Data Mining and machine learning DM Lecture 5

  • Slides: 61
Download presentation
Data Mining (and machine learning) DM Lecture 5: Clustering David Corne, and Nick Taylor,

Data Mining (and machine learning) DM Lecture 5: Clustering David Corne, and Nick Taylor, Heriot-Watt University - [email protected] com These slides and related resources: http: //www. macs. hw. ac. uk/~dwcorne/Teaching/dmml. html

Today • (unsupervised) Clustering – – – What and why A good first step

Today • (unsupervised) Clustering – – – What and why A good first step towards understanding your data Discover patterns and structure in data Which then guides further data mining Helps to spot problems and outliers Identifies `market segments’, e. g. specific types of customers or users ; this is called segmentation or market segmentation • How to do it: – Choose distance measure or a similarity measure – Run a (usually) simple algorithm – We will cover the two main algorithms David Corne, and Nick Taylor, Heriot-Watt University - [email protected] com These slides and related resources: http: //www. macs. hw. ac. uk/~dwcorne/Teaching/dmml. html

What is Clustering? Why do it? Consider these data: Made up; maybe they are

What is Clustering? Why do it? Consider these data: Made up; maybe they are 17 subscribers to a mobile phone services company, and show the mean calls per day and the mean monthly bill for each customer Do you spot any patterns or Structure ? ? Subscriber Calls per day Monthly bill 1 1 3 2 4 7 3 4 8 4 3 5 5 6 1 6 9 3 7 3 5 8 7 2 9 4 7 10 6 3 11 2 5 12 8 4 13 5 3 14 6 2 15 2 4 16 3 6 17 8 3 David Corne, and Nick Taylor, Heriot-Watt University - [email protected] com These slides and related resources: http: //www. macs. hw. ac. uk/~dwcorne/Teaching/dmml. html

What is Clustering? Why do it? Here is a plot of the data, with

What is Clustering? Why do it? Here is a plot of the data, with calls as X and bills as Y Now do you spot any patterns or structure ? ? David Corne, and Nick Taylor, Heriot-Watt University - [email protected] com These slides and related resources: http: //www. macs. hw. ac. uk/~dwcorne/Teaching/dmml. html

What is Clustering? Why do it? Clearly there are two clusters -- two distinct

What is Clustering? Why do it? Clearly there are two clusters -- two distinct types of customer Top left: few calls but highish bills; bottom right: many calls, low bills David Corne, and Nick Taylor, Heriot-Watt University - [email protected] com These slides and related resources: http: //www. macs. hw. ac. uk/~dwcorne/Teaching/dmml. html

So, clustering is all about plotting/visualising and noting distinct groups by eye, right? Not

So, clustering is all about plotting/visualising and noting distinct groups by eye, right? Not really, because: • We can only spot patterns by eye (i. e. with our brains) if the data is 1 D, 2 D or 3 D. Most data of interest is much higher dimensional – e. g. 10 D, 20 D, 1000 D. • Sometimes the clusters are not so obvious as a bunch of data all in the same place – we will see examples. • So: we need automated algorithms which can do what you just did (find distinct groups in the data), but which can do this for any number of dimensions, and for perhaps more complex kinds of groups.

‘Clustering’ is about: - Finding the natural groupings in a data set Often called

‘Clustering’ is about: - Finding the natural groupings in a data set Often called ‘cluster analysis’; Often called (data driven) ‘segmentation’ A key tool in ‘exploratory analysis’ or ‘data exploration’ - Inspection of the results helps us learn useful things about our data – e. g. if we are doing this with supermarket baskets, each group is a collection of typical baskets, which may relate to “general housekeeping”, “late night dinner”, “quick lunchtime shopper”, and perhaps other types that we are not expecting

Quality of a clustering? Why is this: Better than this? David Corne, and Nick

Quality of a clustering? Why is this: Better than this? David Corne, and Nick Taylor, Heriot-Watt University - [email protected] com These slides and related resources: http: //www. macs. hw. ac. uk/~dwcorne/Teaching/dmml. html

Quality of a clustering? A `good clustering’ has the following properties: • Items in

Quality of a clustering? A `good clustering’ has the following properties: • Items in the same cluster tend to be close to each other • Items in different clusters tend to be far from each other It is not hard to come up with a metric – an easily calculated value – that can be used to give a score to any clustering. There are many such metrics. E. g. S = the mean distance between pairs of items in the same cluster D = the mean distance between pairs of items in different clusters Measure of cluster quality is: D/S -- the higher the better. David Corne, and Nick Taylor, Heriot-Watt University - [email protected] com These slides and related resources: http: //www. macs. hw. ac. uk/~dwcorne/Teaching/dmml. html

Let’s try that A B C D E F G H S = [AB

Let’s try that A B C D E F G H S = [AB + AD + AF + AH + BD + BF + BH + DF + DH + FH + CE + CG + EG] / 13 = 44/13 3. 38 D = [ AC + AE + AG + BC + BE + BG + DC + DE + DG + FC + FE + FG + HC + HE + HG ]/15 = 40/15 = 2. 67 Cluster Quality = D/S = 0. 77 David Corne, and Nick Taylor, Heriot-Watt University - [email protected] com These slides and related resources: http: //www. macs. hw. ac. uk/~dwcorne/Teaching/dmml. html

Let’s try that again A B C D E F G H S =

Let’s try that again A B C D E F G H S = [AB + AC + AD + BC+ BD + CD + EF + EG + EH + FG + FH + GH ] / 12 = 20/12 = 1. 67 D = [ AE + AF + AG + AH + BE + BF + BG + BH + CE + CF + CG + CH + DE + DF + DG + DH]/16 = 68/16 = 4. 25 Cluster Quality = D/S = 2. 54 David Corne, and Nick Taylor, Heriot-Watt University - [email protected] com These slides and related resources: http: //www. macs. hw. ac. uk/~dwcorne/Teaching/dmml. html

But what about this? A B C D E F G H S =

But what about this? A B C D E F G H S = [AB + CD + EF + EG + EH + FG + FH + GH ] / 8 = 12/8 = 1. 5 D = [ AC + AD + AE + AF + AG + AH + BC + BD + BE + BF + BG + BH + CE + CF + CG + CH + DE + DF + DG + DH]/20 = = 72/20= 3. 6 Cluster Quality = D/S = 2. 40 David Corne, and Nick Taylor, Heriot-Watt University - [email protected] com These slides and related resources: http: //www. macs. hw. ac. uk/~dwcorne/Teaching/dmml. html

Some important notes • There is usually no `correct’ clustering. • Clustering algorithms (whether

Some important notes • There is usually no `correct’ clustering. • Clustering algorithms (whether or not they work with cluster quality metrics) always use some kind of distance or similarity measure -- the result of the clustering process will depend on the chosen distance measure. • Choice of algorithm, and/or distance measure, will depend on the kind of cluster shapes you might expect in the data. • Our D/S measure for cluster quality will not work well in lots of cases David Corne, and Nick Taylor, Heriot-Watt University - [email protected] com These slides and related resources: http: //www. macs. hw. ac. uk/~dwcorne/Teaching/dmml. html

Examples: sometimes groups are not simple to spot, even in 2 D Slide credit:

Examples: sometimes groups are not simple to spot, even in 2 D Slide credit: Julia Handl David Corne, and Nick Taylor, Heriot-Watt University - [email protected] com These slides and related resources: http: //www. macs. hw. ac. uk/~dwcorne/Teaching/dmml. html

Examples: sometimes groups are not simple to spot, even in 2 D Slide credit:

Examples: sometimes groups are not simple to spot, even in 2 D Slide credit: Julia Handl David Corne, and Nick Taylor, Heriot-Watt University - [email protected] com These slides and related resources: http: //www. macs. hw. ac. uk/~dwcorne/Teaching/dmml. html

Brain Training Think about why D/S is not a useful cluster quality measure in

Brain Training Think about why D/S is not a useful cluster quality measure in the general case Try to design a cluster quality metric that will work well in the cases of the previous slides (not very difficult) David Corne, and Nick Taylor, Heriot-Watt University - [email protected] com These slides and related resources: http: //www. macs. hw. ac. uk/~dwcorne/Teaching/dmml. html

In many problems the clusters are more `conventional’ – but maybe fuzzy and unclear

In many problems the clusters are more `conventional’ – but maybe fuzzy and unclear Slide credit: Julia Handl David Corne, and Nick Taylor, Heriot-Watt University - [email protected] com These slides and related resources: http: //www. macs. hw. ac. uk/~dwcorne/Teaching/dmml. html

And there is a different kind of clustering that can be done, which avoids

And there is a different kind of clustering that can be done, which avoids the issue of deciding the number of clusters in advance Slide credit: Elias Raftopoulos Prof. Maria Papadopouli David Corne, and Nick Taylor, Heriot-Watt University - [email protected] com These slides and related resources: http: //www. macs. hw. ac. uk/~dwcorne/Teaching/dmml. html

How to do it • The most commonly used methods: • K-Means • Hierarchical

How to do it • The most commonly used methods: • K-Means • Hierarchical Agglomerative Clustering David Corne, and Nick Taylor, Heriot-Watt University - [email protected] com These slides and related resources: http: //www. macs. hw. ac. uk/~dwcorne/Teaching/dmml. html

K-Means • If you want to see K clusters, then run K-means. I. e.

K-Means • If you want to see K clusters, then run K-means. I. e. you need to choose in advance the number of clusters. Say K=3 -- run 3 -means and the result is a `good’ grouping of the data into 3 clusters. • It works by generating K points (in a way, these are made-up records in the data); each point is the centre (or centroid) of one cluster. As the algorithm iterates, the points adjust their positions until they stabilise. • Very simple, fairly fast, very common; a few drawbacks. David Corne, and Nick Taylor, Heriot-Watt University - [email protected] com These slides and related resources: http: //www. macs. hw. ac. uk/~dwcorne/Teaching/dmml. html

Let’s see it David Corne, and Nick Taylor, Heriot-Watt University - dwcorne@gmail. com These

Let’s see it David Corne, and Nick Taylor, Heriot-Watt University - [email protected] com These slides and related resources: http: //www. macs. hw. ac. uk/~dwcorne/Teaching/dmml. html

Here is the data; we choose k = 2 and run 2 -means David

Here is the data; we choose k = 2 and run 2 -means David Corne, and Nick Taylor, Heriot-Watt University - [email protected] com These slides and related resources: http: //www. macs. hw. ac. uk/~dwcorne/Teaching/dmml. html

We choose two cluster centres -randomly David Corne, and Nick Taylor, Heriot-Watt University -

We choose two cluster centres -randomly David Corne, and Nick Taylor, Heriot-Watt University - [email protected] com These slides and related resources: http: //www. macs. hw. ac. uk/~dwcorne/Teaching/dmml. html

Step 1: decide which cluster each point is in – the one whose centre

Step 1: decide which cluster each point is in – the one whose centre is closest David Corne, and Nick Taylor, Heriot-Watt University - [email protected] com These slides and related resources: http: //www. macs. hw. ac. uk/~dwcorne/Teaching/dmml. html

Step 2: We now have two clusters – recompute the centre of each cluster

Step 2: We now have two clusters – recompute the centre of each cluster David Corne, and Nick Taylor, Heriot-Watt University - [email protected] com These slides and related resources: http: //www. macs. hw. ac. uk/~dwcorne/Teaching/dmml. html

These are the new centres David Corne, and Nick Taylor, Heriot-Watt University - dwcorne@gmail.

These are the new centres David Corne, and Nick Taylor, Heriot-Watt University - [email protected] com These slides and related resources: http: //www. macs. hw. ac. uk/~dwcorne/Teaching/dmml. html

Step 1: decide which cluster each point is in – the one whose centre

Step 1: decide which cluster each point is in – the one whose centre is closest David Corne, and Nick Taylor, Heriot-Watt University - [email protected] com These slides and related resources: http: //www. macs. hw. ac. uk/~dwcorne/Teaching/dmml. html

This one has to be reassigned: David Corne, and Nick Taylor, Heriot-Watt University -

This one has to be reassigned: David Corne, and Nick Taylor, Heriot-Watt University - [email protected] com These slides and related resources: http: //www. macs. hw. ac. uk/~dwcorne/Teaching/dmml. html

Step 2: We now have two new clusters – recompute the centre of each

Step 2: We now have two new clusters – recompute the centre of each cluster David Corne, and Nick Taylor, Heriot-Watt University - [email protected] com These slides and related resources: http: //www. macs. hw. ac. uk/~dwcorne/Teaching/dmml. html

Centres now slightly moved David Corne, and Nick Taylor, Heriot-Watt University - dwcorne@gmail. com

Centres now slightly moved David Corne, and Nick Taylor, Heriot-Watt University - [email protected] com These slides and related resources: http: //www. macs. hw. ac. uk/~dwcorne/Teaching/dmml. html

Step 1: decide which cluster each point is in – the one whose centre

Step 1: decide which cluster each point is in – the one whose centre is closest David Corne, and Nick Taylor, Heriot-Watt University - [email protected] com These slides and related resources: http: //www. macs. hw. ac. uk/~dwcorne/Teaching/dmml. html

In this case, nothing gets reassigned to a new cluster – so the algorithm

In this case, nothing gets reassigned to a new cluster – so the algorithm is finished David Corne, and Nick Taylor, Heriot-Watt University - [email protected] com These slides and related resources: http: //www. macs. hw. ac. uk/~dwcorne/Teaching/dmml. html

The K-Means Algorithm • Choose k 1. Randomly choose k points, labelled {1, 2,

The K-Means Algorithm • Choose k 1. Randomly choose k points, labelled {1, 2, …, k} to be the initial cluster centroids. 2. For each datum, let its cluster ID be the label of its closest centroid. 3. For each cluster, recalculate its actual centre. 4. Go back to Step 2, stop when step 2 does not change the cluster ID of any point David Corne, and Nick Taylor, Heriot-Watt University - [email protected] com These slides and related resources: http: //www. macs. hw. ac. uk/~dwcorne/Teaching/dmml. html

Simple but often not ideal • Variable results with noisy data and outliers •

Simple but often not ideal • Variable results with noisy data and outliers • Very large or very small values can skew the centroid positions, and give poor clusterings • Only suitable for the cases where we can expect clusters to be `clumps’ that are close together – e. g. terrible in the two-spirals and similar cases David Corne, and Nick Taylor, Heriot-Watt University - [email protected] com These slides and related resources: http: //www. macs. hw. ac. uk/~dwcorne/Teaching/dmml. html

Hierarchical Agglomerative Clustering Before we discuss this: • We need to know how to

Hierarchical Agglomerative Clustering Before we discuss this: • We need to know how to work out the distance between two points: David Corne, and Nick Taylor, Heriot-Watt University - [email protected] com These slides and related resources: http: //www. macs. hw. ac. uk/~dwcorne/Teaching/dmml. html

Hierarchical Agglomerative Clustering Before we discuss this: • And the distance between a point

Hierarchical Agglomerative Clustering Before we discuss this: • And the distance between a point and a cluster ? David Corne, and Nick Taylor, Heriot-Watt University - [email protected] com These slides and related resources: http: //www. macs. hw. ac. uk/~dwcorne/Teaching/dmml. html

Hierarchical Agglomerative Clustering Before we discuss this: • And the distance between two clusters

Hierarchical Agglomerative Clustering Before we discuss this: • And the distance between two clusters ? David Corne, and Nick Taylor, Heriot-Watt University - [email protected] com These slides and related resources: http: //www. macs. hw. ac. uk/~dwcorne/Teaching/dmml. html

Hierarchical Agglomerative Clustering Before we discuss this: • There are many options for all

Hierarchical Agglomerative Clustering Before we discuss this: • There are many options for all of these things; we will discuss them in a later lecture. ? David Corne, and Nick Taylor, Heriot-Watt University - [email protected] com These slides and related resources: http: //www. macs. hw. ac. uk/~dwcorne/Teaching/dmml. html

Hierarchical Agglomerative Clustering • Is very commonly used • Very different from K-means •

Hierarchical Agglomerative Clustering • Is very commonly used • Very different from K-means • Provides a much richer structuring of the data • No need to choose k • But, quite sensitive to the various ways of working out distance (different results for different ways) David Corne, and Nick Taylor, Heriot-Watt University - [email protected] com These slides and related resources: http: //www. macs. hw. ac. uk/~dwcorne/Teaching/dmml. html

Let’s see it 1 5 3 7 9 4 8 2 6 David Corne,

Let’s see it 1 5 3 7 9 4 8 2 6 David Corne, and Nick Taylor, Heriot-Watt University - [email protected] com These slides and related resources: http: //www. macs. hw. ac. uk/~dwcorne/Teaching/dmml. html

Initially, each point is a cluster 1 5 3 7 9 4 8 2

Initially, each point is a cluster 1 5 3 7 9 4 8 2 6 David Corne, and Nick Taylor, Heriot-Watt University - [email protected] com These slides and related resources: http: //www. macs. hw. ac. uk/~dwcorne/Teaching/dmml. html

Find closest pair of clusters, and merge them into one cluster 1 5 3

Find closest pair of clusters, and merge them into one cluster 1 5 3 7 9 4 8 2 6 7 9 David Corne, and Nick Taylor, Heriot-Watt University - [email protected] com These slides and related resources: http: //www. macs. hw. ac. uk/~dwcorne/Teaching/dmml. html

Find closest pair of clusters, and merge them into one cluster 1 5 3

Find closest pair of clusters, and merge them into one cluster 1 5 3 7 9 4 8 2 6 7 93 4 David Corne, and Nick Taylor, Heriot-Watt University - [email protected] com These slides and related resources: http: //www. macs. hw. ac. uk/~dwcorne/Teaching/dmml. html

Find closest pair of clusters, and merge them into one cluster 1 5 3

Find closest pair of clusters, and merge them into one cluster 1 5 3 7 9 4 8 2 6 8 7 93 4 David Corne, and Nick Taylor, Heriot-Watt University - [email protected] com These slides and related resources: http: //www. macs. hw. ac. uk/~dwcorne/Teaching/dmml. html

Find closest pair of clusters, and merge them into one cluster 1 5 3

Find closest pair of clusters, and merge them into one cluster 1 5 3 7 9 4 8 2 6 8 7 93 4 2 David Corne, and Nick Taylor, Heriot-Watt University - [email protected] com These slides and related resources: http: //www. macs. hw. ac. uk/~dwcorne/Teaching/dmml. html

Find closest pair of clusters, and merge them into one cluster 1 5 3

Find closest pair of clusters, and merge them into one cluster 1 5 3 7 9 4 8 2 6 8 7 93 4 2 1 David Corne, and Nick Taylor, Heriot-Watt University - [email protected] com These slides and related resources: http: //www. macs. hw. ac. uk/~dwcorne/Teaching/dmml. html

Find closest pair of clusters, and merge them into one cluster 1 5 3

Find closest pair of clusters, and merge them into one cluster 1 5 3 7 9 4 8 2 6 8 7 93 4 2 1 5 David Corne, and Nick Taylor, Heriot-Watt University - [email protected] com These slides and related resources: http: //www. macs. hw. ac. uk/~dwcorne/Teaching/dmml. html

Find closest pair of clusters, and merge them into one cluster 1 5 3

Find closest pair of clusters, and merge them into one cluster 1 5 3 7 9 4 8 2 6 8 7 93 4 2 1 5 6 David Corne, and Nick Taylor, Heriot-Watt University - [email protected] com These slides and related resources: http: //www. macs. hw. ac. uk/~dwcorne/Teaching/dmml. html

Find closest pair of clusters, and merge them into one cluster 1 5 3

Find closest pair of clusters, and merge them into one cluster 1 5 3 7 9 4 8 2 6 8 7 93 4 2 1 5 6 David Corne, and Nick Taylor, Heriot-Watt University - [email protected] com These slides and related resources: http: //www. macs. hw. ac. uk/~dwcorne/Teaching/dmml. html

Now all one cluster, so stop 1 5 3 7 9 4 8 2

Now all one cluster, so stop 1 5 3 7 9 4 8 2 6 8 7 93 4 2 1 5 6 David Corne, and Nick Taylor, Heriot-Watt University - [email protected] com These slides and related resources: http: //www. macs. hw. ac. uk/~dwcorne/Teaching/dmml. html

The thing on the right is a dendrogram – it contains the information for

The thing on the right is a dendrogram – it contains the information for us to group the data into clusters in various ways 1 5 3 7 9 4 8 2 6 8 7 93 4 2 1 5 6 David Corne, and Nick Taylor, Heriot-Watt University - [email protected] com These slides and related resources: http: //www. macs. hw. ac. uk/~dwcorne/Teaching/dmml. html

E. g. 2 clusters 1 5 3 7 9 4 8 2 6 8

E. g. 2 clusters 1 5 3 7 9 4 8 2 6 8 7 93 4 2 1 5 6 David Corne, and Nick Taylor, Heriot-Watt University - [email protected] com These slides and related resources: http: //www. macs. hw. ac. uk/~dwcorne/Teaching/dmml. html

E. g. 3 clusters 1 5 3 7 9 4 8 2 6 8

E. g. 3 clusters 1 5 3 7 9 4 8 2 6 8 7 93 4 2 1 5 6 David Corne, and Nick Taylor, Heriot-Watt University - [email protected] com These slides and related resources: http: //www. macs. hw. ac. uk/~dwcorne/Teaching/dmml. html

In a proper dendrogram … • The height of a bar indicates how different

In a proper dendrogram … • The height of a bar indicates how different the items are • A dendrogram is also called a binary tree • The data points are the leaves of the tree • Each node represents a cluster – all the leaves of its subtree David Corne, and Nick Taylor, Heriot-Watt University - [email protected] com These slides and related resources: http: //www. macs. hw. ac. uk/~dwcorne/Teaching/dmml. html

The Agglomerative Hierarchical Clustering Algorithm • Decide on how to work out distance between

The Agglomerative Hierarchical Clustering Algorithm • Decide on how to work out distance between two clusters • Initialise: each of the N data items is a cluster • Repeat N-1 times: – Find the closest pair of clusters; merge them into a single cluster (and update your tree representation) David Corne, and Nick Taylor, Heriot-Watt University - [email protected] com These slides and related resources: http: //www. macs. hw. ac. uk/~dwcorne/Teaching/dmml. html

Another common alternative …

Another common alternative …

Another common alternative …

Another common alternative …

Another common alternative … The ‘DBSCAN’ algorithm https: //en. wikipedia. org/wiki/DBSCAN Watch for updates

Another common alternative … The ‘DBSCAN’ algorithm https: //en. wikipedia. org/wiki/DBSCAN Watch for updates …

Next time • Naïve Bayes, and CW 2

Next time • Naïve Bayes, and CW 2