Anomaly Detection for Internet-of-Things Appliances
Budderfly is an energy management company that uses innovative solutions to help businesses and other institutions save energy. One of these solutions is accomplished through the installation and monitoring of energy efficient appliances through Internet of Things (IoT) devices. Given time series data on freezer and bread oven monitors, the goal of our project was to analyze the power levels of these appliances. We first visualized the data to better understand it, and then pre-processed the data to get rid of noise and outliers. Then, using different machine learning models, we were able to detect anomalies among the freezer and bread oven monitors. One method we used was K-Means clustering with soft DTW as the metric, which grouped monitors into clusters and helped us identify which monitors were behaving irregularly. Finally, we used an alternate LSTM machine learning model to detect anomalies and compared this with the K-means clustering.