Terminal Contract Optimization Based On Long Short-term Memory Networks

In recent years,China Unicom's terminal contract business has established a very comprehensive strategic cooperation plan with many companies.The terminal contract mainly refers to a consumer behavior that the user chooses a certain amount of package service through the operator.The terminal contract plan is a measure of China Unicom operators.China Unicom is a marketing plan proposed by China Unicom to increase monthly sales and achieve profitability for the company,allowing users to stay online for a long time and continue to renew their contracts.This topic mainly studies the monthly billing expenses of terminal contract users in Zhengzhou City of Henan Unicom,and applies some related technologies and algorithms such as big data to analyze and model the contract billing expenses.According to the forecast results,it further proposes corresponding measures for this terminal contract of China Unicom and optimizes the corresponding package business.The following is the process of modeling contract billing expenses:Data extraction,statistical analysis and visual operation of user terminal contracts.First,extract the contract data that meets the conditions according to the database language,and then apply the relevant algorithm to preprocess the contract user data,then do a general statistical analysis of the data,and analyze and visualize each attribute through the distribution of histograms,pie charts,etc.Display;use Arc Gis to edit the map to display the distribution of geographic information data sets of different attributes in the contract data,and use this method to observe the statistics of contract package selection in different regions,monthly billing fees,and monthly billing average costs In the end,the relationship between the monthly billing fee of contract users and other attributes is used to obtain the correlation change law with gray correlation and principal component analysis.Establish a neural network model and predict the billing expenses of terminal contract users.Mainly choose BP,RNN and LSTM network models in neural network.Before modeling,first introduce the algorithm and modeling process of each network,and then continuously adjust and optimize the parameters of the prediction model,including the number of neural network layers,the number of iterations epoch,batch size,and the difference in optimization algorithms in the model.select.Finally,the prediction effect map is used to observe the degree of fit between the predicted value and the true value,and the evaluation indicators of these three neural network models are calculated at the same time.After comparative analysis,it can be seen that the error based on the LSTM network model is the smallest,the degree of fit is the highest,and the prediction accuracy is also The highest,so the prediction effect of the model is the best,so the LSTM neural network is used to predict the terminal contract billing fee.The final prediction results are further studied,and the model predictions are combined with the actual operating conditions of the company for analysis,and an optimized terminal contract package plan and a 5G converged terminal contract package plan are proposed.