Jana Nikolovska
Master’s Degree in Artificial Intelligence, University of Bologna
📧: jana.nikolovska@studio.unibo.it
👤: Linkedin profile
The aim of this study is to develop a Dynamic Bayesian Network (DBN) to understand the migration patterns of the white stork, leveraging climate and vegetation indicators to uncover the underlying relationships. The model is constructed using the pgmpy library in Python, fitting it with time-series data gathered from various sources after integration and preprocessing. The fundamental concepts are explored, including Markov blanket, Independence, Sampling, and Inference within the context of the DBN, along with calculated Conditional Probability Distributions (CPDs) and Discretized Factors
InitialDataExploration.ipynb - Exploring the dataset obtained by MoveBank 'Eastern flyway spring migration of adult white storks (data from Rotics et al. 2018).csv'.
- Load full dataset, describe dataset, feature selection based on relevance, transform columns
gather_data.py - Enrich the dataset with NDVI indicator calculated by using Google Earth API and climate indicators calculated using Open-Meteo API
Preprocess_data.ipynb - Analyse and transform the, now, enriched dataset, transforming it into an input for pgmpy's Dynamic Bayesian Network
Build_analyse_DBN.ipynb - Build the Dynamic Bayesian Network, compute CPDs, Discretized factors, Markov blanket, Independencies, Inference, Sampling
utils\utils.py - helper functions for gather_data.py
data\Labels_dictionary_for discretization.txt - dictionary explaining the categories after discretization of the features
For more information about the project read REPORT.pdf