Next, you need to establish whether the information gathered is suitable for the intended audience.
If we have gathered a lot of data, it is unlikely that our analysis will utilise the whole data set. We need to consider the scope of our analysis and what data is required. This is known as feature extraction.
For example, what feature would you use to select the buttons depicted on the left to sew on the shirts depicted on the right?
What feature would you use to select the buttons to sew onto these shirts? Would you still use the same feature?
If you sign up for the Coventry University program ‘AI Technologies for Business and Management’, the next short course (Fundamental Machine Learning for AI) will illustrate how to apply feature extraction to your analysis.
Consider the following extract from the records of a farm:
Which features would you use to analyse the following aspects?
- Average weight of a breed
- Average milk production of a breed
- Most common names for different breeds
- Correlation between weight and daily production
- Correlation between colour and milk production
- Total farm output
- Average output per cow
As you can see, different analyses require different subsets, or features, of the data.
© Coventry University. CC BY-NC 4.0