PDF(Pubmed)
Abstract：
For cancer therapy, the focus is now on targeting the chemotherapy drugs to cancer cells without damaging other normal cells. The new materials based on bio-compatible magnetic carriers would be useful for targeted cancer therapy, however understanding their effectiveness should be done. This paper presents a comprehensive analysis of a dataset containing variables x(m), y(m), and U(m/s), where U represents velocity of blood through vessel containing ferrofluid. The effect of external magnetic field on the fluid flow is investigated using a hybrid modeling. The primary aim of this research endeavor was to construct precise and dependable predictive models for velocity, utilizing the provided input variables. Several base models, including K-nearest neighbors (KNN), decision tree (DT), and multilayer perceptron (MLP), were trained and evaluated. Additionally, an ensemble model called AdaBoost was implemented to further enhance the predictive performance. The hyper-parameter optimization technique, specifically the BAT optimization algorithm, was employed to fine-tune the models. The results obtained from the experiments demonstrated the effectiveness of the proposed approach. The combination of the AdaBoost algorithm and the decision tree model yielded a highly impressive score of 0.99783 in terms of R2, indicating a strong predictive performance. Additionally, the model exhibited a low error rate, as evidenced by the root mean square error (RMSE) of 5.2893 × 10-3. Similarly, the AdaBoost-KNN model exhibited a high score of 0.98524 using R2 metric, with an RMSE of 1.3291 × 10-2. Furthermore, the AdaBoost-MLP model obtained a satisfactory R2 score of 0.99603, accompanied by an RMSE of 7.1369 × 10-3.
摘要：
对于癌症治疗，现在的重点是将化疗药物靶向癌细胞而不损害其他正常细胞。基于生物相容性磁性载体的新材料将用于靶向癌症治疗，然而，应该理解它们的有效性。本文对包含变量x(m)的数据集进行了全面分析，y(m),和U（m/s），其中U表示血液通过含有铁磁流体的血管的速度。使用混合模型研究了外部磁场对流体流动的影响。这项研究的主要目的是构建精确可靠的速度预测模型，利用提供的输入变量。几个基本模型，包括K近邻(KNN)，决策树(DT)，和多层感知器(MLP)，进行了培训和评估。此外,实现了一个名为AdaBoost的集成模型，以进一步提高预测性能。超参数优化技术，特别是BAT优化算法，被用来对模型进行微调。实验结果证明了该方法的有效性。AdaBoost算法和决策树模型的组合在R2方面产生了令人印象深刻的0.99783分，表明了强大的预测性能。此外,该模型表现出较低的错误率，如5.2893×10-3的均方根误差(RMSE)所示。同样，AdaBoost-KNN模型使用R2指标表现出0.98524的高分，RMSE为1.3291×10-2。此外，AdaBoost-MLP模型获得令人满意的R2得分为0.99603，RMSE为7.1369×10-3。
