This repository contains code for an Automated Learning and Data Analysis project focused on predicting picnic weather suitability in Dresden, Europe. The project emphasizes interpretability in weather prediction models and utilizes Decision Tree (DT), Attention Neural Network (Att-NN), and Artificial Neural Network (ANN) models.
The project aimed to predict a day's weather suitability for picnics in Dresden, emphasizing interpretability in weather prediction models. It explored yearly weather patterns and sought insights into influential factors and seasonal weather trends impacting picnic-friendliness.
- Model Performances: Decision Tree model exhibited exceptional accuracy, precision, recall, and F1-score, outperforming neural network models due to dataset simplicity.
- Feature Importance: Maximum temperature and precipitation collectively accounted for 100% importance, highlighting their pivotal role in picnic suitability prediction.
- Yearly Weather Patterns: Analysis using kernel density estimates visualized seasonal variations in weather variables, offering valuable insights.
The dataset used in this project can be found here. Please download the dataset to replicate the experiments and run the code.
models.py: Implements and evaluates the Decision Tree, Attention Neural Network, and Artificial Neural Network models for the classification task.
- Ensure required dependencies are installed.
- Run
models.pyto execute the models and observe their performances.
The project successfully employed machine learning models to classify picnic-friendly weather conditions and provided valuable insights into seasonal weather patterns. Future studies could consider expanding datasets and including additional factors for a more comprehensive analysis.
Feel free to explore the code and adapt it for your own projects!
This project was completed as a course project for the Automated Learning and Data Analysis course (CSC 522) at NC State University.