This research project analyzes global fisheries data across capture production, aquaculture, and per-capita consumption. It combines time-series clustering and predictive modelling to identify country-level patterns and forecast future global sustainable level

The project addresses two related challenges - identifying meaningful similarities between countries with different temporal patterns, and predicting future global trends from a relatively small historical dataset. The analysis therefore explores both country-level time-series structure and global-level predictive modelling.

The first part uses K-Means, OPTICS, DBSCAN, and time-series clustering with TSlearn to group countries based on their temporal patterns. The resulting clusters are then examined using multivariate time-series models such as VAR and VECM. The second part uses the global time series to build a predictive pipeline for future sustainable levels. I compare Linear Regression, Bayesian Ridge, SVR, and Random Forest models, using automated hyperparameter search with GridSearchCV and cross-validation to evaluate model performance and generalization.