Investigation of strategic human resources activity in corporate governance practices: A research with Artificial Neural Networks

Abstract

Corporate governance, which is defined as the laws, rules and elements system controlling the activities in a business and whose principles are fairness, transparency, accountability and responsibility, is of great importance for all companies of national and international nature. The reasons that make corporate governance so important for all market mechanisms are financial crises and large-scale business scandals. Corporate governance, which is constantly regulated by taking into account its historical development and becomes more mandatory every day, is a form of strategic management based on holistic approach and sustainability. This research was interpreted by means of the BIST Corporate Governance Index data, which makes corporate governance arrangements, and the information obtained from the annual reports published on the websites of companies, institutions and organizations. The analysis was carried out with the data of 15 companies, institutions and organizations that can provide common data in the same time series. It is designed to investigate the effectiveness of HRM on 15 companies by applying Artificial Neural Networks (ANN) method, which is one of Artificial Intelligence methods. In the findings of the study, the classification success analysis of the model and the importance of independent variables were determined. It demonstrated that HRM is effective, the most important variable is the fact that it is 100% standardized importance and human resources policy, and 87.3% standardized importance and corporate governance compliance report.

Keywords:

• Artificial Neural Networks
• BIST-Corporate Governance Index
• corporate governance
• corporate governance strategic management human resources management
• human resources management

Notes

1 Artificial Neural Networks (ANN) is one of the artificial intelligence methods developed based on research on how the human brain works and with inspiration from the structure of the biological brain (Çakin & Özdemir, 2019, p. 2019). The first artificial neural network model was developed by Warren McCulloch and Walter Pitts in 1943 (McCulloch & Pitts, 1943). They showed that all kinds of logical situation could be formulated with the artificial neural networks they defined and set the parallel operation principle of artificial neurons to create the first learning rules.

2 The biological neuron consists of four sections including one body, one axone, many nerve endings (dendrite) and thin extensions (synapses) between the axone and the nerve ending of the other neuron. Dendrites transmit incoming signals to the cell. The nucleus gathers the signals received from the dendrite and transmits it to the axone. These gathered signals are processed by the axone and sent to the synapses. The new generated signals in the synapses are transmitted to the other nerve endings (Akdoğan, 2017).

3 The biological neuron consists of four sections including one body, one axone, many nerve endings (dendrite) and thin extensions (synapses) between the axone and the nerve ending of the other neuron. Dendrites transmit incoming signals to the cell. The nucleus gathers the signals received from the dendrite and transmits it to the axone. These gathered signals are processed by the axone and sent to the synapses. The new generated signals in the synapses are transmitted to the other nerve endings (Akdoğan, 2017).

4 Hyperbolic Tangent is similar to the sigmoid function. However, this function is between −1 and 1. It has an S shape. Returning negative value from negative numbers, zero from zero and positive value from positive numbers is one of its advantages. This function is mostly used to separate into two classes. This function is used in feedforward artificial neural networks (Taşkıran, n.d.).

5 The Softmax activation function is used to separate data into three or more classes (Taşkıran, n.d.).

6 Perceptron is based on the principle of one neuron taking multiple inputs and giving one output.

7 The multilayer ANN model receives the inputs from the outside and sends the information to the intermediate layer without processing. Each received information goes to the next layer as it is. The intermediate layers process the information received from the input layer and sends it to the next layer. It processes the information received from the intermediate layers to determine the outputs generated by the network against the given inputs and sends them to the outside. Multilayer networks are used in learning strategies. The network is given both the samples and the outputs that need to be learned from the samples (Kabalci, 2014).