Research On Impact And Prediction Of Contextual Factors On Mobile O2O Advertising

The rapid development of mobile intelligent devices has produced a subversive impetus to the advertising industry.This round of explosion has greatly promoted the development of mobile O2 O advertising.The mobile O2 O advertising created specifically for mobile devices and mobile users’ habits,advertisers can get the excellent effect that only mobile O2 O advertising can bring.How to push the corresponding mobile O2 O advertising to the right users at the right time and in the right geographical location has become an important issue of common concern to advertisers.Therefore,it is of great significance to carry out the research on the influence and prediction of context factors including time and geography on the effect of mobile O2 O advertising.In recent years,the research on influencing factors and prediction of mobile O2 O advertising effect has been gradually carried out,but the research on context factors for mobile O2 O advertising effect is still in its infancy.For the context factor of mobile O2 O advertising,its influence mechanism on the effect of mobile O2 O advertising and how to use the context factor to predict the effect of mobile O2 O advertising have become two core problems to be solved in improving the effect of mobile O2 O advertising.Advertisers are faced with three severe challenges in improving the effect of mobile O2 O advertising.First,how to explore the characteristics of each context factor and its interactive influence mechanism on the effect of mobile O2 O advertising from the actual context environment of mobile O2 O advertising.Second,the real-time and changeable context environment and the blowout growth of mobile O2 O advertising data bring difficulties to the mining of mobile O2 O advertising context data and the improvement of mobile O2 O advertising effect.Third,the interaction among users,context and advertisement is very complex,which makes the traditional low-level interaction methods ineffective and difficult to predict.In the face of big data environment and new technology driven mobile O2 O advertising innovation,new theories and methods are urgently needed to understand the impact mechanism and improve the prediction performance,so as to inject power into its orderly and effective development.The main research work of this paper includes:Firstly,research on the interaction between contextual factors of mobile O2 O advertising.Under the framework of "Time-Space-Connection",based on the theory of context marketing and level of interpretation,the interactive influence model of contextual factors in mobile O2 O advertising scenarios is established.The empirical results show that the interaction between time space association makes the scene marketing of advertising more complex than looking at each element separately.Specifically,the association degree of users and merchants has a negative impact on the redemption time of coupons;the larger the geographical distance is,the longer the redemption time is;that is to say,the longer the distance between users and merchants is,the more sensitive users are to the association degree and the consumption time of coupons The greater the negative impact,and the closer the consumer is to the merchant,the smaller the negative impact of correlation.The research provides an effective reference for merchants in mobile O2 O advertising to balance users’ preference and spatial factors.Secondly,the characteristics of context in mobile O2 O advertising and the influence of its interaction on the effect of mobile O2 O advertising are studied.First,we try to mine context features from mobile O2 O advertising data to expand the feature information.Then,based on this,we analyze the different context features one by one,and propose the interactive influence model of mobile O2 O advertising context factors on the effect of mobile O2 O advertising.Empirical research shows that: the sales volume of merchants in mobile O2 O advertising is very sensitive to time,location and weather conditions.These contextual features have a significant positive correlation with the effect of mobile O2 O advertising,and their interaction has a significant impact on the effect of mobile O2 O advertising.Thirdly,the research on prediction of mobile O2 O advertising based on XGBoost is carried out.In order to improve the traditional prediction method,the first step is to study the prediction method of advertising click through rate based on context factors.In order to study the influence of different time context factors granularity on the prediction model,the model which combines the day level context feature granularity,and the hour level context feature granularity is compared.Finally,through the experimental comparison and analysis,it is found that the model with context factor is better than the existing model with only the basic features of mobile O2 O advertising,which not only guarantees the better prediction performance but also ensures the lower prediction time complexity.In addition,compared with XGBoost mobile O2 O advertising prediction model which integrates day level context features,hour level context features highlight the importance of hour level features for mobile O2 O advertising effect prediction,and further explain the dynamic variability of mobile O2 O advertising in micro context scenarios.Fourthly,aiming at the complex interaction among users,context and advertisement,in order to obtain the best feature representation under the full interaction among them at the same time,based on the feature selection in chapter five,a model for predicting the effect of mobile advertisement with full interaction of bilateral and contextual factors is proposed.Firstly,the basic idea of Wide&Deep framework is analyzed,and its relationship with traditional linear model and Deep Neural Network is analyzed.Then,a model with the ability of full interactive relationship representation of low-order and high-order feature fusion is proposed.Finally,the prediction performance of this model is compared with that of traditional model and deep neural network model.It is found that the XGBDeep FM model,which combines the first-order and the second-order features of mobile O2 O advertising,user and context factors,and the high-order features captured by the deep neural network for full interaction,has better performance than the mobile O2 O advertising prediction model with only one-sided or all-round low-order factors.The model can make more accurate use of context information to predict the personalized delivery of mobile O2 O advertising.In conclusion,this paper takes context factors in mobile O2 O advertising as the research object,based on the context marketing theory,interpretation level theory and preference theory,aiming at the current problems in the promotion of mobile O2 O advertising effect,puts forward the corresponding impact mechanism framework and effect prediction technology scheme,effectively understands the context interaction and improves the advertising prediction performance.