Most data mining algorithms and tools stop at the mining and delivery of patterns satisfying expected technical interestingness. There are often many patterns mined but business people either are not interested in them or do not know what follow-up actions to take to support their business decisions. This issue has seriously affected the widespread employment of advanced data mining techniques in greatly promoting enterprise operational quality and productivity. In this paper, we present a formal view of actionable knowledge discovery (AKD) from the system and decision-making perspectives. AKD is a closed optimization problem-solving process from problem definition, framework/model design to actionable pattern discovery, and is designed to deliver operable business rules that can be seamlessly associated or integrated with business processes and systems. To support such processes, we correspondingly propose, formalize, and illustrate four types of generic AKD frameworks: Postanalysis-based AKD, Unified-Interestingness-based AKD, Combined-Mining-based AKD, and Multisource Combined-Mining-based AKD (MSCM-AKD). A real-life case study of MSCM-based AKD is demonstrated to extract debt prevention patterns from social security data. Substantial experiments show that the proposed frameworks are sufficiently general, flexible, and practical to tackle many complex problems and applications by extracting actionable deliverables for instant decision making.

We studied the minimum latency gossiping (all-to-all broadcast) problem in multihop radio networks defined as follows: Each node in the network is preloaded with a message and the objective is to distribute each node's message to the entire network with minimum latency. We studied this problem in the unit-size message model and the unit disk graph model. The unit-size model means different messages cannot be combined as one message, and the unit disk graph model means a link exists between two nodes if and only if their euclidean distance is less than 1. The minimum latency gossiping problem is known to be NP-hard in these two models. In this work, we designed a gossiping scheme that significantly improved all current gossiping algorithms in terms of the approximation ratio. Our work has approximation ratio 27, a great improvement of the current state-of-the-art algorithm (which has ratio 1,947). We also discussed the single point of failure problem and its impact on our approximation ratio. We designed an amended gossiping algorithm with ratio 27 in case of a nonsource node failure. We also designed an amended gossiping algorithm with ratio 29 in case of source failure.

Minimum-latency beaconing schedule (MLBS) in synchronous multihop wireless networks seeks a schedule for beaconing with the shortest latency. This problem is NP-hard even when the interference radius is equal to the transmission radius. All prior works assume that the interference radius is equal to the transmission radius, and the best-known approximation ratio for MLBS under this special interference model is 7. In this paper, we present a new approximation algorithm called strip coloring for MLBS under the general protocol interference model. Its approximation ratio is at most 5 when the interference radius is equal to transmission radius, and is between 3 and 6 in general.

We studied the minimum-latency gossiping (all-to-all broadcast) problem in multi-hop wireless networks defined as follows. Each node in the network is initially given a message and the objective is to design a minimum-latency schedule such that each node distributes its message to all other nodes. We considered the unit-size message model, in which different messages cannot be combined as one message, and the unit disk graph model, in which a link exists between two nodes if and only if their Euclidean distance is less than 1. This problem is known to be NP-hard in such models. In this work we designed a gossiping scheme that significantly improved all current gossiping algorithms in terms of approximation ratio. Our work has approximation ratio 27, a great improvement of the current state-of-the-art algorithm (which has ratio 1000+).

Our work looks into the reconfiguration of virtual topology for wavelength-routed mesh optical network under dynamic traffic demand. When reconfiguring the optical network, it is important to minimize the number of receivers that need to be retuned. It is also important to minimize the average weighted hop count as it is directly proportional to the throughput of the network. The tradeoff here is between the optimality of the network and the network disruption. In our work we first model this tradeoff as mixed integer linear problem and then present a heuristic algorithm for dynamic reconfiguration of optical network. Our algorithm minimizes the physical resources used, physical hops encountered, and the necessary number of retuning of receivers. It tries to satisfy the traffic demand with minimal network disruption.