An Improved SVM Based Method for Asthmatic Patient Monitoring in Ubiquitous Health GIS

The ever-increasing population in cities intensifies environmental pollution that increases the number of asthmatic patients. Other factors that may influence the prevalence of asthma are atmospheric parameters, physiographic elements and personal characteristics. These affecting parameters can be incorporated into a model to monitor and predict the health conditions of asthmatic patients in various contexts. Such a model is the base for any asthma early warning system.
With the rapid advancement of human knowledge in diverse areas, new science and technology has been offered to aid people in terms of education, food, transportation and health. Ubiquitous computing is one of the newest human developments to enhance individuals’ lives. In recent years, the efficiency of ubiquitous computing in a wide range of applications such as government, health, safety, municipal and transportation have been studied and validated.
This paper introduces a novel ubiquitous health system to monitor asthmatic patients. Ubiquitous systems can be effective in monitoring asthmatic patients through the use of intelligent frameworks. Our paper proposed a model for prediction of asthma conditions in various scenarios. The asthmatic conditions of patients were predicted accurately by a Graph-Based Support Vector Machine (SVM) which functions anywhere, anytime and with any status. Proposed model is an improved version of the common SVM algorithm with the addition of unlabeled data and graph-based rules in a context space.
The study graph was formed by the :union: of the training data (L) and the unlabeled data (U). Afterward, the best kernel type for SVM was estimated, and a multi-class SVM algorithm was performed. Initial classification was carried out and the U dataset was tagged. Next the k-nearest neighbor was determined around each training data item followed by the weighing of each edge of the graph (Wij) based on inverse Euclidean distance. This implies that larger weights were given to any unlabeled data close to the training data. Unlabeled data with high weights were assigned the same label as the reference training data and then the position of the data was varied. In this article, “position” means the location of data point in context space. The positional change was performed to closing the unlabeled data to the training data with respect to its weight. Then, the context space was updated and the SVM algorithm was run. This process continued until an acceptable threshold () was reached. At the end of the process, the final labels were assigned to unlabeled data and the PEF conditions of each patient were predicted.
Based on the stored value for a patient's condition and his/her location/time, asthmatic patients can be monitored and appropriate alerts will be given. Our proposed model was assessed in Region 3 of the city of Tehran, Iran for monitoring 3 different types of asthma. The input data to our asthma monitoring system included air pollution data, the patients’ personal information, patients’ locations, weather data and geographical information for 270 different situations. Our results ascertained that 90% of the system’s predictions were correct. The proposed model also improved the estimation accuracy by 12% in comparison to SVM and ACO methods.