CascadePrediction

A cascade prediction model

  The model is used to predict weibo cascade.

Dataset Description

  In this project, I used WeiboForwards dataset, which contains 15807 weibo and its comments and forwards. An example of the data is like that:

{ "originalMid" : "4099456032803526", "type" : "forward", "originalWeibo" : { "mid" : "4099456032803526", "url" : "http://weibo.com/1953891524/EFJiXbLka", "userurl" : "http://weibo.com/311619000?refer_flag=1001030103_", "name" : "南宁热搜榜", "content" : "\n\t\t#南宁身边事# 【辟谣：对于网传的房屋倒塌砸车事故。】南宁交警反馈：从今天4月22日早上到现在，没有接到大沙田片区有建筑物倾倒砸中公交车和行人的报警。六大队领导又询问了银海建设路口附近岗亭的民警，同时与施工方进行了核实，楼房拆除过程中，没有发生建筑物及砖块砸中公交车和行人的情况。视频因 ​\n\t\t\t...展开全文c\n\t\t", "time" : "1492869945000", "forward" : "32", "comment" : "74", "hotrate" : 0, "isOrigin" : true, "piclist" : [ "http://wx3.sinaimg.cn/large/747604c4gy1fevt4gboedj208w06owey.jpg" ], "userCertify" : 1, "sourcePlatform" : "秒拍网页版", "praise" : "131", "userId" : "1953891524", "userGender" : "female", "userFollowCount" : "1113", "userFanCount" : "230000", "userWeiboCount" : "6253", "userLocation" : "广西", "userDescription" : "微博本地资讯博主（南宁）http://t.cn/RUmMhZ7", "faces" : null, "weiboLocation" : null, "containsMusic" : false, "containsVideo" : false, "containsUrl" : false }, "props" : [ { "mid" : "4100007974570547", "url" : "http://weibo.com/5665979346/F00cRjb0v", "userurl" : "http://weibo.com/5665979346", "username" : "敬軒的小迷妹", "content" : "//@yellowUpBDown：拆房子这么拆？没有隔离设备？那么大的力量石头蹦出来多危险？还有脸说？", "time" : "1493001480000", "forward" : "0", "isComment" : false, "isForward" : true, "originalWid" : "4099456032803526", "faces" : null, "userCertify" : 0, "praise" : "0", "userId" : "5665979346", "userGender" : null, "userFollowCount" : null, "userLocation" : null, "userFanCount" : null, "userWeiboCount" : null, "userDescription" : null }, { "mid" : "4099985140744048", "url" : "http://weibo.com/5175660940/EFX4m37yM", "userurl" : "http://weibo.com/5175660940", "username" : "小蚂蚁的山河", "content" : "转发微博", "time" : "1492996080000", "forward" : "0", "isComment" : false, "isForward" : true, "originalWid" : "4099456032803526", "faces" : null, "userCertify" : 0, "praise" : "0", "userId" : "5175660940", "userGender" : null, "userFollowCount" : null, "userLocation" : null, "userFanCount" : null, "userWeiboCount" : null, "userDescription" : null } ] }

Preprocess flow

1. Use uniq.py to delete the repeat weibos,
2. Use mktrain.py to produce train data
3. Use mkFeatures.py to extract features from it.
4. Use CascadePrediction.py to predict
5. Use dump.js to export data from mongodb, the command is "mongo database dump.js>yourfile.json"