Supervised Machine Learning Supported Time Series Prediction and Analysis of IoT Enabled Physical Location Monitoring
Ajitkumar S. Shitole1, Manoj H. Devare2
1Ajitkumar S. Shitole, Research Scholar, Amity University Mumbai, India.
2Manoj H. Devare, HoI, AIIT, Amity University Mumbai, India.
Manuscript received on 01 June 2019 | Revised Manuscript received on 05 June 2019 | Manuscript published on 30 June 2019 | PP: 2338-2345 | Volume-8 Issue-8, June 2019 | Retrieval Number: H7221068819/19©BEIESP
Open Access | Ethics and Policies | Cite | Mendeley | Indexing and Abstracting
© The Authors. Blue Eyes Intelligence Engineering and Sciences Publication (BEIESP). This is an open access article under the CC-BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/)
Abstract: Internet of Things (IoT) is one of the evolving technologies in the recent days to exchange the information from one device to another using any type of network, at anytime, and at anywhere. With the introduction of IoT and Machine Learning (ML) to monitor physical location in real time fashion is necessary to identify abnormal conditions in the surroundings. The proposed system depicts that different sensors in addition to camera are used to monitor and identify abnormal environment conditions of the same and send alert message to the user to take corrective action to avoid any future loss in the environment. Real time sensor data which is aligned with multimedia data is stored onto local system and ThingsSpeak server as well as it is pushed onto Go Daddy cloud whenever camera detects person to perform systematic and statistical analysis using different supervised machine learning algorithms. This paper presents time series prediction of different sensor values such as temperature, humidity, gas, light dependent resistor, and person prediction using timestamp (day and time) to understand the physical location well in advance to take appropriate decision. Experimental results show that decision tree is the best predictive model to predict person when timestamp is given in the form of date and time. Study also reveals that Decision Tree Regression (DTR) and Random Forest Regression (RFR) give good results with approximately same minimum Root Mean Squared Error (RMSE) to predict different sensor values.
Keywords: Physical Location Monitoring, Time Series Prediction, RMSE, Supervised Machine Learning
Scope of the Article: Machine Learning