income is of great applicat ion val ue.This pape r adopts the t echnology of data mining, mining inner development regularity of telecom busine ss income dat a and the main related factors, three kinds of prediction model, BP neural network, RBF neural network a nd genetic neural ne twor k model, us ing the existing data net work, training with the trained network weights and threshold prediction data, a nd effect of usi ng the actual data to verify, by comparing the predictive results of t he three models to choose t he most appropriate method.The expe riment al r esults show t hat the genet ic algori thm has good global search abilit y, ca n from the sense of probabi lity in a random way to seek the optimal solution to the problem, but the application of genetic algori thm prone t o pr emature phenomenon, local optimization abili ty is poor, and t he combi nation of the genetic neural network can play their respective advantages, to solve the initial weights of neural net work the randomne ss of t he network impact, as well as the traditional neural network easy t o fall into local solution of t he problem, to ma ke more accurate predictions.