DataLearner: A Data Mining and Knowledge Discovery Tool for Android Smartphones and Tablets

Abstract: Smartphones have become the ultimate 'personal' computer, yet despite this, general-purpose data-mining and knowledge discovery tools for mobile devices are surprisingly rare. DataLearner is a new data-mining application designed specifically for Android devices that imports the Weka data-mining engine and augments it with algorithms developed by Charles Sturt University. Moreover, DataLearner can be expanded with additional algorithms. Combined, DataLearner delivers 40 classification, clustering and association rule mining algorithms for model training and evaluation without need for cloud computing resources or network connectivity. It provides the same classification accuracy as PCs and laptops, while doing so with acceptable processing speed and consuming negligible battery life. With its ability to provide easy-to-use data-mining on a phone-size screen, DataLearner is a new portable, self-contained data-mining tool for remote, personalised and learning applications alike. DataLearner features four elements - this paper, the app available on Google Play, the GPL3-licensed source code on GitHub and a short video on YouTube.

• The paper introduces DataLearner, an Android application enabling offline data mining and knowledge discovery on smartphones and tablets using the Weka engine and custom algorithms.
• DataLearner integrates the Weka engine and three custom algorithms (SysFor, ForestPA, SPAARC) to support 40 data mining methods, performing all tasks locally on the device for privacy and offline capability.
• Evaluation shows DataLearner achieves equivalent classification accuracy to desktop Weka with datasets processing in sub-second times, making it a practical tool for education, field research, and privacy-sensitive mobile data analysis.

An Overview of DataLearner: A Data Mining Tool for Android Devices

The paper "DataLearner: A Data Mining and Knowledge Discovery Tool for Android Smartphones and Tablets," introduces DataLearner, a novel application extending the capabilities of data mining to Android mobile devices. This application leverages the Weka data mining engine and incorporates additional algorithms developed by Charles Sturt University, enabling users to perform data mining tasks directly on their smartphones or tablets without requiring cloud connectivity. Below, we explore the design, implementation, evaluation, and implications of DataLearner.

DataLearner fills an existing gap in mobile data mining and knowledge discovery by providing a portable and self-contained tool that operates efficiently on Android devices running at least Android 4.4. This is achieved by incorporating the Weka module and augmenting it with unique algorithms developed in-house, encompassing 40 classification, clustering, and association rule mining methods. The application supports local data model training and evaluation without dependency on external resources, ensuring data privacy and security. This feature is crucial for applications in remote locations or situations where connectivity is limited.

The paper details the meticulous design and implementation strategy for DataLearner, emphasizing ease-of-use for novice users and effective navigation on small screens. The application is segregated into main operations of loading data sets, selecting algorithms, and executing analysis, all managed via user-friendly interfaces. Notably, DataLearner supports the expansion of algorithms by facilitating the addition of new methods into the platform, thus demonstrating its adaptability.

Three additional classification algorithms—SysFor, ForestPA, and SPAARC—developed at Charles Sturt University, were incorporated to enhance DataLearner's capabilities. These algorithms are specifically designed for efficient processing on mobile devices where resource constraints are common. For instance, SysFor constructs a decision forest optimized for low-dimensional datasets, making it ideal for mobile applications with constrained data input sizes.

The evaluation results of DataLearner reveal that it achieves equivalent classification accuracy on Android devices as on traditional personal computers. The tested algorithms showed consistent performance across different devices, with the mobile implementations completing model building tasks efficiently, albeit generally slower than desktop counterparts due to hardware limitations. Despite this, all datasets processed by DataLearner completed within sub-second durations, underscoring its practicality for mobile data mining without excessive time demands or energy consumption.

Examining these results, DataLearner is poised to serve multiple applications, from educational tools supplementing traditional computers to real-time data analysis in field research. The inherent portability and data-locality aspects of DataLearner make it an asset in environments encountering privacy concerns or restricted connectivity. Furthermore, the potential for adaptation, including ongoing development into multi-core processing support, positions DataLearner as a versatile tool for future work in mobile data analysis.

Possible directions for future research and development include optimizing the user interface further, enhancing performance through multi-threading to better utilize mobile hardware, and extending support to additional file formats like .CSV, which are ubiquitous in data science. An exploration into compatibility with iOS devices may also expand DataLearner's reach, making it more universally accessible.

In conclusion, DataLearner represents a significant advancement in the field of mobile data mining by enabling robust, versatile, and secure data analysis directly on Android devices. The application achieves this while maintaining sufficient performance and minimal impact on device resources, marking a forward step in data science as it intersects with mobile computing.