Peer review networks between Bitcoin traders
Lepomäki, Laura (2019)
2019
Teknis-luonnontieteellinen DI-ohjelma - Degree Programme in Science and Engineering
Tekniikan ja luonnontieteiden tiedekunta - Faculty of Engineering and Natural Sciences
This publication is copyrighted. You may download, display and print it for Your own personal use. Commercial use is prohibited.
Hyväksymispäivämäärä
2019-11-12
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi:tuni-201910254091
https://urn.fi/URN:NBN:fi:tuni-201910254091
Tiivistelmä
Bitcoin is a cryptocurrency that can be traded online. Some of the online Bitcoin trading platforms allow traders to give trust ratings to each other. Trust ratings are meant to indicate with whom to trade. Given and received trust ratings between Bitcoin traders form a Bitcoin trader peer review network. Understanding the functionality of Bitcoin peer review networks is crucial due to counter-party risk in Bitcoin transactions. This work studies the social aspects of Bitcoin trading. Trust rating data from two online Bitcoin trading platforms, Bitcoin OTC and Bitcoin Alpha, is used.
Bitcoin trader behaviour in peer review networks is reduced to five behavioural features: attention, reputation, activity, fairness and goodness. The first three are derived from the data in a straightforward way. The last two are determined by using a state-of- the-art algorithm designed for trust/distrust networks. Trader types are extracted by clustering the traders based on the behavioural features. Due to timestamped data it is possible to define how the behaviour of Bitcoin traders evolve over time. Bitcoin peer review networks are represented as chronological aggregated snapshots of the underlying temporal system. Per each aggregated network, traders are clustered based on their behaviour. Cluster transitions provide information about how Bitcoin trader behaviour evolves over time. This work focuses especially on adverse behaviour. Adverse behaviour refers to giving unfair trust ratings to others or being distrusted by other traders, especially fair ones. The impact of receiving unfair ratings on a trader's behaviour is studied. In addition, it is studied if adversely behaving traders form communities. A community is a group of traders who have been rating each other. Behavioural clusters are also studied in relation to the most and the least central traders. The most central traders substantially contribute to the peer review network while the impact of the least central ones is negligible.
The behavioural clusters show clear similarities between the datasets. There are trader types for which behaviour is exceptionally persistent. For well behaving traders it is common to remain as they are. Distrusted traders are likely to remain distrusted or disappear from the network, which can partly be explained by unfair negative treatment. Unfairly negatively rated traders can react to unfair treatment by becoming unfair themselves. Some of the most reputable traders have received their reputation from unfair positive ratings. Active and noticed traders with medium reputation behave in various ways in the future and are likely to stay in the network. In addition, it is observed that communities of unfairness and distrust emerge in Bitcoin peer review networks the same time with a burst of negative trust ratings. Surprisingly, the results on centrality show that the most well behaving traders become the least central. The most central traders in Bitcoin peer review networks are active and noticed traders who do not behave adversely.
Bitcoin trader behaviour in peer review networks is reduced to five behavioural features: attention, reputation, activity, fairness and goodness. The first three are derived from the data in a straightforward way. The last two are determined by using a state-of- the-art algorithm designed for trust/distrust networks. Trader types are extracted by clustering the traders based on the behavioural features. Due to timestamped data it is possible to define how the behaviour of Bitcoin traders evolve over time. Bitcoin peer review networks are represented as chronological aggregated snapshots of the underlying temporal system. Per each aggregated network, traders are clustered based on their behaviour. Cluster transitions provide information about how Bitcoin trader behaviour evolves over time. This work focuses especially on adverse behaviour. Adverse behaviour refers to giving unfair trust ratings to others or being distrusted by other traders, especially fair ones. The impact of receiving unfair ratings on a trader's behaviour is studied. In addition, it is studied if adversely behaving traders form communities. A community is a group of traders who have been rating each other. Behavioural clusters are also studied in relation to the most and the least central traders. The most central traders substantially contribute to the peer review network while the impact of the least central ones is negligible.
The behavioural clusters show clear similarities between the datasets. There are trader types for which behaviour is exceptionally persistent. For well behaving traders it is common to remain as they are. Distrusted traders are likely to remain distrusted or disappear from the network, which can partly be explained by unfair negative treatment. Unfairly negatively rated traders can react to unfair treatment by becoming unfair themselves. Some of the most reputable traders have received their reputation from unfair positive ratings. Active and noticed traders with medium reputation behave in various ways in the future and are likely to stay in the network. In addition, it is observed that communities of unfairness and distrust emerge in Bitcoin peer review networks the same time with a burst of negative trust ratings. Surprisingly, the results on centrality show that the most well behaving traders become the least central. The most central traders in Bitcoin peer review networks are active and noticed traders who do not behave adversely.