本計畫將以??時間?探討整合資?與知?於決策支援的相關問題。隨著資訊硬體技術的效能提升，高速及雲端運算的應用快速蓬勃發展，企業面對的決策問題除?具備大??據的條件外，?需挑戰隨著時間?斷?新的龐大?據計算。因此，大??據以點與點相?方式形成知?網，而大?且多樣化的資?型態?強化知?網中關?性的非線性，目前的知?網建構研究多集中於大??據的??與結構學習，而其中資?的假設則有完整資?學習與加入修補機制的?完整資?學習，但在目前資訊快速?新的環境下，?種假設都無法提供企業即時的決策支援。本計畫第一?將探討因新資?而?斷?新的知?關?性，提出以貝氏網?建構整合新資?與既有知?網的快速學習機制，透過線權重值的?新驅動知?網的結構學習。第二?則以第一?建?的快速?新學習機制為基礎，進一步探討資?與結構的相似性，藉由衡?原始資?集與所建構模式產生的資?集之間的相似性，?檢測所建構的模式是否可以正確的描述資?的真實結構。 Learning in effective way makes enterprise more competitive in the rapid changing world. For the purpose of improving learning effectiveness, combination of user knowledge and statistical data has become an important issue in Bayesian network related researches. This proposal focused on the relevance between random variables, describe an approach combining user knowledge and statistical data for learning Bayesian networks. Bayesian network is a graphical model encoding probabilistic relationships among a set of variables. It’s a knowledge representation tool, based on the new information we get can easily extended or reduced the network to fit the changing knowledge. However, it‘s heavy reliance on domain knowledge while constructing Bayesian network. To address this challenge, this proposal proposes techniques for effective and efficient iterative Bayesian structure learning. In first year, we imply entropy-based link strength technique which represents the belief of the relationship between random variables. Then, combine prior knowledge and statistical data to construction the iterative learning mechanism. Based on the Bayesian network learned in first year, it brings out the issue how to evaluate the similarity between constructed model and the real data-structure. Therefore, in second year, we will construct a data-structure similarity comparative mechanism to evaluate the Bayesian network which combining data and knowledge
