This repository contains my Deep Learning Models, where I have implemented end-to-end pipelines covering data preprocessing, model development and performance evaluation.
In this project, I compared XGBoost and ANN models for customer churn prediction. The results demonstrate the effectiveness of both approaches:
- XGBoost Model: Achieved an accuracy of approximately 0.8515
- ANN Model: Achieved an accuracy of approximately 0.8615
Dataset: https://www.kaggle.com/datasets/shrutimechlearn/churn-modelling
A CNN-based deep learning model that performs multi-task learning to predict both gender and age from images. The model demonstrates strong performance on both gender classification and age regression tasks.
Architecture: Convolutional Neural Network (CNN) with multi-task outputs
Model Performance (Training Results):
| Metric | Training | Validation |
|---|---|---|
| Age MAE | 3.6264 | 6.5056 |
| Age Loss | 3.6260 | 6.5019 |
| Gender Accuracy | 96.11% | 87.73% |
| Gender Loss | 0.0909 | 0.4383 |
| Overall Loss | 3.7174 | 6.9452 |
Key Highlights:
- Excellent gender classification accuracy of 96.11% on training data and 87.73% on validation data
- Low gender loss indicating confident predictions
- Age prediction with MAE of 3.6 years on training and 6.5 years on validation
- Trained for 593 steps per epoch with ~13ms per step
Dataset: https://www.kaggle.com/datasets/jangedoo/utkface-new