Fake Instagram Profile Detection (Assesment Task)

Task Description

This project aims to build robust models to detect fake Instagram profiles using two distinct approaches:

1. Updated Baseline Model: Enhanced Deep Neural Network with advanced hyperparameter tuning and regularization techniques.
2. New Model: Combines Traditional Machine Learning models (Logistic Regression, Random Forest) with an Enhanced Neural Network architecture.

---

Model Overview

1️⃣ Updated Baseline Model

Techniques Used:

• Optuna Hyperparameter Tuning: Optimizes number of layers, units, dropout rate, learning rate.
• Batch Normalization & Dropout: Stabilizes learning, faster convergence, improved generalization.
• EarlyStopping & ReduceLROnPlateau: Prevents overfitting, dynamically adjusts learning rate.
• AdamW Optimizer: Decouples weight decay for better regularization.
• Proper Seeding: Ensures reproducibility of results.

Impact:

• Improved accuracy, generalization, and stability.
• Reduced overfitting and better model convergence.

---

2️⃣ New Model (Traditional ML + Enhanced Neural Network)

Preprocessing:

• Checked for missing values.
• Handled class imbalance using upsampling of minority (fake) class.

Feature Selection Techniques:

1. Low Variance Removal: Eliminated features with near-zero variance.
2. High Correlation Removal: Removed features with correlation > 0.9.
3. Random Forest Feature Importance: Selected significant features (importance > 0.01).

Models Implemented:

• Logistic Regression: Provides baseline performance.
• Random Forest Classifier: Handles feature interactions & non-linearity well.
• Enhanced Neural Network:
  • 4 Dense Layers: 256 → 64 → 64 → 32 → Output.
  • Batch Normalization after each Dense layer.
  • Dropout (0.3) to prevent overfitting.
  • AdamW Optimizer with EarlyStopping & ReduceLROnPlateau.

Evaluation:

• Precision, Recall, F1-score.
• Confusion Matrix.
• Loss & Accuracy Curves.

---

🛠️ How to Run the Code

1️⃣ Install Dependencies:

pip install -r requirements.txt

---

2️⃣ Prepare Data:

Place the following files in your project directory:

• train.csv
• test.csv

---

3️⃣ Run Models:

# For Updated Baseline Model:
python updated_baseline_model.py

# For New Model:
python NewModels.py

📝 Coding Style Guidelines

Feature	Details
PEP8 Compliant	Proper indentation, spacing, and clear naming conventions
Modular Code	Functions for preprocessing, training, evaluation
Reproducibility	Seeded numpy, random, tensorflow for consistent results
Clear Documentation	Comments & docstrings for each block and function
Visualization	Confusion Matrix, Training Curves with Matplotlib/Seaborn

📈 Results Summary

Model	Precision	Recall	F1-Score
Original Baseline Model	88	88	88
Logistic Regression	88	88	87
Random Forest	92	92	92
Enhanced Neural Network	92	92	92
**Updated Baseline **	93	93	92

📄 Files Included

File Name	Description
NewModels.py	New Model pipeline (ML + Enhanced Neural Network)
updated_baseline_model.py	Enhanced Baseline Deep Neural Network Model
train.csv, test.csv	Input datasets
requirements.txt	Required dependencies
README.md	Project overview & instructions