International Journal of Scientific Research and Engineering Development

Early diagnosis of cancer is crucial for the survival and to avoid unnecessary sufferings and complications. Presently, cancer screening involves several investigations, numerous invasive examinations and also relies heavily on the diagnostic imaging facilities in the hospitals. In this work, a machine learning based classification method is implemented on an embedded platform that can detect and classify multiple types of cancers using the sweat bio markers. The system utilizes the structured feature vectors that are composed of various physiological and biochemical parameters such as pH, VOCs, temperature, humidity, pressure and optical fluorescence. Various supervised machine learning algorithms including Decision Tree, Random Forest and a hybrid ensemble model are implemented and tested to classify the type of cancer. The results show that the hybrid ensemble model provided the best classification accuracy for the proposed cancer screening test while being robust to the overlapping bio marker features and also having low computational complexity. Hence, the system is capable of performing real time on device cancer risk level classification and can be highly suitable as a self-help device for the awareness of cancer at an early stage. The proposed work is the first step in the direction of implementing a low cost, highly scalable wearable bio sensing based cancer screening support system. The state of the art biosensors and wearable devices based on sweat biomarkers currently have very limited functionality. They are mostly restricted to simple, qualitative detection.

Farhanas J, Etikala Vishvesha, Asheq Noor N, Bhuvaneshwaran B, Saranya S,"Embedded Machine Learning–Based Model for Automated Multi-Cancer Risk Classification Using Sweat Biomarkers" International Journal of Scientific Research and Engineering Development, V9(2): Page(804-808) Mar-Apr 2026. ISSN: 2581-7175. www.ijsred.com. Published by Scientific and Academic Research Publishing.