What are some design ideas?

What are some ideas for a design project using machine learning, sensors and bigdata?

• I just finished my 3rd of undergrad at a reputed university. I'm taking a year off to work. Next year,we have the opportunity to make anything we want in the form of a design project. I'm presently infatuated with big data and machine learning! Since I'm in computer engineering (as opposed to computer science), my team and I decided to integrate a hardware component. We were thinking along the lines of sensors and a microcontroller to grab data. Sensor based big data seems like a nice and exciting field. Though we don't have any idea on what exactly to do that hasn't been done before. Do you have any cool ideas? Or anything you find cool but don't have the time/resources to implement it. Please let us know. We're ready to change the world :)

• Answer:

  You can try buillding a wireless sensor network (with Arduino [1] and Zigbee [2] for example)  and capture temperature/humidity/moisture/occupancy in a residential building. Send the measurements from different rooms to a gateway server in the same buiding (may be you can use Sheevaplug [3] or something like that, or just someone's desktop with a wireless router and internet connection), and from there to a web server on AWS [4] which would chart the signals using something like D3 [5] for users coming from the web. You can try doing some analysis on the sensor time series [6] data and do simple forecasting, e.g. with linear regression [7], or with something fancier (but start with linear regression), and show on your website the past, current and predicted temperature/humidity for a few rooms of the building. This can be useful for monitoring and predicting energy consumption [8] and optimizing energy usage to save costs. Bonus points: figure out a way to control temperature more efficiently than a plain old thermostat (don't play with the wiring though, it is dangerous) Extra bonus if you get energy usage metrics for the HVAC unit in the building (talk to building superintendant, their might have monthly usage stats and some historical data) and use it in your predictions as one of the signals Extra+ points if you find building/s with existing Building Control System (BCS), and/or Building Management System (BMS) and get permission from the owners to extract and use their data for forecasting, you may find such systems on university campuses-  talk to Facilities Management and Planning office or something similar [1] http://en.wikipedia.org/wiki/Arduino [2] http://en.wikipedia.org/wiki/ZigBee [3] http://en.wikipedia.org/wiki/SheevaPlug [4] http://aws.amazon.com/ [5] http://d3js.org/ [6] http://en.wikipedia.org/wiki/Time_series [7] http://en.wikipedia.org/wiki/Linear_regression [8] http://en.wikipedia.org/wiki/HVAC