Exploring Agent-Based Methods for Activity Network Prediction in Real Time Gross Settlement (RTGS) Based on Clearing House

摘要

This paper presents an exploratory agent-based model for activity network prediction in RTGS based on clearing house. The settlement process, which is developed, is managed by clearing house. The sufficiency values are fulfilled by other banks that are participant banks at clearing house. At this paper, the participant banks are depicted by node. Decision to fulfill settlement process from other banks is done by forming activity network. At this paper, the activity network is depicted by edge. The forming of activity network is depending on information from agents. This paper adopts decentralization paradigm for modeling activity network. The result of this research indicates that amount of nodes are same, more of activity networks between banks hence are more network identified. While degree between 0 to 2, the smaller difference of the health bank level, that will forming network, the greater network are identified. For degree which more than 2, no network are detected, this condition is caused of fixed point behavior. Testing with λ = 0.7, x starts with degradation drastically towards stable value come near 0. λ = 1.8, x starts with degradation drastically value towards stable value, but stability value still not yet come near 0. λ = 2.9, x oscillates about fixed point and converges. While λ = 3.9, x starts to oscillate.