The growth of digital transformation driven by artificial intelligence (AI) techniques has emerged as a significant agent for change in various sectors. Artificial neural networks (ANNs) are one of the major AI techniques many organizations incorporate to act as a strategic technology. ANNs are computational models capable of learning, representing, and computing functions between multivariate data spaces. They are designed based on the neural structure of the human central nervous system and can learn or generalize from experience. Since ANNs can learn complex patterns and make accurate predictions based on large datasets, they are an invaluable strategic decision-making tool for healthcare and business sectors.
Neural networks are designed using numerous interconnected processing elements called artificial neurons, which collaborate to make sense of the information it receives. The principles of neurons involve receiving and generating signals through input data processing to produce an output. The output is not continuous since it is generated when the input surpasses a certain threshold. Thus, the threshold-based function is an activation operation that converts input to output signals.
The activation function plays a transfer role in obtaining the desired output for the challenge designed. Diverse activation functions exist, including logistic regression, bipolar sigmoid, sigmoidal hyperbolic, linear regression, bipolar binary sigmoid, and identity function. ANNs are designed for specific functions, such as pattern identification, binary categorization, and multi-class sorting, in the learning process. They primarily use a weight modification method that requires an activation function that can be differentiated. Optimization algorithms such as backpropagation compute the connection weights (Shahid, Rappon & Berta, 2019). By adjusting the weights, the network minimizes the error and achieves an output close to the correct value.
While numerous functions exist, most ANNs employ sigmoid functions owing to their simple calculations. The neurons, called nodes, combine the inputs and process the sum using a sigmoid function to provide the output value (Han et al., 2018). As depicted in Figure 1, the neuron produces an output value when the sum of inputs A, B, and C surpasses the threshold, and the sigmoid function is activated.
Figure 1. Input and output information from neurons (Han et al., 2018)
Neurons receive input from and send signals to multiple other neurons. The process is depicted in Figure 2 for artificial neural networks. In artificial neural networks, neurons are arranged in layers, and each neuron is connected to multiple other neurons. The input layer is connected to the hidden layer, and the hidden layer is connected to the output layer. The hidden layer is often called a “black box” because it is difficult to understand how the artificial neural network produces a specific output result.
Figure 2. Multiple layers of an ANN (Han et al., 2018)
The healthcare sector is one of the main fields incorporating AI systems to automate decision-making practices. Using AI in medical diagnosis is an active research field, focusing on artificial neural networks and deep learning for computer-aided diagnosis (Shahid, Rappon & Berta, 2019). ANNs are utilized for early cancer detection by examining vast amounts of medical data to discover patterns that may signify the presence of cancer. One of the most common applications of ANNs in cancer detection is in medical imaging, such as mammography, computerized tomography, and magnetic resonance imaging scans (Pandit & Garg, 2021). ANNs are taught on massive datasets of medical images to categorize them as either cancerous or non-cancerous based on various features such as the lesions’ dimensions, structure, and texture. For instance, an ANN trained on mammography images can identify regions of interest in the picture, such as masses or microcalcifications, and categorize them as either benign or malignant. Ongoing advancements in these models are expected to improve their efficacy, accuracy, and reliability in medical diagnosis.
ANNs have become a popular tool in the business industry for various purposes, such as forecasting, risk management, fraud detection, and customer analytics. They can detect anomalies, uncover hidden insights and learn complex patterns from vast amounts of data to make predictions (Alrammahi & Radif, 2019). Supervised learning techniques train ANN models, where the network is fed with large amounts of labeled data to adjust the network’s weights and biases of the neurons. The process is repeated several times while making adjustments until the network can accurately predict outcomes for new data. In business applications, ANNs are often combined with other machine learning techniques, such as regression and clustering, to improve performance. ANNs can be employed in customer analytics to predict customer behavior, preferences, and future purchases based on past transaction data.
Neural networks are a promising technology that could revolutionize many areas of computation and thinking. Since ANNs allow the automation of tasks, their advanced capabilities may eventually surpass the processing capabilities of the human brain. As researchers develop better training techniques and network designs, various fields such as handwriting, speech recognition, and stock market forecasting could significantly improve. Furthermore, current ANN technologies will likely be substantially enhanced in the future.
ANNs are a vital strategic decision-making tool with the potential to transform various sectors by leveraging their ability to learn complex patterns and make accurate predictions based on large datasets. Their mathematical models imitate the brain’s functionality to extract knowledge from historical data for computation and information-processing purposes. As a branch of AI, they are particularly adept at handling complex nonlinear problems due to their self-learning, self-organizing functions, and high-speed computing capabilities. ANNs have been successfully applied to solve diverse, complex issues in healthcare and business, including prediction, estimation, decision-making, optimization, classification, and selection.
References
Alrammahi, A & Radif, M. (2019). Neural networks in business applications. Journal of Physics Conference Series, 1294(4). http://dx.doi.org/10.1088/1742-6596/1294/4/042007
Han, S., Kim, K. W & Kim, S. C. (2018). Artificial neural network: Understanding the basic concepts without mathematics. Dementia and Neurocognitive Disorders, 17(3). https://doi.org/10.12779%2Fdnd.2018.17.3.83
Pandit, A & Garg, A. (2021). Artificial neural networks in healthcare: A systematic review. 2021 11th International Conference on Cloud Computing, Data Science & Engineering (Confluence). IEEE. https://doi.org/10.1109/Confluence51648.2021.9377086
Shahid, N., Rappon, T & Berta, W. (2019). Applications of artificial neural networks in health care organizational decision-making: A scoping review. Plos One. https://doi.org/10.1371/journal.pone.0212356
write