Complex Issues In Artificial Intelligence and How They Are Currently Solved
There is a myriad of complex issues that companies, scientists, and hobbyists, alike, aim to address and solve every day. One of the more recently, popular and rapidly growing fields of study that has had an impact on society across multiple industries to date within the computational sciences is Artificial Intelligence (AI). While there could be a substantive discussion about the impact AI has had, thus far, on all industries, for this discussion, we will emphasize AI’s direct impacts on the medical field.
AI has emerged as a transformative force in the medical field, revolutionizing the way health care is delivered, managed, and diagnosed. One of the primary areas where AI has made a profound impact is disease prediction. In a study published in the Journal of Medical Systems in 2019, titled, Development of Big Data Predictive Analytics Model for Disease Prediction Using Machine Learning Technique, the authors, Venkatesh, Kaliappan, et. al. conducted a study that analyzes a patient’s potential to experience heart disease. The authors acquired their data source from the UCI Machine Learning Repository to build a predictive model that is able to accurately predict a patient’s future health status with a 97.12% accuracy (Venkatesh, Kaliapapan, et. al., 2019).
Techniques such as these prediction models can be instrumental in personalizing treatment plans and improving patient care. Through the analysis of large datasets, AI algorithms can identify patterns and trends that may not be readily apparent to human clinicians. Naturally, this enables providers to tailor treatment based on very specific characteristics such as medical history, and genetic makeup with pinpoint accuracy.
Complex Decision-Making and Branching Logic, An Artificial Intelligence Perspective and Algorithm Analysis
Furthermore, the authors of the aforementioned study used the Naive Bayes machine learning technique as part of their prediction model for determining a patient’s future health status. The Naive Bayes technique is based on Bayes’ theorem which produces independent assumptions based on the defined features of the large data set (Venkatesh, Kaliapapan, et. al., 2019). Traditionally, there are two types of machine learning techniques which include supervised and unsupervised learning. A supervised machine learning technique includes a data set that is labeled while an unsupervised learning technique is unlabeled and allows the algorithm to classify its own data set (Delua, 2021). Some example types of supervised machine learning algorithms are classification and regression modeling problems. The Bayes algorithm would be considered the former (Ventakesh, Kaliappan, et. al, 2019). The implementation and complexity of the Bayes algorithm include several steps which are as follows:
- Training Phase
- Classification Phase
- Naive Assumption
- Handling Zero Probabilities
The training phase, given a labeled data set with features and class labels (i.e. UCI Machine Learning Repository) will calculate the probability of each class and the likelihood of each feature given each class. Once the model is trained, the classification phase will identify and classify new instances. Furthermore, the “Naive” assumption of the Naive Bayse algorithm is the assumption that the presence or absence of one feature does not affect the presence or absence of another feature. Finally, the handling of zero probabilities step is used to smooth out the data to address handling calculating zero probabilities in a real-world-based data set (Ray, 2024). It should be noted, that each step while covered at a relatively high level, has another layer of complexity that is entrenched in statistical and probabilistic algorithms that produce a result set at each step.
A Historical, Current, and a Proposed Approach
Historically, a clinician would need to assess the historical data of a patient, and for distinction and scope, we’ll consider the clinicians’ data to be “local”. This is important so we can emphasize the limited amount of information that the clinician may have about the patient and other patients at any given time. In contrast, a supervised learning algorithm trained on a large data set (i.e. Big Data) may have several petabytes of data it can process at any given time making its reach more “universal”.
Naturally, this historical approach has been effective, as clinicians can train for more than 8 years, hold a vast amount of information about their field, and can identify patterns based on years of experience in their specialties. However, the modern and more current approach is to, in addition to a clinician’s training and experience, leverage machine learning techniques and algorithms to supplement the decision-making process.
As technology advances, there will most certainly be instances where new algorithms are developed and the usage of previous algorithms and techniques will become quickly obsolete. Arguably, AI has proven to, currently, be one of the best solutions to date when considering patient care.
Conclusively, Artificial Intelligence has emerged as a powerful tool for solving numerous challenges in the medical field, from improving diagnostic accuracy and personalized treatment to enhancing patient engagement and access to care. As AI continues to evolve and integrate into healthcare systems worldwide, it holds the potential to revolutionize the way we approach medicine, ultimately leading to better outcomes for patients and a healthier global population.
- Delua, J. (2021, March 12). Supervised vs. unsupervised learning: What’s the difference? IBM Blog. https://www.ibm.com/blog/supervised-vs-unsupervised-learning/
- Ray, S. (2024, February 6). Naive Bayes classifier explained: Applications and practice problems of naive Bayes classifier. Analytics Vidhya. https://www.analyticsvidhya.com/blog/2017/09/naive-bayes-explained
- Venkatesh, R., Balasubramanian, C., & Kaliappan, M. (2019). Development of Big Data Predictive Analytics model for disease prediction using machine learning technique. Journal of Medical Systems, 43(8). https://doi.org/10.1007/s10916-019-1398-y
- Charles, R. (2024, February 11). Unpublished Manuscript. Colorado Technical University.