Skip to content

Latest commit

 

History

History
61 lines (52 loc) · 2.56 KB

README.md

File metadata and controls

61 lines (52 loc) · 2.56 KB

Fake Job Postings Detection

Overview

This project involves detecting fraudulent job postings using machine learning models. The dataset used is fake_job_postings.csv, which contains various attributes related to job postings, such as job title, location, company profile, description, requirements, benefits, and labels indicating whether a job posting is fraudulent or not. The goal is to analyze the data and build models to classify job postings as real or fake.

Project Structure

  1. Importing Data

    • Necessary libraries such as pandas, numpy, seaborn, matplotlib, and sklearn are imported.
    • The dataset is loaded using pandas.read_csv().
    • Missing values are analyzed using missingno.

  2. Data Visualization

    • Distribution of fraudulent vs. non-fraudulent job postings.
    • Distribution of job postings by employment type, required education, and experience.
    • Correlation heatmaps of numerical features.
    • Word cloud visualization of job descriptions.

  3. Data Preprocessing

    • Categorical and numerical feature separation.
    • Handling missing values using SimpleImputer.
    • One-hot encoding categorical features.
    • Splitting data into training and test sets.

  4. Model Training and Evaluation

    • Random Forest Classifier:
      • Trained with different numbers of trees.
      • Achieved an accuracy of 99.48%.
      • Top keywords influencing fraudulent job postings were extracted.
    • Naive Bayes Classifier:
      • Achieved an accuracy of 98.43%.
      • Precision: 0.93, Recall: 0.50, F1 Score: 0.65.
    • Decision Tree Classifier:
      • Achieved an accuracy of 99.42%.
      • Precision: 0.99, Recall: 0.93, F1 Score: 0.96.

  5. Key Findings

    • Fraudulent job postings often contain certain keywords in company profiles and descriptions.
    • Specific industries and job functions are more prone to fraudulent postings.
    • STEM vs. Non-STEM fraudulent job posting analysis showed varying trends.

Usage

  1. Install dependencies: 
    pip install pandas numpy seaborn matplotlib scikit-learn nltk wordcloud missingno
  2. Run the script: 
    python fake_job_detection.py
  3. Modify dataset file path if needed: 
    df = pd.read_csv('fake_job_postings.csv')

Future Work

  • Incorporate NLP techniques like TF-IDF and word embeddings.
  • Experiment with deep learning models for classification.
  • Deploy the model as a web service for real-time job posting verification.

Author

Vedant - GitHub