PlacePredictor
MLA machine learning-based student placement prediction system that provides probability scores and AI-style explanations using a trained Naive Bayes classification model.
Status
Completed
Duration
3 Weeks
Role
ML Engineer
Type
Personal Project
Key Features
High Accuracy
Reliable prediction model using Naive Bayes.
AI Explanations
SHAP-based explanation generation.
Interactive UI
Clean dashboard powered by Streamlit.
Data Handling
SMOTE for imbalanced datasets.
Visualizations
Rich charts with matplotlib and seaborn.
Technology Stack
Challenges Faced
Dataset Imbalance
Handling extreme class imbalance between placed and not placed students using SMOTE.
Explainable AI
Extracting meaningful SHAP-based explanations to provide actionable feedback for students (e.g., suggesting specific skills to improve).
Data Leakage Prevention
Preventing data leakage, such as rigorously removing the salary feature during model training since it is a post-placement value.
Future Improvements
UI/UX Enhancements
Improve the UI/UX of the Streamlit dashboard for a more native feel.
Deep Learning Integration
Incorporate deep learning models for performance comparison.
Batch Processing
Allow batch predictions via CSV upload for examiners and universities.