SSN Demo
From Knoesis wiki
Demonstration
This demonstration section is dedicated towards providing a step-by-step guide on using the interface built for viewing the feature streams in real-time. Figure 1 shows a snapshot of the interface built. The interface would be explained in three steps for ease of understanding.
- Begin Search: The interface provides two options to begin search, marked in figure 1. The options are given below:
- Search Bar: The search bar would allow a user to search for stream of features in a specific location like “Dayton James Cox Airport”. In this case the result would contain the sensors near the location.
- State Selection: The user can search for stream of features for an entire state. This is done by clicking on the specific state. In this case the result would contain all the sensors in the state that are currently active. Figure 2 shows the result of selecting Ohio as the state. The result contains all the sensors in Ohio that are currently active. The snapshot contains few sensors to provide an uncluttered view, however in reality Ohio contains greater than ~100 sensors.
- Feature Selection: In this step the user is given an option to choose the features of interest. A user can choose the features he/she is interested in by adding the features to the Event Bag using a drag and drop option. Figure 2 shows Flurry, RainStorm and RainShower added to the event bag as features of interest. On submit the interface would provide only the sensors that are currently detecting the features of interest. In this case, it would provide all sensors that are currently detecting RainStorm, RainShower or Flurry. For ease, we have provided a unique icon for each feature detected.
- Feature Streams View: Feature selection results in a list of sensors (represented by feature icons) currently detecting a feature of interest (RainShower, RainStorm, Flurry in this case). In order to view the stream of features, the user would have to select a specific sensor of interest. Figure 3 provides a snapshot of the stream of features detected by sensor system KHAO. The last feature detected by sensor system KHAO is RainShower and hence is represented by a RainShower icon. We provide a graphical view, to observe the stream of features that were detected by sensor system KHAO and values that contributed towards the features over time. Figure 3 shows 4 graphs (Temperature, WindSpeed and Precipitation) graph representing the lower-level data values and feature graph that represent the features detected. The four graphs are aligned over time to represent the combination of lower-level values that contribute in feature detection.
Evaluation
- Performance: To evaluate the performance of this system, we collected 120 hours of data for sensors in (and around) Utah between February 2nd to 6th 2003. Figure 4 shows an average of the amount of time (in ms) taken for each phase during feature generation. On average, for each hour, 427 sensors provided data during the evaluation, and produced an average of 1104 observations. 9 flurries, 1 rain shower, and 417 clear features were detected during the evaluation. We found an order of magnitude distinction between the number of observations and feature generated, which means storing only the features (if applicable) would result in massive data reduction. Storage evaluation results can be found below.
- Storage: To evaluate the data reduction, we collected data for sensors in (and around) Utah between February 4th to 6th 2003. Figure 5 shows the total amount of data collected from sensors that resulted in 111,456 observations. The total number of rain storm, rain shower, blizzard and flurry features totaled up to 12,331 which results in 88.94% reduction from the total number of observations. Some applications are interested in “Clear” feature. Hence we stored “Clear” along with other features that totaled up to 37,152 features which still results in 66.67% reduction.
