Placement-Prediction-Data-Science-Model

This project predicts student placement outcomes based on academic and profile data using a trained Random Forest Classifier. It uses a complete machine learning pipeline with preprocessing, label encoding, and model persistence using joblib.

Files and Structure

1. placement.csv # Original dataset
2. input.csv # Automatically generated test data (from StratifiedShuffleSplit)
3. output.csv # Output file with predictions
4. model.pkl # Trained Random Forest model
5. preprocessing_pipeline.pkl # Preprocessing pipeline (numerical + categorical)
6. label_encoder.pkl # Encoder to convert labels to/from numeric

How the Model Works

1. Loads placement.csv as the dataset. (CSV to be downloaded from kaggle)
2. Performs a StratifiedShuffleSplit to ensure class distribution is preserved.
3. Splits the data into training and test sets.
4. Stores the test set as input.csv (used later for inference).
5. Encodes the target labels using LabelEncoder.
6. Separates numerical and categorical columns.
7. Constructs a preprocessing pipeline using:
   • SimpleImputer and StandardScaler for numerical data.
   • OneHotEncoder for categorical data.
8. Trains a RandomForestClassifier on the transformed training data.
9. Saves the trained model, preprocessing pipeline, and label encoder using joblib.

On rerunning the script:

• If model files exist, the script loads the model and pipeline.
• Reads input.csv, applies the pipeline, and predicts outcomes.
• Saves the result to output.csv with an additional column Placement_Prediction.

Requirements

• pandas
• numpy
• scikit-learn
• joblib

Highlights

1. Stratified sampling ensures balanced training.
2. Full pipeline for preprocessing both numerical and categorical features.
3. Label encoding ensures clean conversion between text and numbers.
4. Model persistence using joblib ensures easy inference without retraining.