This is an exploratory data analysis of chocolate ratings data from Kaggle. The dataset contains almost 1800 expert ratings of various chocolate bars from different types of cocoa beans, origins, companies and manufacturing countries. Each rating also contains additional information about the specific bean origin (if known) and information about the percentage of cocoa.

Use of K-Means, Gaussian Mixture Models and HDBSCAN

Use and Comparison of Local and Population-based Search Methods for Concrete Mixture Optimization

Predict the heat capacity, thermal conductivity and density of cow's milk with temperature, water and fat content The objective of this document is to develop statistical models that predict the heat capacity, thermal conductivity and density of milk with various concentrations of fat and water. Various polynomials functions were trained and tested on a training set using best subset regression. The final models selected were assessed on a hold-out test set.

Use various search methods in finding the optimal values of hyperparameters of a machine learning model. The population-based search methods to be tested are genetic algorithms (GA), differential evolution (DE) and particle swarm optimization (PSO). Grid and random search will also be performed and used as benchmarks.

The goal of this project was to develop a prediction algorithm to predict words based on previous text and create an user interface that can be accessed by others. The project is part of the capstone project of the Data Science Specialization by Johns Hopkins University in partnership with Swiftkey.