Amazon Machine Learning vs. Microsoft Azure Machine Learning as Platforms for Sentiment Analysis

Abstract

Recently, there has been an increasing attention towards the use machine learning platforms notably Amazon Machine learning and Microsoft Azure Machine learning (ML) to undertake sentiment analysis. The present experimental study compared Amazon ML against Microsoft Azure ML as platforms for performing sentiment analysis. The evaluation was done d using the evaluation metrics: accuracy, reliability, precision, F-score and Recall. Data was sourced from Twitter a microblogging platform. The sentiment analytics model was created based on Logistic regression analysis algorithms. Results confirmed that Microsoft Azure ML is more accurate and precise than Amazon ML as platforms for Sentiment Analysis and that Azure could be used reliably, accurately and securely to build Sentiment Analysis models for social media sites.

Sentiment Analysis of ASOS Product Reviews Using Machine Learning Algorithms by Comparing Several Models

“…For instance, although libraries like Keras make the development of the neural network fairly simple, sometimes more control is needed over the details of the algorithm. Although modules like Tensorflow provide more opportunities, it is also more complicated, and the development takes much longer [9]. Furthermore, Recurrent Neural Network usually requires more customer review data compared to conventional machine learning algorithms, for example, at least thousands to millions of labeled samples.…”

“…Moreover, the MaxEnt-JABST design also included a maximal entropy classifier for accurately differentiating features or different viewpoints. Furthermore, two machine learning techniques, namely naïve Bayes classification and support vector machine techniques, were used to conduct sentiment analysis on certain customer reviews based on the products bought [9]. An Amazon dataset was utilized to extract opinion lexicons, which included 4783 negative and 2006 positive terms with sentiment ratings for each phrase.…”

“…Another research [8] has asserted that the experimental outcomes demonstrated that the designs surpassed current benchmarks and could comprehend fine-grained characteristics and views, but further improvement was required. A research [9] has affirmed that the support vector machine renders lower accuracy in classification. Moreover, emotion recognition is an extraction process that helps to recognize the mood and perception of the individual; however, emotion mining techniques were not used to classify the reviews and understand the customers' mindset.…”

“…The performance of both the models were evaluated on basis of classification accuracy on one percent data. The dataset chosen by [9] is composed of an ordering log file for three months. The paper states the trial results that were conducted with different feature classification methods in combination with three main classifiers: naïve Bayes, support vector machine, K-nearest neighbors, along with LDA, an unsupervised document topic classifier.…”