本文介绍了一种名为肝癌算法（LCA）的新型仿生优化算法，该算法模拟了肝癌生长和接管过程。它采用了一种进化搜索方法，模拟了肝癌接管肝脏器官时的行为。肿瘤的复制和扩散能力为该算法提供了灵感。LCA算法使用了遗传算子和随机对抗学习（ROBL）策略，以高效平衡局部和全局搜索，并探索搜索空间。该算法的效率在2020年IEEE进化计算大会（CEC'2020）基准函数上进行了测试，并与包括遗传算法（GA）、粒子群优化（PSO）、差分进化（DE）、自适应引导差分进化算法（AGDE）、改进的多算子差分进化（IMODE）、哈里斯鹰优化（HHO）、龙格—库塔优化算法（RUN）、向量加权均值（INFO）和冠状病毒群体免疫优化器（CHIO）在内的七种广泛使用的元启发式算法进行了比较。收敛曲线、箱线图、参数空间和定量指标的统计结果表明，与著名算法相比，LCA算法具有竞争力。此外，LCA算法的通用性超出了数学基准问题的范围。它还成功地应用于特征选择问题，并优化了支持向量机用于各种生物医学数据分类，从而创建了LCA-SVM模型。 LCA-SVM模型在总共12个数据集中进行了评估，其中单胺氧化酶（MAO）数据集表现出色，相比其他数据集，显示出最高的性能。特别是，在MAO数据集上，LCA-SVM模型实现了令人印象深刻的98.704％的准确度。这一出色的结果显示了LCA-SVM方法处理复杂数据集和产生高度准确预测的功效和潜力。实验结果表明，LCA算法超越了其他方法来解决数学基准问题和特征选择问题。版权所有 © 2023 Elsevier Ltd. 保留所有权利。

This paper introduces a new bio-inspired optimization algorithm named the Liver Cancer Algorithm (LCA), which mimics the liver tumor growth and takeover process. It uses an evolutionary search approach that simulates the behavior of liver tumors when taking over the liver organ. The tumor's ability to replicate and spread to other organs inspires the algorithm. LCA algorithm is developed using genetic operators and a Random Opposition-Based Learning (ROBL) strategy to efficiently balance local and global searches and explore the search space. The algorithm's efficiency is tested on the IEEE Congress of Evolutionary Computation in 2020 (CEC'2020) benchmark functions and compared to seven widely used metaheuristic algorithms, including Genetic Algorithm (GA), particle swarm optimization (PSO), Differential Evolution (DE), Adaptive Guided Differential Evolution Algorithm (AGDE), Improved Multi-Operator Differential Evolution (IMODE), Harris Hawks Optimization (HHO), Runge-Kutta Optimization Algorithm (RUN), weIghted meaN oF vectOrs (INFO), and Coronavirus Herd Immunity Optimizer (CHIO). The statistical results of the convergence curve, boxplot, parameter space, and qualitative metrics show that the LCA algorithm performs competitively compared to well-known algorithms. Moreover, the versatility of the LCA algorithm extends beyond mathematical benchmark problems. It was also successfully applied to tackle the feature selection problem and optimize the support vector machine for various biomedical data classifications, resulting in the creation of the LCA-SVM model. The LCA-SVM model was evaluated in a total of twelve datasets, among which the MonoAmine Oxidase (MAO) dataset stood out, showing the highest performance compared to the other datasets. In particular, the LCA-SVM model achieved an impressive accuracy of 98.704% on the MAO dataset. This outstanding result demonstrates the efficacy and potential of the LCA-SVM approach in handling complex datasets and producing highly accurate predictions. The experimental results indicate that the LCA algorithm surpasses other methods to solve mathematical benchmark problems and feature selection.Copyright © 2023 Elsevier Ltd. All rights reserved.