Research And Application On Mobile User Emotion Prediction Based On Time Series Analysis

With the rapid development of mobile communication network technology,people increas-ingly rely on mobile phones and other terminals to carry out network activities,the terminal of the information display tends to be miniaturized,and resulting information overload prob-lem prevents users from obtaining the required data information efficiently.The recommen-dation system is a core technique for the solution of the problem mentioned above,the main-stream recommendation system mainly counts on the user's explicit or implicit feedback to obtain the preference,yet it lacks a full use of data like user's footprint in the mobile network on-line activity and opinions.Sentimental analysis and prediction analysis is the main meth-od of mining the information above,whose essence is through the historical data changes to calculate the future data developing status and understand the user's sentimental feedback on the recommended items.Therefore,combining with sentimental analysis and prediction analysis technology,it can not only calculate the sentimental trend,enrich the recommended basis and assist the recom-mender system in calculating more personalized results,but also a new solution of recom-mendation from another point of view.Based on the thoughts above,this paper import the predictive analysis technology to the field of sentimental analysis,and predicts the future emotion value through the historical results of emotion analysis and the trend of emotion value.Meanwhile,we applied this method to the recommendation system,aiming at user's sentimental level of the preferences to recommend the corresponding content,so it can en-hance the user's experience and the recommendation quality.The main work of this paper includes the following aspects:(1)Under the social network environment,we analyzed the user's on-line behavior and chat-ting language habits and came up with a set of methods to establish users' sentiment diction-ary including punctuation emotion dictionary and modal word emotional dictionary.And we also evaluate the accuracy of the sentimental analysis under this dictionary.(2)We proposed an emotion predicting method for the mobile user based on time series analysis.We also studied the theory and application of time series analysis related algorithms involved.Through analyzing the information context,we can obtain the user's emotion status sequence,then,calculates and predicts the future emotion by time series according to this sequence.(3)Aiming at the existing problem of using emoticons in the social process,we proposed a new algorithm to recommend emoticons,which is characterized by combining the emotional analysis results with the ARIMA algorithm of time series analysis.By analyzing the histori-cal conversation when using emoticons in a user conversation process,the result of the emo-tional analysis is taken as the user's emotional understanding of the emoticons.It will rec-ommend emoticons according to the user's sentimental tendency.(4)We developed a prototype of instant-messaging software and which includes our pro-posed emoticon recommender algorithm.The software uses the MVP architecture,and in-stant messaging cloud as a rapid deployment plan.After testing,the algorithm of this paper can recommend emoticons with similar emotions in the context.It's recommended results is better than frequency-based emoticon recommender algorithm.