This dataset has been created using a novel IoT network with low cost sensor platform, deployed in public buses in a developing country, the first of its kind. There are many points of faults — sensors can be faulty, internet connection can be shaky, buses might be down .... the faults can affect the quantity of data as well as quality. Detecting such anomalies for quick fixes is a necessity. We apply statistical analysis to detect anomalies, which involves many heuristics with manuallytuned thresholds. Our findings can serve as anomaly ground truth for this dataset. Automating this process with ML based methods (instead of manually tuned thresholds) can open up new avenuesof anomaly detection in mobile and IoT networks. ML researchers can try and automate the fault detection process using our dataset and the ground truth anomalies. They can also modify our released code, to change our empirical thresholds for more or less aggressive anomaly definition. Additionally, unsupervised learning methods perhaps would change the anomalies detected manually by us.

click on the below button to start download