Deep Learning Projects

This repository contains deep learning projects focused on neural networks and natural language processing.

Projects

1. ANN_Classification

A customer churn prediction project using Artificial Neural Networks (ANN) for binary classification.

Features:

• Binary classification to predict customer churn
• Data preprocessing with label encoding and one-hot encoding
• Neural network architecture with multiple hidden layers
• Model training with hyperparameter tuning
• Streamlit web application for interactive predictions
• Model artifacts saved for deployment (encoders, scalers, trained model)

Files:

• app.py - Streamlit web application
• experiments.ipynb - Model development and experimentation
• hyperparametertuningann.ipynb - Hyperparameter optimization
• Churn_Modelling.csv - Dataset
• Pre-trained model and preprocessing artifacts (.pkl, .h5 files)

2. Imdb_RNN

A movie review sentiment analysis project using Recurrent Neural Networks (RNN) for text classification.

Features:

• Sentiment analysis of movie reviews (Positive/Negative/Neutral)
• Simple RNN architecture with embedding layers
• Text preprocessing and sequence padding
• IMDB dataset integration with word indexing
• Interactive Streamlit web application
• Real-time sentiment prediction with confidence scores

Files:

• app.py - Streamlit web application
• SimpleRNN.ipynb - Model development and training
• embedding.ipynb - Word embedding experiments
• prediction.ipynb - Model testing and evaluation
• IMDB Dataset.csv - Training dataset
• Simple_RNN_imdb.h5 - Pre-trained RNN model

🚀 Live Demo: https://deeplearning-imdb-simplernn.streamlit.app/

3. CNN - CIFAR-10 Image Classification

A computer vision project using Convolutional Neural Networks (CNN) for multi-class image classification on the CIFAR-10 dataset.

Features:

• Multi-class image classification (10 classes: airplane, automobile, bird, cat, deer, dog, frog, horse, ship, truck)
• CNN architecture with convolutional, pooling, and dense layers
• Image preprocessing and normalization
• Real-time image upload and prediction
• Confidence score display for predictions
• Interactive Streamlit web application

Files:

• app.py - Streamlit web application for image upload and prediction
• Cifar_10_CNN.ipynb - CNN model development and training
• Cifar_10.h5 - Pre-trained CNN model
• requirements.txt - Project dependencies

Classes Supported:

• Transportation: airplane, automobile, ship, truck
• Animals: bird, cat, deer, dog, frog, horse

4. LSTM_Next_word_Prediction

A natural language processing project using Long Short-Term Memory (LSTM) networks for next word prediction and text generation.

Features:

• Next word prediction using LSTM neural networks
• Text sequence generation and completion
• Character-level and word-level tokenization
• Custom text corpus training capability
• Interactive text generation with seed phrases
• Advanced LSTM architecture with embedding layers
• Temperature-controlled text generation for creativity control

Files:

• LSTM_First.ipynb - LSTM model development and training
• next_word_lstm.h5 - Pre-trained LSTM model
• tokenizer.pickle - Trained tokenizer for text preprocessing
• Model supports both single word prediction and full sentence generation

Key Capabilities:

• Predict the next word given a sequence of input words
• Generate coherent text passages based on seed text
• Adjustable prediction confidence and creativity levels
• Support for various text corpora and domains

5. California_Housing_ANN

A regression project using Artificial Neural Networks to predict housing prices in California based on various demographic and geographic features.

Features:

• Regression analysis for continuous price prediction
• Multi-feature input processing (location, population, income, etc.)
• Neural network architecture optimized for regression tasks
• Feature scaling and normalization
• California housing dataset integration
• Statistical analysis and model evaluation metrics

Files:

• HousingANN.ipynb - Neural network development and training for housing price prediction

Key Capabilities:

• Predict house values based on demographic and geographic features
• Analyze feature importance for price prediction
• Regression metrics evaluation (MAE, MSE, R²)
• Data visualization and exploratory analysis

Technologies Used

• TensorFlow/Keras - Deep learning framework
• CNN (Convolutional Neural Networks) - Computer vision and image classification
• LSTM Networks - Advanced RNN architecture for sequence modeling
• Natural Language Processing - Text preprocessing and tokenization
• Computer Vision - Image processing and classification
• Streamlit - Web application framework
• NumPy/Pandas - Data manipulation
• Scikit-learn - Data preprocessing
• Pillow (PIL) - Image processing library
• Pickle - Model and tokenizer serialization
• Python - Programming language

Getting Started

1. Clone the repository
2. Install dependencies: pip install -r requirements.txt
3. Navigate to the desired project folder
4. Run the Streamlit app: streamlit run app.py