The aim of this project was to gain experience in applying machine learning methodology to a real set of data. The focus was on good understanding and justification of steps involving data processing, training and evaluation. The data provided contained various measurements taken from 13,611 instances of 7 different types of dry beans.
The main aim of this project was to gain experience in working with real experimental, imperfect, and limited data which has not been analysed before. Initially the data was read, processed and cleaned from the dataset in a suitable way. Then, a classification model was created to predict the output classes based on a set of inputs, and evaluate its performance. It was particularly important that the evaluation was intricate, discussing limitations of the data and the methods used.
The project involved two datasets containing features extracted from small images (60×60 pixels). The small images were cropped from larger aerial images obtained during seasonal surveys of islands in the North Sea. The task was to classify (Binary and Multi) these images based on what type of seal pup was displayed.