Detecting anomalies in Bitcoin blockchain transactions using machine learning approach and extracting effective features based on Relief algorithm

Anomaly detection is an important technique for recognizing fraudulent activities, suspicious activities, network intrusions, and other unusual events that may be of great importance but difficult to detect. Therefore, the purpose of this research is to investigate and detect anomalies in Bitcoin transactions on the blockchain platform with a machine learning approach, to extract effective features based on Relief algorithm. The steps of the proposed method are as follows: The Bitcoin data set is collected from the Bitcoin Chart site and clustered with cumulative hierarchical clustering, and then feature selection is done using Relief algorithm. In the next step, the data set is analyzed by the method of experimental data analysis and anomalies are detected. In the next step, the data is labeled as normal and abnormal, and a class record is formed. Then, with the support vector machine algorithm classification, the classification and model error is estimated. The evaluation criterion in this research is accuracy and mean square error. The results show that the error of the model with the MSE of the support vector machine was equal to 0.0065. The MSE error value of the proposed model without feature extraction is equal to 0.049, which indicates that feature extraction plays an important role in this model and is one of the new and innovative aspects of this research. Therefore, the proposed method has performed much better by extracting useful features and support vector machine, and the accuracy rate of the model was 91.11%. The accuracy of the proposed method has been compared with other methods, which shows the higher accuracy and better efficiency of this method.