The objective of this project is to develop a comprehensive forecasting system capable of implementing and comparing various time series models (ARIMA, ANN, Hybrid ARIMA-ANN) across different sectors. Additionally, the system includes a user-friendly front-end interface for visualizing data and forecasts.
- Finance Sector: Monthly stock prices from the S&P 500 index.
- Energy Sector: Hourly energy consumption data.
- Environmental Sector: Daily atmospheric CO2 concentrations.
- Cleaning: Identifying and handling missing values.
- Normalization/Standardization: Scaling data to a uniform range.
- Stationarization: Applying differencing and logarithmic transformations for achieving stationarity.
- Backend: Python with Flask for server-side logic and handling API requests.
- Frontend: HTML/CSS for layout and styling.
- Data Science: Python with Pandas, NumPy, Matplotlib, Seaborn, Statsmodels, TensorFlow/Keras.
- Database: SQLite for storing processed data and results, facilitating quick retrieval for visualization.
- Version Control: Git for code management and version control.
- Purpose: Modeling and forecasting time series data showing non-stationarity or seasonal patterns.
- Process: Identifying order of differencing (d), number of autoregressive terms (p), and number of lagged forecast errors (q) using statistical tests like ADF and ACF/PACF plots.
- Purpose: Modeling complex nonlinear relationships and interactions in data.
- Process: Designing neural networks with varying architectures and implementing backpropagation for training.
- SARIMA: Extending ARIMA for seasonal time series data.
- Exponential Smoothing (ETS): Forecasting with weighted averages of past observations.
- Prophet: Handling time series with strong seasonal effects and historical holidays.
- SVR: Modeling nonlinear relationships using support vector regression.
- LSTM: Suitable for sequence prediction problems.
Combining ARIMA and ANN models to leverage their respective strengths and improve forecast accuracy.
Creating an intuitive UI for dataset selection, model comparisons, and forecast viewing.
Implementing interactive charts and graphs for effective data and forecast presentation.
- Select different forecasting models.
- Display time series data, forecasts, residuals, and accuracy metrics dynamically.
- Upload new data and retrain models interactively.
Ensuring reliability and accuracy through historical data validation and techniques like cross-validation.
Confirming proper functionality of both backend and frontend components without errors.
- ARIMA:
- Best model: ARIMA(1, 2, 1) with MSE of 9.9467.
- ANN:
- Finance Dataset: Loss: 298.91, Metric: 10.59.
- Energy Dataset: Loss: 0.00008, Metric: 0.00732.
- Environment Dataset: Loss: 0.000001, Metric: 0.00086.
- ETS:
- Finance Dataset: MSE: 618.79, MAE: 15.11, RMSE: 24.88, R2: 0.9994.
- Energy Dataset: MSE: 213363.86, MAE: 304.03, RMSE: 461.91, R2: 0.9682.
- Environment Dataset: MSE: 0.267, MAE: 0.358, RMSE: 0.517, R2: 0.9996.
- SVR:
- Finance: MAE: 0.124, MSE: 0.035, RMSE: 0.187, R2: 0.967.
- Energy: MAE: 0.777, MSE: 0.957, RMSE: 0.978, R2: 0.041.
- Environment: MAE: 0.072, MSE: 0.007, RMSE: 0.085, R2: 0.993.
- Hybrid:
- Finance: MAE: 4107.04.
- Energy: MAE: 19893.34.
- Environment: MAE: 406.74.
- Clone the repository:
git clone https://github.com/shitbaKashif/forecasting-system.git - Run Flask:
flask runorpython app.py
- Python3
- SQLite
- Libraries for models (Prophet, SARIMA, Exponential Smoothing)
- TensorFlow
- To directly install the requirements from the requirements.txt file, you can use the following command:
pip install -r requirements.txt
Made by Shitba Kashif and Hiba Imran