*8.2. Data Classification Model*

The performance of logistic regression and naive Bayes algorithms could have been influenced by various factors, such as the size and complexity of the data, the hardware

and software utilized, and the specific implementation of the algorithms. Typically, logistic regression has been faster than naive Bayes when working with large datasets, as the naive Bayes algorithm can become computationally demanding as the number of features increases. However, naive Bayes can be faster when working with smaller datasets or when the number of features is relatively limited. Table 5 illustrates the performance of both algorithms after classifying 10 million records of medical data. The table shows the speed and accuracy of both algorithms, which aided in determining which algorithm was more suitable for a specific application. While the speed of an algorithm was an important consideration, accuracy was also taken into account when a compromise between speed and accuracy may be necessary.

**Table 5.** Data classification performance.


Based on the datasets described in Section 7.2, the proposed solution was analyzed to evaluate its classification performance and accuracy.

The results presented in Tables 6 and 7 demonstrated the superior performance of the proposed solution, as compared to the logistic-regression and naive Bayes algorithms. The naive Bayes algorithm is known to be efficient for small datasets, but the proposed solution outperformed both algorithms, even when the dataset size increased. This highlighted the effectiveness of the proposed solution in handling larger datasets, which pose a significant challenge for traditional classification methods. Additionally, the strong performance of the proposed solution, as compared to the standard classification algorithms, such as logistic regression and naive Bayes, further emphasized its potential for practical applications. Overall, the results demonstrated the exceptional performance and potential of the proposed solution.


**Table 6.** Data classification performance (Dataset 1).

**Table 7.** Data classification accuracy of Dataset 1.


One possible reason for the higher accuracy of the proposed solution was that it had been specifically designed to handle larger datasets, which may have been more challenging for traditional classification algorithms. For example, logistic regression and naive Bayes algorithms could have struggled to effectively classify data when the number of features increased significantly, as they can become computationally demanding as the number of features increases. In contrast, the proposed solution used more advanced techniques, including the advanced parallel *k*-means, parallel logistic regression, and the neural engine processor, to effectively classify the large datasets. Additionally, the proposed solution incorporated additional factors and features that were relevant to the classification task, which further improved its accuracy. Overall, the results demonstrated the effectiveness of the proposed solution in handling large datasets and achieving high accuracy in classification tasks.

In order to examine the performance of the proposed solution with the recent advancements in medical data classification, the proposed solution was compared with the three most recent medical-data-classification approaches, which were: [65–67]. All solutions were compared with the proposed solution in terms of classification performance and classification accuracy, as shown in tables below.

As shown in Table 8, the proposed solution significantly outperformed the three compared solutions, while the naive Bayes-based algorithm tended to be slower, the proposed solution was more effective than both the binary logistic regression and the logistic regression. This suggested that the proposed solution was particularly well suited for handling larger datasets, which could be more challenging for traditional classification algorithms. The results demonstrated the strong performance of the proposed solution, as compared to the conventional classification algorithms, indicating that it was an effective and reliable method for classification tasks.


**Table 8.** Data classification speed (min.) when compared with recent approaches.

The accuracy of the proposed solution was compared with previous solutions, and the results, as shown in Table 9, demonstrated its high accuracy. Specifically, the proposed solution outperformed the comparable solutions, achieving an accuracy rate of 99.8% while a novel binary-logistic-regression solution only achieved 98% accuracy. The worst performance in terms of accuracy was observed in the logistic-regression solution designed for the prediction of myocardial infarction disease in [66]. These results suggested that the proposed solution was particularly effective at achieving high accuracy in classification tasks, and that it outperformed other approaches.

**Table 9.** Data classification accuracy when compared with recent approaches.

