*8.1. Operating System Performance*

The proposed solution was designed to be operating system independent and hardware accelerated. This meant that the advanced parallel *k*-means clustering could be executed on any operating system that had two processors: a neural engine processor and a general-purpose processor. However, both iOS and Android operating systems were designed to manage and take advantage of hardware allocation and management that included their neural engine processor. These dedicated operating systems were able to send specific tasks to a particular processor core, enabling the implementation and execution of the advanced parallel *k*-means clustering.

Overall, this type hardware acceleration provides opportunities for future advancements of the operating systems, which is expected since the new M-family MacOS already supports dedicated neural-engine-core assignments.

In order to evaluate the performance of the advanced *k*-means clustering across different operating systems, Table 3 presents two large datasets, each with over 9 million records. These were clustered using the advanced parallel *k*-means clustering algorithm on Windows OS, Android, and iOS systems.

**Table 3.** Clustering datasets.


The performance results, as presented in Table 4, showed that the processing of 11 million records from the Google Play Store dataset doubled in speed with a dedicated ML processor. The next experiment was conducted using the education-sector dataset, and the mobile processor exhibited a performance up to 10-times faster than the desktop OS (Windows 11). Additionally, the performance of iOS was twice as fast as that of the Android OS.

**Table 4.** Clustering performance of big-data sets in minutes.


The performance differences observed between the iOS and Android operating systems, within the context of advanced parallel *k*-means clustering, could be due to several factors. It could be related to the differences in the underlying architectures of the two operating systems. Specifically, iOS was designed to take full advantage of its hardware resources, including the dedicated neural engine cores, which could explain the observed faster performance, as compared to Android.

Additionally, the iOS architecture was based on the use of Objective-C and dispatch queues, which were designed to facilitate concurrent processing and task scheduling. These features provide a more efficient way to execute the parallel *k*-means clustering algorithm, potentially resulting in the observed faster performance.

However, the performance differences observed could have also been influenced by other factors, such as the differences in the hardware configurations of the devices used to test the algorithms, as well as the specific implementation of the parallel *k*-means clustering algorithm on the different operating systems.
