Twitter Sentiment Analysis Using TF-IDF and Machine Learning Classifiers - Research Overview

Aug 3, 2026

Uploaded by Riley

Twitter sentiment analysis extracts live public opinions from microblog text and supports understanding brand perception and social trends. This study builds a machine learning pipeline using TF-IDF for feature extraction and evaluates three classifiers—Logistic Regression, Support Vector Machine, and Random Forest—on the Sentiment140 dataset containing 1,600,000 tweets from the Twitter API. Performance is reported with accuracy and F1 score, achieving an F1 of 0.87 and strong classification performance. The results support more accurate, fine-grained sentiment modeling for informed decisions and trend prediction.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages9

Reading Time23 min

AccessFree to download

Introduction

Sentiment analysis and TF-IDF fundamentals

Challenges: sarcasm and evolving language

Objectives

Tweet sentiment classification goals

Classifier comparison and evaluation metrics

Pre-processing and feature extraction

Motivation and Novelty

Limitations of existing methods

Addressing sarcasm and context dependency

Twitter Sentiment Analysis Using TF-IDF and Machine Learning Classifiers Yash Paul Department of Information Technology,Central University of Kaashmir [mail:](mail:yashpaulcuk@gmail.com)[yashpaulcuk@gmail.com](mail:yashpaulcuk@gmail.com) Seerat ul Nisa Department of Information Technology, Central University of Kashmir e-mail: [seeratulnisaa@gmail.com](seeratulnisaa@gmail.com) Hemah Hussain 1*Department of Information Technology, Central University of Kashmir. [hemahhussain292@gmail.com](hemahhussain292@gmail.com), Rajesh Singh Indian Institute of Foreign Trade,New Delhi [rajesh_phdmp22@iift.edu](rajesh_phdmp22@iift.edu) Abstract—One of the foremost microblogging platforms is Twitter, which serves as a rich repository for the measurement of live user opinions and sentiments. The subject of this study is Twitter sentiment analysis using cutting-edge machine learning techniques. A machine learning pipeline is being constructed that includes three classifiers. Logistic Regression, Support Vector Machine, and Random forest. Also, We utilize Term Frequency-Inverse Document Frequency (TF-IDF) for feature extraction.The data set used in this study is known as the Sentiment140 data set, which comprises 1,600,000 tweets gathered via the Twitter API. These classifiers are measured using accuracy and F1 scores. The results When it comes to sentiment classification, the model is notable for its high accuracy and We are getting an F1 score of 0.87 which is higher than state-of-the-art methods. The findings from this study have important implications for comprehension of public opinion brand perception and societal trends. In the digital age, our scholarly work contributes to the enhancement of machines by improving the accuracy and granularity of sentiment analysis. Notable Learning applications are to be found in the dynamic sphere of social media, in order to reveal the possibilities of informed decision making and trend prediction. Keywords-Frequency-Inverse Document Frequency, Classification, SVM, Regression; Hybrid, Sentiments, Twitter I. INTRODUCTION The process of extracting and examining the sentiment or emotional states expressed in textual data is called sentiment analysis, sometimes referred to as opinion mining [1] . The sky is the limit for what we could learn from the sentiment behind Twitter that was happening, especially TF-IDF and machine learning classifiers' sense-making in social media. Twitter is quite a huge medium where people articulate their thoughts, feelings, and opinions. Everything gets tweeted about, whether it be politics, sports, or even today's happenings, and they're positively staggering numbers. Term Frequency-Inverse Document statistically calculates how important a word is to a document in a collection of documents. More significantly, it also helps in determining which words will form the basis for the analysis of tweets. Let's explain with some examples. When analyzing a tweet, one probably runs into many popular words that say, "the," "and," or ones like "is. " These words or terms don't contribute very much to understanding sentiment. What TF-IDF allows us, however, is to concentrate on words that are particular about describing an emotion. In other words, words like "love," "hate," and "disappointed" carry a lot of weight indicating how someone feels. These words would have relatively higher TF-IDF scores, thereby being more heavily interpreted in the analysis. Now, after keying out these important words, we can then start classifying them with machine learning. There are lots of them: logistic regression, and support vector machines, with an even more complex set of algorithms like neural networks. Sarcasm is an interesting issue to grapple with. Sometimes a person says the opposite of what they mean: that can mess up the best of models. Also, the language is ever-evolving, new slang or trends will pop up with which our models may not recognize would be in

FAQs

What dataset and data size are used for training and evaluation?

The study uses the Sentiment140 dataset containing 1,600,000 tweets collected via the Twitter API.

Which TF-IDF and classifiers are included in the proposed pipeline?

TF-IDF is used for feature extraction, and three classifiers are evaluated: Logistic Regression, Support Vector Machine (SVM), and Random Forest.

How are the models evaluated and what key result is reported?

Accuracy and F1 scores measure performance, with an F1 score of 0.87 reported for sentiment classification.

Aug 3, 2026

Uploaded by Riley