ABSTRACT

This research examines return predictability based on minimized forecast errors of CSR Indices through the grey relational analysis (GRA) and three types of artificial neural networks (ANN) model, namely: back-propagation perceptron (BPN); recurrent neural network (RNN); and radial basis function neural network (RBFNN), to capture non-linear characteristics of CSR indices for better forecasting accuracy. The study finds that the BPN model has the lowest forecast error, outperforming the RNN and RBFNN models. The model is also consistently better in using the 33% testing data. On the other hand, both the RNN and the RBFNN models preferred the 50% testing data. Based on the GRA rankings, the US Dollar Index and the S&P 500 index are the 1st and 2nd ranking variable, respectively. For the BPN and RNN models, the study experienced the lowest mean absolute error and root mean square errors when using the All Variables group.

摘要

本研究通过灰色关联分析（GRA）和三种类型的人工神经网络（ANN）模型，检验基于最小化企业社会责任指数预测误差的回报可预测性，即：反向传播感知器（BPN）；循环神经网络（RNN）；和径向基函数神经网络（RBFNN），捕捉企业社会责任指数的非线性特征，以提高预测精度。研究发现，BPN 模型的预测误差最低，优于 RNN 和 RBFNN 模型。该模型在使用 33% 测试数据方面也始终表现得更好。另一方面，RNN 和 RBFNN 模型都更喜欢 50% 的测试数据。根据GRA排名，美元指数和标准普尔500指数分别是第一和第二排名变量。对于 BPN 和 RNN 模型，