A Catalogue of Machine Learning Algorithms for Healthcare Risk Predictions

Abstract

Extracting useful knowledge from proper data analysis is a very challenging task for efficient and timely decision-making. To achieve this, there exist a plethora of machine learning (ML) algorithms, while, especially in healthcare, this complexity increases due to the domain's requirements for analytics-based risk predictions. This manuscript proposes a data analysis mechanism experimented in diverse healthcare scenarios, towards constructing a catalogue of the most efficient ML algorithms to be used depending on the healthcare scenario's requirements and datasets, for efficiently predicting the onset of a disease. To this context, seven (7) different ML algorithms (Naïve Bayes, K-Nearest Neighbors, Decision Tree, Logistic Regression, Random Forest, Neural Networks, Stochastic Gradient Descent) have been executed on top of diverse healthcare scenarios (stroke, COVID-19, diabetes, breast cancer, kidney disease, heart failure). Based on a variety of performance metrics (accuracy, recall, precision, F1-score, specificity, confusion matrix), it has been identified that a sub-set of ML algorithms are more efficient for timely predictions under specific healthcare scenarios, and that is why the envisioned ML catalogue prioritizes the ML algorithms to be used, depending on the scenarios' nature and needed metrics. Further evaluation must be performed considering additional scenarios, involving state-of-the-art techniques (e.g., cloud deployment, federated ML) for improving the mechanism's efficiency.

Keywords: catalogue; data analysis; healthcare; machine learning; prediction; supervised learning.

Figure 1

Indicative example of BNB steps.

Figure 2

Indicative example of KNN steps.

Figure 3

Indicative example of DT steps.

Figure 4

Indicative example of RF steps.

Figure 5

Indicative example of LR steps.

Figure 6

Indicative example of MLP steps.

Figure 7

Indicative example of SGD steps.

Figure 8

Overall mechanism architecture.

Figure 9

Confusion matrix.

Figure 10

Example of stroke probability form.

Figure 11

Precision results of ML models for each use case.

Figure 12

Recall results of ML models for each use case.

Figure 13

F1-score results of ML models for each use case.

Figure 14

Specificity results of ML models for each use case.

Figure 15

Train–validation–test score for diabetes use case.

Figure 16

Confusion matrix of prediction results for diabetes use case.

Figure 17

Performance comparison in the diabetes use case.

Figure 18

Train–validation–test score for stroke use case.

Figure 19

Confusion matrix of prediction results for stroke use case.

Figure 20

Performance comparison in the stroke use case.

Figure 21

Train–validation–test score for heart failure use case.

Figure 22

Confusion matrix of prediction results for heart failure use case.

Figure 23

Performance comparison in the heart failure use case.

Figure 24

Train–validation–test score for COVID-19 use case.

Figure 25

Confusion matrix of prediction results for COVID-19 use case.

Figure 26

Performance comparison in the COVID-19 use case.

Figure 27

Train–validation–test score for breast cancer use case.

Figure 28

Confusion matrix of prediction results for breast cancer use case.

Figure 29

Performance comparison in the breast cancer use case.

Figure 30

Train–validation–test score for kidney disease use case.

Figure 31

Confusion matrix of prediction results for kidney disease use case.

Figure 32

Performance comparison in the kidney disease use case.

Figure 33

Training performance comparison for each algorithm per dataset.