Design And Implementation Of Prediction Of Potential Paying Users In Smart TV

With the popularity of smart TVs,user payment has become one of the most important sources of profit for TV manufacturers or video content companies Digging potential paying users and prompting users to pay for value-added services has become an urgent issue for enterprises.Enterprise personnel need to understand the user’s payment situation,user’s daily behavior habits,etc.,and at the same time promote the potential paying users in a timely manner to promote more users to pay,need to develop a smart TV potential paying user prediction system to provide system support for this businessOne of the key points in the system for predicting potential paying users for smart TVs is to establish an accurate forecasting model for potential paying users.This thesis first proposes the expansion of log information,feature derivation,and features based on the characteristics of Hisense’s log data,one of the largest domestic TV manufacturers.Extracted solution.Combining the advantages of deep models in high-order abstract feature learning and the advantages of linear models in low-order feature learning,the SWD model is used for training prediction.In order to verify the performance of the SWD model,a comparison was made with the traditional classification model.The experimental results show that the SWD model is better than the traditional classification model in the F1 measure in the prediction of potential paying users of smart TVs,and the F1 value can reach 0.8401.After further analysis of the system development background and current research status at home and abroad,the functions and services of each module are determined through a needs analysis,and Java Web development technologies such as B/S architecture,front-end and back-end separation modes,and Spring MVC are used to complete potential paying users for smart TVs.The background management part of the forecasting system.At the same time,the Spark big data technology is used for statistical analysis of data to provide data preparation for the system.The background management system is divided into five functional modules,including the user’s consumption module today,the user’s historical consumption module,the user behavior analysis module,the potential user prediction module,and the user activity promotion module.Function module,user activity promotion module contains promotion activity and activity setting sub-function module.The system provides real-time data,statistical analysis data,and model prediction data to provide a window for enterprises to quickly and intuitively understand all aspects related to user payments,which can help guide companies to make correct business decisions;and companies can easily make Potential users pay for promotional activities to increase membership purchase conversion rates and thereby increase corporate profits.