Data mining a high-speed bursty stream on a limited buffer in pseudo-stationary states

Mining a high speed bursty data stream is always a challenge on a limited size buffer. Often a relatively cheaper AMS (anytime mining solution) approach may be the only plausible scheme one could rely on at times. Mining task becomes enormously complicated when the first-level buffer has to host several dependent streams. This becomes worse when incoming data streams take time to settle down in their respective steady states. A buffer sharing and capture models are indicated for some simple situations involving multiple streams. These models could be extended to generalize a linear buffer model to a hierarchical model