Liangzhe \Cecilia de Almeida Marques-Toledo, Carolin Marlen Degener,

Liangzhe Chen, K. S. M. Tozammel Hossain, Patrick Butler, Naren Ramakrishnan and B. Aditya Prakash3 proposed HFSTM(Hidden Flu State from Tweet Model) with the hypothesis that a stream of tweets from a particular user can be used to detect the health condition of that user. Using this model state of a user Healthy(S),Infected(I), or Exposed(E) can be determined.So {0.98, 0.02, 0.00} is initially learned by HFSTM, which basically means there is very high probability that starting state will be S and very low probability that starting state will be E and there is no likelihood of the starting state to be infected. After that S is very likely to be in state S, state E is very likely to move to a state I and from the state, I the user either stays at I or recovers and move to state S. As a matter of fact, this model matches exactly with standard epidemic HEIS model and with its intuition.Sangeeta Grover and Gagangeet Singh Aujla4 used Markov-Chain state model to get different state of the user like beginning of epidemic, spread of epidemic and absence of epidemic. After that SEHA(Swine Epidemic Health Algorithm) algorithm was used. Using this algorithm tokenization of useful words are done and a hint score is assigned to each tweet. Once the score id assigned to each tweet using Motif-recovery algorithm time series classification and prediction is done.  \Cecilia de Almeida Marques-Toledo, Carolin Marlen Degener, Livia Vinhal,Giovanini Coelho, Wagner Meira, Claudia Torres Codeço, Mauro Martins Teixeira5 have used GAM(Generalized Additive Model) to model the association between tweet and Dengue.$log(u_t)=f_1(tweets_t)+f_2(week_t)+eyear_t+B0$Where logarithm $log(u_t)$ is the number of dengue cases on given week t.\Highly ambiguous and noisy time series is created from tweets because of a huge number of irrelevant tweets. So giving them proper and specific context is very important5.