CODEWITHSUNDEEP
algorithmattributesartificial-intelligenceclassificationcluster-analysis
Spawnrider
Spawnrider

Reputation: 1773

Profiling / Categorizing algorithms to add people into interest groups

It's not very easy to describe my problem in one sentence (title). I want to find a person's interests by asking them some questions in order to assign to him attributes.

For exemple: In 10 questions (Do you love technology? Are you interested on economics? Are you more food than reading ?), I want to be able to find people's interests (Technology, Books reading, economics, ...) in order to give him attributes like technology, literature, politics, .... I also want that my program learn attributes from users answers.

I am looking for an algorithm which could help me in assigning attributes. For me, it is not a simple binary search (20 questions AI or similar) algorithm but a cluster-like AI.

Do you have any advice on this subject ?

Upvotes: 0

Views: 143

Answers (1)

Daniel
Daniel

Reputation: 754

First, classification is supervised learning while clustering is unsupervised. You can think in supervised learning like this:

I have all this groups already classified and I have a new individual/set of individuals, which group is the most suited for the individual? As you train your model (eg: by hand like marking an email as spam) your individuals are most likely to be classified correctly.

The equivalent problem but in unsupervised learning is called clustering, you have a dataset, you have no model to support on and you want to get an idea (this is best suited for exploratory analysis) on hoy your data is grouped based on some metrics (variance, mean distance between each individual on the same group, so on so forth).

Have you tried using association rule based learning?

Upvotes: 1

Related Questions