Spatial Analysis

• Clustering algorithms divides a dataset into natural groups. Instances in the same cluster are similar to each other, they share certain properties.
• Cluster analysis is used to reduce data (dimension) to a number of multivariate classes.

• Partitioning: divide the data into a given number of clusters. (it uses k-means)
• Hierarchical: create a tree based on the similiraties/ distance of items.
• Density-based: find continguous areas with high density.
• Grid-based: divide the data space into grid cells.

We try to make properties as similar as possible within groups and as different as possible between groups.

• K-medoids: use the centermost cluster member instead of the mean.
• EM: define the cluster by a probability distribution instead of a centroid.
• Guaranteed to find a local optimum.
• Not globally optimal clustering.
• Tend to find circular clusters.

You have to use your common sense and your skill to do the clustering.

• For finding non-spherical clusters.
• Clusters do not necessarily cover all points.

• Goal: find areas with high point desity.
• divide the space to grid cells.
• compute the density of points in each cell.
• use the high-density cells to define clusters.
• more effective than density-based clustering.

• 
• Time is continuous.

A process of dividing a market into homogeneous subgroups. Tourists in the same group are similar to each other, and different from other groups.

• Identifying and characterising groups of tourists.
• Focusing advertising efforts for greater impacts.
• Identifying likely targerts for new tourist products.
• Improving existing products or services.
• Looking for new product services.
• Assessing the impact of a competitor's new offering.
• Establishing a better tourst attraction image.

• Primarily catergorical variables.
• Computing the number of clusters.
• Tending to be a samll number of significant clusters and a number of othes that are not interesting.
• Each tourist is assigned into a certain cluster with certain probability, which govern the set of attribute values of observations in the cluster.