The aim of this dissertation was to generate a tool for automated search and storage of MS spectra from one of the largest publicly available mass spectral libraries. Extracting data from the websites, processing of the collected structural identifiers from the metadata and data on the chromatographic-spectrometric methods used to generated mass spectra, and their storage in a tabular structure has been achieved by using web scraping, Python, Django, Vue.js and Bootstrap, while the complete source code of the web solution is available on Github.
The tool overcomes the problem of analytical data collection by analysts so far, which significantly speeds up the research process.
Taking the importance of timely tumor classification in account, machine learning methods are used for this approach of classifying tumorous skin images. By using features extraction and artificial neural networks, accuracy from 94% is obtained. The accuracy means that 94% of the images from tumorous part of the skin are correctly classified as benign or malignant, accordingly, and that information can be used in the following medical treatment. Working in Python, using the libraries: spicy, cv2, numpy, imageio, sklearn, Google Colab.
Part of Intelligent Information Systems course.
Prediction of newly discovered compounds' molecular masses from ions' intensities obtained by using mass spectrometry; without knowing their molecular formula. Using Artificial Neural Network, Linear Regression, hybrid Machine Learning models, working mostly in Python.
Part of Bioinformatics course.
Preprocessing on data-set containing patients' prescriptions, finding associations between the prescriptions, and measuring medicines' efficiency, depending whether it is acute or chronical therapy.
Part of Data Mining course.
Analysis of cryptocurrencies' data and creating models for future prediction of their values, by using Machine Learning and Statistics methods.
Part of Intelligent Systems course.