Paper Details

— This paper details an attempt at implementing a special SaaS Server that provides Machine Learningservice to client application(s). We first create a Learning Service which is then utilized to predict future actions that theuser may perform. We’ve also developed an ad-hoc control interfacing system for the user, consisting of buttons relatedto various connected devices. This interface is implemented as a webapp that is served up by the controller (for whichwe’re using the Raspberry Pi). It interacts with the LaaS Server to store and analyze the collected user actions.Prediction regarding the next action is requested by the controller to the LaaS Server. This prediction, when received, isapplied instantaneously to the connected devices. Thus, it forms a real-time analysis-and-prediction system. This systemworks by collecting usage statistics from the user and some emulated external factors that may influence the usage of thesystem. These collections are then analyzed for correlation between individual device states and corresponding externalinfluencing factors. This process is repeated every time a new entry is added. The correlation previously established isthen updated in real time at each new entry. This correlation is in the form of a prediction model that is used to generatepredictions. The controller instantaneously acts upon the predictions to change the device states, thus reflecting theuser’s usage pattern and providing a sense of true automation.

Prof. Rukaiya Shaikh, Sameer Shaikh, Ayyaj Attar, Siddik Pathan, “Using Learning-as-a-Service To Enhance Home Automation”, International Journal of Advance Engineering and Research Development (IJAERD), Vol. 4, Issue 5, pp. 704-709, May 2017.