Unit:
Κατεύθυνση Ηλεκτρονικός Αυτοματισμός (Η/Α, με πρόσθετη εξειδίκευση στην Πληροφορική και στα πληροφοριακά συστήματα)Library of the School of Science
Author:
Pelekanou Antonia
Supervisors info:
Άννα Τζανακάκη, Επίκουρη Καθηγήτρια, Τμήμα Φυσικής, ΕΚΠΑ
Διονύσιος Ι. Ρεΐσης, Αναπληρωτής Καθηγητής, Τμήμα Φυσικής, ΕΚΠΑ
Έκτορας Ε. Νισταζάκης, Αναπληρωτής Καθηγητής, Τμήμα Φυσικής, ΕΚΠΑ
Original Title:
Application of Machine Learning techniques in Cloud Services
Translated title:
Application of Machine Learning techniques in Cloud Services
Summary:
Machine Learning and Cloud Computing are gaining increased popularity over that past few years. The reason behind this is the rapid increase of the volume of data and the necessity of their fast processing, in order to extract useful information. The goal of this master thesis is to investigate the application of machine learning techniques in support of cloud services. In this work, we focus our effort to improve the process and analysis of the data of a real railway system (the Reims tramway) by forecasting the train power consumption. We develop a machine learning technique based on Neural Networks to process the dataset composed of various physical quantities related with the Reims tramway, each one measured periodically every second during the period of a day. The measurements were collected by different sensors, which were installed on the train. Initially, we provide the required theoretical background and then we present some relevant experimental results. The theoretical background involves the introduction of the concept of Big Data, the presentation of different Data Mining methods and Machine Learning algorithms that can be used in data processing and a review on Cloud Computing, Internet of Things (IoT) and SiteWhere. In Chapter 2, the problem definition is presented, and the terms of Cloud Computing and Internet of Things are introduced. In Chapter 3, we introduce the notions of Big Data, Data Mining and Machine Learning. In Chapter 4, we begin with the presentation of the architecture and capabilities of SiteWhere and we conclude with the visualization of the train data, using a Graphical User Interface (GUI), referred to as MongoDB Compass. The train dataset was stored in the MongoDB database, which is supported by the SiteWhere platform. Chapter 5 provides a brief overview of Neural Networks and the relevant basic algorithms. In Chapter 6 we introduce the concept of Time Series, implement two different types of Neural Networks, the Multilayer Perceptrons (MLP) and the Long Short-Term Memory (LSTM) on the train dataset and we present a set of relevant results. Chapter 7 provides the summary of the thesis and the conclusions derived and presented in the previous chapters.
Main subject category:
Science
Keywords:
Big Data, Cloud, IoT, Machine Learning, Neural Networks, Forecasting
