International Scientific Journals
of Scientific Technical Union of Mechanical Engineering "Industry 4.0"

This paper presents the comparison of the capabilities of two algorithms for Sentiment Analysis developed in Python. Both Python programs are used on the same Yelp dataset with customer reviews of the quality of the services in USA restaurants. The programs are based on open-source software frameworks and libraries as Python, NTLK, Scikit-Learn, Panda, etc. which are oriented to Machine and Learning and Natural Language Processing. The evaluation of the programs is based on precision of the predicted results and the compactness of the programming code. For model training and prediction, the Multinomial Naïve Bayes and Support Vectors Machines classifiers are applied in both algorithms.

The importance of Big Data and Big Data Mining is growing significantly in recent years. Different kind of e-sources as social networks, e-commerce sites, e-mails, sensors, etc. are generating large amount of structured and unstructured numerical and text data. This data provides valuable information about costumer’s preferences or ratings of products or commodities. This information is essential for making predictions on the base of the sentiment analysis of this data. The sentiment analysis of large amount of text data requires specific big data and machine learning /ML/ libraries. In this paper the implementation of a system for big data sentiment analysis using ML algorithms is proposed. It is based on Naïve Bayes and Support Vector Machines /SVM/ classification ML algorithms for text analysis. The system is implemented in Java and uses Apache Spark ML libraries which are very flexible, fast and scalable. The system is tested with well known Amazon dataset and its performance is measured in form of accuracy. The obtained results approve the effectiveness of big data sentiment analysis algorithms. The System can be applied for recommendation of products and services or predictions of customers’ needs.