Abstract

The ability to foresee financial distress has become an important subject of research as it can provide the organization with early warning. Furthermore, predicting financial distress is also of benefit to investors and creditors. In this paper, we propose a hybrid approach with Multi-Layer Perceptron and Genetic Algorithm for Financial Distress Prediction. There are numerous hyperparameters that can be tuned to improve the predictive performance of a neural network. We focus on genetic algorithm-based tuning of the main four hyperparameters namely Network depth, Network width, Dense layer activation function, and Network optimizer, which can make a difference in the algorithm exploding or converging. The main objective of this study is to tune the hyperparameters of the Multi-Layer Perceptron (MLP) model using an improved genetic algorithm. The prediction performance is evaluated using real data set with samples of companies from countries in MENA region. All the experiments in this study apply the technique of resampling using k-fold evaluation metrics, to get unbiased and most accurate results. The simulation results demonstrate that the proposed hybrid model outperforms the classical machine learning models in terms of predictive accuracy.

中文翻译：

---

遗传算法的多层感知器财务困境预测

摘要

预测财务危机的能力已成为研究的重要课题，因为它可以为组织提供预警。此外，预测财务危机对投资者和债权人也有利。在本文中，我们提出了一种结合多层感知器和遗传算法的财务困境预测方法。有许多超参数可以调整以提高神经网络的预测性能。我们专注于基于遗传算法的主要四个超参数的调整，这四个超参数分别是网络深度，网络宽度，密集层激活函数和网络优化器，它们可以在算法爆炸或收敛方面有所作为。这项研究的主要目的是使用改进的遗传算法来调整多层感知器（MLP）模型的超参数。使用来自中东和北非地区国家公司的样本的真实数据集来评估预测性能。本研究中的所有实验均采用了重采样技术k倍评估指标，以获得无偏见和最准确的结果。仿真结果表明，提出的混合模型在预测准确性方面优于经典的机器学习模型。