A new lexicon learning algorithm for sentiment analysis of big data

Sentiment analysis is the process of identifying opinions expressed in text. It aims to analyze the users' sentiment towards particular topics, products, and other subjects. The aim of this study is to present the parallel version of a method of polarity classification of sentiment for big data. In the original method, named ALGA, a genetic algorithm is incorporated to generate lexicons. Since the fitness calculation of that method is its most time consuming part, especially in the big data, in this research, a new parallel method is presented to efficiently calculate the fitness of ALGA. The experiments are conducted on four datasets in terms of running time, speedup and time complexity. Results show that the proposed method achieves better runtimes than the sequential ALGA when the datasets are big. Therefore, in spite of the sequential ALGA, this method can employ the strength of genetic algorithm for searching the landscapes of big data mining problems.