When using this data set please cite this paper:
"Design, Benchmarking and Explainability Analysis of a Game-Theoretic Framework towards Energy Efficiency in Smart Infrastructure"
Ioannis C. Konstantakopoulos, Hari Prasanna Das, Andrew R. Barkan, Shiying He, Tanya Veeravalli, Huihan Liu, Aummul Baneen Manasawala, Yu-Wen Lin and Costas J. Spanos
arXiv preprint arXiv:1910.07899, 2019


Overview of Social Game Dataset

 

Our experimental environment is comprised of residential housing single room apartments on the Nanyang Technological University campus. We designed a social game such that all single room dorm occupants could freely view their daily room’s resource usage with a convenient interface. In each dorm room we have installed two Internet of Things (IoT) sensors[1] — one close to the desk light and another near the ceiling fan. With the deployment of IoT sensors dorm occupants can monitor in real-time their room’s lighting system (desk and ceiling light usage) and HVAC (ceiling fan and aircon usage) with a refresh interval of up to 1 second.

Dorm occupants are rewarded with points based on how energy efficient their daily usage is in comparison to their past usage before the social game was deployed. The past usage data that serves as our baseline is gathered by monitoring occupant energy usage for approximately one month before the introduction of the game for each semester. Using this prior data, we have calculated a weekday and weekend baseline for each of an occupant’s resources. We bucket data in weekdays and weekends so as to maintain fairness for occupants who have alternative schedules of occupancy (e.g. those who tend to stay at their dorm room over the weekends versus weekdays). We employ a lottery mechanism consisting of several gift cards awarded on a bi-weekly basis to incentivize occupants; occupants with more points are more likely to win the lottery.

Data organization

 

Initially, we have several per minute features like time-stamp, each resource’s status, accumulated resource usage (in minutes) per day, resource baseline, gathered points (both from game and surveys), occupant rank in the game over time, and number of occupant’s visits to the web portal. In addition to these features, we add several external weather metrics like humidity, temperature and solar radiation among others.

After gathering a high dimensional data-set with all of the available features, we propose a pooling picking scheme to enlarge the feature space and then apply Minimum Redundancy and Maximum Relevance (mRmR) feature selection procedure to identify useful features for our predictive algorithms. We pool more features utilizing a subset of the already derived features by leveraging domain knowledge and external information. Specifically, we consider the following features:

Features Description

Below we provide detailed description of all the features. Some features are only applicable in the Spring Semester data set.

General - User related features:

 'owner': name of each player. We use fake names as mean to de-identify the data-set.

'room_code': the label of the each player’s room

'date_time': the date and time at which the data is collected

'period_total_points': the total number of points of each player

'rank': the rank of each player

'period_survey_points': points earned by each player as taking feedback (Likert based) survey

'days_page_views': number of times the player visit the game webpage

'coin_hits': number of times the player visit the game webpage and “clicks” a randomly appeared coin next to resources utilization. It is an indicator of players’ participation - activity. Applicable only at Spring semester data

 

 

Accumulated Resource Usage:

 

'ceiling_fan_status': whether the ceiling fan is turned on or off

'days_fan_minutes': number of minutes ceiling fan is turned on per day (accumulated)

'days_fan_data_minutes': number of minutes of available ceiling fan data. Data can be sparse in some rooms after sensors’ lost of connection 

'days_fan_baseline_minutes': number of minutes for ceiling fan usage (baseline for each player)

'days_fan_points': number of points player earns through ceiling fan usage

 

'ac_status': whether the aircon is turned on or off. Applicable only at Spring semester data

'days_ac_minutes': number of minutes aircon is turned on per day (accumulated). Applicable only at Spring semester data

'days_ac_data_minutes': number of minutes of available aircon data. Data can be sparse in some rooms after sensors’ lost of connection. Applicable only at Spring semester data

'days_ac_baseline_minutes': number of minutes for aircon usage (baseline for each player). Applicable only at Spring semester data

'days_ac_points': number of points player earns through aircon usage. Applicable only at Spring semester data

 

'ceiling_light_status': whether the ceiling light is turned on or off

'days_ceiling_light_minutes': number of minutes ceiling light is turned on per day (accumulated)

'Days_ceiling_light_data_minutes': number of minutes of available ceiling light data. Data can be sparse in some rooms after sensors’ lost of connection 

'Days_ceiling_light_baseline_minutes': number of minutes for ceiling light usage (baseline for each player).

'days_ceiling_light_points': number of points player earns through ceiling light usage

'ceiling_light_brightness': measured brightness in the room from IoT sensor at the ceiling (in lux)

'ceiling_temp': measured temperature in the room from IoT sensor at the ceiling (in fahrenheit)

'ceiling_humidity': measured humidity in the room from IoT sensor at the ceiling

 

'desk_light_status': whether the desk light is turned on or off

'days_desk_light_minutes': number of minutes desk light is turned on per day (accumulated)

'Days_desk_light_data_minutes': number of minutes of available desk light data. Data can be sparse in some rooms after sensors’ lost of connection 

'Days_desk_light_baseline_minutes': number of minutes for desklight usage (baseline for each player).

'days_desk_light_points': number of points player earns through desk light usage

'desk_light_brightness': measured brightness in the room from IoT sensor at the desk light (in lux)

'desk_temp': measured temperature in the room from IoT sensor at the desk light (in fahrenheit)

'desk_humidity': measured humidity in the room from IoT sensor at the desk light

 

External weather metrics:

 

'external_solar_radiation': the solar radiation outside the building

Watts per Square Meter (W/m2)

'external_rain_rate': raining rate outside the building

Inches per Hour (in/hr)

'external_temp': temperature outside the building

Degrees Fahrenheit (°F)

'external_humidity': humidity outside the building

Relative Humidity Percent (%)

'external_pressure': air pressure outside the building

Inches of Mercury (inHg)

'external_wind_speed': wind speed outside the building

Miles per Hour (mph)

'external_wind_direction': wind direction outside the building

 Degrees (°)

 

 

 

 

College schedule dummy features-indicators:

 

'break': dichotomous indicator of whether it is school break

'midterm': dichotomous indicator of whether it is midterm school period

'final': dichotomous indicator of whether it is final school period

'reading': dichotomous indicator of whether it is reading school period

 

 

Seasonal dummy features-indicators:

 

'Morning_label': dichotomous indicator of whether it is in the morning (between 5am & 11:59am)

'Afternoon_label': dichotomous indicator of whether it is in the afternoon (between 12pm & 4:59)

'Evening_label': dichotomous indicator of whether it is in the evening (between 5pm & 8:59)

'Night_label': dichotomous indicator of whether it is at night (between 9pm & 4:59am)

'Weekday_label': dichotomous indicator of whether it is weekdays period

'Weekend_label': dichotomous indicator of whether it is weekend period

 

Resources status continuous features:

 

'fan_proportion_on': proportion of time ceiling fan is turned on per day (actual switches and their accumulated number)

'ceiling_fan_frequency': how often ceiling fan is turned on

'fan_baseline_ratio': ratio of per minute accumulated ceiling fan usage over ceiling fan baseline

 

'ceiling_light_proportion_on': proportion of time ceiling light is turned on per day (actual switches and their accumulated number)

'ceiling_light_frequency': how often ceiling light is turned on

'ceiling_light_baseline_ratio': ratio of per minute accumulated ceiling light usage over ceiling light baseline

'desk_light_proportion_on': proportion of time the desk light is turned on per day (actual switches and their accumulated number)

'desk_light_frequency': how often desk light is turned on

'desk_light_baseline_ratio': ratio of per minute accumulated ceiling fan usage over ceiling fan baseline

 

For demonstrations of our infrastructure and for downloading de-identified high dimensional data-sets, please visit our web-site [2]

 

 

 

Licenses

 

The data are licensed under CC BY 4.0 license.

 

 

 

[1] Sensor Tag: http://www.ti.com/ww/en/wireless_connectivity/sensortag/

[2] smartNTU demo web-portal: https://smartntu.eecs.berkeley.edu