Research On The Prediction Method Of Game Information Service Adoption Based On Independent Subspace

With the rapid development of the Internet,the information service industry also develops rapidly.There are increasingly companies providing information services,the competition in the information service industry is becoming increasingly fierce,and the market environment is becoming more and more severe.To be successful,companies must not only develop information services that can meet user needs,but also accurately predict the adoption of information services.The adoption of information services is an intuitive response to its commercial success.Predicting the adoption of information services and exploring the factors affecting users' adoption of information services can guide information service developers to formulate R&D plans and marketing plans,as well as follow-up information services.Provide reference for the update work.Since game development is the most competitive in the information service industry,and the number of newly released game information services is huge each year,this thesis selects game information services as the research object,explores the factors that influence users' adoption of game information services,and optimizes existing Information service forecasting methods.This article combs the existing prediction methods and finds that among the existing prediction methods,the prediction performance of linear prediction algorithms based on statistical learning is often inferior to non-linear prediction algorithms based on machine learning,but prediction algorithms based on machine learning cannot evaluate input features.Therefore,it is impossible to explore the influencing factors of users' adoption of game information services.Therefore,this thesis introduces an independent subspace to optimize the existing prediction algorithms,hoping that the proposed prediction method can accurately assess the importance of input features while also having good prediction performance.First of all,this article mainly uses the LSTM emotional polarity classification model and the LDA topic model to extract the predicted feature indicators of the game information service.The emotion classification of online word-of-mouth reviews of game information services is carried out through LSTM,and user emotions and user ratings are used as the online word-of-mouth characteristics of game information services.Then use the LDA topic model to cluster online word-of-mouth comment topics,dig out the user's attention points,and organize the generated topics as user attention features.Then,this thesis constructs a forecasting model based on time series,and uses ISM integrated forecasting method,RSM integrated forecasting method and single forecasting method to predict the adoption of game information services and evaluate the importance of features.After conducting an empirical study on one hundred mobile game applications newly released by Xiaomi Game Center,the research results show that the ISM integrated prediction method can accurately evaluate the features of the input,and the prediction performance has also been improved.Secondly,after the emotional classification and topic mining of online word-of-mouth comments on game information services,the results show that users tend to make positive or neutral comments before the release of game information services.After the information service is released,the number of negative comments will increase slowly.In addition,when users make online reviews of mobile game applications,they mainly focus on the four aspects of mobile game play,art,dubbing and payment.Finally,after using the ISM integrated prediction method to evaluate the importance of the input features,the results show that the initial stage of the information service release by users is mainly based on intuitive factors to adopt information services,such as information service developers,spokespersons,etc.As the release time increases,users will pay more attention to factors such as gameplay,payment,and comment emotions.At the theoretical level,this thesis uses independent subspaces to optimize the existing prediction algorithms,the optimized ISM integrated algorithm can evaluate the importance of features and has good prediction performance.It also provides a new forecasting method for the same type of product forecasting.In addition,in the feature selection,this thesis incorporates the emotions of online reviews of game information services,and mines the user's concerns through topic clustering,and fully considers user factors.It provides new ideas for feature extraction and variable selection in product forecasting research.At the practical level,the forecasting method proposed in this thesis helps game developers to predict the adoption of game information services and explores the factors that influence users' adoption of game information services and provides references for the R&D and marketing of enterprises.