The Whale Optimization Algorithm (WOA) is a recently developed algorithm that is based on the chasing mechanism of humpback whales. Benefiting from the unique structure, WOA has virtuous global search capability. One of the drawbacks of this algorithm is the slow convergence rate that limits its real-world application. In resolving complicated global optimization problems, without any exertion for adequate fine-tuning preliminary constraints, Teaching-learning-based optimization (TLBO) is smooth to plunge into local optimal, but it has a fast convergence speed. By given the features of WOA and TLBO, an active hybrid WOA-TLBO algorithm is proposed for resolving optimization difficulties. To explore the enactment of the proposed WOA-TLBO algorithm, several experimentations are accompanied by regular benchmark test functions and compared with six other algorithms. The investigational outcomes indicate the more magnificent concert of the proposed WOA-TLBO algorithm for the benchmark function results. The proposed method has also been applied to the Facial Emotion Recognition (FER) functional problem. FER is the thought-provoking investigation zone that empowers us to classify the expression of the human face in everyday life. Centered on the portions’ actions in the human face, the maximum of the standard approaches fail to distinguish the expressions precisely as the expressions. In this paper, we have proposed FER’s productive process using WOA-TLBO based MultiSVNN (Multi-Support Vector Neural Network). Investigational outcomes deliver an indication of the virtuous enactment of the proposed technique resolutions in terms of accurateness.

Whale Optimization Algorithm (WOA) 是最近开发的一种算法，它基于座头鲸的追逐机制。受益于独特的结构，WOA具有良性的全球搜索能力。该算法的缺点之一是收敛速度慢，限制了其实际应用。在解决复杂的全局优化问题时，无需付出足够的微调初步约束，基于教学的优化（TLBO）可以平滑地陷入局部最优，但它具有较快的收敛速度。针对WOA和TLBO的特点，提出了一种主动混合WOA-TLBO算法来解决优化难题。为了探索所提出的 WOA-TLBO 算法的制定，一些实验伴随着常规的基准测试功能，并与其他六种算法进行了比较。研究结果表明，所提出的 WOA-TLBO 算法对于基准函数结果的效果更加出色。所提出的方法也已应用于面部情绪识别（FER）功能问题。FER 是发人深省的调查区，它使我们能够对日常生活中的人脸表情进行分类。以人脸部分的动作为中心，标准方法的最大值无法准确区分表情和表情。在本文中，我们提出了使用基于 WOA-TLBO 的 MultiSVNN（多支持向量神经网络）的 FER 生产过程。