1. FAKE NEWS PREDICTION MODEL
GARIMA SAIGAL
2021A1R094
Under the guidance of
- Parmveer Nandal
2. OBJECTIVES,
NEED, HOW?
● FAKE NEWS: False presented
as factual, misleading.
● IMPACT: Misinformation,
manipulation, influence public
opinions.
● NEED: Identify and counteract
fake news in real time.
● PREDICITON: NLP & ML
● HOW: Analyze patterns,
identify suspicious sources,
uncover misinformation.
3. STUDY WORK
EXISTING SYSTEMS EXPLORED:
https://www.kaggle.com/code/forgetabhi/fake-news-prediction-97
https://www.kaggle.com/code/sravankumargatla/fake-news-prediction
https://www.kaggle.com/code/daniilkrasnoproshin/99-accuracy-fake-news-predictor-log-regr
4. COMPARATIVE ANALYSIS OF EXISTING SYSTEMS
● Model-Centric Focus
● Lack Visual representations
● Simplicity Focused
● Limited Interpretability
5. WHAT’S BETTER IN OUR SYSTEM?
CONVIENCE
VISUAL
REPRESENTATIONS
SIMPLE PREDICTING
SYSTEM
HIGH ACCURACY
7. DIVE INTO KEY CONCEPTS:
● Logistic regression :effective method for news categorization.
● Logistic regression : binary classification to predict falsity of news.
● Train-Test Technique: Handles over fitting, increases efficiency
● Confusion Matrix: Visualizes model performance by comparing predicted vs.
actual labels.
8. CONFUSION MATRIX
● TP, TN, FP, FN metrics.
● Calculates accuracy and precision.
● Provides insights for effectiveness.
9. TECHNOLOGY STACK
● :Python
● -Libraries and Frameworks: Pandas, Seaborn, Matplotlib, NumPy, NLTK, Scikit-Learn, WordCloud
● -Machine Learning Models: Logistic regression
● - Data Preprocessing:** Text cleaning, stopword removal, TF-IDF vectorization
● -Visualization Tools: Word Clouds, Confusion Matrix Display
● Data:News articles dataset (real and fake)
● Google Collab notebook