Sentiment analysis is critical for extracting insights from vast user-generated content. Amazon reviews are one example where a large amount of hidden information can be extracted. Traditional techniques often struggle with the complexity of natural language, leading to lower accuracy and causing scalability issues. In contrast, the development of AI Techniques leads machine learning (ML) techniques to offer significant advantages, learn from large datasets, and enhance accuracy. In this context, we collected an Amazon review dataset from Kaggle, pre-processed it by handling missing values, and merged review titles and texts. Initially, we randomly selected four algorithms, i.e., Naive Bayes (NB), Logistic Regression (LR), Support Vector Machine (SVM), and Random Forest (RF) algorithms. Then we applied two feature extraction methods, i.e., Count Vectorizer (CV) and Term Frequency-Inverse Document Frequency (TF-IDF) and measured the performance of all four algorithms. Our experimental results reveal that Random Forest and SVM provide the highest accuracy of 93% over NB (92%) and LR (90%) using both CV and TF-IDF feature extraction methods. Further, we fine-tuned the Random Forest with randomized search and SVM using the Nu-SVM algorithm. Our experimental results conclude that the accuracy of the RF model reduces from 93 to 90% and slows down the entire process. But, when we fine-tuned the SVM using Nu-SVM, it yielded the highest accuracy of 94%. So, our experimental results conclude that Nu-SVM is the most effective model for sentiment analysis of Amazon reviews, offering a robust method for extracting valuable customer insights.