【摘要】 连续型属性的离散化问题是机器学习中的关键问题 ,是一个NP难题 .该文针对决策表 ,在NaiveScaler算法的基础上 ,给出了一种直观、有效和易于理解的离散化方法 .该方法从整个属性空间的角度来考虑属性的离散化问题 ,可有效地保证决策表中原有分类结果的不变性 .更多还原

【Abstract】 Discretization of continuous type of attributes is a key issue in machine learning, it is a NP puzzle. This paper, aiming at decision tables, presents an effective and easily understandable discretization method based on Naive Scaler algorithm. it considers simultaneously, not separately, all attributes in decision tables, consequently, the break-points obtained are much less than that obtained by Naive Scaler algorithm, while the original classification property of the decision table keeps invariant.更多还原