Forecasting of Twitter Hashtahg Temporal Dynamics Using Locally Weighted Projection Regression
[ X ]
Tarih
2017
Dergi Başlığı
Dergi ISSN
Cilt Başlığı
Yayıncı
IEEE
Erişim Hakkı
info:eu-repo/semantics/closedAccess
Özet
Popularity of social networks opens great opportunities for market such as advertisement. Using hashtags increasingly used in twits helps us to realize popular topics on the internet. Since most of new hashtags become popular and then fade away quickly, there is a limited time to predict the trend. Therefore, this paper proposes a fast incremental method to forecast the rate of the used hashtags in hour like time series. Two main parts for forecasting system are applied Preprocessing and Supervised Learning. Normalization is one of most popular preprocessing of dataset also proposed to have larger dataset. Moreover, the efficiency of the system under changing number of input (number of past hours from hashtag history) and output (number of next hours which is going to be predicted) are evaluated. Locally Weighted Projection Regression as one of the most powerful machine learning methods with no meta-parameter are applied in this paper as real-time learning method. The performance of the system is verified by implementation of Volume Time Series of Memetracker Phrases and Twitter Hashtags. The results show that the errors of forecasting system are good enough to understand the trend of the hashtag.
Açıklama
International Conference on Engineering and Technology (ICET) -- AUG 21-23, 2017 -- Akdeniz Univ, Antalya, TURKEY
Anahtar Kelimeler
Hashtag, Twitter, Machine Learning, Supervised Learning, Locally Weighted Projection Regression, Time Series Forecasting
Kaynak
2017 International Conference on Engineering and Technology (Icet)
