Consumption Intention Mining Based On Microblog

As a new kind of social media, microblog gains mass size of users and data. It‘seasy for people to post on microblog. Also, the posts spread fast and can be easily seenby many other users. For the reasons above, users post various and large numbercontents on microblog. Among these posts, we find numerous posts that express authors‘purchase wish for a certain product, in other words, consumption intention. These postswith consumption intention are of great research value and commercial value.We conduct the research about consumption intention in three aspects:(1) Acquiring, preprocessing and classifying consumption intention text data. Wediscuss the acquirement method of initial consumption intention corpus, and buildcrawlers to fetch raw consumption intention text data from Taobao QA, Jingdong andmicroblog. Then, we regard consumption intention mining as a binary classificationproblem, and extract multi-dimensional features from the raw corpus. Finally, wepropose models based on SVM, Na ve Bayes and deep learning to solve theconsumption intention classification problem. The experimental result shows that thedeep learning based method achieves the highest F-measure.(2) Discussing the problem of conversion from consumption intention toconsumption behavior. In our previous experiments, we obtain microblogs withconsumption intention by the means of annotation. Although we have already set upstrict standards for data annotation, the annotation result of different people remainsdifferent. Moreover, as there are various factors which can affect users‘consumptionbehavior, users‘consumption intention doesn‘t always convert into consumptionbehavior. To solve this problem, we propose an automatic survey method on socialmedia. This method enable us to collect massive consumption behavior data withoutany annotation. The consumption behavior data is used to evaluate our model forconsumption intention mining, and also used to build a consumption behaviorclassification model.(3) Applying consumption intention mining to prediction tasks in social media.This paper discusses consumption intention towards a certain kind of product, i.e. movie, and uses movie consumption intention as an important feature in box office prediction.We combine consumption intention with traditional features used in the problem of boxoffice prediction, and achieve a of0.95which outperforms previous work of thisproblem We build a system based on linear regression which automatically predictsmovies‘total box office and opening weekend box office one day prior to the movie‘srelease date.