Diabetes is a chronic disease that affects millions of people throughout the world and studies show that there is a gap between research and smartphone-based solutions. This paper presents an initial investigation into the possibility and viability of using smartphones to predict hypoglycaemia in diabetic patients accurately. This investigation involved the identification of smartphone compatible algorithms and numerous tests for accuracy and performance of the proof-of-concept application. This is achieved through a three-stage approach where the Literature was investigated, followed by the implementation of two selected algorithms – J48 Tree and Fuzzy Lattice Reasoning – and tests to compare them to their smartphone implementation, watching for mobile-based metrics such as CPU, memory usage and execution time. The results show that the mobile implementations of the selected machine-learning algorithms maintain an accuracy rate as high as reported in their studies with regular use of CPU and Memory, and shorter execution time when algorithms are initially trained. Consequently, this paper delivers a proof-of-concept smart application that will serve as a basis for future applications aiming at a broader diabetic population who use smartphones.
|Number of pages||8|
|Journal||Procedia Computer Science|
|Early online date||21 May 2019|
|Publication status||Published - 2019|
- smart applications
- machine learning
- blood glucose