■ TechHorizon Internship – Machine Learning Projects
■ Objective:
As a Machine Learning Intern at TechHorizon, you will work on practical and impactful ML problems using
real-world datasets. This internship is designed to strengthen your foundations, improve your problem-solving
skills, and give you hands-on project experience.
■ General Instructions:
•
Complete all 3 tasks within the internship period.
•
Submit your work in a well-documented Jupyter Notebook or Python script.
•
Include a short report (1–2 pages) summarizing: - Your approach for each task - Challenges faced - Key
learnings and model performance
■ TASKS
Task 1: Fake News Detection
Goal: Build a machine learning model to classify news articles as fake or real.
Dataset: Fake and Real News Dataset (https://www.kaggle.com/clmentbisaillon/fake-and-real-news-dataset)
•
Preprocess text (removal of stopwords, lemmatization/stemming).
•
Use classifiers like Logistic Regression, Naïve Bayes, or Random Forest.
•
Evaluate your model using accuracy, precision, and F1-score.
•
(Optional): Create a basic Streamlit app for testing news snippets.
Task 2: Customer Segmentation Using Clustering
Goal: Segment customers based on demographic and spending data using unsupervised learning.
Dataset: Mall Customer Segmentation Data
(https://www.kaggle.com/vjchoudhary7/customer-segmentation-tutorial-in-python)
•
Perform EDA and feature scaling.
•
Apply K-Means or Hierarchical clustering.
•
Visualize clusters and explain business insights.
Task 3: Sentiment Analysis of Movie Reviews
Goal: Create a model that classifies movie reviews as positive or negative.
Dataset: IMDb Movie Reviews Dataset
(https://www.kaggle.com/lakshmi25npathi/imdb-dataset-of-50k-movie-reviews)
•
Text preprocessing: tokenization, removing stopwords, etc.
•
Train an ML model (SVM, Naïve Bayes, or Logistic Regression).
•
Evaluate with accuracy and F1-score.
•
(Optional): Add a basic interface to test custom reviews.
■ Submission Guidelines
•
Code: Submit clean, well-commented Python scripts or Jupyter Notebooks.
•
Report: 1–2 pages covering approach, challenges, and outcomes.
•
Deployment (Optional): Share a link or instructions if you deployed any app.