📰 Fake News Detection

A Machine Learning-based project to classify news articles as True or Fake using natural language processing and a combination of statistical and semantic features. The model leverages pre-trained Word2Vec embeddings to extract meaningful linguistic characteristics, and it employs supervised learning algorithms to deliver high-accuracy predictions.

---

👨‍💻 Team Members

• Arnab Biswas
• Maddi Pranav Reddy
• Mohan Nishantam

---

📌 Problem Statement

Classify a given news text as True or Fake using supervised machine learning algorithms. Utilize the Google News Word2Vec (300-d) pre-trained embeddings for feature extraction and semantic representation.

---

🔍 Approach

1. Data Preprocessing

• Text cleaning: removal of punctuation, numbers, brackets, and stop words.
• Lemmatization: focusing primarily on nouns (NN, NNS) as they hold key semantic value.
• Conversion to lowercase for normalization.

2. Feature Engineering

• Sentence Vectors: Sum of all word vectors in a sentence.
• Frobenius Norm: Captures the volume of the text’s hyperspace.
• Word Similarity: Dot product between individual word vectors and the sentence vector.
• Text Length: Used as a proxy for editorial standards.
• Named Entity Recognition (NER): Counts of entity types (person, organization, location, etc.).

3. Modeling

• Tried and tested models: Logistic Regression, Decision Tree, and Random Forest.
• Hyperparameter tuning using GridSearchCV.
• Evaluation metric: F1 Score (balances precision and recall, ideal for binary classification with roughly balanced classes).

---

📈 Exploratory Data Analysis (EDA)

• Text length distributions differ significantly between true and fake news.
• Word clouds and N-grams highlight focus areas:
  • Fake news: more personal references, vague justifications, and visual proof.
  • True news: more institutional terms, verifiable sources, and official tone.
• NER patterns: Fake news focuses disproportionately on persons; true news has a balanced distribution across types.

---

🧠 Best Model

✅ Logistic Regression

• (solver: liblinear, regularization: l1, penalty: 0.1)
• Accuracy: 93.06%
• Precision: 92.47%
• Recall: 93.03%,
• F1 Score: 92.75%

✔️ Outperformed Decision Trees and Random Forest on most metrics

---

📁 Files Included

All files are provided inside the ZIP archive:
Fake_News_Detection_ArnabBiswas_MaddiPranavReddy_MohanNishantam.zip

• Fake_News_Detection.ipynb – Jupyter notebook containing the full implementation.
• Fake_News_Detection_Report.pdf – Final project report with analysis, EDA insights, and model evaluations.