Opinion mining, otherwise known as sentimental analysis, is to interpret the whole idea of online written text. This work involves implementation of sophisticated machine learning algorithms for sentiment polarity classification (positive, negative, or neutral context) using texts. We investigate the efficacy of two popular algorithms for sentiment analysis tasks: A multinomial Naive Bayes (MNB) and support vector classification (SVC) with a linear kernel. The study applies the comparative approach on a widely used benchmarking dataset in order to assess the efficacy of sentiment analysis on SVC and MNB. The feature engineering methods that we use are aimed at converting the textual input into a numeric representation, which is comprehensible for machine learning algorithms. The precision is examined by applying the following typical evaluation metrics, such as accuracy, precision, recall and F1-score. The following discussion reveals the outcomes of the study in terms of the strengths and limitations of the models used for sentiment analysis. As well as identifying performance value-impacting factors, the study also contributes to research of how different algorithms differ in average power to categorize sentiment.