Bindu Gupta bindu.gupta2@tcs.com

Aditeya Pandey aditeya.pandey@tcs.com

Gunjan Sehgal sehgal.gunjan@tcs.com

Karamjit Singh karamjit.singh@tcs.com

Geetika Sharma geetika.s@tcs.com

Gautam Shroff gautam.shroff@tcs.com

Student Team:  NO

Tools Used:

List of tools used. 

Excel

Eclipse

Approximately how many hours were spent working on this submission in total?

A total of approximately 300 hours were spent on this submission

May we post your submission in the Visual Analytics Benchmark Repository after VAST Challenge 2016 is complete? YES

Video

Provide a link to your video.  Example:

https://youtu.be/IPiz6qSANxc

Questions

MC3.1 – Describe the unusual or unexpected patterns and anomalies you observed during the first two days of streaming data (June 14-15). Please describe up to ten notable patterns or anomalies.

We have developed models of normal behaviour for sensors and employees using the static sensor and prox data of MC2. We tag deviations from these models as anomalies

Sensor Model

In order to discover sensor anomalies we use a Gaussian model for each sensor. We compute the mean and standard deviation for all sensors at all locations for each hour. If the value of a sensor reading is outside the range mean +/- std deviation we tag it as an anomaly.

Prox Model

For each employee, we compute the probability of him/her being at a particular zone in 15 minute time bins. In order compute the probabilities locations, we have taken into account the support of each employee that is the number of days each employee is present in office.

Our dashboard is divided into 4 parts, as shown below. The top left part shows the anomalies in prox and sensor streams for the current hour for each floor. These are shown on radial charts each with twelve 5-minute bins. The top right part shows the three most anomalous prox ids and sensors in the last 15 minutes over all floors. Top three prox anomalies are computed using the probablity of the location where the prox id is detected and the time spent in that location and top three sensor anomalies are based on count of anomalies. The bottom right part shows the most anomalous floor in the last 15 mins. This is computed using the sum of sensor and prox anomalies for each floor. A prox zone heatmap shows where prox anomalies are taking place, while sensor glpyhs overlaid at sensor zone locations indicate which sensors are anomalous. Finally, the bottom left part serves as a catch up view with a heatmap of floor-wise sensor and prox anomaly counts per hour for the selected day.
Following were the unusual patterns observed during the first two days:
1. On June 14, from 2 am onwards, sensor anomalies were detected in the stream. Floor 1 zone 1 and zone 2 had anomalies in Thermostat temperature and thermostat heating setpoint. Floor 2 was seen to be the most anomalous with Thermostat heating setpoint, Thermostat temperature and supply inlet temperature reporting anomalies in multiple zones. Some anomalies were also seen on Floor 3. From 5 am onwards, Floor 3 also started recording significant anomalies. Floor 2 remained most anomalous with additional sensors such as Supply inlet mass flow rate and vav reheat damper position also recording anomalies. most sensor anomalies were found from sensors of floor 2, as can be seen from the heatmap below.






2. Thermostat sensors were amongst the most abnormal sensors in floor 2.

MC3.2 – Which additional sensor did you choose to add to your data stream? What was the rationale for your selection? Did it provide additional insight?

We chose to add a Hazium sensor on Floor 2 zone 11.
We had been seeing Hazium anomalies on Floor 2 zones 2 and 4 since 2am on June 16. The list of location options for the additional sensor included conference rooms on Floors 1 & 3. As there is a possibility that conference rooms may remain unutilised from 8am-12pm, we ruled these locations out. The other two options were Floor 2 zones 11 and 15. Zone 11 is near zones 2 and 4 which were already showing anomalies and it is at a dense office area on floor, while zone 15 is near the staircase.
A number of Hazium anomalies where reported from Zone 11 in the remainder of the stream.

MC3.3 – Describe anomalies or unusual events you saw in the last four hours of the data stream. 

a.      Describe specific anomalies or unusual events

b.      Which of these anomalies may be of greatest concern? What is your rationale?

a. We observed a high concentration of Hazium in the building. All Hazium sensors were showing extremely high Hazium levels. Can we show catch up of each floor??
Mechanical Ventilation Mass Flow rate decreases suddenly at 9am.
A number of prox card anomalies take place. We observe a number of security personnel from Floors 1 and 3 going to Floor 2, zones 7 and 3 at around 8:30am.

People in the zones with Hazium sensors on Floors 2 and 3 and seen leaving their zones and going to another zone, sometimes even to Floor 1.

The most anomalous hour is 10am-11am with maximum prox and sensor anomalies on each floor.

The prox id pyoung001 is detected in the building along his usual path leading up to the Server Room on floor three and coming down to floor 2. His last prox card entry is floor 2 zone 1.

b. The anomaly of biggest concern is the increased level of Hazium. It is clear that the high Hazium levels made people uncomfortable in their zones and they had to move to another zone.

MC3.4 – Mini-Challenge 3 asks you to develop ways of understanding developments in streaming data that take place even while you are away from the data.

a.      Describe how your team approached the challenge of catching up on events that took place while you were not monitoring the stream. What features of your software helped you to review past events and catch up on things you missed? 

b.      How could these features be used to help you reconsider recent data in light of new events?

Our dashboard includes a heatmap in the bottom left panel which shows the floor-wise count of sensor anomalies for the hours of the current day that have passed. This enables the user to identify which floor was most anomalous. Using the date selection option, heatmaps of previous days can also be viewed.

Further, when a user clicks on a particular bin of the heatmap, selecting a sensor or prox floor and hour, the heatmap on the bottom right panel shows the selected floor's plan with anomalies for that hour.

For example in this case floor 1 of sensor was selected