【摘要】 提出了一种结合随机蛙跳与主动学习的木糖含量预测模型。针对木糖分离提纯过程中存在的高维度和冗余特征问题，先采用随机蛙跳算法筛选出对预测性能影响较大的特征子集并构建初步回归模型，然后利用主动学习方法查询边缘样本优化模型性能，再使用加权集成策略合成子模型预测结果。为验证算法的有效性，将不同模型预测性能进行对比，结果显示该模型相较于其他方法能够更好地捕捉木糖含量的变化。研究表明，结合随机蛙跳算法和主动学习的方法能够有效优化木糖含量预测模型，具有较强的泛化能力和应用前景。更多还原

【Abstract】 A xylose content prediction model that has the random leapfrog algorithm and active learning integrated was proposed. Aiming at high dimensionality and redundant features in separating and purifying the xylose, having the random leapfrog algorithm employed at first to select a subset of features that significantly influencing the prediction performance to construct an initial regression model, and then, making use of active learning query marginal samples for further optimization of the model performance, including employing a weighted ensemble strategy to synthesize the predictions of the sub-models. With a view to validating the effectiveness of the proposed algorithm, a comparative analysis of different models’ prediction performance was conducted to demonstrate that the proposed model outperforms traditional methods in capturing variations in xylose content; combining the random frog leap algorithm with active learning can effectively optimize the xylose content prediction model and it has strong generalization ability and promising application prospects.更多还原