Data
Analysis
Data Processing
Big
Data
Cloud Computing
Internet of Things
Information Systems
Review and analysis of methodologies and techniques of machine learning and deep learning techniques in electronic health records
The objective of the thesis will be the study, recording, and comparison of both commercial and research techniques and methodologies of engineering and deep learning that exist and are applied in the electronic health records. The ultimate goal is to record the most common machine learning and in-depth learning architectures for the analysis of health records and the implementation of one of them as well as the identification of future opportunities in the field of machine learning in electronic health records.
Required Skills: Python
Implementation of data prioritization techniques
The objective of the thesis will be the study, development and configuration of data prioritization techniques based on certain criteria and requirements, in order to categorize them into high and low priority data.
Required Skills: Java/ Python
Comparative analysis of AI Explainability tools
The objective of the thesis will be to analyse and compare the various tools and methodologies that exist around from the Explanation of Machine Learning Algorithms as well as the interpretation of their results. A comparison will be made between tools such as LIME, Skater, SHAP, ELI5, etc.
Required Skills: Python, Machine Learning
Analysis of Feature Importance techniques from datasets
The objective of the thesis will be to analyse and implement techniques for extracting important features which help to better understand the data sets as well as optimize the Machine Learning algorithms.
Required Skills: Python, Machine Learning
Sentiment Analysis in financial news feed
The objective of the thesis will be the study and comparison of various natural language processing models that will perform accurate sentiment analysis in financial news. The ultimate goal will be to develop a micro-service that will first pre-process the input text and then predict its class (ie negative, neutral, positive) using the best one of the examined models.
Required Skills: Python, Machine Learning
AI-driven workflows for implementing composite serverless applications
The objective of the thesis focus on the design and implementation of algorithmic models regarding machine and deep machine learning (ML / DL) in order to determine the workflow of applications developed in serverless environments. Recommended tools: Apache OpenWhisk (OW composer, OW action conductors), OpenFaas, AWS lambda, Google (commercial product) Cloud functions (commercial product).
Required Skills: Python, Machine Learning, Cloud Platforms
Distributed ML/DL training using FaaS
The objective of the thesis focus on the re-design of the topology and implementation of distributed training regarding existing machine and deep learning machine models, by exploiting the Function-as-a-Service (FaaS) model. Recommended tools: Apache OpenWhisk OpenFaas, AWS lambda, Google Cloud functions. Additionally, it is advisable to perform a comparative analysis among different cloud platform tools to assess the performance of training in terms of accuracy and time.
Required Skills: Python, Machine Learning, Cloud Platforms
Dynamic resource allocation strategies in edge environments using AI
The objective of the thesis focus on the testing and implemention of machine and deep learning models and techniques (Reinforcement and Federated Learning), in order to allocate -optimally- the edge infrastructure resources (because the edge is a sensitive and constraint environment in terms of resources) for retaining the performance of applications in an acceptable level. (Prerequisite: an edge computing infrastructure i.e. Raspberry Pis, for experimental purposes).
Required Skills: Python, Machine Learning, ΙοΤ
Development of a data-driven approach to facilitate the development of personalized value adding services for business customers of a Bank
The objective of the thesis will be the creation of a Business Financial Managemen (BFM) tool that will be comprised of a Transaciton Categorizaiton, a Transaciton Monitoring, a Budget Prediciton Engine, a KPI Engine and Benchmarking Engine and will be presented in a user friendly Android applicaiton.
Required Skills: Python, ML/DL, Android
Comparative Study of algorithms and techniques establised for “small data” situations
Target of this thesis will be the study and presentation of modern techniques and algorithms developed specifically for situations where the data available are limited in numbers. These techniques include algorithms to enhance and proliferate data such as data enrichment and synthetic data creation. In adiition algorithms such as Few Shot Learning will be addressed that are developed to specifically address the lack of big volumes of data.
Required Skills: Python, Machine Learning, Neural Networks
Automated Machine Learning for Time Series data
This study will explore how Automated Machine Learning, the domain of automated algorithm selection and hyperparameter tuning, can be applied on time series data. Due to the nature of time series data, data the come in certain time intervals such as stock predictions, certain challenges will arise such as when do you need to change the model or its parameters that are used for the prediction implementig concepts that take into account certain aspects such as drifts.
Required Skills: Python, Machine Learning, Optimization
Comparative studies for quality of service for microservice
This thesis is about the compative study of different method for evaluating the compliance of the application’w owner expectations and real application performance.
Required Skills: Java, Python, Go, Docker
Comparative study of techniques and approaches for achieving Entity-Level Sentiment Analysis (ELSA) on Tweets.
The main goal of this thesis is to analyse and compare various approaches and methodologies that can be implemented and utilized for achieving Entity-Level Sentiment Analysis on tweets.
Required Skills: Python, Machine Learning, Deep Learning, Neural Networks
Comparative study of the utilization of Ontologies and Knowledge Graphs for achieving Semantic Interoperability on Electonic Health Records (EHRs)
The main goal of this thesis is to analyse and compare various tools and approaches from the domains of Semantic Web, such as Ontologies, RDFs, JSON-LD, in order to achieve high Semantic Interoperability and Expressivity in the healthcare domain and more specific on patiets’ Electronic Health Records (EHRs)
Required Skills: Python, Machine Learning, Semantic Web, SparkQL