This project focuses on forecasting weather conditions using the Facebook Prophet model. The primary objective is to predict future weather patterns based on historical weather data, with a specific emphasis on optimizing forecast accuracy and computational efficiency for large datasets. We focus on data obtained for NY city, first attempting the analysis on a smaller dataset and then on a larger dataset (https://www.kaggle.com/datasets/danbraswell/new-york-city-weather-18692022).
The analysis is performed on a comprehensive dataset of weather observations, including parameters like temperature, humidity, pressure, and precipitation. The dataset used is extensive, making the computation of forecasts particularly challenging.
-
Data Preprocessing: The dataset undergoes rigorous preprocessing, including cleaning, normalization, and transformation to fit the requirements of the Prophet model.
-
Model Training and Forecasting: The Facebook Prophet model is trained on the historical weather data. Special attention is given to parameter tuning and model optimization to handle the large size of the dataset.
-
Performance Evaluation: The model's performance is evaluated using cross-validation techniques, with a focus on optimizing the computational efficiency of these operations.
-
Forecast Analysis: The forecasts generated by the model are analyzed and compared against actual weather data to assess accuracy and reliability.
- Handling a large dataset efficiently.
- Optimizing the Prophet model parameters for better performance and faster computation.
- Ensuring accurate and reliable forecasts.
- Python
- Pandas and NumPy for data manipulation.
- Facebook Prophet for forecasting.
- Matplotlib and Seaborn for data visualization.