IDRF: An Improved Dynamic Random Forest Approach for Blockchain Time Series Data Classification

Recently, blockchain time series data has been widely studied throughout the communities of machine learning and data mining. However, Blockchain time series data dynamic class maintenance is still challenging. Existing works on blockchain time series data classification have shown serious accuracy and class maintenance limitations. Therefore, this paper proposes a novel framework called Improved Dynamic Random Forest (IDRF). The proposed framework includes two components as follows: initial classification and class maintenance. For classification, the proposed approach generates an initial set of classes. When new blockchain data arrive, we further proposed an incremental classification approach for maintaining the existing classes dynamically. Experiments on a real world dataset called “Bitcoin Heist Ransom Ware Address” verify the efficiency and effectiveness of the proposed blockchain time series data classification and maintenance approaches in terms of accuracy, execution time and RMSE.