Big Data Case Study: TV Usage and Power Consumption


Want to challenge yourself before you go to apply the Data Scientist/Machine Learning/Big Data jobs?

You are given the data of total (aggregated) power consumption of two days in a household. This household uses TV in these two days. The task is to come up with ideas to identify times when the TV is “ON” using the total power consumption of future days. The data of the future days is not provided. Please describe what kind of data, information, and techniques can be used and possible issues. You can also provide a script to identify the period which TV is “ON”, but it is not mandatory.

The data is stored in the attached text file. The first column is the time stamp. The second column is the TV power consumption per time stamp. The third column is the aggregated (total) power consumption per time stamp.

power-consumption-tv-usage Big Data Case Study: TV Usage and Power Consumption case study interview questions machine learning

power-consumption-tv-usage

Data Download: Data of TV Usage and Power Consumption (around 3 MB)

Some Thoughts

This is an supervised learning problem where the data is labelled i.e. you know from the data that when the TV is ON by checking if the power of TV consumption (second column) is larger than zero.

So the task is to find a relation between total (aggregated) power consumption noted as p and the status of TV. We can map the real values of p using the logistic functions:

tex_c774db2e90bcf816fa2466c807aaf9e1 Big Data Case Study: TV Usage and Power Consumption case study interview questions machine learning

so that the power consumption falls between 0 and 1. Then we need to train the given dataset by using the logistic regression equation:

tex_8280b3d9d73e585f241cec684eaad8d6 Big Data Case Study: TV Usage and Power Consumption case study interview questions machine learning

where tex_c27bbbb5c3c9bffc613af848e4d3eb00 Big Data Case Study: TV Usage and Power Consumption case study interview questions machine learning is the bias or intercept term and tex_73276333a6e751a6a87eabee40eecd4d Big Data Case Study: TV Usage and Power Consumption case study interview questions machine learning is the coefficient for the single input value x – which is the logistic regression value of p

Then, depending on the future value of total power consumption p’, we can predict if the TV is on if tex_84f59266f07d601a5c785ee0e24e0053 Big Data Case Study: TV Usage and Power Consumption case study interview questions machine learning or off otherwise.

One possible issue may be that the dataset size is not large enough, therefore, the model may be underfit. Another issue is that there may be power consumption due to other electric compliances, and we can filter out those noisy points when the TV is not on by setting a threshold of power consumption, which will improve the accuracy of the prediction model.

–EOF (The Ultimate Computing & Technology Blog) —

GD Star Rating
loading...
694 words
Last Post: Understanding The Math Before Investing In Stocks
Next Post: Low PHP-FPM max_children value causes Website Slow to Response + CloudFlare 524 Error

The Permanent URL is: Big Data Case Study: TV Usage and Power Consumption

Leave a Reply