• cryptocurrency
• machine learning
• cryptocurrency exchange
• data analysis

The thesis aims to conduct a data analysis regarding cryptocurrency exchanges using statistical tests often used in financial data such as Benford's Law, trade size clustering on round numbers and analysis of the correlation relationship between trading volume and transaction data.
We also propose a machine learning approach to detect anomalies in trading volumes with convolutional autoencoders using unsupervised learning, as well as an attempt to normalize the trading volumes reported using web traffic data of the exchanges' websites.
The techniques employed can be used as good indicators of potentially fraudulent activities such as wash trading and volume inflation which benefit dishonest cryptocurrency exchanges at the expense of regulated cryptocurrency exchanges which follow mandatory regulatory compliance and by creating a false perception of the cryptocurrency market for customers.