Data Mining for Detecting Bitcoin Ponzi Schemes
这篇文章使用不同的监督式机器学习算法检测比特币庞氏骗局,并评测出最优算法,随机森林算法(交叉验证中的测试集中庞氏-非庞氏比例为1:20)。并且找到了庞氏骗局具有识别度的多个特征。 首先爬取庞氏骗局地址,并使用聚类分析算法找到了相应32个地址的子地址。将识别转化为二元分类问题(庞氏-非庞氏),其中对于类不均衡问题(class imbalance problem)进行处理,这也设计到了两个算法,基于样本的方法(sample-based)和成本敏感(cost-sensitive)的方法。 检测模型使用了RIPPER、贝叶斯网络、随机森林。 评测的参数包括真假阳性、真假阴性和基于这4个参数的计算等等。 数据测试则是采用交叉验证的方法,并对测试数据集中的庞氏骗局比例进行处理。 最后找出最优算法,随机森林(1:20),并基于这个算法对其他庞氏骗局的数据进行了计算,并找出了最具辨识度的多个特征。
Ponzi scheme data mining
Predicting the Price of Bitcoin Using Machine Learning
This paper introduces a deep learning model based on long short term memory network to analyze the bitcoin price in USD, which trumps the popular time series forecasting models like ARIMA. The results of the experiment is laudable, and has practical reference value to data mining using deep learning in the blockchain field.
Bitcoin data mining
Inferring the interplay between network structure and market effects in Bitcoin
This paper used Principal Component Analysis to extract important features of the time variation of Bitcoin transaction network and investigated the relation between those features and the exchange price of bitcoins. The experiment results showed high correspondence between the first singular vector of the graph of long-term bitcoin users and the market price of bitcoins, which are interesting and important for digital assets value estimation.
Bitcoin Transaction Graph data mining
Using Time-Series and Sentiment Analysis to Detect the Determinants of Bitcoin Prices
This paper tried to shed light on the factors determining the price of Bitcoins in the short-run as well as in the long-run. The authors built an empirical model incorporating multiple economic and technological variables but also extended the existing literature by taking Twitter sentiment into account. The experimental results show that some variables have a positive effect on the price of Bitcoins, and on the contrary, some variables have negative effects. This paper helps the researchers to batter understanding the factors that influence the price of Bitcoin.
Bitcoin Cryptocurrency data mining
T-EDGE: Temporal WEighted MultiDiGraph Embedding for Ethereum Transaction Network Analysis
This paper proposed a novel Temporal Weighted Multidigraph (TWMDG) method to perform node embedding on Ethereum transaction network. TWMDG incorporated temporal and weighted information of financial transaction networks into node embeddings and obtain better experimental results against baselines on two tasks: link prediction and node classification. It was the first work to understand Ethereum transaction records via graph embedding.
Transaction Graph Ethereum data mining