AI Domains and Predictive Techniques

AI works in three main domains (Computer Vision, Natural Language Processing, Data Science) and predicts with classification, regression and clustering.

Key ideas

Computer Vision works with images and video, such as reading number plates or naming a plant from a photo.
Natural Language Processing (NLP) works with text and speech, such as translating or answering questions.
Data Science finds patterns in numbers and records, such as rainfall trends or what a shop sells most.
Classification puts things into labelled groups; regression predicts a number; clustering finds groups with no labels.
A computer sees a black-and-white picture as a grid of pixels, each 1 (black) or 0 (white).
A nearest-neighbour model labels a new item like the closest example it already knows.

Remember

Regression with a steady change	value at step k = first value + (k − 1) × change
Example	5, 8, 11, 14 cm in weeks 1–4 → week 8: 5 + 7 × 3 = 26 cm
Distance used here	difference in the first numbers + difference in the second
Two clusters	sort the numbers and split at the biggest gap
Pixels	percentage of 1s = black pixels ÷ all pixels × 100
NLP first step	count how often each word appears

Watch out

Calling “predict tomorrow’s temperature” classification: a number to predict means regression.
Using clustering when labels are given: with labels it is classification.
Using k instead of k − 1 steps when continuing a pattern from step 1.

Worked example

A clustering method groups these times (in minutes) that children took to finish a puzzle into two clusters of close values: 63, 17, 19, 69, 22, 67, 66, 21. How many numbers are in the same cluster as 63 (counting 63)?

Working: in order: 17, 19, 21, 22, 63, 66, 67, 69; the biggest gap splits {17, 19, 21, 22} from {63, 66, 67, 69}

Answer: 4