Towards successful aging classification using machine learning algorithms

System architecture

Figure 1 depicts the main steps of the proposed system architecture. Data pre-processing, feature selection, model construction, performance measurements, and successfully classifying aging are among the phases. The original dataset, which is unstructured and not beneficial for the design of the model, is created during the data pre-processing stage by sorting out variables from the electronic medical records of elderly patients. Additionally, feature selection was utilized to separate the features that are redundant from the features that support accurate prediction in the system and to extract the significant variables that are relevant to successful aging. Thirdly, Support Vector Machine, Naive Bayes, and Artificial Neural Network were among the prediction models developed for this study. During the developmental process, particular hyperparameters for the SVM and NB models were carefully selected. Thereafter, the preprocessed data, was inputted into the model, using it to train the classification model. A better learning pattern for predicting successful aging is provided by the ANN model in combination with a dropout strategy. The combined technique offers increased classification accuracy, specificity, and sensitivity when compared to the other models. ReLu and sigmoid activation functions are used in the network’s hidden and output layers during the training and testing phases, respectively. Initially, three core classification algorithms, namely artificial neural network (ANN), support vector machine (SVM), and naive Bayes (NB) models, underwent training with the purpose of identifying whether an individual possessed the status of SA or non-SA. Then, the best hyperparameters and training ratios are used to increase the models’ predictive accuracy.

Figure 1. Proposed system architecture.

Study parameters

The proposed study will be a retrospective analysis focusing on the sociodemographic, clinical, behavioural, and psychological parameters of specific patients. It will access the medical records of patients aged 60 years or older.²⁹ Multiple variables have been identified based on preliminary data gathered. A machine learning model will be used to analyze the data, and the most useful features will be taken out and used as input parameters. In addition, the factors associated with successful aging will be determined through interactions with gerontology experts and analyses of relevant literature.

The sociodemographic factors that will be considered are age, sex, reading proficiency, marital status, occupation, and income level. The clinical factors include diseases such as hypertension, heart disease, kidney, liver, bone, and muscular disorders, as well as depression, eye and eyelid disorders, diabetes, and cancer. The behavioural and psychosocial factors include the ability to carry out activities of daily living (ADLs), life satisfaction, quality of life (QOL), a healthy lifestyle, interpersonal relationships, nutrition, physical activity, illness prevention strategies, and stress management. The sociodemographic and clinical information was taken from the medical records of elderly individuals.

The outcome variable will be divided into two classes: SA-related (coded 1) and non-SA-related (coded 0). To quantify SA, Rowe and Khan’s approach.¹⁶ SA which consists of maintaining good mental and physical function, ongoing involvement in life, and lack of disease and disease-related disability, will be used.

Data collection and pre-processing

This study involves a dataset of 2000 older individuals who were included in the Afe Babalola University Multi-System Hospital’s electronic medical record in Nigeria. The research committee of the hospital approved the study (Approval Number: AMSH/REC/AVA/133) with the requirement of adhering to all international guidelines and regulations. The data collected for this research was retrospective and did not involve interviews with participants. All data used in this research were anonymized and did not reveal the patient’s identity in any form. The dataset was collected from the electronic medical records of the Afe Babalola University Teaching Hospital between January 2019 and April 2023 in Nigeria and was pre-processed using various techniques to create optimal models. The dataset had some missing values, but these were not excluded as they could contain important information to influence predictions. Instead, the missing values were filled in using the mean of the corresponding feature in the dataset. Another issue was the imbalance of data class distribution. This was corrected to make the dataset more effective for machine learning algorithms.³⁰ After pre-processing and balancing, the dataset included 2000 records of both successfully aged and unsuccessfully aged persons. The data was then categorized as SA, based on the absence of two or fewer diseases, quality of life, nutrition, and the capacity for daily activities. After categorization, there were 1100 SA and 900 Non-SA instances in the dataset.

Feature selection

Feature selection is a crucial technique in data mining that helps to eliminate redundant and irrelevant features, thus reducing the dataset dimension and enhancing machine learning (ML) performance. It is an essential method that filters out duplicate and unrelated features from the dataset. By using statistical techniques to reduce the dataset dimension, feature selection can help to improve comprehension, avoid algorithm overfitting, boost processing power, and enhance mining performance.^31,32 Our study involved working with gerontologists to identify the features that contribute to successful aging in the elderly population. Using Rowe and Khan’s model for successful ageing, we sorted features based on their criteria for the absence of disability (met when adults have no disability and the number of chronic diseases is less than or equal to two), quality of life (based on continued engagement with life through employment, social, religious, and volunteering activities), and ability to carry out daily activities (physical and mental functions). After careful consideration, we determined that the most relevant features for defining successful aging in elderly people from the dataset are age, hypertension/CVD, renal illness, liver disease, neuromuscular disease, depression, eye disease, diabetes, cancer, ADLs, nutrition status, and QOL. To improve the precision, sensitivity, and accuracy of all produced models, we removed information such as marital status, occupation, educational level, gender, and hospital department from the dataset. Figure 2 shows the correlation between selected features in the extracted dataset, and a maximum correlation is observed between renal disease and the ability to carry out daily activities.

Figure 2. Correlation between features of the dataset.

Figure 3 denotes the variables in the dataset including the feature name, feature type, and concept codes. In the dataset, NMD represents all categories of neuromuscular diseases, while QOL refers to the quality of life of each individual, and lastly ADLs represent the ability of aged people to carry out daily activities regularly.

Figure 3. Cross-section of the input dataset.

Development of classification models

For SA prediction, three supervised learning approaches were used. The elderly individuals were classified as either SAs or non-SAs using the ANN, Naïve Bayes, and Support Vector Machine algorithms.

ANN: An artificial neural network (ANN) is a technique in machine learning that emulates the natural information processing mechanisms of the human brain. Numerous processing units (neurons) make up the neural network’s structure, and they communicate with one another via weights. A nonlinear mechanism in the neural network enables parallel processing, learning, and decision-making. An ANN modifies its weighted connections by using a variety of learning cases. One of the process’s outcomes is to modify the network’s settings so that it can be retrained in a different environment.³³

Naïve Bayes: This algorithm’s characteristics are assumed to be unrelated to one another. It is the Bayesian theorem generalized. NB calculates the likelihood that a data sample belongs to a specific class as part of its process because it is a probabilistic model.^34,35 NB is another name for independent or simple Bayes. The development of this technique is straightforward and does not necessitate difficult initial parameter estimates. As a result, it offers tremendous accuracy and speed when handling large datasets and can be applied to enormous amounts of data. However, this technique has problems with conditional class independence and access to data probabilities.^36,37

Support Vector Machine: According to Zach,³⁸ the Support Vector Machine (SVM) is a supervised learning technique utilized for classification and regression tasks. Its primary purpose is to identify the most effective classifier that can divide a given dataset into two distinct classes. When dealing with datasets that can be linearly separated, the SVM employs a linear function to determine a hyperplane that passes through the middle of the two classes. Notably, there exist multiple potential hyperplanes for separation in such cases. However, the SVM ensures the identification of an optimal function by maximizing the margin between the two classes. The margin, as described in Ref. 39, refers to the degree of separation between the classes, which is represented by the hyperplanes. In the field of machine learning, hyperparameters refer to the parameters used to regulate the learning process. They are predetermined values that are set before the model begins learning, with the aim of enhancing the learning outcomes. It is worth noting that not all hyperparameters carry equal significance. The effectiveness of the ML algorithm is more significantly impacted by some hyper-parameters than others. Table 1 lists the hyperparameters that we employed in our models.

Design of classification models

The training and hyperparameters tuning processes are part of the classifier design process. The Support Vector Machine, Naive Bayes, and Artificial Neural Network are the classification models investigated. For all generated models, training was typically done on between 50% and 70% of the dataset. Table 1 lists the optimized hyperparameters used in various iterations of the Naive Bayes and Support Vector Machine models.

Initially, epochs of 100 and 250 were utilized to train the ANN model; however, later in the experiment, a batch size of 32 and 500 epochs was used for adequate training. To find the best model configuration, we trained our models on various train/test ratios (40/60, 50/50, 70/30, and 60/40). The 70/30 and 50/50 configurations produced the best accuracy with the least amount of over-fitting and were thus presented as the ideal models in this study. In addition, the Adam optimizer was used to train the model, and network hyperparameters such as the number of hidden layers and activation functions were modified. The ReLu activation was then used in all hidden layers, while the output layer for classification utilized a sigmoid activation function. The dropout technique was applied to reduce over-fitting and improve the model’s accuracy.⁴⁰ The proposed framework was developed in Python using the Anaconda IDE and built-in modules including Scikit-Learn, Keras, Tensorflow, Numpy, and Matplotlib. All experimentation was conducted on a Dell Precision WorkStation Computer equipped with 8GB RAM, 1TB HDD, and a 3.3GHz Core i5 processor. The test dataset was used to evaluate the prediction performance of our model, and 20% of the training dataset was used for validation.

Evaluation of the machine learning models

Cross-validation was not used in this investigation as the dataset had sufficient samples for training and testing. Instead, the holdout validation technique was employed. After training the SVM, NB, and ANN models, their performance was evaluated using the testing dataset. To assess the performance of the models, metrics such as accuracy, precision, sensitivity, specificity, and F1 score were used. Accuracy is an indicator of the model’s classification or validation accuracy. The confusion matrix was used to estimate the number of true positives, true negatives, false positives, and false negatives, which helped evaluate the model’s effectiveness. Table 2 provides the formulas required to calculate the measures mentioned above. Accuracy measures the proportion of accurate predictions to all other predictions. Precision, on the other hand, assesses the model’s accuracy in predicting positive outcomes. Sensitivity measures how accurately the model has identified the positive instances, while specificity gauges how many true negatives the model accurately detected. The F1 score is the weighted harmonic mean of precision and sensitivity. The dataset is divided into four main groups in each experiment: True Positive (TP), True Negative (TN), False Positive (FP), and False Negative (FN).

Experimental results for SVM, NB, and ANN models

Experimental results on 50% of training

bookmark://f5 In this experiment, the samples were divided into two halves – 1000 samples for testing and 1000 samples for training. Figures 4, 5, and 6 show the confusion matrix used to evaluate the performance of three ML algorithms – SVM, NB, and ANN. The performance measures were calculated based on this confusion matrix. The SVM model accurately predicted 550 cases of SA out of all cases, with no misdiagnosis, and also correctly predicted 183 non-SA cases. However, it misdiagnosed 267 non-SA cases. The model achieved a 73.3% accuracy and a 67.3% specificity. However the sensitivity of the model was 100%, and the F1 score was 57.8%, indicating that the model’s efficiency may decrease when faced with unseen data on a large scale.

Figure 4. Confusion matrix for SVM50 classifier.

Figure 5. Confusion matrix for NB50 classifier.

Figure 6. Confusion matrix for ANN50 classifier.

The NB model, on the other hand, accurately predicted 561 cases of SA and 371 non-SA cases, with only 68 non-SA cases mispredicted, achieving a precision of 84.5% and an accuracy of 93.2%. It outperformed the SVM model by a significant margin and is expected to be useful in predicting successful aging in a new dataset. The confusion matrix for the ANN model showed that it correctly predicted 550 SA cases and 450 non-SA cases. It achieved a 6.8% increase in accuracy as compared to the NB model and fared better due to the dropout technique.

Experimental results on 70% of training

In this experiment, 70% of the samples (1400 samples) were assigned for training and the remaining 30% (600 samples) for testing. The evaluation of the three machine learning methods (SVM, NB, and ANN) was done using the confusion matrix, which is shown in Figures 7, 8, and 9. The confusion matrix was used to calculate the performance measures. Figure 8 shows that the SVM model can predict 330 SA instances accurately, with no cases of SA misdiagnosis. The model also predicted 178 non-SA cases correctly but mispredicted 92 non-SA cases. The SVM model generated an accuracy of 84.7% and a precision of 65.9%. The NB model, as shown in Figure 8, predicted 332 SA instances accurately, with no cases of SA misdiagnosis. The model also predicted 224 non-SA cases accurately but mispredicted 44 non-SA cases. The NB model generated an accuracy of 92.7% and a precision of 83.6%. The NB model outperformed the SVM model by a significant margin and is expected to be effective in predicting successful aging in a new dataset. The ANN model’s confusion matrix, trained for 70% of the time, is shown in Figure 9. The model correctly predicted 330 and 270 instances of SA and non-SA, respectively. The ANN model with dropout surpassed the NB model in precision, with a percentage gain of 16.4%.

Figure 7. Confusion matrix for SVM70 classifier.

Figure 8. Confusion matrix for NB70 classifier.

Figure 9. Confusion matrix for ANN70 classifier.

ANN model accuracy and loss plots

In Figure 10, the accuracy plot of the ANN model is displayed. The model’s training accuracy increased from 0.8 to 0.9 between 50 and 100 epochs. At 300 epochs, the validation accuracy was 0.92, and it continued to grow from there. The reported average training accuracy was 100%. Additionally, the accuracy and loss plots of the ANN model using other train/test ratios were consistent. The graphic suggests that the ANN model is not overfitting, which indicates that the model can be used to predict aging successfully in the event of new data. Furthermore, Figure 11 shows the training and validation losses of the model. The loss function reached its minimum after 500 epochs, as seen in the graph. The low loss function of the ANN model has significance in providing accurate predictions for successful ageing, as supported by research.

Figure 10. Training and validation accuracy of ANN50 model.

Figure 11. Training and validation loss of ANN50 model.

Performance of all the ML models

The results of all the Machine Learning (ML) models created are listed in Table 3. As part of the precision criterion, the number of individuals whose classifier has assigned them to a positive class (SA) and who are positive is calculated. The proposed ANN model performs best according to this criterion, while the SVM model is the least efficient. The NB50 model has a precision value of 93.2%, which is 0.5% higher than that of the NB70 model. The sensitivity criteria have the best value for the ANN/SVM/NB algorithms. Sensitivity criteria are crucial for detecting every SA-positive person in the dataset. When it comes to identifying everyone who does not have SA, the ANN50/ANN70 model outperforms other ML models. This suggests that these models demonstrate the highest level of success in terms of specificity when they achieve a value of 100%. An algorithm is considered highly efficient when it can strike a favourable balance between sensitivity and specificity. The optimal design suggested in this research has established the best balance between these two requirements. The F1-score, which considers both sensitivity and precision parameters, has a value of 100% in ANN-based models, and it is higher than that of the NB50/NB70 models (91.6%/91.1%).

Accuracy is the most fundamental and direct indicator of a classifier’s performance, and it typically manifests itself in the accurate detection of samples. Additionally, the SVM model has the lowest classification accuracy (73.3%), and the best classifier among the other models is the ANN-based approach, which has a value of 100%. Figure 12 is a bar chart that contrasts the machine learning algorithms designed according to their F1 score, specificity, sensitivity, accuracy, and precision.

Figure 12. Bar graph showing performance of machine learning models.

Authors contribution

JEZ – Conceptualization, Methodology, Software, Result Analysis and Writing (Original draft preparation, final review, and editing, VAO – Software, Data Curation, and Original Draft, AJO – Investigation, Visualization, Writing (final review and editing), ASO – Software, Data Curation, Methodology, Writing (final review and editing). All authors read and approved the manuscript.