宁波大学学报(理工版)
寧波大學學報(理工版)
저파대학학보(리공판)
JOURNAL OF NINGBO UNIVERSITY(NSEE)
2013年
3期
40-44
,共5页
贝叶斯网%混合模型%数据挖掘%升学预测
貝葉斯網%混閤模型%數據挖掘%升學預測
패협사망%혼합모형%수거알굴%승학예측
bayesian networks%cooperative model%data mining%examination predicting
给出了基于MAP和MDL混合机制的贝叶斯网络结构学习算法,新算法吸取了两种方法各自的特点,具有计算简单、收敛速度快且能综合利用先验知识及专家知识的优点。并结合A校研究生的海量数据进行实验,结果表明:新的预测模型准确率可达84%,且推理高效合理。
給齣瞭基于MAP和MDL混閤機製的貝葉斯網絡結構學習算法,新算法吸取瞭兩種方法各自的特點,具有計算簡單、收斂速度快且能綜閤利用先驗知識及專傢知識的優點。併結閤A校研究生的海量數據進行實驗,結果錶明:新的預測模型準確率可達84%,且推理高效閤理。
급출료기우MAP화MDL혼합궤제적패협사망락결구학습산법,신산법흡취료량충방법각자적특점,구유계산간단、수렴속도쾌차능종합이용선험지식급전가지식적우점。병결합A교연구생적해량수거진행실험,결과표명:신적예측모형준학솔가체84%,차추리고효합리。
Based on the Maximum Aposterior Probability (MAP) and Minimum Description Length (MDL), this paper presents an algorithm of Bayesian networks structure learning. The algorithm effectively combines the characteristics of two methods, and takes the full advantage of simplistic computation, rapid convergence and priori knowledge and expert system. The experiment with mass data collected from a University graduate entrance examination shows that the model-based algorithm can efficiently be used for prediction and the success rate reaches 84%.