Recently, many studies indicated that the minimum spanning tree (MST) network whose metric distance is de?ned by using correlation coe?cients have strong implications on extracting infor- mation from return time series. However in many cases researchers may hope to investigate the strength of interactions but not the directions of them. In order to study the strength of interaction and connection of ?nancial asset returns we propose a modi?ed minimum spanning tree network whose metric distance is de?ned from absolute cross-correlation coe?cients. We had investigated 69 daily ?nancial time series, which constituted by 3 types ?nance assets (29 stock market indica- tor time series, 21 currency futures price time series and 19 commodity futures price time series). Empirical analyses show that the MST network of returns is time-dependent in overall structure, while same type ?nancial assets usually keep stable inter-connections. Moreover each asset in same group show similar economic characters. In other words, each group concerned with one kind of traditional ?nancial commodity. In addition, we ?nd the time-lag between stock market indicator volatility time series and EUA (EU allowances), WTI (West Texas Intermediate) volatility time series. The peak of cross-correlation function of volatility time series between EUA (or WTI) and stock market indicators show a signi?cant time shift (> 20days) from 0.