mrharrison
Posts: 2
Joined: Fri Oct 19, 2018 3:35 pm

'Data Science' relatable MSc project sought

Fri Oct 19, 2018 4:45 pm

Dear all.

I've been putting out the feelers lately to see if anyone had any ideas or leads towards a potential research project. I'm looking for something creative (or not) to base my Msc around. Pi ideas are always full of potential and a great community. I appreciate that it's a slightly scatter-gun approach, but i'm stuck and would prefer to base it around something that excites me rather than pull one of the off-the-shelf uni projects. Has anyone thoughts on potential for a project, a larger data set or similar that might warrant an interesting looking further at? I've spent a bit of time with the sensehat and thought that there is potential perhaps in that for data generation/analysis... maybe something like a risk assessor for driving based on live data collection from the device. Has that been done ?

My background is in education, but getting education related data to work with is going to be nigh on impossible.

Its been great to see the progress and recent builds for a Hadoop cluster based around the Pi. Not yet experienced enough to know if there might be further scope in those for me to work on, and not yet been able to try them out but an option perhaps.

anyway, thanks for any suggestions you might have, just thought that i would at least ask.
Rob

vinnieb
Posts: 21
Joined: Wed Feb 19, 2014 6:44 pm

Re: 'Data Science' relatable MSc project sought

Fri Oct 19, 2018 6:21 pm

Definitely a good idea to pick a topic that motivates you, plus more sense of ownership.

I wonder if the driver data is done in the models that insurance companies use when offering reduced premiums if you use their black box in your car?

Personally I'm trying to get a bit of time to look at the data set generated by my home data. Home assistant can run on a pi, and can gather alot of data from IoT stuff in the house. Or, anything you can write a quick MQTT interface to. I've got alot of pis around the house with various sensors plugged in, plus hue, hive, TV, presence detection etc, all logging. I wonder if as creatures of habit there might be some surprising patterns in that data. And perhaps those patterns could be used for some bigger benefit e.g. energy reduction, hot water requirement prediction. But equally, anomalies could be interesting too (maybe some deep learning algorithm approach there) and could trigger further analysis if something unexpected is going on.

mrharrison
Posts: 2
Joined: Fri Oct 19, 2018 3:35 pm

Re: 'Data Science' relatable MSc project sought

Sat Oct 20, 2018 10:50 pm

Thanks for the responses. I agree, there has to be scope for finding patterns in home data. That sounds pretty interesting as a concept aswell. The telematics boxes... i actually thought that they only tracked distances, journey times and more general GPS related, but having looked it seems as though such as braking patterns are part of the collected data.

User avatar
Joel_Mckay
Posts: 141
Joined: Mon Nov 12, 2012 10:22 pm
Contact: Website

Re: 'Data Science' relatable MSc project sought

Sun Oct 21, 2018 11:23 am

The first steps in real scientific inquiry is asking "why are you doing it", and "how it will benefit civilization".
Take it from someone that spent years of their life reading reported nonsense-findings by the optimistic:
http://www.academiaobscura.com/still-not-significant/
;-)

Julia is like R, octave, and Python... but actually scales in a cluster...
https://julialang.org/
https://www.raspberrypi.org/blog/julia- ... pberry-pi/

A dynamic sensor network that used something like Queryverse to look for abstract event correlations in a robust distributed manner could prove interesting. Project artifacts could include a reusable library that others used for traffic-monitoring cities, earth and ocean sciences.
(i.e. keep it simple and useful... people would likely cite it because it saves time, and by design automatically finds correlated relationships.)

Cheers,
J

Heater
Posts: 9957
Joined: Tue Jul 17, 2012 3:02 pm

Re: 'Data Science' relatable MSc project sought

Sun Oct 21, 2018 11:59 am

Joel_Mckay,
The first steps in real scientific inquiry is asking "why are you doing it", and "how it will benefit civilization".
That can't be right. At least not all the time. Sure ask yourself "Why am I doing this" but if starting out on any line of inquiry demanded that it looks like it has some probability of being useful at some time in t h future then most science would never get started.

My impression is that a lot of great science is done just because people are insatiably curious and want to find out how something works. They have a passion for discovery, any thought of practical use is just a happy outcome sometimes.

Anyway, thanks for the link to the article. "only marginally nearly insignificant", brilliant, made me chuckle.

Return to “General discussion”