Online Tweet Recommendation Using Extreme Learning Machine

Abstract

In this report, we propose a scheme to provide online tweet recommendation for the movies. The method is based on user behavior patterns and interactions (based on retweets and favorites) performed on the tweets expressing the users' ratings. Traditional recommendation algorithms develop a model offline using training data and this model is then used to generate online recommendations. Training in traditional recommender systems is a very time consuming and hence cannot be used in domains which generate data very fast. To address this problem, we propose a novel recommendation algorithm, based on Extreme Learning Machine (ELM). We call our approach Extreme Learning Machine based Recommendation technique (ELMR). ELM is a learning system like single layer feed-forward neural network (SLFNN) which significantly reduces the amount of time needed for training of the system. It has been widely used for many applications requiring generalization, classification and prediction. The proposed ELMR method performs parameter learning with an intention to optimize recommendations on user engagement with an item (in this case tweet). The initial training and updating of ELMR are very fast and can be finished in real time. The experimental results show that the mean recommendation time of ELMR is shorter than MLP/ANN and naïve based classification algorithms reported in other reports with better accuracy.

Agrupamento de dados baseado em predições de modelos de regressão: desenvolvimentos e aplicações em sistemas de recomendação

“…novas avaliações feitas pelos usuários, novos usuários, novos itens, etc). Visando melhorar a acurácia do sistema, o modelo de predição é atualizado de forma incremental utilizando o algoritmo Online Sequential Learning for ELM (OS-ELM) (LIANG et al, 2006) (BEDI et al, 2014) foi desenvolvido para fornecer recomendações de filmes online (via Twitter). A abordagem é baseada em padrões de comportamento dos usuários e suas interações (baseadas em tweets e retweets) que expressam as avaliações dos usuários sobre os filmes.…”