fatmahussein252/COVID-19-Daily-New-Cases-Prediction

This project applies an LSTM neural network to forecast daily new COVID-19 cases in Egypt based on past case data. # LSTM Time Series Forecasting: COVID-19 Daily New Cases in Egypt ## Overview This project applies a **Long Short-Term Memory (LSTM)** neural network to forecast **daily new COVID-19 cases in Egypt** based on past case data. It uses time series modeling techniques to capture temporal dependencies and predict future case counts. **Dataset**: `egypt_daily_covid.csv` **Columns:** - `date`: Date of record - `country`: Egypt - `daily_new_cases`: Number of newly reported cases The dataset is sorted chronologically and split into: - **80% training set** - **20% testing set** --- ## Steps and Workflow ### 1. **Data Preprocessing** - Handle missing values using forward fill. - Scale values to [0, 1] using `MinMaxScaler`. - Transform data into supervised learning format — use the previous 5 entries to predict the next. ### 2. **Model Architecture** - Sequential LSTM model with: - Input layer - Two LSTM layers - Dense hidden layer - Dense output layer - Compiled using the **Adam optimizer** and **Mean Squared Error (MSE)** loss. - Uses **EarlyStopping** callback to prevent overfitting. ### 3. **Training and Evaluation** - Trained the model on the training set for 8 epochs and validates it on the test set. - Evaluates model performance using **MSE** and visualizes predicted vs. actual values. ### 4. **Results** - mae: 0.0372 - mse: 0.0034 - val_mse: 7.0946e-04 - val_mae: 0.0224